Field release of genetically modified Pseudomonas putida WCS358r

Transcription

1 Field release of genetically modified Pseudomonas putida WCS358r Molecular analysis of effects on microbial communities in the rhizosphere of wheat Introductie van genetisch gemodificeerde Pseudomonas putida WCS358r in het veld Moleculaire analyse van effecten op de microbiële gemeenschappen in de rhizosfeer van tarwe (Met een samenvatting in het Nederlands en in het Duits) Proefschr ift Ter verkrijging van de graad van doctor aan de Universiteit Utrecht op gezag van de Rector magnificus, Prof. Dr. W.H. Gispen, ingevolge het besluit van het College voor Promoties in het openbaar te verdedigen op dinsdag 26 april 2005 des middags te 12:45 door Mareike Viebahn geboren op 19 juni 1958 te Keulen

2 Promotor: prof. dr. ir. L.C. van Loon Section of Phytopathology, Faculty of Biology, Utrecht University Co-promotoren: dr. P.A.H.M. Bakker Section of Phytopathology, Faculty of Biology, Utrecht University dr. E. Smit National Institute for Public Health and the Environment, Bilthoven Print: Layout and cover: PrintPartners IPSKAMP B.V., Enschede M. Kortbeek-Smithuis The research described in this thesis was performed at the Section of Phytopathology, Faculty of Biology, Utrecht University (Sorbonnelaan 16, 3584 CA Utrecht, The Netherlands) and at the National Institute for Public Health and the Environment (RIVM, Antonie van Leeuwenhoeklaan 9, 3720 BA Bilthoven, The Netherlands). The reproduction of the color prints was partly sponsored by BD Biosciences-Clontech. ISBN

3 Sometimes you just have to take the leap, and build your wings on the way down Kobi Yamada voor James en voor mijn ouders

4

5 C o n t e n t s Chapter 1 General introduction and outline 9 Chapter 2 Repeated introduction of genetically modified 25 Pseudomonas putida WCS358r without intensified effects on the indigenous microflora Chapter 3 Response of bacterial communities in the rhizosphere 45 of field-grown wheat to repeated introductions of genetically modified Pseudomonas putida WCS358r Chapter 4 Assessment of differences in ascomycete communities 59 in the rhizosphere of field-grown wheat and potato Chapter 5 Ascomycete communities in the rhizosphere of 75 field-grown wheat are not affected by introductions of genetically modified Pseudomonas putida WCS358r Chapter 6 Microarray analysis and suppression subtractive 93 hybridization to identify shifts in rhizosphere bacterial communities Chapter 7 General discussion 113 References 125 Summary 139 Samenvatting 143 Zusammenfassung 147 Dankwoord 153 List of publications 157 Curriculum vitae 159

6

7 List of abbreviations ABC ATP-binding cassette AMF arbuscular mycorrhizal fungi ANOVA analysis of variances ARDRA amplified ribosomal DNA restriction analysis ATP adenosintriphospate BCA biological control agent 3CB 3-chlorobenzoate CCRO Coordination Commission Risk Research cfu colony forming units cv cultivar DAPG 2,4-diacetylphloroglucinol datp 2 -deoxyadenosine 5 -triphosphate dctp 2 -deoxycytidine 5 -triphosphate dgtp 2 -deoxyguanosine 5 -triphosphate DNA deoxyribonuleic acid dttp 2 -deoxythymidine 5 -triphosphate DGGE denaturing gradient gel electrophoresis Ggt Gaeumannomyces graminis var. tritici GMM genetically modified microorganism ITS internal transcribed spacer KB King s medium B km kanamycin KNMI Royal Netherlands Meteorological Institute M reference marker 4MB 4-methyl benzoate MM mismatch NPA nitrifying potential activity OTU operational taxonomic unit P probability PCA phenazine-1-carboxylic acid PCR polymerase chain reaction PDA potato dextrose agar PGPR plant growth-promoting rhizobacetrium PM perfect match RFLP restriction fragment length polymorphism rif rifampicin SSCP single-stranded confirmation polymorphism SSH suppression subtractive hybridization SSU small subunit TAD take-all decline T-RFLP terminal restriction fragment length polymorphism TSM Trichoderma-selective medium U unit UPGMA unweighted pair-group method using average linkages

8

9 Chapter 1 General introduction and outline B i o l o g i c a l c o n t r o l o f p l a n t d i s e a s e s Losses of food and feed crops to plant diseases are common. The amount of a single crop lost to plant diseases worldwide can be as high as 22% (Agrios, 1997). Control measures are, for instance, the breeding of resistant varieties, crop rotation, and application of chemical pesticides. However, not all plant pathogens can be controlled in such ways. The appearance of resistant pathogens, the banning of several chemical pesticides from the market, and public concern have led to the development of alternative strategies. As a result, biological control of plant pathogens is becoming increasingly important. Biological control is the reduction of the amount of inoculum or disease-producing activity of a pathogen accomplished by one or more organisms other than man (Cook and Baker, 1983). Mechanisms of biological control include antibiosis (Handelsman and Stabb, 1996; Weller et al., 2002), competition for nutrients and space, competition for iron through the production of siderophores (Bakker et al., 1988), the production of lytic enzymes such as cellulases, chitinases, glucanases,

10 1 0 C h a p t e r 1 and proteases, and induced systemic resistance (Pieterse et al., 2001; Van Loon et al., 1998a). Biological control has been developed based on the occurrence of natural soils that are suppressive to diseases. In such disease-suppressive soils resident non-pathogenic microorganisms have been found to be responsible for the condition that a plant pathogen cannot establish itself or cannot cause disease. General suppressiveness relies on the total microbial biomass and is not transferable to non-suppressive soils. In contrast, specific suppressiveness depends on selected groups of microorganisms (Van Loon, 1998a). It has been demonstrated that specific suppressiveness can be eliminated by heat or certain chemical treatments of the soil, and that it can be transferred from one field to another (Weller et al., 2002). One of the earliest examples of a biological control agent released into the field was Trichoderma harzianum, which effectively controlled root and stem rot of tomato caused by Sclerotium rolfsii (Wells et al., 1972). The best understood example of biological control is take-all decline (TAD) in wheat. Take-all is caused by the ascomycetous pathogen Gaeumannomyces graminis var. tritici (Ggt). After several years of crop monoculture and at least one severe outbreak of take-all, the disease severity decreases spontaneously in subsequent years. Take-all has been studied intensively because of the economic importance of wheat, the difficulties in breeding for resistance against Ggt, and the limited chemical control of the pathogen. Raaijmakers et al. (1997) provided strong evidence that antibiotic-producing pseudomonads were the key factor in TADsoils. The decrease or prevention of deleterious effects caused by phytopathogenic organisms often results in enhanced plant growth. Microorganisms that promote plant growth are called plant growth-promoting rhizobacteria (PGPR), and may act either through a direct stimulation of plant growth or through suppression of deleterious microorganisms. They are associated with the roots of many plant species (Cook et al., 1995; Glick et al., 1999), and comprise genera such as Azospirillum, Azotobacter, Bacillus, Burkholderia, Enterobacter, Erwinia, Pseudomonas and Streptomyces (Glick et al., 1999; Weller, 1988). Studies in biological control have focused mainly on fluorescent pseudomonads, because of their metabolic versatility, their excellent root colonization ability, and their capacity to produce a wide range of antimicrobial metabolites (O Sullivan and O Gara, 1992). These secondary metabolites include, among others, 2,4-diacetylphloroglucinol (DAPG), phenazines, pyoluteorin, and pyrrolnitrin. The best-studied antibiotics with respect to biological control are DAPG and phenazine-1-carboxylic acid (PCA) (Fig. 1). DAPG, a phenolic compound, is produced mainly by Pseudomonas species and has antifungal, antibacterial, antiviral, antihelminthic, and phytotoxic activities (Raaijmakers and Weller, 1998; Thomashow and Weller, 1988). Its mode of

11 G e n e r a l i n t r o d u c t i o n 1 1 COOH O OH O N CH 3 CH 3 N OH OH PCA DAPG Fig. 1. Chemical structure of the antibiotics phenazine-1-carboxylic acid (PCA) and 2,4- diacetylphloroglucinol (DAPG). PCA is naturally produced by P. fluorescens 2-79, DAPG is naturally produced by P. fluorescens Q2-87. action is largely unclear. Phenazines are pigmented, nitrogen-containing, heterocyclic compounds that are produced by different bacterial species such as Brevibacterium, Burkholderia, Pseudomonas, and Streptomyces. Currently, more than 50 naturally occurring phenazines have been described, and as many as ten different types can be synthesized by one organism. Almost all phenazines exhibit antagonistic activity against a wide range of bacteria and fungi. PCA is the common precursor of all other phenazines. Despite numerous studies on candidate biological control agents and their modes of action, only a very limited number has been developed for commercial applications. One possible explanation is their inconsistent performance under field conditions. Reasons for that, among others, are (i) the biological control agent has lost its ecological competence; (ii) root colonization is variable; and (iii) the antibiotic production, when required, occurs too late or amounts are insufficient (Lugtenberg et al., 2001; Weller, 1988). Production of many bacterial antibiotics is regulated by quorum sensing. Quorum sensing is a cell-density dependent gene regulation process, which allows microorganisms to express certain genes only when they reach high population densities (von Bodman et al., 2003). The performance of biological control may be improved by avoiding quorum sensing. Microorganisms genetically modified to produce antibiotics constitutively, thus being independent of environmental variations, can enhance the efficacy of biological control agents in the field. G e n e t i c a l l y m o d i f i e d b a c t e r i a f o r a g r i c u l t u r a l p u r p o s e s Current knowledge on microbial genetics allows us to construct strains with novel capabilities. Such genetically modified microorganisms (GMMs) can be applied in agriculture to improve crop production (Amarger, 2002). An example of genetic engineering is the modification of the DAPG-gene cluster in Pseudomonas fluorescens CHA0, which led to overproduction

12 1 2 C h a p t e r 1 of this antibiotic (Girlanda et al., 2001). The recombinant strain proved to suppress Pythium spp. on cucumber more efficiently than the wild type. GMMs could also be employed for bioremediation, to degrade polluting compounds in the environment (Dutta et al., 2003), and as biosensors to determine concentrations of pollutants in soil and water (Belkin, 2003). Constraints currently preventing large-scale environmental applications of GMMs are not only the public concern of possible negative effects on ecosystems, but also our lack of knowledge on survival, dispersal, and transfer of genetic material to indigenous microorganisms, as well as effects of the introduced strains on microbial biodiversity (Leung et al., 1994). M i c r o b i a l b i o d i v e r s i t y In the past, researchers considered describing the composition of microbial communities an intractable problem. The enormous and largely unknown diversity and heterogeneity of bacteria and fungi in soil and their refract ability to cultivation have certainly contributed to this notion. They also have the ability to create new diversity through mutation and exchange of genetic information within short time periods. Microbial biodiversity, in other words, the variety of different types of microorganisms occurring together in a biological community (Atlas and Bartha, 1998), can be immense. A typical soil community usually contains 10 8 to 10 9 bacterial cells per gram dry weight of soil. For a soil in Norway, Torsvik et al. (1990) measured the heterogeneity of extracted DNA as a measure of the total number of genetically different bacteria. They concluded that the bacterial DNA isolated from the soil was very heterogeneous and corresponded to about 4000 completely different genomes of soil bacteria. The diversity differs widely depending on the habitat. Curtis et al. (2001) estimated 163 different prokaryotic taxa in seawater, based on the extrapolation of species-abundance curves. In soil the estimated taxa were The structural complexity of soil and sediments and their steep gradients in substrate concentrations, redox potential, and ph contribute to the formation of microhabitats favoring the diversification of microorganisms. Also fungi are an important part of the microbial communities. Hawksworth and Rossman (1997) estimated that about 1,5 million fungal species exist, of which only 5% have been described. These numbers are derived by extrapolation of data on well-known fungi from well-studied regions and host plants. After the introduction of molecular techniques such as DNA/DNA hybridization, polymerase chain reaction (PCR), ribosomal (r)rna sequencing, and DNA re-association kinetics, the question about the microbial diversity has been addressed by many research groups. Knowledge about the microbial community composition is valuable in many respects. It can lead to the development of more efficacious biological control agents, as well as an

13 G e n e r a l i n t r o d u c t i o n 1 3 improved efficacy of waste water treatment systems, and enhancement of feed efficiency of ruminant animals (Tiedje, 1995). As stated by Torsvik et al. (1996), knowledge about microbial community structure and diversity is necessary to understand the relationship between environmental factors and ecosystem functioning. Such knowledge can then be used to assess effects of environmental stress and perturbation, such as pollution, agricultural exploitation, or the introduction of biological control agents on ecosystems. S u r v i v a l o f g e n e t i c a l l y m o d i f i e d b a c t e r i a Cells introduced into the environment will encounter both biotic and abiotic stresses that affect their survival. The fate of bacteria is determined by an intricate interplay between the physiological state of the bacteria and the environmental conditions. As a reaction to these conditions bacteria can revert to different physiological states. From being in a normal culturable state, cells can become more stress resistant or form dwarf cells (van Overbeek et al., 1995), they can produce exopolysacharides for protection (Ophir and Gutnick, 1994), they can enter a non-culturable but viable state (Colwell and Grimes, 2000), form spores, and some are able to establish associations with plants. Soil factors such as a high clay content, high ph, relatively high humidity and the presence of plant roots generally have a positive effect on bacterial survival (Van Elsas et al., 1986). Factors that have a negative effect are dry periods, the presence of a competing microflora, predation by protozoa and nematodes, and lysis by bacteriophages (Ashelford et al., 2000; Eberl et al., 1997; Smit et al., 1996). In general, applications of bacteria in agriculture involve microorganisms that are naturally associated with crop plants and particularly those that colonize plant roots. Most PGPR are well adapted to the rhizosphere. While most reports on the survival of Pseudomonas spp. in soil demonstrated a fairly rapid decline, populations on plant roots tend to increase in numbers (Bailey et al., 2000; Bashan et al., 1995; Lugtenberg et al., 2001; Raaijmakers and Weller, 1998; Raaijmakers and Weller, 2001; Rosado et al., 1996). It seems reasonable to assume that genetically modified bacteria will survive in a similar fashion as their wild-type parents. However, expression of the inserted genes requires additional energy, which may reduce their environmental fitness (Lenski, 1993). Moreover, the genetic modification may disrupt unknown functions, debilitating the strain s competitiveness. There are several studies, in which no difference in survival between GMMs and their parent strains was detected. In the late 1980s the first field release of a laczy-engineered strain of Pseudomonas aureofaciens 3732 RN- L11 took place (Kluepfel, 1993). Both parental and transgenic strains were monitored over three cropping cycles. On a first wheat crop the population

14 1 4 C h a p t e r 1 density rose to approximately colony forming units (cfu) g -1 root ten days after inoculation. Nine months after harvest, the population had decreased below the detection limit. On the following crop, soybeans, P. aureofaciens L11 reached a density of less than 100 cfu g -1 root, and on the third crop, again winter wheat, L11 was not detected at all any more. Gagliardi et al. (2001) likewise did not find any difference in survival between wild-type parent and genetically modified Pseudomonas chlororaphis and P. fluorescens in five different soils. The laczy genes were inserted into the chromosome of both GMMs. Schwieger et al. (2000) investigated survival of two isogenic, luciferase marker gene (luc)-tagged Sinorhizobium meliloti strains, a recombinant-intact L33 (reca + ) and a recombinant-defective L1 (reca - ) strain, in field lysimeters. Both strains declined from 10 6 to 10 4 cfu g -1 soil in lysimeters seeded with alfalfa in the first year. However, the reca + strain survived significantly better. Pseudomonas putida WCS358r modified to produce different amounts of PCA or DAPG declined at the same rate as the wild type after introduction as a seed coating into the rhizosphere of wheat (Glandorf et al., 2001; Viebahn et al., 2003). In other studies, GMMs survived less well than their non-modified parent strains (Brockman et al, 1991, Van Elsas et al., 1991; Wang et al., 1991). Results from De Leij and co-workers (De Leij et al., 1998) showed that the presence of a number of constitutively expressed marker genes (laczy, aph-1, xyle), inserted singly or in combination in P. fluorescens SBW25, had a negative effect on survival in competition with the wild-type strain. The site of insertion into the chromosome did not affect the outcome. Since the study indicated that this effect did not occur under nutrient-rich conditions, it seems that it was purely the metabolic load that was responsible for the decreased fitness observed. The method of detection is of crucial importance for interpreting bacterial survival data, since cells that have become non-culturable could escape detection by cultivation-based methods. In several studies GMMs introduced into soil were shown to have become non-culturable (England et al., 1995; Kluepfel, 1993). The ecological significance of the presence of viable but nonculturable cells detected with molecular techniques remains largely unsolved and will definitely require further investigation. E c o s y s t e m e f f e c t s o f g e n e t i c a l l y m o d i f i e d b i o c o n t r o l b a c t e r i a Possible effects of the introduction of GMMs on the natural microbial ecosystem range from perturbations such as the input of organic substrate, displacement of species and changes in population structure, to possible loss of certain functions (Smit et al., 1992). Detection of such effects may be

15 G e n e r a l i n t r o d u c t i o n 1 5 difficult, and the relationship between microbial diversity and ecosystem functioning is not well understood. A number of biological control agents have been genetically modified to enhance their biocontrol properties, including the increase of the levels of bioactive metabolites and the extension of the metabolic repertoire by the insertion of novel genes. Quite a number of studies describe survival or target effects, and the potential use of genetically modified biocontrol strains (Jones et al., 1991; Moënne-Loccoz et al., 2001; Van Dillewijn et al., 2001; Völksch and May, 2001; Wilson et al., 2002b). However, only a few studies focus on the effects on the indigenous microflora (De Leij et al., 1995; Girlanda et al., 2001), and only some describe field introductions of biocontrol agents in a natural environment (Glandorf et al., 2001; Schwieger and Tebbe, 2000; Viebahn et al., 2003; Winding et al., 2004). Studying the effect of biocontrol bacteria in in vitro assays might be useful but results in vitro cannot be translated to predict their impact in the field. Johansen and co-workers (Johansen et al., 2002) discovered that a large percentage of a collection of Cytophaga-like and Pseudomonas isolates was sensitive to the biocontrol strain P. fluorescens CHA0 in in vitro assays. However, in soil even the most sensitive isolate was not affected by the biocontrol strain. Therefore, field tests are necessary to evaluate the effectiveness of biocontrol agents. Natsch et al. (1997) studied the effect of a genetically modified derivative of P. fluorescens CHA0 on the diversity of resident pseudomonads in the rhizosphere of cucumber. The modification consisted of the insertion of an extra copy of a housekeeping gene encoding sigma factor σ 70, resulting in increased production of the antibiotics DAPG and pyoluteorin. Several days after inoculation both the parent strain and the GMM reduced the number of resident pseudomonads in the rhizosphere, but the impact of the modified strain was more significant, and persisted for more than a month. However, the composition of the Pseudomonas population was more strongly affected by the developmental stage of the cucumber roots than by the bacterial inoculants, and the effects of both the modified and non-modified strains appeared to be small and transient. In microcosm studies Delany et al. (2001) compared the naturally DAPGproducing P. fluorescens F113 with two GMMs modified to produce elevated levels of DAPG. All strains inhibited growth of Pythium ultimum in vitro, and when introduction into the rhizosphere of sugar beet plants the GMMs showed enhanced control of Pythium damping-off. Non-target effects of genetically modified Pseudomonas were described by Glandorf et al. (2001). P. putida WCS358r was genetically modified with the biosynthetic locus phz from P. fluorescens 2-79, resulting in constitutive PCA production. Two derivatives, one with a low and one with a high level of

16 1 6 C h a p t e r 1 PCA production, were selected for study in two small-scale field experiments during two subsequent years (Glandorf et al., 2001; Leeflang et al., 2002). The strains were introduced as a coating on wheat seeds at a density of 10 7 cells per seed. Over the first three months after sowing, the strains decreased below the detection limit. Monitoring of various soil ecosystem functions, such as substrate-induced soil respiration, cellulose decomposition, and nitrification potential activity did not reveal effects of the introduction of any of the strains. Effects of the GMM that produced the highest amount of PCA had a transient effect on the culturable fungal microflora. The total fungal microflora, including non-culturables, was analyzed using 18S rdna amplified ribosomal DNA restriction analysis (ARDRA). Introduction of both the wild type and the GMMs changed the composition of the total fungal rhizosphere microflora transiently. Notably, the effects of the genetically modified strains were distinct from those of the parental strain and persisted for a longer time. A different approach is studying the impact of GMMs on specific ecologically important non-target microbial taxa. Enhanced antibiotic production by pseudomonads could result in adverse effects on e.g. mycorrhizae. Mycorrhizae are stable, mutualistic associations between a fungus and the root of a plant (van der Heijden, et al.,2002). The fungus improves the uptake of nutrients by the host plants and protects the plants against drought and root pathogens, whereas the plant provides the fungus with simple sugars and an appropriate environment. Barea et al. (1998) studied the impact of GMMs on mycorrhizae in experiments with Pseudomonas sp. F113 and two genetically modified derivatives. Wild type F113, a natural DAPG-producer, and two derivatives that either produced no DAPG, or had enhanced DAPG production, were used. Colonization of tomato roots by the arbuscular mycorrhizal fungus Glomus mossae, an endomycorrhiza living mainly inside the plant roots, was not affected by any of the bacterial treatments. Moreover, enhanced DAPG production by the bacterium did not inhibit spore germination or mycelium development of the fungus. Apparently, production of DAPG by Pseudomonas sp. F113 does not affect G. mossae. From the forgoing it appears that non-target effects of deliberate introductions of (genetically modified) microbial inoculants are transient and relatively small. Moreover, they are most apparent using cultivation-independent methodology. During the last decades molecular tools have been developed that not only aid in studying the total microflora, but also led to the discovery of new groups of microorganisms. In the next section the different molecular approaches available for microbial ecological studies will be described and evaluated.

17 G e n e r a l i n t r o d u c t i o n 1 7 M o l e c u l a r a p p r o a c h e s u s e d i n m i c r o b i a l e c o l o g y Previously, microbiologists were limited in their ability to characterize non-culturable microorganisms. This was due partly to a lack of knowledge about complex nutrient requirements of specific microorganisms, and partly to missing sophisticated methodologies to identify microorganisms. It is estimated that more than 99% of all microbial organisms are not yet culturable (Atlas and Bartha, 1998; Hugenholtz and Pace, 1996). Only since the advent of molecular techniques, researchers are capable of determining the composition of microbial communities, including the non-culturable microflora. The foundation was laid by Carl Woese and Norman Pace (Pace et al., 1986; Woese, 1987). They determined and compared the DNA sequences of the small subunit (SSU) rrna genes of many different microorganisms to understand their phylogenetic relationships (Pace et al., 1986; Woese, 1987). These SSU rdna molecules are particularly suitable for studies on microbial communities, because (i) they are universal in Bacteria, Archaea, and Eukarya, (ii) they contain both conserved and variable regions, which provide target sites for general and specific primers, (iii) they are large enough to provide specific information, and (iv) they are easily amplified and sequenced (Woese et al., 1990; Woese, 1987). Thus, a suitable cultivation-independent method is the construction and analysis of clone libraries containing SSU rdna, which can provide detailed phylogenetic information about members of the resident microbial communities (Kent and Triplett, 2002). However, when spatial and temporal changes in microbial community structure are addressed, the analysis of numerous samples becomes cost- and time-intensive. Suitable alternatives are fingerprinting techniques, which provide banding patterns after electrophoretic separation of nucleic acids, and which allow comparison of numerous samples simultaneously. F i n g e r p r i n t i n g Te c h n i q u e s Until now, fingerprinting techniques such as ARDRA, denaturing gradient gel electrophoresis (DGGE), restriction fragment length polymorphisms (RFLP), and single strand conformation polymorphism (SSCP) are preferred compromises between the number of samples to be analyzed and the information obtained. Since there exist several excellent, comprehensive reviews on molecular techniques (Hill et al., 2000; Muyzer and Ramsing, 1995; Vaneechoutte, 1996), mainly methods used in this study are briefly described. R F L P This method has been modified for application to microbial ecology by extracting DNA from environmental samples and amplifying bacterial 16S rdna. It provides a genetic fingerprint of the PCR-amplified DNA upon

18 1 8 C h a p t e r 1 digestion with a rare cutting restriction enzyme, followed by fragment length analysis by standard gel electrophoresis. The method is based on the principle that the restriction sites are conserved according to phylogenetic patterns. A much better resolution is achieved with terminal restriction fragment length polymorphism (T-RFLP), in which one primer is fluorescently labeled. This yields a mixture of amplicons with a fluorescent label at one end. After purification, the amplicon mixture is digested with a restriction enzyme, which generates fragments of different sizes. These are separated through gel or capillary electrophoresis. A laser reader detects the labeled fragments and generates a profile based on fragment lengths. Buckley and Schmidt (2001) used T-RFLP on PCR-amplified 16S rdna from soil to examine the effect of plant community composition and land-use history on microbial communities. The sites sampled were part of the Long Term Ecological Research project in agricultural ecology at the W.K. Kellogg Biological Station of Michigan State University (Hickory Corners, MI, USA). Microbial communities were similar in plots that shared a long-term history of agricultural management. In contrast, they differed significantly from fields that had never been cultivated. A R D R A ARDRA is RFLP of amplified rrn genes of the ribosomal RNA operon (rdna) (Vaneechoutte, 1996). The amplified genes are of similar sizes, but contain small differences in nucleotide composition. One way of detecting these differences is to digest the PCR products with a restriction endonuclease. The resulting fragments are separated on an agarose or polyacrylamide gel. The main advantage is that this technique is convenient, since no particular conditions for separation are necessary. The main limitation lies in the choice of restriction enzyme, which is essential for obtaining optimal resolution. ARDRA has been successfully used to differentiate between closely related bacterial isolates by comparing fingerprints with restriction patterns of reference organisms (Dijkshoorn et al., 1998; Vaneechoutte et al., 1992). It has also been used for analyzing mixed microbial populations (McSpadden Gardener and Weller, 2001; Smit et al., 1997). Smit et al. (1997) used ARDRA to detect distinct differences in bacterial communities between copper-contaminated and uncontaminated soil. D G G E Originally used as a method to detect mutations, DGGE is nowadays a common tool to analyze microbial communities. PCR-amplified DNA molecules are subjected to electrophoresis on a denaturant gradient gel. The DNA remains double-stranded until it reaches the denaturant concentration at which the double-stranded molecules start melting. Partial melting of the DNA causes branch structures, which lead to a mobility shift in the gel.

19 G e n e r a l i n t r o d u c t i o n 1 9 Since the melting behavior depends on the nucleotide sequence, separation of individual members of the community is possible. A great advantage of DGGE is that additional information can be obtained from sequencing excised bands. However, when using SSU rdna amplicons, a major limitation is that the numbers of rrna operons vary among taxonomic groups, and sequence heterogeneity occurs within a single cell (Nübel et al., 1996). This implies that one organism can be represented by more than one band. DGGE has been used successfully in various studies to assess the diversity of bacterial and fungal communities. Smalla et al. (2001) analyzed bacterial rhizosphere communities of three host plants of the fungal pathogen Verticillium dahliae. The DGGE fingerprints showed plant-dependent shifts in the relative abundance of bacterial populations in the rhizosphere. Although generally DNA fragments of bp are separated, recently DNA fragments of 1650 bp have been used to analyze fungal communities (Gomes et al., 2003). Schabereiter-Gurtner et al. (2001) identified the fungal diversity of historic church glass windows. The authors constructed an 18S rdna clone library from samples taken off the glass to identify fungi that might play a role in biodeterioration. Sequence analysis of clones after DGGE analysis revealed a high fungal diversity, with some genera that had not been detected on historic glass earlier. H y b r i d i z a t i o n t e c h n i q u e s DNA hybridization techniques are powerful tools in environmental microbiology to assess the distribution and relatedness of nucleic acid sequences and to track microorganisms in various environments. Factors affecting hybridization are the degree of mismatches between target and probe, the length and concentration of target and probe sequences, hybridization time, salt concentration, and temperature (Sayler and Layton, 1990). If the population structures of complex environmental communities are to be analyzed, enhanced specificity and sensitivity are needed to detect organisms at low abundance or organisms that present only a small percentage of the total microbial community. Notably measurements of microbial population structure in soil are difficult, because of the heterogeneity and complexity of the soil system (Ritz et al., 1997). S u b t r a c t i v e H y b r i d i z a t i o n Subtractive hybridization provides higher specificity compared to general hybridization techniques. The principle of subtractive hybridization is the removal of sequences shared by two DNA populations, resulting in isolation of those sequences unique to one DNA population. It was used as early as 1966 to identify specific sequences from the bacteriophage T4 genome by hybridization of DNA isolated from appropriate deletion mutants (Bautz and Reilly, 1966).

20 2 0 C h a p t e r 1 Single subtractive methodologies are limited in their usefulness for analyzing complex communities. The reason is that the enrichment required to purify target sequences is very high (in the order of fold). If sequence complexity of the DNA is too high, the DNA will not sufficiently hybridize to completion (Lisitsyn and Wigler, 1995). The addition of oligonucleotide linkers to the DNA increases the specificity and allows selective amplification. Multiple rounds of hybridization using different adaptors can be carried out. S u p p r e s s i o n s u b t r a c t i v e h y b r i d i z a t i o n Suppression subtractive hybridization (SSH) is a further advancement developed by Diatchenko et al. (1996). The method is based on the suppression PCR effect, which implies the preferential amplification of unique target sequences. Long inverted terminal repeats that are attached to the DNA can selectively suppress amplification of non-target sequences in PCR procedures. The advantage of this adaptation is that only one round of selective hybridization is necessary without any physical separation of singleand double-stranded DNA. Now, commercial kits are available. Subtracted fragments can be excised from the gel, labeled and used as hybridization probes for further screening of subtraction clone libraries, which makes this technique versatile and powerful. Bogush et al. (1999) applied SSH to identify differences between the genomes of Escherichia coli and Salmonella typhimurium. About 60% of the differential clones identified by SSH were present in one of the genomes and absent from the other. D N A M i c r o a r r a y s The historical origin of microarray technology reaches back as far as the 1960s (Ekins and Chu, 1999). During that time immunoassays, for example, relied on antibodies attached to a solid surface, and analogous approaches have been applied to DNA measurements later. This led in the 1980s to the development of microarrays for immunodiagnostic purposes (Ekins and Chu, 1999). Only recently have microarrays been adapted to microbial ecology research (Loy et al., 2002; Wilson et al., 2002a). DNA microarrays are small, solid supports onto which known DNA sequences from thousands of genes are immobilized in an arrayed order at fixed locations. The supports themselves are glass microscope slides, silicon chips or nylon membranes. The DNA is printed, spotted, or actually synthesized directly onto the support. Nucleic acid samples to be analyzed are fluorescently labeled and hybridized to the array, allowing the parallel analysis of thousands of genes. The microarrays used in this study (Chapter 6) were Affymetrix-fabricated chips with 25-mer oligonucleotides synthesized in situ. Molecular ecology makes use of a wide range of techniques, because each method is prone to various limitations, such as preferential amplification of

21 G e n e r a l i n t r o d u c t i o n 2 1 specific DNAs, or sequence heterogeneity within ribosomal rrn genes. They should not be considered as substitutes for more conventional methods such as culturing populations or activity studies, but as complementary methods that allow investigation of both culturable and non-culturable microorganisms in their natural habitat. The great value lies in the speed and possibility to compare different dominant members present in a community. Considering the complexity of the microbial world, not only in soil, it cannot be expected that all microorganisms in samples will be accessible with the same efficiency. O u t l i n e o f t h i s t h e s i s In 1997, the Coordination Commission Risk Research (CCRO), the Netherlands, initiated a field study to answer questions regarding possible nontarget effects of GMMs. Glandorf et al. (2001) performed two field experiments (1997 and 1998), in which the biological control strain P. putida WCS358r and two transgenic derivatives that constitutively produced PCA, were introduced into the rhizosphere of wheat. Both the wild type, WCS358r, and the PCAproducing GMMs caused a transient shift in the composition of the fungal rhizosphere microflora. The effects of the parent strain and the GMMs were differential and related to the amount of PCA produced by the GMMs. The present study is a follow-up on the field experiments described by Glandorf et al. (2001) and based on the following questions: 1) Is the observed effect specific for PCA or also applicable to other antimicrobial metabolites, and 2) will repeated introduction of GMMs, as envisaged in commercial agricultural applications, lead to enhanced non-target effects? To this end, the same Pseudomonas strains (wild type and GMM #8 used by Glandorf et al., 2001) were introduced yearly into the same field for 4 consecutive years ( ). To answer the first question, a second GMM, which produced DAPG, was included in the present study. Since both antibiotics are toxic for a broad range of microorganisms, effects on fungi as well as bacteria were investigated. To evaluate the significance of changes caused by the GMMs, those were compared to effects caused by a crop rotation from wheat to potato, a common agricultural practice, which was expected to have a profound impact on microbial communities. In Chapter 2 the results of the first two years of the field study (1999 and 2000) are described. The indigenous bacterial and fungal communities were studied using colony counts and ARDRA to assess both culturable and non-culturable microorganisms, as used in the previous studies (Glandorf et al., 2001; Smit et al., 1997). In addition, effects on nitrifying bacteria were determined by measuring soil nitrifying potential activity. In the second year

22 2 2 C h a p t e r 1 the impact of the GMMs on the microbial community was compared to effect of crop rotation. In Chapter 3 effects of wild type and genetically modified WCS358r on the rhizosphere bacterial communities were studied. Since previous results revealed an effect of crop rotation with DGGE, but not with ARDRA, rhizosphere samples from all four seasons were analyzed by DGGE, including those previously analyzed by ARDRA. The effects were compared to those of direct and long-term effects of a crop rotation of wheat and potato. In Chapter 4 DGGE was adapted for the study of the composition of ascomycete communities in the rhizosphere of wheat and potato. Since the relevance of perturbations caused by GMMs needs to be assessed relative to disturbances by common agricultural practices, a base line of microbial diversity was generated. Effects of the two crops on the ascomycetes were compared to positional and seasonal effects. Possible effects of the antibiotic-producing pseudomonads on the ascomycete communities are described in Chapter 5. Rhizosphere samples were collected during the four seasons and analyzed by DGGE. In addition, DNA clone libraries were constructed and sequenced for samples taken in 1999 and 2000 in order to identify members of the microflora that are sensitive to perturbations. However, the method proved very laborious and not suitable for analyzing large sets of samples. Whereas ARDRA and DGGE provide conclusive data on the magnitude of changes in microbial communities, they offer little information about the identity of the microorganisms affected by the introduced pseudomonads. Because cloning and sequencing is laborious and time-consuming, a different approach to identify perturbations was followed in Chapter 6. Samples from two time points in 2001 and 2002 and four different treatments were subjected to microarray analysis. The Affymetrix GeneChips used contained arrays of oligonucleotide probes targeting the SSU rrna of almost all bacteria represented in the Ribosomal Database Project (Maidak et al., 2001). Clustering of the treatments from the microarrays was compared with clustering of the same samples obtained by DGGE. In addition, the suitability of suppression subtractive hybdridization to detect the identity of changes in the microflora was evaluated. In Chapter 7 the results described in Chapters 2 to 6 are discussed with reference to current knowledge about ecosystem effects of GMMs.

23

24

25 Chapter 2 Repeated introduction of genetically modified Pseudomonas putida WCS358r without intensified effects on the indigenous microflora of field-grown wheat M. Viebahn, D.C.M. Glandorf, T.W.M.Ouwens, E. Smit 1, P. Leeflang 1, K. Wernars 1, L.S. Thomashow 2, L.C. van Loon, and P.A.H.M. Bakker 1 2 National Institute of Public Health and the Environment, Bilthoven, The Netherlands USDA, Washington State University, Pullman, WA, USA Present address: National Institute of Public Health and the Environment, Bilthoven, The Netherlands Applied and Environmental Microbiology, 69, (2003)

26 2 6 C h a p t e r 2 A b s t r a c t To investigate the impact of genetically modified, antibiotic-producing rhizobacteria on the indigenous microbial community, Pseudomonas putida WCS358r and two transgenic derivatives were introduced as a seed coating into the rhizosphere of wheat in two consecutive years (1999 and 2000) in the same field plots. The two genetically modified microorganisms (GMMs), WCS358r::phz and WCS358r::phl, constitutively produced phenazine-1-carboxylic acid (PCA) and 2,4-diacetylphloroglucinol (DAPG), respectively. Introduced bacteria in all treatments decreased from 10 7 colony forming units (cfu) per g root soon after sowing to less than 10 2 cfu after harvest 132 days after sowing. The phz and phl genes remained stable in the chromosome of WCS358r. The amount of PCA produced in the wheat rhizosphere by WCS358r::phz amounted to about 40 ng g -1 root in the first application of The DAPGproducing GMMs caused a transient shift in the indigenous bacterial and fungal microflora in 1999, as determined by amplified ribosomal DNA restriction analysis (ARDRA). However, after the second application of the GMMs in 2000 no shifts in the bacterial or fungal microflora were detected. To evaluate the importance of the effects induced by the GMMs, they were compared with those induced by crop rotation by planting wheat in 1999 followed by potatoes in No effect of rotation on the microbial community structure was detected. In 2000 all bacteria had a positive effect on plant growth, supposedly due to suppression of deleterious microorganisms. Our research suggests that the natural variability of the microbial communities can surpass effects of GMMs. I n t r o d u c t i o n Antagonistic bacteria can suppress plant diseases caused by microbial pathogens (Walsh et al., 2001; Weller et al., 2002). A major mechanism of control of soilborne plant pathogens by fluorescent pseudomonads is the production of antibiotics (Fujimoto et al., 1995; Haas et al., 2000; Maurhofer et al., 1992; Weller et al., 2002). Many naturally occurring antibiotics such as 2,4- diacetylphloroglucinol (DAPG), phenazine, pyoluteorin and pyrrolnitrin have been identified (Handelsman and Stabb, 1996; Walsh et al., 2001). Application of biocontrol agents to suppress soilborne diseases is often unsuccessful because of the inconsistency in their performance under field conditions. Explanations for this inconsistency include variable plant root colonization by the biocontrol agent, insufficient concentration of the antibiotic, instability or degradation of the antibiotic, and genetic diversity of the pathogen (Mazzola et al., 1992a; Mazzola et al., 1992b; Weller, 1988). One approach to overcome some of these difficulties is to genetically modify biocontrol strains for enhanced and/or constitutive biosynthesis of antibiotics. Before genetically modified microorganisms (GMMs) can be commercially used as biocontrol agents, field studies must be performed to attain information about their possible impact on non-target organisms. So far, effects of introduced wildtype strains and GMMs on the soil ecosystem have been studied mainly in microcosms (Hase et al., 1999; Thirup et al., 2000). In most field studies attention has been focused primarily on the fate of the introduced GMMs (Van Overbeek et al., 1997; Völksch and May, 2001). Only a few studies

27 R e p e a t e d i n t r o d u c t i o n o f P. p u t i d a o n w h e a t 2 7 describe effects on the indigenous microbial community (De Leij et al., 1995; Schwieger and Tebbe, 2000). Particularly GMMs with an enhanced capacity to produce antibiotic compounds are likely to affect non-target microorganisms. Glandorf et al. (2001) demonstrated that introduction of the root-colonizing bacterium Pseudomonas putida WCS358r, genetically modified to produce the antimicrobial compound phenazine-1-carboxylic acid (PCA), caused a differential, but transient shift in the fungal rhizosphere microflora of wheat plants, as compared to the parental strain. In the present study strain WCS358r, modified to constitutively produce PCA or DAPG, was introduced in two consecutive years into a field to monitor possible long-term effects on the rhizosphere microflora. The objectives were to determine whether the effects on the microbial community observed by Glandorf et al. (2001) are specific for PCA or apply also to another antibiotic. To simulate commercial application protocols, we introduced WCS358r and its GM derivatives in the same field plots during two consecutive years in order to investigate whether a second application of the strains would cause intensified effects on the microflora. Population dynamics and dispersal of the introduced bacteria were also studied. A general method to study effects on the microbial community is plate count enumeration of culturable microorganisms from rhizosphere samples. Because only 0.1-1% of the total microflora can be cultured (Atlas and Bartha, 1998; Hugenholtz and Pace, 1996), molecular techniques based on direct extraction of nucleic acid from the samples are more suitable (Hill et al., 2000; Schmidt, 1994; Smit et al., 1999). Here, we used a PCR-based amplified ribosomal DNA restriction analysis (ARDRA) to detect restriction fragment length polymorphisms of the microbial rrna genes. The impact of the GMMs on the microflora was compared with effects resulting from a common agricultural practice, crop rotation, an often-used strategy for control of soilborne plant diseases (Hill et al., 2000). In our study microbial communities in field plots cultured with wheat for two consecutive years were compared with plots planted to wheat in the first year and potato in the second year. M a t e r i a l s a n d M e t h o d s B a c t e r i a l s t r a i n s The bacterial strains used are listed in Table 1. WCS358r, a spontaneous rifampicin-resistant mutant of plant growth-promoting P. putida strain WCS358, was used as wildtype and was described previously (Bakker et al., 1986; Geels and Schippers, 1983; Glandorf et al., 1992). A constitutively PCA-producing derivative contained the phzabcdefg gene cluster on the

28 2 8 C h a p t e r 2 Table 1. Bacterial strains used in this study. Strain Relevant characteristic a Reference Escherichia coli 1936 JLS vector putkm::phz Pseudomonas fluorescens Q2-87 Phl + Bangera and Thomashow Phz + Weller and Cook 1983 Pseudomonas putida WCS358r Rif r Geels and Schippers 1993, Glandorf et al WCS358r::phz8 Rif r Kan r Phz + Glandorf et al WCS358r::phl6 Rif r Kan r Phl + this study a Phl + production of DAPG; Phz + production of PCA, Rif r resistance. rifampicin resistance; Kan r kanamycin disarmed mini Tn5 transposon-based vector putkm (Herrero et al., 1990) inserted into the chromosome. This strain, WCS358r::phz, was identical to the one designated GMM 8 used in the study by Glandorf et al. (2001). A DAPG-producing GMM was constructed by insertion of the phlfacbde genes from P. fluorescens Q2-87 (Bangera and Thomashow, 1996), a naturally DAPG-producing strain, into the chromosome of WCS358r, and designated WCS358r::phl. The disarmed mini Tn5 transposon including the nptii gene coding for kanamycin (km) resistance contained in putkm was used as vector. In this construct the phlf repressor gene is disrupted, resulting in constitutive production of DAPG. WCS358r was cultured on King s medium B (KB) agar (King et al., 1954), supplemented with 150 μg ml -1 rifampicin. The transconjugants were cultured on KB in the presence of 150 μg ml -1 rifampicin and 30 μg ml -1 kanamycin (KB rif/km). All bacteria were incubated at 28ºC for 2 days. S e e d t r e a t m e n t Wheat seeds (Triticum aestivum cv. Baldus) were treated with a 1:1 mixture of washed bacterial suspensions (WCS358r, WCS358r::phz, WCS358r::phl) and 3% methylcellulose, as described by Glandorf et al. (2001). For the control treatment the bacterial suspension was replaced by 10 mm MgSO 4. Coated seeds were air-dried overnight and sown the next day. Coating resulted in approximately 10 7 cfu per seed, as determined from plate count enumerations. E x p e r i m e n t a l F i e l d Experiments were performed in 1999 and 2000 on an experimental field located near the Botanical Garden of Utrecht University, The Netherlands.

29 R e p e a t e d i n t r o d u c t i o n o f P. p u t i d a o n w h e a t 2 9 The field had a planting history of grass and consisted of clay soil (12%) with an organic matter content of 4% and a ph (KCl) of 5.0. In 2000 the plots were fertilized one day before sowing (P 2 O 5 35 g per m 2, K 2 O 57 g per m 2, NO 3-8 g per m 2 ). The field was divided into two halves, each containing 18 plots of 1 m 2. For each half a random block design with 6 treatments, each in 3 replicates, was used. The plots were separated by unplanted, 50 cm wide buffer zones, and the field was surrounded by an additional buffer zone of 1.8 m width. The site was surrounded by straw mats and a fence, and covered with a net to prevent rabbits and birds from entering the site. Wheat seeds were sown and potatoes were planted in April and harvested in August. The treatments were seeds coated with either WCS358r, WCS358r::phz, WCS358r::phl, or a 1:1 mixture of WCS358r::phz and WCS358r::phl, a control with uninoculated seeds, and a rotation plot with uninoculated wheat seeds in 1999 and with non-treated potatoes (Solanum tuberosum L. cv. Modesta) in Per plot 70g of wheat seeds were sown in 11 rows of 1 m length at a depth of 2-3 cm. In the rotation plots in the second year 40 potatoes were planted in 8 rows, each containing five seed tubers per row. Plant and rhizosphere soil samples were taken during the growing season at 5 to 8 different time points. P o p u l a t i o n d y n a m i c s o f t h e i n t r o d u c e d s t r a i n s On each sampling date during the growing season plant roots with adhering soil were harvested from three randomly selected spots within each plot. Prior to dilution plating samples were prepared according to Glandorf et al. (2001). Population densities of WCS358r and the PCA- and DAPGproducing GMMs were determined by plating on KB + (Geels and Schippers, 1983) supplemented with 150 μg ml -1 rifampicin (KB + rif). The numbers of rifampicin resistant cfu were determined after incubation for 2 days at 28ºC. Wildtype and GMMs were also differentiated by fluorescence under UV light (366 nm). PCA and DAPG are UV-absorbing compounds (Pierson III and Thomashow, 1992) and their production results in weaker fluorescence of the GMMs on KB + agar compared to the wildtype. Single colonies of the GMMs grown on KB + rif were also transferred to KB + rif/km to verify the presence of the gene cluster. Soil samples to check for dispersal of GMMs were collected between the plots and in the surrounding buffer zones at two time points in both seasons. P r o b e s a n d c o l o n y h y b r i d i z a t i o n To distinguish WCS358r::phz and WCS358r::phl in the rhizosphere of plants treated with the mixture of the GMMs, specific probes were developed.

30 3 0 C h a p t e r 2 Escherichia coli 1936 JLS containing the phz gene locus was grown on Luria Bertani agar medium (Sambrook et al., 1989) supplemented with kanamycin (30 μg ml -1 ), and grown for 1 day at 37ºC. P. fluorescens Q2-87 containing the phl gene locus, was grown on KB agar plates for 2 days at 28ºC. Colonies were lysed, 1:50 diluted and 1 μl was used as template DNA for a PCR reaction. Primers Phl2a (5 -GAGGACGTCGAAGACCACCA-3 ) and Phl2b (5 -ACCGCAGCATCGTGTATGAG-3 ) (Raaijmakers et al., 1997) amplify a 745 bp fragment within the sequence of the phld gene. Primers Phen1 (5 -CCCCTGTTGACAATTAATCATCGG-3 ) and Phen2 (5 - ACCTTGACGTTGTACCATTCCCAA-3 ) target sites within the sequences of the phza and phzb genes and amplify a 1,014 kb fragment. Both primer sets were synthesized by Eurogentec, Maastricht, The Netherlands. The 50 μl reaction mixture contained 5 μl of diluted heat-lysed cells, PCR buffer (Amersham Pharmacia Biotec), 200 μm dntps, 200 μm phl and phz primers, and 1.5 U Taq DNA polymerase (Amersham Pharmacia Biotec). The PCR program consisted of 4 min at 92ºC and 30 cycles of 92ºC for 30 s, 56ºC for 30 s and 72ºC for 60 s. The size of the PCR products were checked on a 0.7% agarose gel, excised from the gel and purified with QIAEX II Agarose Extraction Kit (Qiagen, Hilden, Germany) according to the manufacturer s protocol. The probes were then labeled with alkaline phosphatase from the Gene Images TM AlkPhos Direct TM labeling and detection system (Amersham Pharmacia Biotech) according to the protocol provided by the manufacturer. To determine the relative numbers of the PCA- and DAPG-producing GMMs in the treatments with the mixture of both GMMs, at each sampling time 120 bacterial colonies were randomly selected from KB + rif/km plates containing colonies from plants treated with the mixture. The colonies were transferred to Hybond TM -N + nylon membranes (Amersham Pharmacia Biotech), bacterial cells were lysed and cell debris was washed off by standard methods (Sambrook et al., 1989). The DNA was fixed to the filters by exposure to UV light (365 nm) for 2 min (Hoefer Scientific Instruments, San Fransisco, USA). Hybridization with the Gene Images TM AlkPhos Direct TM labeling and detection System (Amersham Pharmacia Biotech) and detection with the CDP-Star TM chemiluminescent detection reagent (Amersham Pharmacia Biotech) were performed as described in the supplier s protocol. P C A e x t r a c t i o n f r o m t h e w h e a t r h i z o s p h e r e To determine whether PCA was produced by WCS358r::phz in the rhizosphere, samples were collected 18 days after sowing. Samples from 3 replicate plots were pooled, yielding 2 replicates per treatment. The determination was carried out as described by Bonsall et al. (1997), and modified by Glandorf et al. (2001). The detection limit of PCA in roots or in soil was 15 ng per sample.

31 R e p e a t e d i n t r o d u c t i o n o f P. p u t i d a o n w h e a t 3 1 E n u m e r a t i o n o f t h e c u l t u r a b l e m i c r o f l o r a Selected groups of bacteria and fungi in the wheat rhizosphere were quantified by plating on different media. The aerobic heterotrophic bacterial population was determined on tryptic soy agar (TSA, Difco Laboratories, Detroit, MI, USA) containing 100 μg ml -1 cycloheximide after incubation at 20ºC for 8 days. To determine the total number of Bacillus spores, dilutions of rhizosphere samples were incubated for 15 min at 80ºC to inactivate vegetative cells prior to plating on TSA agar and incubation at 28ºC for 2 days. Plate count enumerations of pseudomonads were determined on KB + after incubation at 28ºC for 2 days. The predominant fungal community was estimated on onefourth strength potato dextrose agar (PDA, Difco) containing 2 μl ml -1 Triton X-100 and 200 μg ml -1 aureomycine. Fusarium spp. were determined on Komada s agar medium (Komada, 1975); fungi of the order Mucorales were identified on PDA supplemented with 50 μg ml -1 benomyl (Bollen, 1972), and Trichoderma species on one-fourth strength PDA and on Trichoderma-selective medium (TSM) (Elad et al., 1981). All fungi were incubated at 20ºC for 5 days. D e t e r m i n a t i o n o f t h e c o m p o s i t i o n o f p r e d o m i n a n t f u n g a l a n d b a c t e r i a l c o m m u n i t i e s b y A R D R A A cultivation-independent approach based on the amplification of the small subunit rrna was used to compare the composition of the bacterial and fungal communities between treatments. Three replicates of each treatment were pooled to produce two replicate samples for analysis per treatment. Samples were treated as described earlier (Glandorf et al., 2001) and total DNA was extracted using a bead beater as described (Smalla et al., 1993). PCR on the DNA extracts was performed with fungal and bacterial specific primers. The primer pair EF3 (5 -TCCTCTAAATGACCAAGTTTG-3 ) and EF4 (5 -GGAAGGG[G/A]TGTATTTATTAG-3 ) (Smit et al., 1999) amplify a 1.4 kb DNA fragment of fungal 18S rdna. The eubacterial primer set 338F (5 -ACTCCTACGGG[A/G][G/C]GCAGC-3 ) (Amann et al., 1990) and 1492R (5 -GGTTACCTTGTTACGACTT-3 (Embley et al., 1988) were used to amplify a fragment of 1.1 kb of bacterial 16S rdna. Primer conditions for primer pair EF3/EF4 were the following: 5 min at 94 C (1 cycle), 1 min at 94 C, 1 min at 48 C, 3 min at 72 C (40 cycles), and 10 min at 72 C. The PCR conditions for the primer pair 338F/1492R differed in the annealing temperature set at 60 C. Fungal PCR products were digested with TaqI, and bacterial PCR products with HinfI. The samples were loaded on precast polyacrylamide gels (GeneGelExcel 12.5, Amersham Pharmacia Biotech), bands were separated on a GenePhor horizontal electrophoresis unit (Amersham Pharmacia Biotech)

32 3 2 C h a p t e r 2 and silver-stained in a Hoefer Automated Gel Stainer using a DNA Silver Staining Kit (Amersham Pharmacia Biotech). Gel images were digitalized using the GeneGenius Bio Imaging System (SYNGENE, UK). The Dice coefficient was used for calculating the similarities of the banding pattern. The algorithm for clustering the resulting DNA patterns was the unweighted pair-group method using arithmetic averages (UPGMA) (BioNumerics program version 2.0, Applied Maths, Belgium). Clusters were defined by a cut-off similarity value of 60%. P l a n t g r o w t h Plant growth was determined by measuring height, fresh and dry weight of 10 to 15 plants per plot at each sampling date. Plant dry weight was determined after drying plant shoots at 70ºC for 1-3 days. After harvest, 20-ear and 100- seed weights were determined for each plot. N i t r i f y i n g P o t e n t i a l A c t i v i t y ( N PA ) To determine effects of the GMMs on the nitrifying microorganisms, we determined the nitrtrifying potential activity during both seasons. At each sampling date 10g root-free rhizosphere soil from each plot was suspended in 25 ml assay medium (Bodelier, 1997; Stienstra et al., 1994), and the mixture was shaken at 200 rpm for 48h at 25ºC. Samples of 1 ml were withdrawn and centrifuged for 5 min at 13,000 g. Then 0.5 ml of the supernatant was mixed with an equal amount of 2 M KCl to stop the nitrification reaction. Samples were stored at -20ºC until analysis of NO 3 - on an autoanalyzer (San Plus System, Interface 8708/16, Skalar). S t a t i s t i c a l A n a l y s i s All data, except those obtained with ARDRA, were statistically analyzed using the SPSS software package (version 10.0). Results for the populations of WCS358r and its GMMs, the culturable microflora and plant yield were determined, for each sampling date, by one-way analysis of variance (ANOVA), followed by a Bonferroni correction. A significant interaction between time and treatment was analyzed with repeated measurements ANOVA. Before performing analysis of variances, homogeneity of variances and normal distribution were verified. In cases of heterogeneity of variances or nonnormal distribution, the non-parametric test of Kruskal-Wallis was used. For the ARDRA-dendrograms the Dice similarity indices were calculated based on the comparison between the banding patterns (BioNumerics program version 2.0, Applied Maths, Belgium) and statistically analyzed using a permutation test with random sampling. In all cases the confidence interval was 95%.

33 R e p e a t e d i n t r o d u c t i o n o f P. p u t i d a o n w h e a t 3 3 R e s u l t s C o n s t r u c t i o n o f W C S r : : p h l The rhizobacterium P. putida WCS358r was genetically modified by insertion of the phlfacbde genes from P. fluorescens Q2-87 (Bangera and Thomashow, 1996), encoding production of DAPG. The resulting strain, WCS358r::phl, contained one copy of the gene cluster in the chromosome and produced the same amount of DAPG in vitro as the donor strain (data not shown). P o p u l a t i o n d y n a m i c s o f P. p u t i d a W C S r a n d i t s G M M s After planting of inoculated seeds in 1999, rhizosphere populations of both the wild type WCS358r and the GMMs decreased from over 10 7 cfu g -1 root at 5 days post-sowing to less than 10 2 cfu g -1 root at 100 days after sowing (Fig.1 A). In the year 2000 cell numbers dropped more rapidly, from about 10 6 cfu g -1 root at 11 days after sowing to g -1 root at 60 days (Fig. 1 B). During the next 40 days the cell numbers remained relatively constant, after which an additional but variable decrease was observed at 130 days after sowing. Between treatments no difference was apparent, except at two time points in the second year. Eleven days after sowing cfu of the GM strains were significantly lower than of the parental strain (WCS358r::phz, P = 0.028, WCS358r::phl, P = 0.049, WCS358r::phz + WCS358r::ph, P = 0.004). At 39 days the cfu of WCS358r::phz were significantly lower than of the parental strain (P = 0.018). To investigate if the effect of one antibiotic is enhanced by the presence of a second antibiotic, we applied a mixture of the PCA- and the DAPG-producing GMMs in a 1:1 ratio on the wheat seeds. During the season the proportions of both GMMs in these plots were calculated from colony hybridization data (Table 2). In 1999 the percentage of the DAPG-GMM decreased to about 30% in the course of the growing season. In 2000 the percentage of the DAPG-GMM in the combination treatment varied between 37% at the beginning and 67% during the field trial. These latter results were compared to results obtained from plate count enumeration. Because of their different morphology WCS358r::phz and WCS358r::phl could be easily distinguished on KB + rif/km agar plates. The colonies of WCS358r::phl appeared later and were smaller in size. Using these two independent assays we obtained similar results for the average percentage of the DAPG-GMM in 2000: 44 ± 26% by plate count enumeration and 53% ± 15% by colony hybridization analysis. In the control treatment no cfu were detected on KB + amended with rifampicin. Early in the season single colonies of rifampicin- and kanamycinresitant strains were found outside the plots, into which they were introduced

34 3 4 C h a p t e r # A log cfu g -1 root phz phl phl + phz wt # B log cfu g -1 root * * Days after sowing Fig. 1. Population dynamics of WCS358r ( ), its PCA- and DAPG-producing derivatives WCS358r::phz ( ) and WCS358r::phl ( ), and the combination of both derivatives (π) in the wheat rhizosphere in 1999 (A) and 2000 (B). Colony forming units (cfu) were determined on King B + agar supplemented with rifampicin. The asterisks indicate a significant difference between the parent strain and WCS358r::phz, WCS358r::phl or a mixture of both GMMs 11 days after sowing (P = 0.028, 0.049, 0.004) and between the parent strain and WCS358r::phz 39 days after sowing in 2000 (P = 0.018). # indicates the number of cfu per seed at the beginning of the field experiment. (data not shown), indicating dispersal of the introduced bacteria. Later in the season neither the wildtype nor the GMMs were detected outside the plots. S t a b i l i t y o f t h e p h z a n d p h l g e n e s Under field conditions, the stability of the mini Tn5 transposon in the GMMs, which, besides the phz and phl genes, contains the nptii gene coding for kanamycin resistance, was examined by determining kanamycin resistance in the GMMs. At each sampling date rhizosphere samples were plated on KB + rif agar medium with and without kanamycin, after which the numbers of colonies on both media were compared. No significant differences between numbers of colonies on both media were observed in either season. This indicates that the constructs containing the phl or the phz genes remained stable in the chromosome of WCS358r. D e t e c t i o n o f P C A i n t h e r h i z o s p h e r e Roots with adhering soil were sampled 18 days after sowing, and PCA was extracted and analyzed using reversed-phase HPLC. In ng and 37 ng

35 R e p e a t e d i n t r o d u c t i o n o f P. p u t i d a o n w h e a t 3 5 Table 2 Percentage and standard deviation of the numbers of rifampicin-resistant colony forming units of the DAPG-producing GMM isolated from the rhizospheres of plants. Seeds were coated with a 1:1 mix of the PCA- and DAPG-producing GMMs. Colony hybridizations were done at 5 sampling dates during the field trials of 1999 and colonies were tested per sampling date. Plate counts to morphologically distinguish PCA- from DAPG-producing GMMs were performed only in 2000 n.d.: not determined. days after sowing colony blots colony blots plate counts 0 n.d. 37 ± 8 17 ± 11 13/ ± 11 n.d. 16 ± ± 5 67 ± ± ± 6 64 ± ± 20 59/ ± 8 45 ± 7 54 ± 46 mean 37 ± ± ± 26 1 : 13 and 59 days after sowing in 1999, 11 and 54 days after sowing in 2000 PCA per gram roots were detected in the field plots where the PCA-producing GMMs were introduced, alone or in combination with the DAPG-GMM, respectively (data not shown). No PCA was detected in the other plots. In 2000 PCA was detected only in the plots where both GMMs were introduced. Quantification was not possible because of amounts near the detection limit (Thomashow et al., 1997). E f f e c t s o f t h e G M M s o n t h e i n d i g e n o u s c u l t u r a b l e m i c r o f l o r a Effects of WCS358r and the GMMs on numbers of component groups of the culturable indigenous microflora were determined by plating rhizosphere samples on selective media. Besides aerobic heterotrophic bacteria, fluorescent pseudomonads, actinomycetes, Bacillus spp. spores, filamentous fungi and Fusarium spp., in 1999 members of the order Mucorales and Trichoderma species were also included. In neither season was a significant effect of the introduced bacteria on fungal or bacterial communities detected. A single exception concerned the filamentous fungi at 59 days after sowing in 2000, as shown in Fig. 2. There was a significant difference between counts of filamentous fungi from the rhizospheres of the control plants and those from plants treated with the DAPG-GMM (P = 0.032, one-way ANOVA). In both seasons the population of filamentous fungi increased from 10 4 cfu g -1 root to about cfu g -1 root after harvest. E f f e c t s o f t h e G M M s o n t h e p r e d o m i n a n t f u n g a l a n d b a c t e r i a l m i c r o f l o r a Total DNA from rhizosphere samples of each plot was extracted and analyzed using ARDRA. The field was laid out in such a way that three replicates of each treatment were located on the left half of the field, the other

36 3 6 C h a p t e r 2 7 A log cfu g -1 root phz phl phl + phz wt contr 3 7 B log cfu g -1 root * Days after sowing Fig. 2. Population dynamics of total filamentous fungi (log cfu g -1 root) in the rhizosphere of wheat plants treated with WCS358r ( ), its PCA- and DAPG-producing derivatives (, ) and a combination of both (π) in the field season of 1999 (A) and 2000 (B). In the control treatment ( ) seeds were only coated with methylcellulose (no bacteria). Colony forming units (cfu) were determined on potato dextrose agar medium. The asterisk indicates a significant difference between the control treatment and the treatment with WCS358r::phl (P = 0.032). three replicates on the right half. For analysis samples of the three plots from each half were pooled, resulting in two replicates per treatment. A particular partitioning for all dendrograms was obtained by cutting the dendrograms at a similarity index of 60%. Fig. 3 shows the dendrograms based on the ARDRA patterns of the fungal community in the wheat rhizosphere in Throughout the season one distinct cluster was apparent, which contains fungal communities from plants treated with WCS358r::phl, either alone or in combination with WCS358r::phz. Other clusters contained fungal communities from plants treated with the control, WCS358r, and WCS358r::phz. At 25 days samples from plants treated with WCS358r::phl and with WCS358r::phz + WCS358r:: phl clustered separately at the 60% similarity cut-off value, though at the 48% break point they cluster together, apart from all other treatments. At 96 days the cluster contained one rhizosphere sample of the WSC358r::phz treatment. The distinct clustering was most evident at the end of the season at 132 days. Only at that time point the two control samples clustered together with a similarity of more than 80%. The clustering at 132 days suggests that also the WCS358r and the WCS358r::phz treatments affected the fungal community.

37 R e p e a t e d i n t r o d u c t i o n o f P. p u t i d a o n w h e a t 3 7 A: day 13 D: day 96 % similary % similary phz a wt b phz b phz + phl a phl b phl a phz + phl b contr a wt a contr b phz + phl a phz + phl b phl b phz b phl a phz a contr b wt b contr a wt a B: day 25 E: day 132 % similary % similary phl b phl a phz + phl b phz + phl a wt a phz a contr a wt b contr b phz b phl a phl b phz + phl b phz + phl a phz a wt b phz b wt a contr a contr b C: day 40 % similary phz + phl a phz + phl b phl b wt a phl a phz a phz b wt b contr b contr a Fig. 3. Dendrograms representing the percentages similarity of rhizosphere fungal communities of field grown wheat plants in 1999 that were untreated (contr), treated with P. putida WCS358r (wt), with WCS358r::phz (phz), WCS358r::phl (phl), or with a combination of both GMMs (phz+phl). Similarities are based on ARDRA patterns generated from amplified 18S rdna digested with TaqI. Samples were taken at 13 (A), 25 (B), 40 (C), 96 (D) and 132 (E) days after sowing. Samples from 3 plots were pooled resulting in 2 replicates per treatment belonging to the left (a) or right (b) half of the field. Percentages of similarities are shown above each dendrogram. In the bacterial microflora the transgenic WCS358r::phl also caused the same distinct clustering as in the fungal communities 13 and 25 days after sowing (data not shown). Later in the season the banding pattern affected by the DAPG-producing GMMs, alone or in the combination treatment, clustered in more than one group. However, at 132 days the samples treated with the DAPG-producing GMMs clustered together, including one replicate of the WCS358r::phz treatment. The dendrogram of the bacterial community showed larger differences than that of the fungal community (data not shown). The

38 3 8 C h a p t e r 2 similarity indices within treatments were compared with the average similarity indices between treatments. To this end we determined all pairwise similarity indices, and compared the average within-treatment similarities with the average pairwise similarities between different treatments, using a permutation test with random sampling. Significant differences were detected for the fungal community at 25 (P = 0.014) and 132 days (P = 0.004), and for the bacterial community at 25 days (P = 0.008) (data not shown). In the field experiment of 2000 rhizosphere samples of the six treatments grouped in a different way compared to At the beginning of the field experiment both the fungal (Fig. 4) and the bacterial communities (data not shown) clustered mainly according to their position in the field (rhizosphere samples belonging to the left half or the right half of the field) regardless of the treatment. This was most evident for the fungal (Fig. 4) and bacterial community (data not shown) 11 days after sowing. During the growing season the differences between the fungal communities in the plots increased and clustered at the 60% level in eight groups, whereas the bacterial community clustered in three, six, and five groups at 25, 39, and 59 days, respectively (data not shown). Each group contained mostly rhizosphere samples from either the left or the right half of the experimental field. The heterogeneity between samples for the fungal and bacterial community was high, comparable with that found in 1999 for the bacterial community. No effect of the introduced strains was detected. In 2000 potatoes were planted in the rotation plots. We anticipated that different crops would carry different microbial communities, as described in other studies (Olsson and Alström, 2000). However, ARDRA, as shown in Fig. 4, did not reveal a distinct clustering of the rotation plots. At no time point did similarity indices analyzed with the permutation test reveal significant differences. E f f e c t s o n n i t r i f y i n g p o t e n t i a l a c t i v i t y Nitrifying bacteria, such as Nitrosomonas and Nitrobacter, are an important group of soil microorganisms, which oxidize ammonium to nitrite or nitrate (Atlas and Bartha, 1998). One way to study the nitrifying activity in soil is to measure the nitrifying potential activity (NPA) (Bodelier, 1997). The NPA was determined at each sampling date. In 1999 the field was not fertilized and the NPA remained between 0.4 and 2.6 nmol NO 3 - g -1 dry soil h -1 during a period of 54 days after sowing (data not shown). In 2000 the field was fertilized and the NPA of all treatments varied between 2 and 8 nmol NO 3 - g -1 dry soil h -1 before decreasing to 0.5 nmol NO 3 - g -1 dry soil h -1 at 59 days after sowing (data not shown). Neither in 1999 nor in 2000 introduction of WCS358r or

39 R e p e a t e d i n t r o d u c t i o n o f P. p u t i d a o n w h e a t 3 9 A: day 11 % similary contr b phz b wt b phl b wt a phz a contr a potato a phz + phl a phl a B: day 25 % similary potato b phz + phl b wt b phz b phl b potato a contr b phz + phl a phl a wt a phz a contr a C: day 39 D: day 59 % similary % similary phz + phl b phz b wt b phz + phl a phl b contr b potato a phl a potato b phz a contr a wt a phl a phz a phz + phl a wt a potato b phz b contr b phz + phl b wt b phl b potato a contr a Fig. 4. Dendrograms representing the percentages similarity of rhizosphere fungal communities of field grown wheat plants in 2000 that were untreated (contr), treated with P. putida WCS358r (wt), with WCS358r::phz (phz), WCS358r::phl (phl), or with a combination of both GMMs (phz+phl). Similarities are based on ARDRA patterns generated from amplified 18S rdna digested with TaqI. Samples were taken at 11 (A), 25 (B), 39 (C) and 59 (D) days after sowing. Samples from 3 plots were pooled resulting in 2 replicates per treatment belonging to the left (a) or right (b) half of the field. Percentages of similarities are shown above each dendrogram. the GMMs significantly affected the soil nitrifying potential activity (P > 0.05) (data not shown). P l a n t y i e l d To determine if the introduced bacteria affected plant development, fresh and dry weight of shoots from each plot per sampling date were determined. In addition, after harvest 100-seed and 20-ear weights were measured. In 1999 there was no effect of the introduced bacteria on plant growth and yield (data not shown). However, in 2000 all introduced bacteria had a positive effect on shoot fresh and dry weight. Dry weight was increased by 41% in all bacterial treatments compared to the control at harvest, 102 days after sowing (P < 0.04) (Fig. 5). The 20-ear weight had increased by 14 to 21% with a significant difference between the control plants and the GMMtreated plants (P < 0.04), the 100-seed weight had increased by 5 to 8% with a

40 4 0 C h a p t e r avg. plant dry weight (g) * phz phl phl + phz wt contr Days after sowing Fig. 5. Dry weight (g) per shoot of wheat plants in 2000 treated with P. putida WCS358r ( ), its PCA- and DAPG-producing derivatives (, ) and the combination of both derivatives (π). In the control treatment ( ) seeds were coated with methylcellulose (no bacteria). Values represent means of 10 plants per treatment and sampling date. The asterisk indicates a significant difference between control plants and the bacterial treated plants (P < 0.04). significant difference between control and bacterized plants (P < 0.035), except for the WCS358r::phz-treated plants (P = 0.131) (data not shown). However, no significant difference in plant growth yield between the parent strain and the GM derivatives was detected. Discussion P. putida WCS358r is a well-studied plant growth-promoting rhizobacterium that suppresses Fusarium wilt in carnation and radish based on the production of its fluorescent siderophore pseudobactin 358 (Duijff et al., 1994; Raaijmakers et al., 1995). In Arabidopsis thaliana it suppresses Fusarium wilt by induction of systemic resistance (Van Wees et al., 1997). Glandorf et al. (2001) modified WCS358r with the phz gene locus resulting in constitutive production of PCA. WCS358r and two PCA-producing derivatives were introduced on wheat seeds in a one-time application in two separate field experiments in 1997 and In both years a transient effect on the indigenous fungal microflora was detected (Glandorf et al., 2001). In our field study we released two times into the same field WCS358r and its PCA-producing derivative WCS358r::phz, as well as a DAPG-producing derivative, WCS358r::phl, alone and in combination with WCS358r::phz. We simulated commercial application protocols by applying the strains as seed coating in two consecutive years. Moreover, a one-time application of a biological control agent is unlikely to be fully effective in controlling disease (Cook, 1993). Production of PCA or DAPG is an important trait in biological control agents, which can lead to suppression of plant root diseases (Weller et al., 2002). A single copy of a disarmed Tn5 transposon containing the genes

41 R e p e a t e d i n t r o d u c t i o n o f P. p u t i d a o n w h e a t 4 1 for the production of DAPG was inserted into the chromosome of WCS358r, resulting in WCS358r::phl. Chromosomal insertions are more stable and are less frequently subject to horizontal gene transfer than extra-chromosomal elements (Sengeløv et al., 2001). Thus, the possibility that the Tn5 transposon including the DAPG gene cluster is transferred to other bacteria is minimized, an important aspect when releasing microorganisms into the environment. After introduction of the wild type and the GMMs into the same field in 1999 and 2000, cell numbers of the inoculated bacteria dropped more rapidly during the first 60 days in 2000 compared to Successful colonization is a process that depends, among other things, on environmental conditions such as weather. There was no major difference in the average temperature (14.2ºC in 1999, 14.7ºC in 2000). However, a total rainfall of 85 mm was registered at the nearby Royal Netherlands Meteorological Institute (KNMI) in May 2000 compared to only 52 mm in May In both years population dynamics in the wheat rhizosphere showed no significant differences between the wild type and the GMMs. This indicates that the GMMs were as fit as the wild type under the conditions tested. These results confirm previous observations that the chromosomal insertion of the phz genes into Pseudomonas had no negative effect on its fitness in the wheat rhizosphere (Glandorf et al., 2001; Timms-Wilson et al., 2000). In two strains, P. fluorescens 2-79 and P. chlororaphis 30-84, production of phenazine antibiotics contribute to their ability to survive in and colonize the wheat rhizosphere in competition with the indigenous soil microflora (Mazzola et al., 1992b). However, genetic modification to produce antibiotics can also lead to either a reduced ecological fitness or to defects in competitive or survival abilities (Cook, 1993). Since DAPG is also effective against bacteria (Keel et al., 1992), effects on the fungal microflora (Glandorf et al., 2001), but also effects on the bacterial microflora were expected. Introduction of WCS358r, WCS358r::phz, and WCS358r::phl had no effect on the numbers of culturable bacterial and fungal microflora in either the first or the subsequent introduction. In vitro assays indicated that, compared to WCS358r, the PCA- and DAPG-producing GMMs had an increased inhibitory effect against saprophytic fungi isolated from the same field soil (data not shown). This confirms that results obtained under laboratory conditions can differ from those obtained under field conditions due to complex environmental conditions (Sharifi-Tehrani et al., 1998). In 1999, ARDRA revealed that introduction of the DAPG-GMM, in the single treatment or in combination with the PCA-producing GMM, affected the composition of the rhizosphere microflora. It is reasonable to expect that a repeated introduction of the bacterial inoculants would increase their effect. In 2000, however, rhizosphere samples clustered according to their position in

42 4 2 C h a p t e r 2 the experimental field, belonging to the left or right half, independent of the seed treatment. It is possible that changes in environmental conditions caused a difference between the left and right half of the experimental field, causing the distinct clustering in the second year, and that effects of the bacteria did not exceed this variability. Despite the fact that PCA was produced in the rhizosphere of wheat, there was no effect of the PCA-producing GMM. Glandorf et al. (2001), however, found that the same strain, WCS358r::phz (GMM 8), after a one-time introduction did affect the fungal community. In our study the transient effect on the fungal and bacterial microflora resulted from the introduction of the DAPG-producing GMM. A primary benefit of crop rotation relies on the activity of the indigenous microorganisms in the soil, which deplete the energy sources for pathogenic fungi after crop harvest (Cook, 1992). A one-year rotation is sufficient, for instance, to control root pathogens of barley and wheat (Cook, 1992). Miethling et al. (2000) compared the effect of crop specificity, soil origin and a bacterial inoculant on the establishment of microbial communities. With the methods applied, community level physiological profile, fatty acid analysis, and temperature gradient gel electrophoresis, they showed that crop species had the most pronounced effect on the composition of rhizosphere microorganisms. Tomato and flax had a selective influence on populations of fluorescent pseudomonads, as demonstrated by Lemanceau et al. (1995). In this study phenotypic and taxonomic characteristics of type strains and wild type isolates from soil and plants were compared. The bacteria were grouped on the basis of their enzyme activities and their ability to utilize 147 substrates. Comparison of lipopolysaccharide and cell envelope protein patterns of pseudomonads isolated from potato, grass, and wheat revealed 30 distinct patterns for each crop, of which the majority was not observed for the other crops (Glandorf et al., 1993). Differences in root exudates of different plant species and varieties are known to affect the microbial communities (Kremer et al., 1990). In our field study no effect of crop rotation was detected with the methods applied. The overall variability of microbial communities of both the wheat and the potato rhizospheres may have been too high to detect crop specificity under the environmental conditions encountered, using ARDRA. ARDRA has been used successfully previously to detect shifts in microbial communities (Glandorf et al., 2001; Schwieger and Tebbe, 2000). Contradictory to our expectations we could not demonstrate an enhanced effect on the microbial communities after a second introduction of the bacterial inoculants or an effect of crop rotation. This may be partly due to the limitations of the technique we applied (DNA extraction hampered by humic acids in soil, preferential amplification of dominant members of the microbial community,

43 R e p e a t e d i n t r o d u c t i o n o f P. p u t i d a o n w h e a t 4 3 low discriminatory power of ARDRA). However, since there was a significant effect of the DAPG-producing GM derivatives in the first year, it is likely that the high variability between samples in 2000 exceeded possible effects caused by the GMMs. Only a few microbial genera are able to utilize energy derived from nitrification, a process that can be severely affected by environmental stress (Atlas and Bartha, 1998). We anticipated that the activity of nitrifying bacteria would be easily affected by the introduction of bacterial inoculants into the soil and we therefore determined the NPA. However, in both seasons we did not detect an effect of the introduced microorganisms on the NPA, indicating that the production of the antibiotics did not adversely affect the capacity for nitrification in the soil. In 2000 all bacterial treatments resulted in increased plant growth, compared to the control. Supposedly, the main mechanism in plant growth promotion by rhizobacteria is suppression of deleterious microorganisms (Schippers et al., 1987). These microorganisms commonly colonize the root system of crops such as wheat, potato and sugar beet (Fujimoto et al., 1995). The siderophores of the rhizobacterium WCS358r, produced under iron-limited conditions, have a higher affinity for iron than those of deleterious microorganisms (Schippers et al., 1987). Consequently, deleterious microorganisms are outcompeted by WCS358r in the rhizosphere, resulting in higher growth of plants treated with WCS358r or its GMMs. However, no significant difference in plant growth occurred between wild type and GMM treated plants, suggesting that production of either PCA or DAPG is not important for the observed plant growth stimulation. This effect was only detected in the second year, and not in 1999 or in the one-year field experiments described earlier (Glandorf et al., 2001), despite similar population dynamics of the introduced bacteria in the different years. As for the effect of the genetic modification, our study showed that the DAPG-producing GMM caused a transitory shift in the fungal and bacterial microflora. A repeated introduction of the GMMs did not lead to the expected intensified effects. So far, we found no evidence that GMMs modified for enhanced biological control activity had a major impact on the environment. A c k n o w l e d g e m e n t This study was financed by the Dutch Ministry of Housing, Spatial Planning and the Environment. We thank Hans van Pelt and Ientse van der Sluis for excellent technical assistance, Niko Nagelkerke for assistance with the statistical analysis, and Marjolein Kortbeek-Smithuis for editing Fig. 3 and 4. We also thank Bas Valstar, Fred Siesling and Jeroen van Schaik (Botanical Garden, Utrecht University) for constructing and maintaining the experimental field site.

44

45 Chapter 3 Response of bacterial communities in the rhizosphere of field-grown wheat to repeated introductions of genetically modified Pseudomonas putida WCS358r Mareike Viebahn, Karel Wernars 1, Leendert C. van Loon, Eric Smit 1, and Peter A.H.M. Bakker 1 National Institute of Public Health and the Environment, Bilthoven, The Netherlands

46 4 6 C h a p t e r 3 A b s t r a c t In previous studies, amplified ribosomal DNA restriction analysis (ARDRA) revealed changes in the indigenous microbial communities in the rhizosphere of wheat upon a onetime introduction of antibiotic-producing genetically modified microorganisms (GMMs). However, no changes were apparent upon a second introduction in the same field plots. In this study samples were reanalyzed using denaturing gradient gel electrophoresis (DGGE), and the study was extended to four years to monitor time-dependent effects. Each year ( ) Pseudomonas putida WCS358r and two transgenic derivatives constitutively producing phenazine-1-carboxylic acid (PCA) or 2,4-diacetylphloroglucinol (DAPG) were introduced as a seed coating. Effects caused by the bacterial inoculants were compared to changes in the bacterial microflora in a crop rotation of wheat and potato. In the first year, all bacterial treatments transiently affected the composition of the bacterial communities, without differences between the wild type and the GMMs. After reintroduction, the DAPG-producing GMM influenced the bacterial microflora significantly different from the wild type in the following years. The PCAproducing derivative of WCS358r showed an effect different from the parental strain only after four consecutive introductions. Cropping potato in alternation with wheat had a lasting effect on the bacterial communities compared to continuous wheat cropping. Throughout this long-term field trial, effects of the GMMs never exceeded those of the crop rotation. I n t r o d u c t i o n Several root-colonizing Pseudomonas spp. exhibit antagonistic effects towards plant pathogenic fungi, and have, therefore, attracted much attention for their potential as biological control agents (Chin-A-Woeng et al., 2003). The production of secondary antimicrobial metabolites has been recognized as a major mechanism for disease suppression. Effective biological control of soilborne pathogens has been demonstrated for e.g. take-all in wheat caused by the fungus Gaeumannomyces graminis var. tritici (Weller and Cook, 1983). However, biological control is often hampered by inconsistency under field conditions. The indigenous microflora as well as the plant can influence the expression of antibiotic biosynthetic genes in pseudomonads (Haas and Keel, 2003). For instance, genes encoding phenazine production in P. aureofaciens strain have been shown to be regulated by quorum sensing (Wood and Pierson, 1996). As a result, antibiotic production may come too late or amounts be insufficient (Lugtenberg et al., 2001; Weller, 1988). Therefore, it can be advantageous to genetically modify biocontrol strains such that antibiotics are produced constitutively, making production independent of environmental conditions. As an example, the biocontrol strain P. fluorescens F113 can antagonize Pythium ultimum, the causative agent of damping-off of sugar beet. Its biocontrol activity was increased through constitutive overproduction of 2,4-diacetylphloroglucinol (DAPG), and was then comparable to that of a commercial fungicide (Delany et al., 2001).

47 R e s p o n s e o f b a c t e r i a l c o m m u n i t i e s t o G M M s 4 7 However, public concerns exist about the environmental release and potential ecological effects of genetically modified microorganisms (GMMs), such as direct and indirect effects on non-target organisms (Timmis and Pieper, 1999). For that reason we conducted field experiments to monitor changes in the natural bacterial communities in the rhizosphere of wheat upon application of such genetically modified strains. The plant growth-promoting rhizobacterium P. putida WCS358r and two antibiotic producing recombinant derivatives were introduced as a seed coating into the rhizosphere of wheat. The recombinant strains constitutively produced phenazine-1-carboxylic acid (PCA) or DAPG, both secondary metabolites with a broad-spectrum activity against bacteria and fungi. In our previous studies the PCA- and DAPG-producing derivatives significantly influenced the fungal and the bacterial microflora (Glandorf et al., 2001; Viebahn et al., 2003) after a one-time introduction, as studied by amplified ribosomal DNA restriction analysis (ARDRA). When the GMMs were introduced into the same plots a second time (Viebahn et al., 2003), we expected the effects to become more prevalent. Similarly, in a crop rotation from wheat to potato, a significant shift in the microbial communities was expected, as differences in rhizosphere microbial populations between plant species are well established (Lemanceau et al., 1995; Oehl et al., 2003; Olsson and Alström, 2000). However, after the second application of the GMMs, no shifts in the bacterial or fungal microflora were detected, and no differences between potato and wheat were observed (Viebahn et al., 2003). In a study to monitor changes in microbial communities due to antropogenic impact, Torsvik et al. (1998) compared the microbial diversity obtained by different molecular techniques with DNA melting profiles and reassociation kinetic studies. Their conclusion was that fingerprinting methods such as ARDRA are too sensitive and that the resolution is too high for a reliable characterization of complex microbial communities, as the result cannot be reliably assessed. Lower resolution methods such as denaturing gradient gel electrophoresis (DGGE) (Fischer and Lerman, 1979) are better suited to monitor changes at the community level. Liesack and Dunfield (2002), on the other hand, argued that DGGE is often too sensitive for the determination of total community fingerprints. Apparently, it is difficult to conclude which technique yields the best resolution and sensitivity. Since preliminary experiments using DGGE revealed an effect of the rotation plots, we reanalyzed the rhizosphere soil samples from 1999 and 2000 by this technique. Effects on fungal communities are discribed in Chapter 4 and 5.

48 4 8 C h a p t e r 3 M a t e r i a l a n d M e t h o d s B a c t e r i a l s t r a i n s The bacterial strains used in this study are P. putida WCS358r (Geels and Schippers, 1983) and two transgenic derivatives, WCS358r::phz and WCS 358r::phl, that constitutively produce PCA and DAPG, respectively (Glandorf et al., 2001; Viebahn et al., 2003). The wild type strain was grown on King s Medium B (KB) agar (King et al., 1954) supplemented with 150 μg rifampicin ml -1, the transgenic derivatives on KB with 150 μg rifampicin ml -1 and 30 μg kanamycin ml -1. All strains were routinely grown at 28ºC for 2 days. S e e d t r e a t m e n t a n d e x p e r i m e n t a l f i e l d Wheat seeds (Triticum aestivum cv. Baldus) were sown and seed potatoes (Solanum tuberosum L. cv. Modesta) were planted in four consecutive years (1999 to 2002) in an experimental field located near the Botanical Garden of Utrecht University, The Netherlands, as described by Viebahn et al. (2003). The field was divided into two halves, each containing 18 1-m 2 plots. For each half a random block design with six treatments, each with three replicates, was used. Wheat seeds were treated with a 1:1 mixture of washed bacterial suspensions and 3% methylcellulose, as described by Glandorf et al. (2001). The bacterial treatments used were WCS358r, WCS358r::phz, WCS358r::phl, or a 1:1 mixture of WCS358r::phz + WCS358r::phl. For the control treatment the bacterial suspension was replaced by 10 mm MgSO 4. Coated seeds were airdried overnight and sown the next day. Coating resulted in approximately 10 7 cfu per seed, as determined from plate count enumerations. A sixth treatment consisted of a rotation plot, in which in the first (1999) and third year (2001) non-bacterial treated wheat seeds were sown, and in the second (2000) and fourth year (2002) non-treated potatoes were planted. The same treatments were repeated every year on the same plots, and rhizosphere soil samples were taken each year at three to six time points. D N A e x t r a c t i o n a n d P C R From each plot 3-5 g of roots with adhering soil were sampled at each time point during the growing season and mixed with 10 ml sodium phosphate buffer (120 mm, ph 8.0). One g of gravel was added and samples were vortexed for 30 s. The supernatant was decanted into a new tube. Appropriate dilutions were plated on different media to enumerate actinomycetes, Bacillus spp. spores, fluorescent pseudomonads and aerobic heterotrophic bacteria, as described previously (Viebahn et al., 2003). One ml of the supernatant was used to extract total DNA with the FastDNA SPIN Kit for Soil (Bio 101, Biogene, Vista, CA, USA) in combination with a Ribolyser (Hybaid, Ashford, UK) (Smit

49 R e s p o n s e o f b a c t e r i a l c o m m u n i t i e s t o G M M s 4 9 et al., 2003). The extracts were suspended 1:100 in 100 µl Millipore-filtered distilled water before purification with the Wizzard DNA Clean-Up System (Promega, Madison, WI, USA) according to the manufacturer s protocol. Three replicates from the left half of the field were pooled and contributed one sample, and pooled three replicates of the right half of the field represented a duplicate sample. rdna from the two samples of each treatment was amplified with the primer pair F-968-GC (5 -CGCCCGGGGCGCGCCCCGGGCG GGGCGGGGGCACGGGGGGAACGCGAAGAACCTTAC-3 ) and 1491R (5 -CGGTGTGTACAAGACCC-3 ) (Nübel et al., 1996). The primer pair amplified the segment of the eubacterial 16S rdna from nucleotide 968 to nucleotide 1491 [E. coli numbering (Brosius et al., 1978)]. F-968-GC contained a 40 bp GC-clamp to stabilize the melting behavior of the DNA fragments for DGGE analysis (Sheffield et al., 1989). The PCR was performed in 10x PCR buffer 2 (ph 9.2), containing 2.25 mm MgCl 2 (Roche Diagnostics, Mannheim, Germany), 250 μm of each of the four deoxynucleoside triphosphates, 200 nm of each primer, 2.5 U Expand Long template enzyme (Roche, Diagnostics, Mannheim, Germany) and 1 μl of appropriately diluted template DNA in a total volume of 50 μl. The PCR conditions used in the thermocycler (Hybaid, Ashford, UK) were: 5 min 94 C, followed by 35 cycles of 1 min 94 C, 1 min 60 C, and 3 min 72 C, and finally 10 min 72 C. D G G E PCR fragments were separated on a denaturing gradient polyacrylamide gel containing 0.5xTAE (20 mm Tris-acetate, 10 mm sodium acetate, 0.5 mm EDTA). The gel consisted of 8% acrylamide (acrylamide/bisacrylamide in a ratio of 37.5:1) with a denaturant gradient of 30-60%. One hundred % denaturing solution contained 7 M urea, and 40% formamide. Up to 20 μl of PCR product per lane were loaded, and gels were run for 17 h at 80 V at a constant temperature of 60 C in a DCode Universal Mutation Detection System (Bio-Rad Laboratories, Veenendaal, The Netherlands). For comparison, a reference DNA marker set was added in triplicate on each gel. After electrophoresis gels were stained for 30 min in 1:10000 diluted SybrGold (Molecular Probes, Leiden, The Netherlands), and viewed under a blue light transilluminator (Clare Chemical Research, Dolores, CO, USA). Images were digitalized using the GeneGenius Bio Imaging System (Syngene, Cambridge, UK). S t a t i s t i c a l a n a l y s i s The bacterial community fingerprints of the DGGE gels were analyzed with the BioNumerics program vers. 3.5 (Applied Maths, Sint-Martens-Latem, Belgium). Pearson s correlation coefficient was used to calculate the similarity

50 5 0 C h a p t e r 3 index between each banding pattern, and clusters were calculated with the unweighted pair-group method using average linkages (UPGMA). Significant clusters were determined by using the statistical point-bisectional correlation method (BioNumerics Manual 3.5). R e s u l t s In a previous study (Viebahn et al., 2003) we determined the number of culturable actinomycetes, Bacillus spp. spores, fluorescent pseudomonads, and aerobic heterotrophic bacteria in the rhizosphere of wheat treated with P. putida WCS358r or the GMMs. No significant differences were found, indicating that the introduced bacteria had no effect on the numbers of these specific members of the culturable bacterial community. To reassess the changes in the total microflora, DGGE was applied to the same samples as analyzed previously with ARDRA. The resulting banding patterns were complex (Fig. 1) and could only be computationally analyzed. The resulting data were used to construct dendrograms based on percentage similarity between treatments (Fig. 2). At 13, 25, 40, 54, 96 and 160 days after sowing the number of significant clusters varied between two and three. Up to and including the fourth sampling date, one distinct cluster was apparent that contained mainly samples from non-treated wheat plants. Samples from plants treated with the wild-type strain, WCS358r::phz, WCS358r::phl, or the combination of both GMMs grouped together in a second cluster, which at 40 days split into two subclusters. One subcluster contained both duplicate samples of the wild-type treatment and one replicate of the PCA treatment. The other subcluster contained the other replicate of the PCA treatment and all replicates from the DAPG and the combination treatments. Up to the fourth sampling date the similarity between the two main clusters varied between 24 and 78%. Later in the season, at 96 days and after harvest at 160 days, all samples became more similar (> 85%). The two clusters that were obtained contained mainly samples either from the left (a) or the right (b) half of the field. S h i f t s i n t h e b a c t e r i a l c o m m u n i t i e s f r o m t o Similar dendrograms were constructed for samples taken in 2000, 2001 and The overall similarity of samples from all treatments varied between 61 and 92% in 2000, between 46 and 96% in 2001, and between 42 and 71% in 2002 (data not shown). In contrast to our earlier analysis with ARDRA that did not reveal differences (Viebahn et al., 2003), the samples from 2000 analyzed with DGGE did show significant effects (see below). To compare the main effects per year, we constructed composite dendrograms. Composite

51 R e s p o n s e o f b a c t e r i a l c o m m u n i t i e s t o G M M s 5 1 M contr a b rp wt phz phz phl phz+phl a b a b a M b a b a b M Fig. 1. DGGE profiles showing the bacterial community structure of rhizosphere samples of field grown wheat plants 40 days after sowing in Wheat seeds were treated with P. putida WCS358r (wt), WCS358r::phz (phz), WCS358r::phl (phl), or with a combination of both (phz+phl). In the control treatment (contr) seeds were coated with methylcellulose (no bacteria). rp = samples taken from a rotation plot, in which wheat was grown in M = Reference marker. dendrograms represent the average of the first three to four time points per year, because the most pronounced effect of the bacterial treatments occurred at these early time points. The composite dendrogram for the data obtained in 1999 at 13, 25, 40 and 54 days is depicted in Fig. 3A. In accordance with the analysis presented in Fig. 2, two significant clusters were obtained, with one cluster containing only samples from the rhizosphere of untreated wheat plants from both the control plots and the rotation plots. The other cluster contained samples derived from plants treated with the bacteria, the wild type as well as the transgenic strains. Thus, the effect of the introduced bacteria was evident, but the effect of the genetically modified derivatives was not different from that of the wild-type strain. Following the second introduction of the bacterial strains in 2000 there were again shifts in the bacterial communities (Fig. 3B). Two clusters were obtained, of which one contained two subclusters. The main cluster contained samples from the control plants, from plants treated with the wild-type strain or the PCA producer, and one replicate from plants treated with the DAPG producer. The two other subclusters contained samples from plants treated with the DAPG producer in the single and the combination treatment, on the one hand, and both samples from the rhizosphere of potato, on the other hand. These results indicate an impact of the DAPG-producing derivative, as well as of crop rotation on the bacterial microflora. Results obtained from the samples taken in 2001 (Fig. 3C) showed two clusters, of which one contained two subclusters. One of these subclusters

52 5 2 C h a p t e r 3 A: day 13 B: day 25 C: day 40 % similarity % similarity % similarity contr b rp b rp a contr a phz+phl b phz a phz b phz+phl a phl a phl b wt a wt b contr a contr b rp a phl a phl b phz b phz+phl a wt b phz a phz+phl b rp b wt a contr b rp a contr a rp b phz a wt b wt a phl b phz+phl a phl a phz b phz+phl b D: day 54 E: day 96 F: day 160 % similarity % similarity % similarity contr a rp a contr b phl a phz b phz+phl a rp b phl b wt b phz a wt a phz+phl b contr a contr b phl a phz a wt a phz+phl a rp a rp b wt b phz b phz+phl b phl b rp a wt a phz a contr a phl a phz+phl b contr b phz b phl b phz+phl a rp b wt b Fig. 2. Dendrograms based on the genetic similarity of the bacterial communities of fieldgrown wheat plants in 1999 at 13 (A), 25 (B), 40 (C), 54 (D), 96 (E) and 160 (F) days after sowing. For treatments see legend to Fig. 1. Samples from three field plots were pooled, resulting in two replicates per treatment, from the left (a) and from the right half of the field (b). Similarities are based on DGGE patterns generated from 16S rdna fragments using Pearson s correlation coefficient. Cluster analysis was done with UPGMA. Significant (grey lines) and non-significant clusters (black lines) were calculated by the point-bisectional cutoff method. The levels of similarity are shown above the dendrograms. Notice that in panel 3 (C) the scale of the similarity index ranges from % instead of %. contained both replicate samples from the control treatment, and the other subcluster contained those from root samples of wheat plants treated with the parental strain WCS358r and the PCA-producing derivative WCS358r::phz. The second cluster contained all samples from plants treated with the DAPG producer, whether in combination with the PCA producer or not. Despite the fact that wheat was grown in the rotation plots in 2001, these samples clustered apart from the continuous wheat plots together with samples from plants treated with the DAPG-producer, in the single and the combination treatment. This shows that the potato crop in 2000 had a lasting effect on the bacterial community.

53 R e s p o n s e o f b a c t e r i a l c o m m u n i t i e s t o G M M s 5 3 A: 1999 % similarity B: 2000 % similarity contr a rp a contr b rp b phz+phl b wt b phz a phz+phl a wt a phl a phl b phz b phz b wt b phl a phz a contr b contr a wt a phl b phz + phl a phz + phl b rp a rp b C: 2001 D: 2002 % similarity % similarity contr a contr b phz a phz b wt a wt b phl a phz+phl b rp a phl b phz+phl a rp b contr a contr b wt a wt b phl a phl b phz a phz b phz+phl a phz+phl b rp b rp a Fig. 3. Composite dendrograms representing the genetic similarity of the bacterial communities of field-grown wheat plants in 1999 (A), 2000 (B), 2001 (C), and 2002 (D). Dendrograms are derived from the combined analysis of samples taken at the first three or four sampling dates. For further explanation see legend to Fig. 1 and Fig.2. In the fourth year (2002) the samples of the control treatment and the treatment with the wild type clustered together. All treatments with the transgenic strains and samples from the rhizosphere of potato grouped together in the second cluster, of which a subcluster contained one of the replicates of the potato rotation plots (Fig. 3D). The effects of both transgenic derivatives on the bacterial microflora were comparable, and appear to differ only slightly from the effect of cropping potato instead of wheat. Over the four-year field trial period, samples from all treatments became more similar. The similarity of the main significant clusters increased from 72% in 1999 to 80% and 82% in the following years. The impact of the introduced transgenic bacteria never exceeded that resulting from alternating a wheat and potato crop.

54 5 4 C h a p t e r 3 D i s c u s s i o n P. putida WCS358r is a well-studied biological control strain. Its fluorescent siderophore pseudobactin 358 is responsible for suppression of Fusarium wilt in carnation and radish (Duijff et al., 1994; Raaijmakers et al., 1995) and increases in potato tuber yield (Bakker et al., 1986). To increase its biological control properties, it was genetically modified to constitutively produce PCA or DAPG. These antibiotics play a major role in antagonism of plant pathogens such as Gaeumannomyces graminis var. tritici, the causative agent of take-all disease, and Pythium ultimum, which causes damping-off in sugar beet (Handelsman and Stabb, 1996). Previously, ARDRA was used to demonstrate that the PCA-producing GMM affects the fungal microflora after a one-time introduction into the rhizosphere of wheat (Glandorf et al., 2001). Using the same technique to detect changes in the fungal and bacterial communities as affected by the PCA and/or the DAPG-producing derivatives after a one-time introduction, it was found that only the DAPG-producing GMM had an effect. However, after a second introduction effects of the introduced GMMs were no longer apparent (Viebahn et al., 2003). Neither did the ARDRA technique reveal significant differences between the microbial communities in the rhizosphere of potato and of wheat. This result was unexpected, since it is well established that plant species have a pronounced effect on the composition of the microbial communities (Lemanceau et al., 1995; Oehl et al., 2003; Olsson and Alström, 2000). For this reason we reanalyzed the samples from 1999 and 2000 and extended the analysis for a further two years. DNA was extracted directly from rhizosphere soil samples and subjected to DGGE instead of ARDRA. The DGGE banding patterns obtained by using universal 16S rdna primers were complex and suggested a high diversity of the bacterial rhizosphere microflora of wheat. However, it must be taken into account that also with this technique, mainly dominant members of the bacterial communities are assessed, and band numbers and intensity only partially reflect the actual numbers of the microbial species. A single organism can produce more than one DGGE band (Nübel et al., 1996), and one band can result from more than one organism (Stephen and Kowalchuk, 2002). After introducing the bacterial strains into the field in the first year (1999), all bacterial treatments were found to affect the composition of the indigenous bacterial microflora. However, by the end of the growing season, the effect diminished: the bacterial microflora of all treatments became more similar, which was most evident at the latest time point, taken after harvest. This transient effect of the bacterial treatments can be explained by a decline

55 R e s p o n s e o f b a c t e r i a l c o m m u n i t i e s t o G M M s 5 5 with time of the numbers of colony forming units of the introduced strains, as described earlier (Viebahn et al., 2003). After the second introduction in 2000, the DAPG-producing GMM in the single and the combination treatments had a significant effect on the bacterial community, and samples from wheat treated with the wild-type strain and the PCA-producing GMM clustered together with the control treatment. There was also a clear difference between the bacterial microflora of the rhizosphere of wheat and potato. The composition of root exudates such as organic acids, amino acids and sugars varies with plant species and within species with plant age, and influences the microbial community composition (Lugtenberg et al., 2001; Normander and Prosser, 2000). Maloney et al. (1997) found a difference in the microbial community structure of lettuce and tomato using a physiological approach. Crop species had a bigger effect on microbial communities than soil factors or bacterial inoculants (Miethling et al., 2000). Indeed, using DGGE we demonstrate that a crop rotation from wheat to potato had a clear and lasting effect on the rhizosphere bacterial community compared to continuous wheat cropping. Apparently, a second introduction of the DAPG-producing GMM resulted in an effect on the bacterial microflora of wheat. However, the magnitude of this effect was comparable to that of the change from wheat to potato cropping. During the extension of the study, in 2001 all bacterial strains had an effect. The DAPG-producing GMM in the single and the combination treatments affected the bacterial community differentially with respect to the impact of the wild-type strain WCS358r and the PCA-producer. The effect of the potato crop in the previous year was still apparent, although now again wheat had been planted in the rotation plots. In 2002, not only the DAPGproducing GMM, but also the PCA-producing GMM, had an effect on the bacterial microflora. Again, this effect did not exceed the effect of crop rotation. Interestingly, no enhanced effect of the repeated introduction of the DAPGproducing GMM was demonstrated, not even after four years. For the PCAproducing GMM an effect on the bacterial microflora was only apparent after the four consecutive introductions. An additional effect of the two antibiotics in the treatment with both GMMs was never observed. In spite of the effects of the GMMs, during this four-year field trial the rhizobacterial communities of all treatments appeared to become more similar, with similarities between the main clusters increasing from 74% in the first year to 82% in the fourth year. Various microorganisms have developed defense mechanisms against antimicrobial metabolites such as PCA and DAPG, and different sensitivities towards these antibiotics have been reported, notably for plant pathogenic fungi (Mazzola et al., 1995). Mechanisms involved are the activation of efflux pumps, development of tolerance, and degradation of the antimicrobial compound (Duffy et al., 2003). Rhizosphere bacterial populations that come into contact

56 5 6 C h a p t e r 3 with introduced antibiotic producing GMMs are likely to develop such defense mechanisms, causing a shift in the microbial communities. This may explain the lack of an enhanced effect of repeatedly introducing the DAPG-producing GMM. Besides its antimicrobial activity, DAPG has been shown to induce systemic resistance to Peronospora parasitica in Arabidopsis (Iavicoli et al., 2003). The state of induced systemic resistance may affect the physiology of the plant such that the root-associated bacterial community is changed. Risk assessment studies can provide valuable information about the impact of biological control agents on the environment. Here, we showed that GMMs modified to produce antimicrobial compounds had effects on the bacterial microflora. However, these effects were not increasing with repeated introductions, and did not exceed the impact of crop rotation. The nature of the effects has not been clarified and needs to be further investigated (Griffiths et al., 2004). Therefore, our next goal is to identify bacterial species that are affected by the introduction of the GMMs (Chapter 6). A c k n o w l e d g e m e n t This study was financed by the Dutch Ministry of Housing, Spatial Planning and the Environment. We thank Christiaan Veenman and Rogier Doornbos for excellent technical assistance, and Bas Valstar and Fred Siesling (Botanical Garden, Utrecht University) for constructing and maintaining the experimental field site.

57

58

59 Chapter 4 Assessment of differences in ascomycete communities in the rhizosphere of field-grown wheat and potato Mareike Viebahn, Christiaan Veenman, Karel Wernars 1, Leendert C. van Loon, Eric Smit 1, and Peter A.H.M. Bakker 1 National Institute of Public Health and the Environment, Bilthoven, The Netherlands FEMS Microbiology Ecology, in press

60 6 0 C h a p t e r 4 A b s t r a c t To assess effects of plant crop species on rhizosphere ascomycete communities in the field, we compared a wheat monoculture and an alternating crop rotation of wheat and potato. Rhizosphere soil samples were taken at different time points during the growing season in four consecutive years (1999 to 2002). An ascomycete-specific primer pair (ITS5-ITS4A) was used to amplify internal transcribed spacer (ITS) sequences from total DNA extracts from rhizosphere soil. Amplified DNA was analyzed by denaturing gradient gel electrophoresis (DGGE). Individual bands from DGGE gels were sequenced and compared with known sequences from public databases. DGGE gels representing the ascomycete communities of the continuous wheat and the rotation site were compared and related to ascomycetes identified from the field. The effect of crop rotation exceeded that of the spatial heterogeneity in the field, which was evident after the first year. Significant differences between the ascomycete communities from the rhizospheres of wheat in monoculture and one year after a potato crop were found, indicating a long-term effect of potato. Sequencing of bands excised from the DGGE gels revealed the presence of ascomycetes that are common in agricultural soils. I n t r o d u c t i o n The fungi are a diverse group of microorganisms, which play a fundamental role in terrestrial ecosystems. They are involved in nutrient cycling, transportation of nutrients to plants, plant growth stimulation, and antagonism against plant pathogens (Christensen, 1989; Thorn, 1997). The majority of plant-associated fungi show a beneficial or neutral interaction with the host plant. Additionally, some fungi are used as biological control agents against plant pathogenic fungi. On the other hand, they also comprise a large number of plant pathogens (Agrios, 1997). It is estimated that only 5% of all fungal species are known (Hawksworth and Rossman, 1997) and that only 17% of the known species can be grown in culture (Bridge and Spooner, 2001). This implies that the diversity of fungi is largely unrecognized. Moreover, cultivating fungi from soil is not without bias, because they can be dormant, occur as spores, as mycelium, or both, and are often closely associated with other organisms (Bridge and Spooner, 2001). Therefore, in recent studies cultivationindependent molecular approaches have been used to determine the diversity of fungal communities through analysis of ribosomal (r) RNA genes (Ranjard et al., 2001; Schabereiter-Gurtner et al., 2001; Smit et al., 1999). The rrna genes are well suited for analyzing microbial diversity, because they are present in all known organisms, contain conserved as well as variable regions, and their sequences are collected in large public databases (Muyzer and Ramsing, 1995). While comparison of the small subunit (SSU or 16S) rrna is a wellaccepted tool in bacterial community studies, the SSU or 18S rrna genes of fungal communities are less informative and therefore less suitable for fungal community studies (Prosser, 2002). Previous research has shown that the internal transcribed spacers (ITS) of the nuclear rrna genes are better targets for the molecular analysis of fungal

61 A s c o m y c e t e s i n w h e a t a n d p o t a t o r h i z o s p h e r e 6 1 communities than the 18S rrna genes (Lord et al., 2002; White et al., 1990). The ITS are variable, non-coding regions located between the 18S and 5.8S subunit (ITS1) and between the 5.8S and 28S subunit (ITS2) of the rrna (Kowalchuk, 1999). Ribosomal RNA operons in fungi are often found as tandem repeats of up to a hundred copies, which makes it possible to amplify specific fragments even from small environmental samples (Buchan et al., 2002). Several primers have been designed to target this region. Primers ITS1 to ITS5 cover the two spacer regions and partial sequences of the 18S and 28S rdna (White et al., 1990). Recently, Larena et al. (1999) developed a primer with enhanced specificity for ascomycetes. With this primer, in combination with ITS1 or ITS5, only the ascomycetes were successfully amplified from a mixture of ascomycetes, basidiomycetes, oomycetes, zygomycetes, and mitosporic fungi. Ascomycetes are the largest group of the true fungi (Larena et al., 1999). Most of them are saprophytic and live on dead organic material, which they help decompose. However, ascomycetes also cause plant diseases, varying from powdery mildews to rots, cancers, and vascular wilts (Agrios, 1997). In the present study we compared the rhizosphere microflora in a long-term field trial between a crop rotation of wheat and potato and continuous wheat cultivation. In our first attempt we tried to analyze the total fungal community. However, PCR amplicons of 18S rdna could not satisfactorily be separated on a denaturant gel. Therefore, we decided to analyze a subgroup. The ascomycete communities were compared by performing an rdna fingerprint analysis of the ITS region of the rrna operon by denaturing gradient gel electrophoresis (DGGE) (Fischer and Lerman, 1979; Muyzer and Smalla, 1998). Ascomycetes were chosen because of their preponderance in the soil and because of the availability of specific primers. Finally, some of the predominant ascomycete bands revealed by DGGE were identified by cloning and sequencing. M a t e r i a l s a n d M e t h o d s A s c o m y c e t e i s o l a t e s Ascomycete isolates used to investigate the specificity of DGGE were grown on potato dextrose agar at 25ºC for 3-7 days. Alternaria brassicicola was kindly provided by Dr. J. Ton (Utrecht University, The Netherlands). Gaeumannomyces graminis var. tritici was kindly provided by Dr. J. M. Raaijmakers (Phytopathology, University Wageningen, The Netherlands). Genomic DNA of Trichoderma harzianum B28 and Verticillium longisporum was kindly provided by Dr. J. Postma (Plant Research International, Wageningen, The Netherlands). All other isolates were derived from the experimental field

62 6 2 C h a p t e r 4 and have been identified by Dr. W. Gams (Centralbureau voor Schimmelcultures, Utrecht, The Netherlands). R h i z o s p h e r e s o i l s a m p l i n g Rhizosphere soil was sampled from an experimental field located at the Botanical Garden of Utrecht University, the Netherlands, as described earlier (Bakker et al., 2002, Glandorf et al., 2001; Viebahn et al., 2003). The soil contained 12% clay, has an organic matter content of 4% and a ph (KCL) of 5, and was formerly grassland for more than 20 years. From 1999 to 2002 wheat (Triticum aestivum cv. Baldus) was grown continuously in six 1 m 2 plots, while in another six plots wheat was grown in 1999, followed by potato (Solanum tuberosum L. cv. Modesta) in 2000, wheat in 2001, and potato in Wheat was sown and potatoes were planted in early spring every year, and samples were taken 11, 25, 39 and 54 days later (Glandorf et al., 2001; Viebahn et al., 2003). The field was laid out in such a way that three replicates of each treatment were located on the left half of the field, the other three replicates on the right half. From each plot three random samples of plant roots with adhering soil were sampled at each time point, and excess soil was removed. The upper 10 cm of the roots (3-5 g) with tightly adhering soil were mixed with 10 ml sodium phosphate buffer (120 mm, ph 8). To t a l D N A e x t r a c t i o n One g of gravel was added and samples were vortexed for 30 s. The supernatant was decanted into a new tube. One ml of the supernatant was used to extract total DNA with the FastDNA SPIN Kit for Soil (Bio 101, Biogene, Vista, CA, USA) in combination with a Ribolyser (Hybaid, Ashford, UK) (Smit et al., 2003). The extracts were diluted 1:100 with Millipore-filtered distilled water before purification with the Wizzard DNA Clean-Up System (Promega, Madison, WI., USA) according to the manufacturer s protocol. DNA from the ascomycete isolates was extracted by using the Fast DNA Spin Kit (Bio 101) according to the supplier s manual, and subsequently purified with the Wizzard DNA Clean-Up System. P C R a m p l i f i c a t i o n o f t h e I T S r e g i o n Polymerase chain reaction (PCR) on purified DNA extracts was performed with primers specific for ascomycetes amplifying the ITS1, 5.8 S, and ITS 2 region of the fungal ribosomal RNA operon. The forward primer ITS5 (5 G GAAGTAAAAGTCGTAACAAGG-3 ) (White et al., 1990) with the reverse primer ITS4A (5 -CGCCGTTACTGGGGCAATCCCTG-3 ) (Larena et al., 1999) generate amplicons of about 700 bp. ITS4A was extended with a 40- base-pair G+C rich sequence to stabilize the melting behavior of the DNA

63 A s c o m y c e t e s i n w h e a t a n d p o t a t o r h i z o s p h e r e 6 3 fragments for DGGE analysis (Sheffield et al., 1989). Primers were synthesized by Eurogentec, Maastricht, The Netherlands. The 50 μl reaction mixture contained 1 μl of appropriately diluted DNA extracts, PCR buffer 2 (ph 9.2) containing 2.2 mm MgCl 2 (Roche Diagnostics, Mannheim, Germany), 200 μm each of datp, dttp, dgtp and dctp, 200 nm primers ITS5 and ITS4A, and 1.5 U Expand Long Template enzyme (Roche Diagnostics, Mannheim, Germany). Amplification was carried out in a Hybaid Thermocycler (Thermo Hybaid, Ashford, UK). The initial denaturing step was done at 94ºC for 5 min, followed by 35 cycles of 94ºC for 1 min, 60ºC for 1 min and 72ºC for 3 min. The reaction was terminated at 72ºC for 10 min. To check the size of the amplicons, PCR products were separated on a 1% agarose gel in TAE buffer (0.04 M Tris-acetate, M EDTA), run for 40 min at 150 V, and stained with the nucleic acid stain SYBR Green (Molecular Probes Inc., Leiden, The Netherlands) for 30 min. The gels were viewed under a blue light transilluminator (Clare Chemical Research, Dolores, CO, USA), and digitalized using a charged-coupled device (CCD) camera (Syngene, Cambridge, UK). D G G E PCR amplicons of the expected size from rhizosphere samples and ascomycete isolates were analyzed by DGGE. Denaturing gels were prepared with the gradient former Bio-Rad Model 230 (Bio-Rad Laboratories, Veenendaal, The Netherlands) at a speed of 4.5 ml min -1 as described earlier (Myers et al., 1987). The resulting gels consisted of 8% (wt / vol) polyacrylamide (0.5 TAE buffer, 37.5 : 1 ratio of acrylamide/bisacrylamide, cm) and had a gradient of 20 to 50% denaturant. A 100% denaturing acrylamide gel contained 7 M urea and 40% formamide. Up to 25 μl of PCR product per lane was loaded and gels were run for 17 h at 80 V at a constant temperature of 60 C in a DCode Universal Mutation Detection System (Bio-Rad Laboratories, Veenendaal, The Netherlands). For comparison of DNA patterns a reference marker was added in triplicate to each gel. After electrophoresis gels were stained and digitalized as described above for the agarose gels. Gels were analyzed with the BioNumerics program vers. 3.5 (Applied Maths, Sint-Martens-Latem, Belgium). After normalizing the gel and background subtraction, Pearson s correlation coefficient was calculated for each lane, followed by the UPGMA cluster analysis (unweighted pair group method using arithmetic averages). Dendrograms were generated from samples at a single time point, or combined from samples taken at three to five time points (composite dendrograms). The significance of the clusters was statistically analyzed using a permutation test with random sampling, which was designed specifically for that purpose. The distribution of the similarities was considered under the null hypothesis that the ascomycete communities of the wheat and

64 6 4 C h a p t e r 4 potato rhizosphere were identical. This null distribution was then calculated using 1000 random permutations. The confidence interval was 95%. C l o n i n g a n d s e q u e n c i n g o f D G G E b a n d Twelve dominant bands from gels loaded with samples taken 11 days after sowing / planting were excised from DGGE gels and eluted in 20 μl Milliporefiltered distilled water. One μl was used for re-amplification with the primers mentioned above using the following conditions: one cycle of 94ºC for 5 min, followed by 35 cycles of 94ºC for 1 min, 63ºC for 1 min and 72ºC for 3 min. The reaction was terminated at 72ºC for 10 min. After re-amplification the purity of the excised DNA fragments was checked in a final DGGE under the conditions described above. Appropriate PCR products were ligated into the pgem -T Easy Vector (Promega, Madison, WI, USA) and transformed into Escherichia coli Ultra Competent Cells (Stratagene, Cambridge, UK) as described by the manufacturers. Clones were directly resuspended in the PCR mixture and amplified with primer pair ITS5/ITS4A. The amplified DNA fragments were identified by DGGE as described. Only clones with a migration comparable to the excised band were used for further sequence analysis. Prior to sequencing, the PCR products were purified with the QIAquick PCR Purification Kit (Qiagen, The Netherlands), according to the manufacturer s protocol. The sequencing reaction was carried out with the BigDye Terminator v.3.1 Cycle Sequencing Kit in an AB3700 Sequencer (Applied Biosystems, Nieuwerkerk, The Netherlands). Two clones per band were sequenced in both directions with primers ITS5 and ITS4A, respectively. Sequences obtained were checked for chimeras using the program CHIMERA_CHECK ver. 2.7 (Cole et al., 2003). Sequences that most likely did not contain chimeric regions were compared with known ascomycete sequences of relevant databases with the Standard nucleotide Blast program (Altschul et al., 1997) provided by the National Center for Biological Information (NCBI, USA). Sequences of the clones were submitted to GenBank with accession numbers AY to AY C o m p a r i n g D G G E g e l s w i t h a d e f i n e d m a r k e r Sequencing of the twelve excised DGGE bands revealed matches for ten bands with ascomycete sequences from relevant databases, and these bands were used as a reference marker. DGGE was performed with rhizosphere samples taken 11, 25, 39 and 54 days after sowing wheat in the experimental field in Prior to PCR samples from three replicate plots were pooled, yielding two replicates per treatment, and samples from six continuous wheat plots were compared with samples from six wheat-potato-wheat rotation plots. DNA was extracted and DGGE was performed as described above.

65 A s c o m y c e t e s i n w h e a t a n d p o t a t o r h i z o s p h e r e 6 5 M M M Fig. 1. DGGE profile of ITS rdna fragments of ascomycete species generated by PCR with ascomycete specific primers. Lane 1: Alternaria brassisicola, lane 2: Acremonium murorum, lane 3: Gaeumannomyces graminis var. tritici (G.g.t.), lane 4: G.g.t. (PCR product 1:10 diluted), lane 5: Chaetomium bostrychodes, lane 6: Trichoderma harzianum B28, lane 7: Verticillium longisporum, lane 8: Muratella elegans, lane 9: Monocillium mucidum. M = marker. R e s u l t s D G G E a n a l y s i s o f a s c o m y c e t e i s o l a t e s a n d p r e d o m i n a n t a s c o m y c e t e c o m m u n i t i e s i n t h e w h e a t a n d p o t a t o r h i z o s p h e r e s To evaluate the specificity of DGGE, DNA extracted from different ascomycete isolates were analyzed. Fig. 1 shows the separation of the ITS amplicons of eight different ascomycetes. Most isolates showed a single discrete band. However, the bands in lane 2, 8, and 9 were fuzzy. Total DNA from rhizosphere samples of six wheat plots and six rotation plots was extracted and analyzed by DGGE. For analysis samples of the three plots from each half were pooled, resulting in two replicates per treatment. Fig. 2 shows the results from the cluster analysis of the four consecutive years (1999 to 2002). In 1999 wheat was sown in both the control and the rotation plots. Samples from the left half of the field (marked a) clustered apart from samples of the right half (b) at a similarity index of only 30% (Fig. 2A). The similarity of both samples within each cluster was 40 and 52%, respectively. In 2000, with potato planted in the rotation plots, there was a clear differentiation of the ascomycete communities in the rhizospheres of wheat and potato (Fig. 2B). All samples from the rhizosphere of potatoes clustered separate from those of wheat. The 2001 samples clustered in the same way as in 2000, even though now wheat was sown again in the rotation plots (Fig. 2C). The similarity between the two clusters was only 2%. In 2002 the dendrogram from the wheat and potato rhizosphere samples was again comparable, with only about 2% similarity between the potato and the wheat clusters (Fig. 2D). The ascomycete communities from the continuous wheat plots and the rotation plots were

66 6 6 C h a p t e r 4 A: 1999 B: 2000 % similarity % similarity wheat b (cw) wheat b (rp) wheat a (cw) wheat a (rp) wheat a (cw) wheat b (cw) potato a (rp) potato b (rp) C: 2001 D: 2002 % similarity % similarity wheat a (cw) wheat b (cw) wheat a (rp) wheat b (rp) wheat a (cw) wheat b (cw) potato a (rp) potato b (rp) Fig. 2. Dendrograms representing the percentage genetic similarity of rhizosphere ascomycete communities in continuous wheat ( ) and in a crop rotation of wheat and potato ( ). In the rotation plots wheat was grown in 1999 (A) and 2001 (C), and potatoes were planted in 2000 (B) and 2002 (D). Each dendrogram is a combination of multiple dendrograms from samples taken at the earliest three to four time points per year. Similarities were based on DGGE patterns generated from ITS1 / ITS2 rdna fragments using Pearson s correlation coefficient. Cluster analysis was done with UPGMA. Samples from three of six plots of continuous wheat culture and a wheat-potato rotation, respectively, were pooled, resulting in two replicates per cultivar, from the left (a) and from the right half of the field (b). rp = rotation plots. significantly different ( > P < 0.036) in 2000 and in the years thereafter. In 1999 the spatial heterogeneity had a significant effect on the ascomycete communities (P = 0.047). The pronounced differences in the ascomycete communities of the rhizospheres of wheat and potato led us to a more thorough investigation of samples taken at 11 and 25 days after sowing in Fig. 3A and B show the dendrograms based on the DGGE patterns of all six replicate samples of the rhizosphere of wheat and potato. After 11 days, three main clusters were defined at a cut-off value of 50% similarity. One cluster contained only the rhizosphere samples from wheat (Fig. 3A). Five of the six replicates from the wheat rhizosphere were at least 65% similar. The remaining two clusters contained only the samples from the potato rhizosphere. The three replicates from the left half of the field (potato 1, 2, 3) formed one cluster and were at least 63% similar. The three replicates from the right half (potato 4, 5, 6) formed the other cluster and differed by no more than 15%. Analysis of samples taken at the later time point revealed a largely comparable clustering (Fig. 3B). Again, overall similarity was less than 10%. At a cut-off value of 50% five clusters were distinguished. One cluster contained only one replicate from wheat, with almost no similarity to any of the other samples. The other clusters each contained only rhizosphere samples from either wheat or potato.

67 A s c o m y c e t e s i n w h e a t a n d p o t a t o r h i z o s p h e r e 6 7 A: day 11 % similarity wheat 3 wheat 4 wheat 5 wheat 6 wheat 1 wheat 2 potato 1 potato 2 potato 3 potato 5 potato 6 potato 4 B: day 25 % similarity wheat 1 wheat 3 wheat 6 wheat 5 wheat 2 wheat 4 potato 3 potato 6 potato 4 potato 5 potato 1 potato 2 Fig. 3. Dendrograms representing the percentage genetic similarity of the ascomycete communities from the rhizosphere of wheat ( ) and potato ( ) 11 (A) and 25 (B) days after sowing in In Fig. A the corresponding gel is shown. Samples originated from plots of continuous wheat culture and a wheat-potato rotation. Cluster analysis was based on the UPGMA linkage of Pearson s correlation coefficients of DGGE profiles obtained from six replicate samples of wheat and potato. The arrows indicate the similarity level, at which clusters were defined. Furthermore, DGGE gels profiling the ascomycete communities revealed only a small number of bands per treatment, varying between four and twelve. Statistically, the ascomycete communities of wheat and potato were significantly different at both days after sowing (P = and P = for samples taken at 11 and 25 days after sowing, respectively).

68 6 8 C h a p t e r 4 M W W P P M * * * * 1 2 * 3 * 4 * * * 5 6* * * Fig. 4. DGGE profile of DNA fragments from rhizosphere samples using ascomycete specific primers. DNA was directly extracted from the rhizosphere of wheat and potato 11 days after sowing in Excised and sequenced bands are labeled (*): W1-W6 refer to DNA isolated from wheat after 4-year monoculture, bands P7-P12 refer to DNA isolated from potato after alternating wheat-potato culture. M = marker S e q u e n c e a n a l y s i s o f I T S f r a g m e n t s r e c o v e r e d f r o m D G G E g e l s To further characterize the differences in the ascomycete communities of the wheat and potato rhizospheres, twelve differentially migrating DGGE bands were excised (Fig. 4), re-amplified and cloned into a vector. At least four clones per band were subjected to DGGE under the same standardized conditions, and their migration compared to that of the excised band. Two clones per band with exactly the same position in the DGGE gel were used for further sequence analysis. Two of the twelve bands were not successfully recovered and excluded from further analysis. Table 1 shows the best match of the ten sequences of the recovered bands with known ascomycete sequences. Two sequences (bands 9, 11) did not match any of the known sequences and, therefore, could not be identified. The other sequences all represent ascomycetes that are common in agricultural soils (Domsch et al., 1980). Six bands showed high similarities (at least 97%) with the genera Issatchenkia, Plectosphaerella, Podospora, and Verticillium. Two bands showed a modest similarity of 95 and 93% to Chaetomium sp. and Ericoid mycorrhiza, respectively. In two occasions two bands with different melting behavior and, thus, different sequences resembled the same species. V. nigrescens sequences 1 and 2 differed by 1% of their sequence, P. cucumerina sequences 1 and 2 by 3%.

69 A s c o m y c e t e s i n w h e a t a n d p o t a t o r h i z o s p h e r e 6 9 Table 1. Characterization of partial ITS sequences recovered from bands of DGGE profiles with DNA extracted from wheat and potato rhizosphere 11 days after sowing in Band Most similar sequence GenBank Similarity no. a Accession no. (%) 1 n.d. b 2 Verticillium nigrescens Seq. c 1 AJ Podospora sp. AF Verticillium nigrescens Seq. 2 AJ Plectosphaerella cucumerina d Seq. 1 AJ Issatchenkia orientalis e AF Plectosphaerella cucumerina d Seq. 2 AJ n.d. b 9 unidentified 10 Chaetomium sp. AJ unidentified 12 Ericoid mycorrhiza AY a b c d e Band no. refers to the excised DGGE bands. Band no. 1 to 6 originated from the wheat rhizosphere, no. 7 to 12 from the potato rhizosphere. n.d. = not determined Seq. = Sequence the current valid name is Monographella cucumerina (Lindf.) Arx the current valid name is Candida krusei (Castell.) Berkhout A s c o m y c e t e c o m m u n i t i e s o f t h e w h e a t r h i z o s p h e r e a f t e r m o n o c u l t u r e a n d r o t a t i o Ascomycete communities from the rhizospheres of wheat from the preceding three-year monoculture and from the wheat-potato-wheat rotation were compared to the ten ascomycetes identified from the 2002 samples (Table 1). Given the fact that the samples were derived from the same field, there is a high probability that corresponding bands represent the same species. Based on this assumption, Table 2 shows the different ascomycetes from the continuous wheat and the wheat-potato-wheat rotation plots in the season of 2001 that were identified according to their position in the DGGE gel in comparison with the characterized bands. Two species, Chaetomium sp. and I. orientalis, were found only in the rotation plots. V. nigrescens sequence 2 and one unidentified species did not occur at all prior to All other species were detected in both treatments. Within the season some changes became apparent. In the continuous wheat plots P. cucumerina sequence 1 was detected only at 11 days, one of the unidentified species (band 9) at the first two sampling dates, and the Ericoid mycorrhiza at 11, 39 and 54 days after sowing. In the rotation plots Podospora occurred at the first sampling date only, the Ericoid mycorrhiza at the last two sampling dates, and P. cucumerina sequence 2 and V. nigrescens sequence 1 intermittently at 11, 39 and 54 days after sowing.

70 7 0 C h a p t e r 4 Table 2. Ascomycetes of the rhizosphere of wheat as identified by comparison of DGGE bands. Samples were taken at 11, 25, 39 and 54 days after sowing during the season of 2001 from continuous wheat plots (cw) and from rotation plots of wheat-potato-wheat (wpw) cultivation. w and p indicate that samples, from which the excised DGGE bands were recovered, originated from the rhizosphere of wheat (w) or potato (p) in Band no. Ascomycetes days after sowing cw wpw w I. orientalis x x x x 10 p Chaetomium sp. x x x x 5 w P. cucumerina Seq. a 1 x x x x x 11 p unidentified x x x x x x 7 p P. cucumerina Seq. 2 x x x x x x x 2 w V. nigrescens Seq. 1 x x x x x x x 12 p Ericoid mycorrhiza x x x x x 3 w Podospora sp. x x x x x 4 w V. nigrescens Seq. 2 9 p unidentified a Seq. = Sequence D i s c u s s i o n Currently, various genetic fingerprint techniques are used in microbial ecology. They have provided patterns of the microbial community diversity in different habitats, but it is not clear, which technique is yielding better resolution or sensitivity. Torsvik et al. (1998) compared the microbial diversity obtained by different fingerprinting methods with DNA melting profiles and reassociation kinetic studies. They found that ARDRA must be considered too sensitive for a reliable characterization of complex microbial communities, which contain too much information to be reliably assessed, and that fingerprinting methods such as DGGE are better suited for that purpose. Liesack and Dunfield (2002), on the other hand, argued that DGGE is often too sensitive for the determination of total community fingerprints. Previously, it has been demonstrated that crop species have a pronounced effect on the composition of the microbial communities (Berg et al., 2002; Smalla et al., 2001; Oehl et al., 2003; Olsson and Alström, 2000). In a recent study, using ARDRA, we could not detect that a crop rotation from wheat to potato influenced the predominant bacterial and fungal communities (Viebahn et al., 2003). However, by re-examining the samples using DGGE for comparing genetic fingerprints of PCR-amplified ITS regions of ascomycete communities, distinct clusters were evident, one comprising all samples from the potato rhizosphere and one containing all wheat rhizosphere samples. Ascomycete communities were analyzed since the separation of 18S

71 A s c o m y c e t e s i n w h e a t a n d p o t a t o r h i z o s p h e r e 7 1 rdna amplicons on a denaturant gradient did not lead to proper gels. It demonstrates that the two plant species did have an effect on the composition of the rhizosphere communities. It also suggests that use of different molecular techniques can lead to different conclusions. In the present study, using DGGE, it was shown that in the first year of cultivating wheat the ascomycete communities differed according to their location in the field. Apparently, the soil and its environment are not homogeneous, which caused a differentiation between ascomycete communities from the rhizosphere of the same plant species, positioned in the left or the right half of the field. Crop rotation from wheat to potato in the second year led to a marked difference between the ascomycete communities of the two plant species. This effect was still evident after cultivating again wheat in the rotation plots, suggesting that a single change of wheat to potato had a long-term impact on the microflora. Thus, the effect of crop rotation clearly exceeded that of spatial heterogeneity in the field. In addition, DGGE was performed on six non-pooled replicate rhizosphere samples from wheat and potato taken 11 and 25 days after sowing in Results from both DGGE analyses were consistent and in agreement with the earlier results. Although DGGE can be used successfully to assess the diversity of fungal communities (Gomes et al., 2003; Schabereiter-Gurtner et al., 2001; Kowalchuk et al., 1997; Vainio and Hantula, 2000), the method provides no information about the species composition. This information can be obtained by sequencing individual bands of the DGGE gels and comparison with known sequences from the public databases. Sequencing of DGGE bands revealed only members of the phylum Ascomycota, confirming the specificity of the primers. Ten of the 12 excised bands showed unambiguous results and were included in the analysis. Two of the ten bands could not be identified. One likely explanation is that the number of sequences of the ITS region in the public databases are limiting. Two clearly distinct bands in the DGGE gel resembled V. nigrescens, whereas two other bands resembled P. cucumerina. The sequences differed by 1 and 3%, respectively. This indicates that the described species can be heterogeneous or that the names deposited in the databases are inconsistent. All identified ascomycetes are common in agricultural soils. Three of these ascomycetes are reported to be potentially pathogenic to plants. Chaetomium species are often found as decomposers of herbaceous and lignified plant material and have been recognized as the causal agent of soft rot in wood (Domsch et al., 1980). P. cucumerina has been almost exclusively associated with arable soil and has been isolated from the rhizosphere of flax, grass, potato, sugar beet and wheat (Domsch et al., 1980). It can produce necrotic spots on potato stalks and cause wilting in beets (Domsch et al., 1980). V. nigrescens is

72 7 2 C h a p t e r 4 a common soil fungus in arable and non-cultivable soil. It has been found to attack spearmint and peppermint (Domsch et al., 1980). Using DGGE we compared patterns of the ascomycete communities of the continuous wheat and the rotation sampling sites with ascomycete reference markers resembling species identified from the same sites, but in different years. Bands that correspond to bands of the markers were assumed to be the ascomycete represented by the marker band. We are aware that this is a very simplified method to assess differences in the microbial communities. However, it seems to be justified, since all material was sampled from the same site, and only clones with unambiguous positions in the gel before and after cloning and with high sequence identity were used as marker bands. Two species, I. orientalis and Chaetomium sp., were detected only in the rhizosphere of wheat after crop rotation, but not in the rhizosphere of wheat monoculture. The other ascomycetes occurred in both sampling sites at one or more time points. V. nigrescens sequence 2 and an unidentified strain did not occur in 2001 but were found in Comparable investigations support the assumption that crop rotation practices in agriculture influence the composition of the microbial communities. Olsson and Alström (2000) analyzed the fatty acid profiles of rhizobacterial populations on barley roots from monoculture and from an 8-year crop rotation (fallow, winter rape, winter wheat, peas, spring barley, ley, ley and oat). The fatty acid profiles of the microbial populations from the monocropping were significantly different from that of the rotation treatment. Oehl et al. (2003) Investigated the impact of land use on the diversity of arbuscular mycorrhizal fungi (AMF). The highest AMF species number and diversity was found in the low input, organically managed arable land under crop rotation. They identified 18 AMF species from land under crop rotation, and 8-13 AMF species from land under maize monocropping using trap cultures. It has also been shown that plants have a selective influence on bacterial populations (Glandorf et al., 1993; Lemanceau et al., 1995). Van Elsas et al. (2002) used DGGE analysis to show a clear effect on the bacterial and fungal microflora of arable land under oat-maize rotation, maize monoculture, and permanent grassland. The microbial diversity was higher in arable land under crop rotation than under monoculture, and higher in grassland than in arable land. We report differences in diversity of ascomycete communities in wheat monoculture and in a wheat-potato rotation system in a long-term field trial. This information will aid in assessing the relative importance of possible changes of microbial rhizosphere communities induced by perturbations of ecosystems.

73 A s c o m y c e t e s i n w h e a t a n d p o t a t o r h i z o s p h e r e 7 3 A c k n o w l e d g e m e n t Financial support of M.V. by the Dutch Ministry of Housing, Spatial Planning and the Environment is gratefully acknowledged. We thank Niko Nagelkerke for assistance with the statistical analysis and Bas Valstar and Fred Siesling (Botanical Garden, Utrecht University) for constructing and maintaining the experimental field site.

74

75 Chapter 5 Ascomycete communities in the rhizosphere of field-grown wheat are not affected by introductions of genetically modified Pseudomonas putida WCS358r Mareike Viebahn, Rogier Doornbos 1, Karel Wernars 1, Leendert C. van Loon, Eric Smit 1, and Peter A.H.M. Bakker 1 National Institute of Public Health and the Environment, Bilthoven, The Netherlands Environmental Microbiology, in press

76 7 6 C h a p t e r 5 A b s t r a c t A long-term field experiment ( ) was conducted to monitor effects on the indigenous microflora of Pseudomonas putida WCS358r and two transgenic derivatives constitutively producing phenazine-1-carboxylic acid (PCA) or 2,4-diacetylphloroglucinol (DAPG). The strains were introduced as seed coating on wheat into the same field plots each year. Rhizosphere populations of ascomycetes were analyzed using denaturing gradient gel electrophoresis (DGGE). To evaluate the significance of changes caused by the genetically modified microorganisms (GMMs), they were compared to effects caused by a crop rotation from wheat to potato. In the first year, only the combination of both GMMs caused a significant shift in the ascomycete community. After the repeated introductions this effect was no longer evident. However, cropping potato significantly affected the ascomycete community. This effect persisted into the next year when wheat was grown. Clone libraries were constructed from samples taken in 1999 and 2000, and sequence analysis indicated ascomycetes of common genera to be present. Most species occurred in low frequencies, distributed almost evenly in all treatments. However, in 1999 Microdochium occurred in relatively high frequencies, whereas in the following year no Microdochium species were detected. On the other hand, Fusarium-like organisms were low in 1999, and increased in Both the DGGE and the sequence analysis revealed that repeated introduction of P. putida WCS358r had no major effects on the ascomycete community in the wheat rhizosphere, but demonstrated a persistent difference between the rhizospheres of potato and wheat. I n t r o d u c t i o n Worldwide losses of cultivated crops due to plant diseases amount to 12% (Agrios, 1997). In many cases plant pathogens can be controlled effectively by chemicals. However, not all plant diseases can be controlled chemically. Moreover, noxious pesticides will be taken from the market in the near future, e.g. methyl bromide (Martin, 2003). Therefore, development of alternative methods of crop protection is necessary. One alternative is the use of biological control agents (BCAs) to suppress plant pathogens. In this respect, fluorescent Pseudomonas spp. have received special attention, because of their excellent root-colonizing ability, and their potential to produce a wide variety of antimicrobial metabolites (O Sullivan and O Gara, 1992). Selected strains have been shown to suppress various soilborne plant pathogens, such as Fusarium oxysporum, Gaeumannomyces graminis var. tritici, Phytophtora cinnamoni, and Rhizoctonia solani (Van Loon and Bakker, 2003; Weller et al., 2002). Despite decades of research, the commercial use of BCAs is still limited. Their commercial development is being hampered by failure or inconsistent performance of the BCAs under field conditions. Reasons for the latter include variable root colonization, loss of ecological competence, and low or late production of disease-suppressing metabolites (Weller, 1988). To circumvent these limitations, genetically modified BCAs with enhanced biocontrol activity can be constructed. Despite the potential benefits of such genetically modified microorganisms (GMMs), their use is often associated

77 A s c o m y c e t e c o m m u n i t i e s n o t a ff e c t e d b y P. p u t i d a W C S r 7 7 with public concern. There are several potential effects of introducing BCAs in the environment that may, or may not, be linked to an engineered trait: (i) displacement of non-target organisms, (ii) allergenicity to humans and animals, (iii) toxigenicity, and (iv) pathogenicity to non-target organisms (Cook et al., 1996). Effects of BCAs on bacterial community structure have been repeatedly found (Bakker et al., 2002; Glandorf et al., 2001; Sigler et al., 2001; Viebahn et al., 2003; Winding et al., 2004). However, reports on effects of introduced BCAs on the fungal microflora are scarce. In this study effects of genetically modified pseudomonads on rhizosphere ascomycete communities were monitored in a long-term field experiment. P. putida WCS358r and two genetically modified derivatives were introduced into the rhizosphere of field-grown wheat in four consecutive years ( ). The recombinant strains constitutively produced either phenazine-1- carboxylic acid (PCA) or 2,4-diacetylphloroglucinol (DAPG) (Viebahn et al., 2003), secondary metabolites of low molecular weight with a broad-spectrum activity against soilborne plant pathogens. In previous studies, a one-time field introduction of the PCA- and DAPG-producing derivatives of WCS358r significantly influenced the fungal and bacterial microflora, as detected by amplified ribosomal DNA restriction analysis (ARDRA) (Viebahn et al., 2003). Here, we focus on the group of ascomycetes, since an important member of this group, Fusarium, was shown to be affected by the introduction of WCS358r and its derivatives (Glandorf et al., 2001; Leeflang et al., 2002). We describe the application of denaturing gradient gel electrophoresis (DGGE) for analysis of genetic fingerprints derived from the amplification of the internal transcribed spacer regions (ITS) of rdna with ascomycete-specific primers. Possible effects caused by the introduction of the GMMs were compared to effects caused by crop rotation of wheat and potato. Crop rotation is commonly used in agriculture to control plant pathogens (Cook, 1992). It is well accepted that different crops establish different microbial communities in their rhizospheres (Lemanceau et al., 1995; Oehl et al., 2003; Olsson and Alström, 2000). In addition to DGGE analysis of ITS rdna amplicons, the same amplicons were used to construct clone libraries, which provide taxonomic information reflecting the ascomycete community composition (Kent and Triplett, 2002). M a t e r i a l a n d M e t h o d s B a c t e r i a l s t r a i n s The bacterial strains used in this study are P. putida WCS358r (Geels and Schippers, 1983), and two transgenic derivatives (Glandorf et al., 2001; Viebahn

78 7 8 C h a p t e r 5 et al., 2003). The PCA-producing derivative, WCS358r::phz, contained the phzabcdefg gene cluster on the disarmed mini Tn5 transposon-based vector putkm (Herrero et al., 1990) inserted into the chromosome. The DAPG-producing GMM was constructed by insertion of the phlfacbde genes from P. fluorescens Q2-87 (Bangera and Thomashow, 1996), a naturally DAPG-producing strain, into the chromosome of WCS358r, and designated WCS358r::phl (Viebahn et al., 2003). The wild-type strain WCS358r was grown on King s Medium B (KB) agar (King et al., 1954) supplemented with 150 µg rifampicin ml -1, and the transgenic derivatives on KB with 150 µg rifampicin ml -1 and 30 µg kanamycin ml -1. All strains were routinely grown at 28ºC for 2 days. S e e d t r e a t m e n t a n d e x p e r i m e n t a l f i e l d Wheat seeds (Triticum aestivum cv. Baldus) were sown in four consecutive years (1999 to 2002) in an experimental field located near the Botanical Garden of Utrecht University, The Netherlands, as described by Viebahn et al. (2003). Permission for the deliberate field release of GMMs was granted by the Dutch Ministry of Housing, Spatial Planning and the Environment, The Netherlands (BGGO 98/10). Wheat seeds were treated with a 1:1 mixture of washed bacterial suspensions and 3% methylcellulose, as described by (Glandorf et al., 2001) The bacterial treatments used were WCS358r, WCS358r::phz, WCS358r::phl, or a 1:1 mixture of WCS358r::phz + WCS358r::phl. For the control treatment the bacterial suspension was replaced by 10 mm MgSO 4. Coated seeds were air-dried overnight and sown the next day. Coating resulted in approximately 10 7 colony forming units per seed, as determined from plate count enumerations. An additional treatment was a rotation plot, in which in the first (1999) and third year (2001) non-treated wheat seeds were sown, and in the second (2000) and fourth year (2002) non-treated potatoes (Solanum tuberosum L. cv. Modesta) were planted. The bacterial strains were introduced every year on seeds in the same plots, and rhizosphere soil samples were taken each year at three to six different time points. D N A e x t r a c t i o n a n d P C R From each plot 3-5 g of roots with adhering soil were sampled at the different time points during the growing season and mixed with 10 ml sodium phosphate buffer (120 mm, ph 8). One g of gravel was added and samples were vortexed for 30 s. The supernatant was decanted into a new tube. One ml of the supernatant was used to extract total DNA with the FastDNA SPIN Kit for Soil (Bio 101, Biogene, Vista, CA, USA) in combination with a Ribolyser (Hybaid, Ashford, UK) (Smit et al., 2003). The extracts were resuspended in 100μl Millipore-filtered distilled water and purified with the Wizzard DNA Clean-

79 A s c o m y c e t e c o m m u n i t i e s n o t a ff e c t e d b y P. p u t i d a W C S r 7 9 Up System (Promega, Madison, WI, USA.) according to the manufacturer s protocol. The six replicate DNA samples of each treatment were pooled into two samples, each containing three replicates, and DNA was amplified with the forward primer ITS5 (5 -GGAAGTAAAAGTCGTAACAAGG-3 ) (White et al., 1990) and the reverse primer ITS4A-GC (5 -CGCCCGCCGCGCCCCG CGCCCGGCCCGCCGCCCCCGCCCCGCCGTTACTGGGGCAATCC CTG-3 ) (Larena et al., 1999). The primer pair amplified a fragment of about 700 bp of the ITS1/ITS2 region of ascomycete rdna. ITS4A-GC contained a 40 bp GC-clamp to stabilize the melting behavior of the DNA fragments for DGGE analysis (Sheffield et al., 1989). The PCR was performed in 10x PCR buffer 2 (ph 9.2) containing 2.25 mm MgCl 2 (Roche, Diagnostics, Mannheim, Germany), 250 µm of each of the four deoxynucleoside triphosphates, 200 nm of each primer, 2.5 U Expand Long template enzyme (Roche, Diagnostics, Mannheim, Germany), and 1 µl of appropriate diluted template DNA in a total volume of 50 µl. The PCR conditions used in the thermocycler (Hybaid, Ashford, UK) were: 5 min 94 C, followed by 35 cycles of 1 min 94 C, 1 min 60 C, and 3 min at 72 C, and finally 10 min 72 C. Amplified PCR fragments were checked for the correct size on a 1% agarose gel. D G G E The PCR fragments were separated on a denaturing gradient polyacrylamide gel containing 0.5 TAE buffer (20 mm Tris-acetate, 10 mm sodium acetate, 0.5 mm EDTA). The gel consisted of 8% (wt vol -1 ) polyacrylamide (acrylamide/bisacrylamide in a ratio of 37.5:1) with a denaturant gradient of 20 to 50%. One hundred % stock solution contained 7 M urea and 40% formamide. After correction for concentration up to 25 µl per lane of PCR product were loaded, and gels were run for 17 h at 80 V at a constant temperature of 60 C in a DCode Universal Mutation Detection System (Bio-Rad Laboratories, Veenendaal, The Netherlands). For comparison a reference DNA marker, consisting of 18S rdna amplicons of 7 different ascomycete species, was run in triplicate on each gel. After electrophoresis gels were stained for 30 min in 1:10000 diluted SybrGold (Molecular Probes, Leiden, The Netherlands) and viewed on a blue-light transilluminator (Clare Chemical Research, Dolores, CO, USA). Images were digitalized using the GeneGenius Bio Imaging System (Syngene, Cambridge, UK). C l u s t e r A n a l y s i s The ascomycete community fingerprints of the DGGE gels were analyzed with the BioNumerics program vers. 3.5 (Applied Maths, St. Marten-Latem, Belgium). After normalization and background subtraction Pearson s correlation coefficient was used to calculate the similarity between each banding pattern.

80 8 0 C h a p t e r 5 Patterns were grouped into clusters by the unweighted pair-group method using average linkages (UPGMA). Significant clusters in the dendrograms were determined by calculating the cut-off value that produced the highest point-bisectional correlation (BioNumerics manual, version 3.5). Briefly, a line is drawn through the dendrogram at a certain similarity level, and from the resulting number of clusters defined by that line, a new similarity matrix is created, in which all within-cluster values are 100%; all between-cluster values are 0%. Then, the correlation between this new matrix and the original matrix is calculated, which is called the point-bisectional correlation. The same is done for different cut-off similarity levels, and the level with the highest pointbisectional correlation is the one defining the significant clusters. C l o n i n g a n d s e q u e n c i n g Clone libraries were constructed from samples taken in 1999 from the left and the right half of the field, and from samples taken in 2000 from the right half of the field. Two µl of the PCR products (see above) was ligated to the pgem -T Easy Vector (Promega, Madison, WI, USA) and transformed into E.coli JM109 Ultra Competent Cells (Stratagene, Cambridge, UK) according to the manufacturer s protocol. The bacteria were plated on LB agar supplemented with 50 µg ml -1 ampicillin, 80 µg ml -1 X-gal (5-bromo-4-chloro- 3-indolyl-β-D-galactopyranoside) and 20 mm IPTG (isopropyl-1-thio- β-dgalactopyranoside). For each sample 60 white colonies were resuspended in the PCR mixture and amplified directly using the primer pair ITS5 and ITS4A under the conditions as described above. PCR products of each clone were analyzed on a 1% agarose gel and fragments of the correct size (approximately 700 bp) were purified using the QIAquick PCR Purification Kit (Quiagen, Venlo, The Netherlands) according to the manufacturer s protocol. The sequencing reaction was carried out with the BigDye Terminator v.3.1 Cycle Sequencing Kit in an AB3700 Sequencer (Applied Biosystems, Nieuwerkerk, The Netherlands). Clones were sequenced in both directions with the reverse primer ITS5 and the forward primer ITS4A. Both sequences from each clone were aligned with the program Kodon vers. 2.0 (Applied Maths, St. Marten- Latem, Belgium) and sequences were compared with known ascomycete sequences from GenBank with the Standard nucleotide Blast program (Altschul et al., 1997) provided by the National Center for Biological Information (NCBI, USA). R a r e f a c t i o n Rarefaction (Krebs, 1989) was performed to estimate the expected number of ascomycete species in random samples of a given size. Based on the number of different species a rarefaction curve was calculated from

81 A s c o m y c e t e c o m m u n i t i e s n o t a ff e c t e d b y P. p u t i d a W C S r 8 1 rhizosphere samples taken from wheat plants treated with WCS358r in The Analytical Rarefaction Program vers. 1.3 used was made available by Steve Holland, University of Georgia Startigraphy Lab ( uga.edu/~strata /software). R e s u l t s To determine shifts in the ascomycete community upon introduction of the GMMs, DGGE was applied to rhizosphere samples of wheat plants that were grown from seeds treated with wild type or genetically modified P. putida WCS358r. A typical DGGE gel with banding pattern of ascomycete ITS1/ITS2 amplicons from differentially treated wheat rhizosphere samples (2000, day 59) is shown in Figure 1. The dendrograms (Fig. 2) represent the percentage similarity of the cluster analysis of the patterns from rhizosphere samples taken at 13, 25, 40, and 54 days after sowing in Samples from the different treatments were grouped in either three or four clusters throughout the growing season. However, significant variability was evident, as indicated by the different clustering at different times. Moreover, differences between the samples throughout the season were high, as indicated by the low similarity indices (0-20%). At the beginning of the growing season mainly samples of the non-treated wheat rhizospheres clustered together (Fig. 2A). All other samples grouped at a similarity of 47%, with the exception of one replicate of the M contr a b wt phz phl phl phz+phl rp a b a b a M b a b a b M Fig. 1. DGGE profiles showing ascomycete communities of rhizosphere samples of field-grown wheat plants 59 days after sowing in Plants were grown from seeds that were untreated (contr), treated with P. putida WCS358r (wt), its PCA- and DAPG-producing derivatives (phz, phl), and a combination of both (phz+phl). Two replicate samples (a, b) were analysed. rp = samples taken from a rotation plot, in which potatoes were planted in M = reference marker.

82 8 2 C h a p t e r 5 treatment with the combination of the GMMs. At day 25 the two samples from the combination treatment with both GMMs formed a distinct cluster (Fig. 2B). All other samples clustered at a similarity of 30%, with one replicate of the DAPG-treatment and one replicate of the rotation plot each representing a distinct cluster. At 40 days two main clusters could be distinguished with only 14% similarity, while similarity within the clusters was around 30% (Fig. 2C). One cluster contained all replicates from the treatments with the DAPGproducing derivative of WCS358r, including those from the treatment with the combination of the GMMs, and two samples from non-treated wheat. The second cluster contained samples of the treatment with wild type WCS358r A: day 13 B: day 25 % similarity % similarity phz+phl b wheat b (rp) contr b wheat a (rp) phz b wt b phl b contr a phz a wt a phl a phz+phl a phz a wt b contr a wt a contr b phz b phl a wheat a (rp) wheat b (rp) phl b phz+phl a phz+phl b C: day 40 D: day 54 % similarity % similarity wt b wheat b (rp) phz b contr b wt a phz a phl a wheat a (rp) contr a phl b phz+phl a phz+phl b contr a wt a phl a phz b phz+phl a phl b phz+phl b wheat b (rp) phz a contr b wt b wheat a (rp) Fig. 2. Dendrograms based on the genetic similarity of the ascomycete communities of fieldgrown wheat plants in 1999 at 13 (A), 25 (B), 40 (C), 54 (D) days after sowing. For treatments see legend to Table 1. Samples from three plots were pooled, resulting in two replicates per treatment, from the left (a) and from the right half (b) of the field. rp = samples taken from a rotation plot, in which wheat was grown in Similarities are based on DGGE patterns generated from amplified sequences of the ITS1/ITS2 regions of rdna using Pearson s correlation coefficient. Cluster analysis was done with UPGMA. Significant clusters are indicated by the grey lines and were separated by the point-bisectional cut-off method. The level of similarities is shown above the dendrograms.

83 A s c o m y c e t e c o m m u n i t i e s n o t a ff e c t e d b y P. p u t i d a W C S r 8 3 and with the PCAproducer, and the other two samples from non-treated wheat. At 54 days one main cluster was observed at a similarity level of 68% (Fig. 2D). In this cluster samples from all treatments were present. Only three samples, two from non-treated wheat and one from the treatment with the wild type WCS358r, fell outside the cluster, but were also significantly different from each other. Thus, treatments with the combination of the PCA- and the DAPG-producing derivatives clustered away from the other treatments, indicating a significant effect on the ascomycete community after a one-time introduction. Neither the treatment with the PCA producer, nor that with the DAPG producer alone had an effect that exceeded the variation in the control treatments. A: 1999 B: 2000 % similarity % similarity phz a wt b contr a phz b wt a phl a phl b contr b wheat b (rp) wheat a (rp) phz+phl a phz+phl b contr a wt a phz a phz+phl a phl a contr b phz+phl b wt b phl b phz b potato a (rp) potato b (rp) C: 2001 D: 2002 % similarity % similarity wt b phz b contr b phl b phz+phl b contr a phl a wt a phz a phz+phl a wheat a (rp) wheat b (rp) wt a phl a phl b phz+phl b phz b phz+phl a contr a contr b wt b phz a potato a (rp) potato b (rp) Fig. 3. Composite dendrograms representing the genetic similarity of the ascomycete communities of field-grown wheat plants in 1999 (A), 2000 (B), 2001 (C), and 2002 (D). Dendrograms are derived from the combined analysis of samples taken at three to four sampling dates. For further explanation see legend to Table 1 and Fig. 2.

84 8 4 C h a p t e r 5 Similar dendrograms were constructed from samples taken at four time points in 2000 and 2001, and three time points in 2002 (data not shown). The field trial was destroyed by activists in June 2002, resulting in only three sampling dates. The overall similarity of the samples varied between 5 and 38% in 2000, between 0.3 and 9% in 2001, and between 0.4 and 4% in To visualize the overall effects in each year, we constructed composite dendrograms. Composite dendrograms represent the average of the dendrograms of all samples in one year. Figure 3 shows the composite dendrograms of each of the four years. The number of significant clusters varied between two and three. The composite dendrogram for the data obtained in 1999 at 13, 25, 40, and 54 days is presented in Fig. 3A. In agreement with the analysis shown in Fig. 2, a distinct cluster, containing the rhizosphere samples from the treatment with the combination of both GMMs, stands out (Fig. 2A). With the exception of one of the samples from non-treated wheat, all other samples grouped at a similarity of about 40%. In the second year the effect of cropping potato instead of wheat was predominant (Fig. 3B). Both samples from the rotation plots formed one significant cluster, whereas all other samples, from wheat, formed the second significant cluster at a similarity level of 50%. The effect of growing potato in 2000 was still significant in 2001, although wheat was now again sown in the rotation plots (Fig. 3C). Again, samples from the rotation plots formed a cluster distinct from all other samples. Interestingly, samples from the left (a) and samples from the right half (b) of the field tended to form separate clusters, although differences were too small to be significant. After the fourth trial in 2002, in which again potatoes had been planted in the rotation plots, the samples from the rotation treatment clustered again separate from the samples of all other treatments (Fig. 3D). During the four-year field trial only an effect of the combination treatment of both GMMs after the first introduction was significant. Once potato has been planted instead of wheat, the effect of this crop rotation was maintained during the following years. C l o n e l i b r a r y s c r e e n i n g PCR amplified ITS1/ITS2 fragments from rhizosphere samples taken in 1999 and 2000 were cloned and 960 clones were sequenced. All sequences revealed similarities to ascomycetes except one that was most similar to an arbuscular endomycorrhizal fungus. Technical failures (including sequences with a similarity lower than 90%) ranged from 12-35%. To determine if the number of clones would sufficiently represent the ascomycete diversity in the rhizosphere soil, a rarefaction curve was constructed. In this curve the number of clones analyzed (from samples treated with the wild type in 2000) were

85 A s c o m y c e t e c o m m u n i t i e s n o t a ff e c t e d b y P. p u t i d a W C S r Expected number of species Number of samples Fig. 4. Rarefaction curve of the library from rhizosphere samples treated with the wild- type in 2000 using the program Analytical Rarefaction 1.3 made available by Steve Holland, University of Georgia Startigraphy Lab ( uga.edu/~strat /software). The ordinate gives the expected number of species for any given sample size. plotted against the cumulative number of ascomycete genera detected for that number of clones (Fig. 4). This relationship provides an estimate of the sampling efficiency of any given sample size. The fact that the rarefaction curve almost reached a plateau and that the range of the standard deviation decreased with increasing sample size suggested that the analysis of clones per treatment provided reasonable coverage of the fungal diversity. Genera that were identified from the sequence analysis comprised common ascomycetes such as Fusarium, Giberella, Nectria, Penicillium, Trichoderma, and Verticillium (Table 1). Most genera occurred in low frequencies, and were distributed almost evenly in all treatments. Sequences obtained in 1999 from samples from the left half of the field indicated that only three ascomycete genera occurred frequently. Forty-three percent of the clones from the control treatment showed similarity to Microdochium, and from the treatment with the PCA-producing GMM 50% revealed similarity with Paecilomyces, and 20% with a leaf litter ascomycete. In the right half of the field more than 40% of the clones from the control treatment, the treatment with the wild type, and the treatment with the combination of both GMMs were indicative of Microdochium. In samples taken in 2000, the percentage of Fusarium species was about 20% in all treatments, including the rotation plots, compared to only 5% in The percentage of Nectria also increased in 2000, but this fungus was not recovered from the potato rhizosphere. Twenty-seven percent of the sequences obtained from the control treatments revealed similarity with Trichoderma, compared to only 2 or 3 % in samples from the other treatments. In the potato rhizosphere two ascomycetes occurred in higher numbers than in rhizosphere of wheat, namely Plectosphaerella and Verticillium.

86 8 6 C h a p t e r 5 Table 1. Ascomycete genera with ITS-rDNA sequences most similar to clones isolated from the rhizosphere of field-grown wheat originated from the left and right half in 1999, and from the right half in Plants were grown from seeds that were untreated (contr), treated with P. putida WCS358r (wt), its PCA- and DAPG-producing derivatives (phz, phl), and a combination of both (phz+phl). Ascomycete sequences were identified with BLASTn Search (NCBI). Numbers are the percentage out of a total of 60 clones per treatment. rp = rotation plot, in which potatoes were planted in Abundant genera are indicated in bold type left half right half right half Ascomycete contr wt phz phl phz+phl contr wt phz phl phz+phl contr wt phz phl phz+phl rp Acremonium Alternaria Ampelomyces 2 Antarctomyces 2 Aphanocladium 2 Arthrinium 2 Arthrobotrys 2 Aspergillus 2 Aureobasidium 2 Auxarthon 2 Bionectria 2 2 Calonectria Candida Capronia 2 Cenococcum 2 Chaetomium 2 2 Chaetosphaeria 5 Cladia 2 Cladosporium Coniothyrium Cordyceps Cryptococcus 2 2 Cylindrocarpon Ectomycorrhizal 2 Entrophospora (Gloeromycota) 2 Epicoccum Ericoid mycorrhizal Ericoid endophyte 3 2 Eupenicillium 2 Exophiala Fusarium Gibberella Glomerella 2 Helminthosporium 2

87 A s c o m y c e t e c o m m u n i t i e s n o t a ff e c t e d b y P. p u t i d a W C S r left half right half right half Ascomycete contr wt phz phl phz+phl contr wt phz phl phz+phl contr wt phz phl phz+phl rp Hypocrea Iceman fungal clone Leaf litter ascomycete Lewia 2 Massarina 0 2 Mariannaea Microdochium Mycorrhizal Nectria Neonectria Paecilomyces Penicillium Phaeoacremonium 3 2 Phaeococcomyces 2 Phaeoramularia 3 Phialemonium Phialophora Phoma Phomopsis 2 2 Pichia Plectospaerella Podospora Pseudeuotium 3 Rhodotorula 2 Scoleobasidium 2 Sepedonium 2 2 Stachybotrys 2 Stilbella 2 Strumella Terfezia 2 Tetracladium Thielavia 2 Trichoderma Trimmatostroma 2 Truncatella Verticillium Xanthoria 2 uncultured fungus 2 unidentified ascomycete

88 8 8 C h a p t e r 5 D i s c u s s i o n Molecular engineering to construct BCAs with enhanced biocontrol activity may lead to increased interest in commercial use of GMMs. Several studies showed that such introductions of genetically modified BCAs could have an influence on resident microbial communities. However, most of these studies were performed in microcosm and pot experiments, and do not necessarily reflect the complex situation in natural field environments. Possible long-term effects and effects of repeated introductions of GMMs should be studied preferentially in the field. In the present study genetically modified derivatives of P. putida WCS358r, which produced either PCA or DAPG, were introduced into the rhizosphere of field-grown wheat for four consecutive years (1999 to 2002). The genetically modified derivatives expressed the inserted genes constitutively, making the production of the antimicrobial compounds independent of environmental variables. Both production of antibiotics and siderophoremediated competition for iron have been implicated as mechanisms of suppression of plant diseases by fluorescent pseudomonads (Bakker et al., 1991). The production of PCA and DAPG was demonstrated to be important in suppression of take-all in wheat and barley by P. fluorescens strains 2-79 (Thomashow and Weller, 1988) and Q2-87 (Bangera and Thomashow, 1999). For P. putida WCS358r production of its fluorescent siderophore pseudobactin 358 is essential for suppression of Fusarium wilt in carnation and radish (Duijff et al., 1994; Raaijmakers et al., 1995). Combining two modes of action, either in one strain or by using combinations of different strains, can improve their efficacy (De Boer et al., 2003; Van Loon, 1998). The naturally DAPGproducing P. fluorescens Q8r1-96, which was genetically modified to produce PCA in addition, has been shown to be more suppressive to Rhizoctonia root rot than the wild-type strain (Huang et al., 2004). For the PCA-producing strain used in this study it was demonstrated that it provides improved control of take-all in wheat (Bakker et al., 2002). After the first year of introducing the GMMs in 1999, an effect of seed treatment with the combination of the DAPG- and the PCA-producing derivatives of WCS358r on the ascomycete community was apparent. Application of either the wild type or the DAPG and PCA producers alone did not affect the ascomycetes as detected by DGGE analysis. Whereas one would expect enhanced non-target effects after repeated introduction of the GMMs, no such effects were detected. This lack of effects of the GMMs cannot be ascribed to insensitivity of the DGGE technique, since a significant difference between the ascomycete communities of wheat and potato was detected. The effect of crop rotation was still evident after cultivating wheat again in the

89 A s c o m y c e t e c o m m u n i t i e s n o t a ff e c t e d b y P. p u t i d a W C S r 8 9 rotation plots in the third year, suggesting that a single cropping of potato had a long-term effect. Also in 2002 samples from the rotation plot, in which potato has been planted, were significantly different from all other treatments. In two previous field experiments, a single introduction of the PCAproducing derivative of WCS358r significantly influenced the fungal microflora as determined by ARDRA (Glandorf et al., 2001). Similarly, Viebahn et al. (2003) described a significant effect of the DAPG-producing derivative of WCS358r on the fungal community after a one-time introduction. However, after a second introduction in the same field plots this effect was no longer evident. Therefore, we conclude that the effects of the GMMs on the fungal microflora that were observed previously (Glandorf et al., 2001; Viebahn et al., 2003) cannot be explained by a major perturbation of the ascomycete community. Recently, Schouten et al. (2004) screened 117 Fusarium oxysporum strains isolated from various cultivated soils for their sensitivity to DAPG. Seventeen percent of the strains were highly tolerant to DAPG. The tolerance was independent of formae speciales, geographic location, genetic background, or production of fusaric acid, which is known to negatively effect antibiotic production in P. fluorescens CHA0 (Notz et al., 2002). It was suggested that deacetylation of DAPG to monoacetyl-phloroglucinol and phloroglucinol is one of the mechanisms involved in tolerance. The study focused on Fusarium, but it is quite possible that other ascomycetes possess similar defense strategies. Preliminary evidence indicate that Fusarium species isolated from the rhizosphere of wheat treated with the DAPG-producing Pseudomonas strain were able to degrade DAPG (Dr. J.M.Raaijmakers, pers. communication). Whether these strains were also able to tolerate PCA needs to be investigated. The activation of efflux ATP-binding cassette (ABC) transporters, which are common among eukaryotes (Higgins, 1992) have been reported to protect certain fungi from toxicants. Schoonbeek et al. (2001) demonstrated that Botrytis cinerea became sensitive to reservatrol by disruption of the ABC transporter gene BcatrB. Tolerance to, or degradation of, the antibiotic and/ or expression of drug transporter genes might explain why no effect of the GMMs on the ascomycete communities was detected. In addition to fingerprinting the ascomycete community by DGGE, we constructed clone libraries of samples taken in 1999 and 2000 to detect specific Ascomycota genera. The coverage of the ascomycetes was estimated by constructing a rarefaction curve, which suggested a sample size of 60 clones per treatment to be sufficient to cover the ascomycete diversity. Similar coverage has previously been reported by Anderson et al. (2003). They tested different primer pairs to assess the fungal diversity in natural grassland, and found 50 clones per library to be sufficient to cover the fungal diversity.

90 9 0 C h a p t e r 5 In our study, with only one exception, all ITS1/ITS2 region sequences amplified with the ITS5/ITS4A primer pair showed high similarities to ascomycetes in public databases. No obvious effects of the different bacterial treatments on the composition of the ascomycete community structure were detected. The genera most affected appeared to be Microdochium, Fusarium, and Trichoderma. Seasonal effects were observed for all three genera. Whereas in 1999 Microdochium was prevalent in samples from several of the treatments, no Microdochium species were detected in the following year. On the other hand, the frequency of Fusarium-like sequences was low in all treatments in 1999, but increased in Likewise, Trichoderma was detected at low frequencies in 1999, however, in 2000 the frequency increased, be it only in the control treatment. Recently, Simpson et al. (2004) described a competition experiment with F. culmorum and M. nivale. Mixed inoculations of both ascomycetes on wheat seedlings led to reduced levels of M. nivale, indicating that Microdochium was suppressed by Fusarium. This may explain the concurrent disappearance of Microdochium and appearance of Fusarium in the samples taken in The difference in the ascomycete communities of wheat and potato, which has been demonstrated with DGGE, was also reflected by the cloning approach. Nectria did not occur in the potato rhizosphere, in contrast to most samples of the wheat rhizosphere. On the other hand, Plectosphaerella and Verticillium occurred more often in the rhizosphere of potato than of wheat. Overall, the results obtained by cloning/sequencing were largely in agreement with those obtained by DGGE. Both techniques revealed that a repeated introduction of P. putidawcs358r had no major effects on the ascomycete community in the wheat rhizosphere, but demonstrated a difference between the rhizospheres of potato and wheat. At least, possible effects of the genetic modification on ascomycete fungi are smaller than those observed for rotating wheat with potato, the long term effect of a one-time potato crop, seasonal effects, and positional effects in the field. A c k n o w l e d g e m e n t This study was financed by the Dutch Ministry of Housing, Spatial Planning and the Environment. We thank Bas Valstar and Fred Siesling (Botanical Garden, Utrecht University) for constructing and maintaining the experimental field site.

91

92

93 Chapter 6 Microarray analysis and suppression subtractive hybridization to identify shifts in rhizosphere bacterial communities Mareike Viebahn, Rogier Doornbos 1, Karel Wernars 1, Leendert C. van Loon, Eric Smit 1, Gary L. Andersen 2, Todd Z. DeSantis 2, Peter A.H.M. Bakker 1 2 National Institute of Public Health and the Environment, Bilthoven, The Netherlands Center for Environmental Biotechnology, Lawrence Berkeley National Laboratory, Berkeley, CA, USA

94 9 4 C h a p t e r 6 A b s t r a c t Different molecular techniques were applied to analyze shifts in bacterial communities in the rhizosphere of wheat as a result of the introduction of genetically modified microorganisms. Samples originated from an experimental field in which Pseudomonas putida WCS358r and two genetically modified derivatives of this strain were introduced as a seed coating. The transgenic strains constitutively produced phenazine-1-carboxylic acid (PCA) or 2,4-diacetylphloroglucinol (DAPG), two broad-spectrum antimicrobial compounds important for suppression of soilborne plant diseases. Bacteria-specific amplicons of DNA isolated from the rhizosphere samples were generated and subjected to denaturing gradient gel electrophoresis (DGGE) or hybridized to an Affymetrix GeneChip containing 16S rdna sequences of approximately 9000 operational taxonomic units (OTUs). Cluster analysis of DGGE banding patterns and GeneChip data gave similar results, and indicated a response of the indigenous bacteria to the genetically modified pseudomonads. Bacterial species present in the rhizosphere were identified using the GeneChip data. Whereas it was expected that introduction of the bacterial strains would reduce biodiversity, species diversity was actually increased by all bacterial treatments. Some species were stimulated only by the DAPG-GMM. The bacteria-specific 16S rdna amplicons from rhizosphere samples of plants treated with the wild type or the DAPG-producing strain were also analyzed by suppression subtractive hybridization (SSH) to identify species unique to one sample. By enriching rhizosphere samples with test sequences, i.e. DNA derived from digested phage ΦX174, it was demonstrated that SSH can, indeed, be used on samples obtained from rhizosphere soil, as phage DNA was recovered. However, additional non-specific sequences were not removed, indicating that hybridization of sequences present in both samples was not complete. When SSH was applied to rhizosphere samples of plants treated with the wild type or the DAPG-producing strain, no obvious differences were detected, suggesting that as yet this technique is not sufficiently optimized to identify subtle shifts in microbial communities. I n t r o d u c t i o n Plant roots release many compounds, such as sugars, amino acids and organic acids, that stimulate microorganisms to grow in the rhizosphere (Lugtenberg et al., 2001). The plant growth-promoting rhizobacterium Pseudomonas putida WCS358r can suppress soilborne plant diseases by effectively competing for iron with the pathogen in the rhizosphere of a variety of crops (Bakker et al., 1986). It produces siderophores to chelate iron, making it unavailable for other rhizosphere microorganisms, including plant pathogens. To increase the effectiveness of disease suppression by this strain, it was genetically modified to produce phenazine-1-carboxylic acid (PCA) (Glandorf et al., 2001) or 2,4- diacetyl-phloroglucinol (DAPG) (Viebahn et al., 2003), secondary metabolites with broad-spectrum antimicrobial activities. However, field release of such genetically modified microorganisms (GMMs) may result in perturbations of populations of non-target organisms. In a series of field releases, we have studied the impact of genetically modified P. putida on the indigenous microflora of the rhizosphere of wheat. In earlier work, Glandorf et al. (2001) demonstrated that PCA-producing derivatives of P. putida WCS358r transiently affected the

95 M i c r o a r r a y a n a l y s i s a n d S S H o f b a c t e r i a l c o m m u n i t i e s 9 5 composition of the fungal microflora after a single introduction. Repeated introductions of the same GMM did not lead to intensified effects on the fungal microflora (Viebahn et al., 2003). Viebahn et al. (2005b) were unable to detect changes in the ascomycete communities, but did observe effects on the bacterial microflora upon introduction of the PCA-producing, as well as a DAPG-producing transconjugant (Chapter 3). The techniques used earlier (amplified ribosomal DNA restriction analysis [ARDRA], and denaturing gradient gel electrophoresis [DGGE]) resulted in complex banding patterns that did not allow identification of individual members of the rhizosphere microbial communities that were affected by the bacterial treatments. However, microarrays allow the simultaneous analysis of hundreds or thousands of gene sequences in parallel on small surface areas. They were developed in the mid- to late 1980s (Ekins and Chu, 1999), and are now routinely used in clinical diagnostic, functional genomic, and genetic analysis (Lander, 1999). Although they have been adapted to the detection of microorganisms in their natural environment, their application to microbial ecology has been limited so far. Guschin et al. (1997) introduced a new hybridization format for rrna-targeted oligonucleotides. These authors were able to identify unequivocally key genera of nitrifying bacteria, such as Nitrosomonas and Nitrosovibrio, by gel array microchip analysis of either DNA or rrna extracted from pure cultures. Loy et al. (2002) developed the SRP-PhyloChip, an oligonucleotide microarray with 132 probes targeting all known lineages of sulfate-reducing prokaryotes (SRP). This chip was evaluated with 41 reference cultures of SRPs and applied to determine SRP diversity in environmental and clinical samples. Microarrays can provide a picture of the species within the resident bacterial communities in as far as those are represented on the array. The photolithography Affymetrix GeneChips used in the present study contain an array of oligonucleotide probes targeting the prokaryotic small subunit (SSU) rrna (Affymetrix Inc., Santa Clara, CA, USA). Probes were designed based on subalignment of the SSU rrna sequences in the database of the Ribosomal Database Project (Maidak et al., 2001), comprising almost all possible 25- mer probes for every single ribosomal sequence in the database (DeSantis et al., 2003), and representing approximately 9000 operational taxonomic units (OTUs). However, it is likely that many more species are present in the rizosphere samples than are represented on the GeneChip. Therefore, an alternative technique was sought that would allow differences between communities to be identified on the sequence level. The PCR-based technique suppression

96 9 6 C h a p t e r 6 subtractive hybridization (SSH) is a novel approach to compare microbial communities, and was originally developed to study differential gene expression in eukaryotes (Diatchenko et al., 1996). It is based on the selective amplification of DNA sequences that are not shared with a reference sample. The key to this strategy is its suppression effect: internal inverted terminal repeats attached to the DNA can selectively suppress amplification of shared sequences in PCR procedures. In a study on bacterial populations in steer rumen it allowed the identification of bacterial strain-specific DNA sequences present in one sample, defined as the tester, but absent in another, defined as the driver (Galbraith et al., 2004). SSH was applied to specifically enrich sequences of bacteria that are affected by the GMMs in a subset of rhizosphere samples. To identify bacterial species that were affected by the introduction of the GMMs in earlier field trials (Chapter 3; Glandorf et al. 2001; Viebahn et al. 2003), both microarray and SSH were applied in a complementary approach. This combined use of microarray analysis and SSH was expected to be particularly useful in identifying bacterial species that are sensitive to the introduction of the GMM(s). M a t e r i a l s a n d M e t h o d s B a c t e r i a l s t r a i n The bacterial strains used were P. putida WCS358r (Geels and Schippers, 1983) and its two transgenic derivatives, WCS358r::phz and WCS358r::phl, that produce PCA and DAPG, respectively (Glandorf et al. 2001; Viebahn et al., 2003). The wild-type strain WCS358r was grown on King s Medium B (KB) (King et al., 1954) supplemented with 150 μg rifampicin ml -1, the transgenic derivatives on KB with 150 μg rifampicin ml -1 and 30 μg kanamycin ml -1. All strains were routinely grown at 28ºC for 2 days. S e e d t r e a t m e n t, e x p e r i m e n t a l f i e l d, a n d s a m p l i n g Wheat seeds were treated with a 1:1 mixture of washed bacterial suspension and 3% methylcellulose, as described by Glandorf et al (2001). For the control treatment the bacterial suspension was replaced by 10 mm MgSO 4. The coated seeds were sown every year in the same plots in four consecutive years (1999 to 2002) in an experimental field located near the Botanical Garden at Utrecht University, The Netherlands, as described by Viebahn et al. (2003). The field consisted of 36 plots in two rows of 18 plots each. In each row, each treatment was represented by three replicate plots. For this study we used samples obtained from one row [the left (Viebahn et al., 2003)] within the

97 M i c r o a r r a y a n a l y s i s a n d S S H o f b a c t e r i a l c o m m u n i t i e s 9 7 field. The three replicates of the control treatment and the bacterial treatments with WCS358r, WCS358r::phz and WCS358r::phl from 2001 and 2002 were analyzed by microarray analysis and DGGE. The additional treatments with the combination of both GMMs and crop rotation (Chapter 3; Viebahn et al., 2005a; Viebahn et al. 2005b) were not included in the present study. Soil samples from the three replicate plots were analyzed individually, in contrast to our previous analyses (Chapter 3; Viebahn et al., 2005a; Viebahn et al. 2005b), in which samples were pooled before ARDRA or DGGE analysis. Three to 5 g of roots with adhering soil, sampled at 25 days after sowing during the growing seasons of 2001 and 2002, were mixed with 10 ml sodium phosphate buffer (120 mm, ph 8). One g of gravel was added and samples were vortexed for 30 s. The supernatant was decanted into a new tube. One ml of the supernatant was used to extract total DNA with the FastDNA SPIN Kit for Soil (Bio 101, Biogene, Vista, CA, USA.) in combination with a Ribolyser (Hybaid, Ashford, UK), as previously described (Smit et al., 2003). The extracts were suspended 1:100 in 100 μl Millipore-filtered distilled water before purification with the Wizzard DNA Clean-up System (Promega, Madison, WI, USA) according to the manufacturer s protocol. D N A a m p l i f i c a t i o n a n d D G G E Polymerase chain reaction (PCR) amplification of the rdna for DGGE was done with the primer pair F-968-GC (5 -CGCCCGGGGCGCGCCC CGGGCGGGGCGGGGGCACGGGGGGAACGCGAAGAACCTTAC-3 ) and 1401R (5 -CGGTGTGTACAAGACCC-3 ) (Nübel et al., 1996). This primer pair amplified the segment of the eubacterial 16S rdna comprising the variable regions V6 to V9 (Brosius et al., 1978). F-968-GC contained a 40 bp GC-clamp to stabilize the melting behavior of the DNA fragments for DGGE analysis (Sheffield et al., 1989). The PCR was performed in 10 PCR buffer 2 (ph 9.2) containing 2.25 mm MgCl 2 (Roche Diagnostics, Mannheim, Germany), 250 μm of each of the four deoxynucleoside triphosphates, 250 nm of each primer, 2.5 U Expand Long template enzyme (Roche, Diagnostics, Mannheim, Germany), and 1 μl of appropriately diluted template DNA. The PCR conditions used in the thermocycler (Hybaid, Ashford, UK) were: 5 min at 94 C, followed by 35 cycles of 1 min at 94 C, 1 min at 60 C, and 3 min at 72 C, and finally 10 min at 72 C. PCR fragments were separated on a denaturing gradient polyacrylamide gel consisting of 0.5 TAE (20 mm Tris acetate, 10 mm sodium acetate, 0.5 mm EDTA) and 8% acrylamide (acrylamide/bisacrylamide at a ratio of 37.5:1) with a denaturant gradient of 30-60%. One hundred % denaturant solution contained 7 M urea and 40% formamide. Up to 25 μl of PCR product per lane were loaded and gels

98 9 8 C h a p t e r 6 were run for 17 h at 80 V at a constant temperature of 60 C in a DCode Universal Mutation Detection System (Bio-Rad Laboratories, Veenendaal, The Netherlands). For comparison of DNA patterns, a reference marker was added in triplicate on each gel. After electrophoresis gels were stained for 30 min in 1:10000 diluted SybrGold (Molecular Probes, Leiden, The Netherlands), and viewed under a blue light transilluminator (Clare Chemical Research, Dolores, CO, USA). Images were digitalized using the GeneGenius Bio Imaging System (Syngene, Cambridge, UK). C l u s t e r a n a l y s i s The bacterial community fingerprints on the DGGE gels were analyzed with the BioNumerics program vers. 3.5 (Applied Maths, Sint-Martens- Latem, Belgium). After normalization and background subtraction Pearson s correlation coefficient was used to calculate the similarity between each banding pattern. Patterns were grouped into clusters by the unweighted pairgroup method using average linkages (UPGMA). Significant clusters in the dendrograms were determined by calculating the cut-off value that produced the highest point-bisectional correlation (BioNumerics manual, version 3.5). Briefly, a line is drawn through the dendrogram at a certain similarity level, and from the resulting number of clusters defined by that line, the program creates a new similarity matrix, in which all within-cluster values are 100%, all between-cluster values are 0%. Then, the correlation between this new matrix and the original matrix is calculated, which is called the point-bisectional correlation. The same is done for other cut-off similarity levels, and the level with the highest point-bisectional correlation is the one defining the significant clusters. D N A a m p l i f i c a t i o n f o r t h e G e n e C h i p a s s a y Amplification of rdna was carried out using the primer pair 27F (5 - AGAGTTTGATCCTGGCTCAG-3 ) and 1492R (5 -GGTTACCTTGTTAC GACTT-3 ) (Lane, 1991), which amplify the 16S rrna gene of a wide range of members of the domain Bacteria [nucleotide 27 to nucleotide 1492 Escherichia coli numbering (Brosius et al., 1978)]. The PCR was performed as described above, except that the concentration of each of the primers was 1 μm. The PCR conditions were: 3 min at 95 C, followed by 35 cycles of 30 s at 95 C, 30 s at 53 C, and 1 min at 72 C, and finally 7 min at 72 C. Two μg DNA amplicons were prepared containing One-Phor-All Buffer Plus and 0.2 U DNAse I per μg DNA (Invitrogen, Carlsbad, CA, USA) in a 50 μl solution, and fragmentation was performed on a PE 9600 thermocycler for 10 min at 25 C, followed by 10 min at 98 C. The fragmented products were

99 M i c r o a r r a y a n a l y s i s a n d S S H o f b a c t e r i a l c o m m u n i t i e s 9 9 biotin-labeled for 1 h at 37 C using the Enzo BioArray Terminal Labeling Kit (Affymetrix, Inc., Santa Clara, CA, USA, P/N ) according to the manufacturer s protocol. M i c r o a r r a y h y b r i d i z a t i o n, s c a n n i n g a n d a n a l y s i s Hybridization was done according to the GeneChip expression analysis technical manual as previously described (Masuda and Church, 2002). The hybridization solution contained 100 mm MES (N-morpho-linoethanesulfonic acid), 1 M NaCl, 20 mm EDTA, and 0.01% Tween 20, ph 6.6 (referred to as 1 MES). In addition, the solution contained 0.1 mg ml -1 herring sperm DNA (Promega, Madison, WI, USA), 0.5 mg ml -1 acetylated bovine serum albumin (BSA; Invitrogen, Carlsbad, CA, USA), and 50 pm control oligonucleotide B2 (Affymetrix Inc., Santa Clara, CA, USA). Hybridization was carried out at 48 C for 16 h with mixing on a rotary shaker at 60 rpm. Following hybridization, the sample solution was removed and the array was washed and stained in a GeneChip FS400 fluidics station (Affymetrix, Inc., Santa Clara, CA, USA). In brief, to enhance the signals, 10 μg ml -1 streptavidin (Vector Laboratories, Burlingame, CA, USA) and 2 mg ml -1 BSA in 1 MES were used as the first staining solution. After the streptavidin solution was removed, an antibody mix was added as the second stain, containing 5 μg ml -1 biotinylated anti-streptavidin antibody (Vector Laboratories, Burlingame, CA, USA), 0.1 mg ml -1 goat immunoglobulin G (Sigma-Aldrich, St. Luis, MO, USA), and 2 mg ml -1 BSA in 1 MES. Nucleic acid was fluorescently labeled by incubation with 10 μg ml -1 streptavidin R-phycoerythrin conjugate (Molecular Probes, Eugene, OR, USA) and 2 mg ml -1 BSA in 1 MES. The arrays were scanned at 570 nm with a resolution of 3 µm with a GeneArray scanner (Affymetrix, Inc., Santa Clara, CA, USA). These Affymetrix chips contain 16S rdna sequences representative of approximately 9000 operational taxonomic units (OTUs). Each OTU is represented by at least one probe set consisting of twenty to twenty-eight 25- mer oligonucleotides in perfect match (PM) and mismatch (MM) sequences. In case at least 95% of the probes in the probe set is positive, the OTU represented by the probe set is considered to be present in the sample. A probe is considered positive when (i) the fluorescence of the PM is at least 1.25 times higher than the fluorescence of the MM; and (ii) the fluorescence of the PM minus the fluorescence of the MM is at least 8 times higher than background and noise (Wilson et al., 2002a). Replicate samples 1 and 2 of all treatments were processed separately from replicate sample 3. Data analysis was performed by using custom software, which scaled each array such that the average of the probes responding to a

100 1 0 0 C h a p t e r 6 set of spike-in controls had an average of 2500 fluorescent intensity units. Cluster analysis was performed with GeneMaths vers. 2.1 (Applied Maths, Sint-Martens-Latem, Belgium). After log transformation similarity between treatments was determined by Pearson s correlation coefficient. Similarity between bacterial species was calculated using the Euclidian distance. Dendrograms were constructed based on Ward s averaging. S u p p r e s s i o n s u b t r a c t i v e h y b r i d i z a t i o n SSH, as schematically outlined in Fig. 1, was performed using the Clontech PCR-Select Bacterial Genome Subtraction Kit (BD Biosciences-Clontech, Palo Alto, CA, USA), with slight modifications as described below. Briefly, DNA was extracted from rhizosphere samples. Then, 16S rdna was amplified with the primers identical to primers used for the microarrays, purified with the QIAquick PCR Purification Kit (Quiagen, Venlo, The Netherlands), and digested with the restriction endonuclease Rsa I. The cleaved tester sample, which presumably contains unique sequences, was divided in two, and different oligonucleotide adaptors provided in the kit were ligated to each half. The two tester DNA samples were denatured and hybridized to excess driver DNA, after it had also been digested with Rsa I. Most tester sequences will form heterohybrids with driver sequences. Fragments with identical adaptors on both ends form panhandle structures, which do not allow further amplification. Only tester-specific sequences selfhybridize, and can then be enriched by PCR with primers complementary to the ligated adaptors. Unsubtracted fragments of tester DNA, which contain the adaptors but are not subjected to hybridization, are amplified also. For each subtraction two PCR amplifications were performed. The primary PCR was conducted with PCR primer 1: 2 min at 72 C, followed by 25 cycles of 30 s at 94 C, 30 s at 66 C, and 1.5 min at 72 C. The secondary PCR was conducted with nested primers 1 and 2R: 15 cycles of 30 s at 94 C, 30 s at 68 C, and 1.5 min at 72 C. To determine if the procedure was applicable to soil samples, 16S rdna was obtained from one rhizosphere sample after amplification with an appropriate primer pair and used both as driver and as tester. The tester, rhizosphere DNA extracted from samples treated with P. putida WCS358r, was spiked with phage ΦX174 DNA digested with Hae III, to provide unique sequences to the tester DNA. A subtraction of tester and driver both spiked with Hae III-digested ΦX174 DNA, was performed to test if the phage fragments were removed after subtraction. A control without rhizosphere DNA consisted of E. coli DNA spiked with the phage DNA or not. After PCR, amplicons were subjected to electrophoresis on a 2% agarose gel. Then, the method was applied to rhizosphere samples treated with the wild type P. putida WCS358r or with the DAPG-GMM.

101 M i c r o a r r a y a n a l y s i s a n d S S H o f b a c t e r i a l c o m m u n i t i e s Tester cdna with adapter 1 Driver cdna (in excess) Tester cdna with adapter 2R First hybridization a b c d a, b, c, d + e Second hybridization: mix sample, add freshdenatured driver, and anneal Fill in the ends a b c d e Add primers Amplify by PCR a, d no amplification a b' no amplification c linear amplification 5' 3' and 3' 5' e exponential amplification Rsa l-digested tester cdna Rsa l-digested driver cdna Outer portion of adaptors/pcr primer sequence Inner portion of Adaptor 1/Nested PCR Primer 1 Inner portion of Adaptor 2R/Nested PCR Primer 2R Fig. 1. Schematic diagram of suppression subtractive hybridization. Solid lines represent RSA I-digested tester and driver DNA. Black boxes represent the outer part of adaptor 1 and corresponding PCR primer 1 sequence. Shaded boxes represent the outer part of adaptor 2 and corresponding PCR primer 2 sequence. White boxes represent the inner part of the adaptors and the corresponding nested PCR primer 1 and R2 sequences. Courtesy of BD Biosciences-Clontech.

102 1 0 2 C h a p t e r 6 R e s u l t s D G G E o f t h e b a c t e r i a l c o m m u n i t i e s In previous work (Viebahn et al., 2003), the same samples have been analyzed, but in that study the samples from the replicate plots within a row were pooled. In our current analysis, the replicate plots were analyzed separately to detect the variability between individual plots. The rhizosphere samples from 2001 grouped into two significant clusters (1 and 2) (indicated by the grey lines in Fig. 2A), of which cluster 1 contained two subclusters (Fig. 2A: 1a and 1b). All samples were at least 60% similar. Cluster 2 contained one replicate of each of the PCA- and the DAPG-treatments. Cluster 1 contained one replicate of a control treatment in subcluster 1b, and all other treatments in subcluster 1a. No consistent effect of the different treatments was evident. A comparison of DGGE banding patterns of replicate samples from 2002 is shown in Fig. 2B. There was a clear differentiation between the bacterial microflora of samples from the control and the treatment with the wild type on the one hand (cluster 1), and of samples from the treatments with the GMMs on the other hand (cluster 2). All profiles were over 64% similar. These results indicate that in 2002 there was a significant effect of the PCA- and DAPGproducing transgenic strains on the bacterial communities, but no effect of the wild type. B a c t e r i a l s p e c i e s d e t e c t e d w i t h t h e m i c r o a r r a y a n a l y s i s To determine in how far individual bacterial species were affected by the introduction of P. putida WCS358r and its two genetically modified derivatives, DNA of the three rhizosphere samples was extracted, labeled, and hybridized to the Affymetrix GeneChips. Presence or absence of OTUs was scored and represented in Fig. 3 by a color scale ranging from green (absent) to red (present). The effects of the treatments are depicted in the upper part of Fig. 3A (2001) and B (2002). Rhizosphere samples from 2001 grouped in two clusters (Fig. 3A). Cluster 1 contained two replicates (1 and 3) of the control, and the third replicates of the treatments with the wild type and with the PCA-producing GMM. Cluster 2 contained samples of each of the GMM treatments in subcluster a, and all other samples in subcluster b. No clear effect of the treatments on the bacterial community was apparent, in line with the results obtained by DGGE (Fig. 2A). Rhizosphere samples from 2002 also grouped in two clusters (Fig. 3B). Cluster 1 contained the third replicate of all treatments. Cluster 2 contained rhizosphere samples treated with both GMMs in subcluster a, and mainly rhizosphere samples from the control and the treatment with the wild type

103 M i c r o a r r a y a n a l y s i s a n d S S H o f b a c t e r i a l c o m m u n i t i e s A: 2001 % similarity a b wt 2 wt 3 wt 1 contr 2 phl 2 phz 1 phl 3 contr 3 phz 2 contr 1 phz 3 phl 1 B: 2002 % similarity contr 1 wt 3 contr 2 wt 1 contr 3 wt 2 phz 3 phl 2 phz 1 phl 1 phl 3 phz 2 Fig. 2. Dendrograms based on the genetic similarity of the bacterial communities of field-grown wheat plants at 25 days after sowing in 2001 (A) and 2002 (B) with the corresponding gels. Wheat seeds were treated with P. putida WCS358r (wt), WCS358r::phz (phz), or WCS358r::phl (phl). For the control treatment (contr) seeds were coated with methylcellulose (no bacteria). Three replicates per treatments (1, 2, 3) were analyzed. Similarities are based on DGGE patterns generated from 16S rdna fragments using Pearson s correlation coefficient. Cluster analysis was done with UPGMA. Significant (grey lines) and nonsignificant clusters (black lines) were separated by the point-bisectional cutoff method. The levels of similarities are shown above the dendrograms. in subcluster b. Apparently, in this year both transgenic strains had an impact on the bacterial communities. This is in agreement with the results obtained by DGGE, which also revealed an effect of the transgenic strains in 2002 (Fig. 2B). Grouping of the most relevant OTUs is shown in the lower part of Fig. 3A and B. Only 361 probe sets met the stringent conditions to be considered represented in at least one sample (data not shown). Out of these 361 OTUs, only 130 in 2001 (Fig. 3A) and 171 in 2002 (Fig. 3B) showed differences between the treatments. The separate processing of sample sets 1 and 2 versus sample sets 3 probably affected the results, since sample sets 3 of all treatments

104 1 0 4 C h a p t e r 6 A: 2001 B: a b a b contr 3 wt 3 contr 1 phz 3 phz 2 phl 3 contr 2 wt 1 plz 1 wt 2 phl 1 phl 2 phz 3 wt 3 contr 3 phl 3 phz 1 phz 2 phl 2 contr 1 wt 1 wt 2 contr 2 phl 1 Fig. 3. Presentation of microarray data from samples of 2001 (A) and 2002 (B) using GeneMaths (Applied Maths, Belgium). Dendrograms at the upper part shows the clustering of the different treatments (cf. legend to Fig. 2). Three sample sets per treatment (1, 2, 3) were analyzed. The lower part shows a grouping of the operational taxonomic units (OTUs) identified. The scale ranged from green (absent) to red (present). Different color shades represent different fluorescence intensities.

105 M i c r o a r r a y a n a l y s i s a n d S S H o f b a c t e r i a l c o m m u n i t i e s M M 900 bp 700 bp 500 bp a b c a b c a b c 900 bp 700 bp 500 bp Fig. 4. Results of suppression subtractive hybridization as analyzed by agarose gel electrophoresis. Lane 1: subtracted rhizosphere DNA, of which the tester was spiked with Hae III-digested bacteriophage ΦX174 DNA. Fragments a - c correspond to the spiked phage DNA. The remaining bands (3, 4, and 5) presumably correspond to unspecific tester DNA fragments. Lane 2: unsubtracted rhizosphere DNA, corresponding to lane 1; bands 1 and 2 are presumably unspecific tester fragments. Lane 3: subtracted rhizosphere DNA, of which both tester and driver were spiked with digested phage DNA. Lane 4: unsubtracted rhizosphere DNA, corresponding to lane 3. Lane 5: subtracted tester E. coli DNA, spiked with Hae III-digested bacteriophage ΦX174 DNA. Lane 6: unsubtracted E. coli DNA, corresponding to lanes 5 and 7. Lane 7: ΦX174 DNA after subtraction from spiked E. coli DNA, as provided in the commercial kit as a reference. Lane 8: negative control. M = 100 bp DNA markers from 100 to 3000 bp. clustered mainly apart. For this reason only results from sample sets 1 and 2 were used for further analysis. All OTUs that were either present or absent in all treatments were not considered further. For 2001, only 12 OTUs showed differential presence/ absence in both sample sets of a particular treatment. As shown in Table 1, controls lacked ten OTUs that were present in each of the bacterial treatments, whether wild type or the GMM, and two that were present only in the samples treated with the DAPG-GMM. Of these 12 OTUs, eight belonged to the Proteobacteria, two to the Gram positives, one to the Planctomyces and relatives, and one to an Environmental clone opb2 subgroup (Table 1). Of the 171 OTUs in the samples taken in 2002, none was present in both replicates of each treatment. S u p p r e s s i o n s u b t r a c t i v e h y b r i d i z a t i o n o n t e s t s a m p l e s Since, to our knowledge, SSH has never been applied to DNA extracted from soil, a proof of principle was conducted on a rhizosphere sample spiked with a marker bacteriophage DNA (Fig. 4). Secondary PCR products of subtracted spiked E. coli DNA as a control contained mainly DNA fragments corresponding to the phage DNA (Fig. 4, lanes 5, 7; bands a, b, and c). The secondary PCR products of unsubtracted E. coli DNA contained a smear (lane 6), indicating that in this unsubtracted sample no specific enrichment took place. Results from the rhizosphere samples are shown in Fig. 4, lanes 1 to 4. Subtractions of the tester sample spiked with the phage DNA resulted in an enrichment of phage DNA (lane 1). The upper three bands (a, b and c)

106 1 0 6 C h a p t e r 6 Table 1. List of OTUs represented on the Affymetrix Genechip that were affected by the different treatments of rhizosphere samples taken in Wheat seeds were untreated (contr) or treated with P. putida WCS358r (wt), and its PCA- and DAPG-producing derivatives WCS358r::phz (phz) and WCS358r::phl (phl). Two replicate samples were used. Numbers are the fractions of positive probe sets. The threshold value at which a probe set was called positive, was When the threshold value was at least 0.95, the OTU was considered to be present (indicated in bold). OTU Representative organisms contr 1 contr 2 wt 1 wt 2 phz 1 phz 2 phl 1 phl 2 Env. Clone opb2 group Environmental clone opb2 group clone GOUTB Planctomyces & Rel. Chlamydophila pneumoniae c α -Proteobacteria Rhizobium leguminosarum c α -Proteobacteria Agrobacterium tumefaciens c α -Proteobacteria Rhizobium galegae c β -Proteobacteria Azoarcus indigens subgroup beta clone JG37-AG γ -Proteobacteria Calyptogena symbionts subgroup unnamed organism δ -Proteobacteria Desulfomicrobium baculatum Gram Positives Dialister invisus Gram Positives Lachnospira multipara subgroup clone p-1594-c γ -Proteobacteria Halomonas subglaciescola subgroup γ UMB18C γ -Proteobacteria Vibrio aestuarianus

107 M i c r o a r r a y a n a l y s i s a n d S S H o f b a c t e r i a l c o m m u n i t i e s A B M M bp bp bp 700 bp 500 bp bp Fig. 5. SSH of rhizosphere samples treated with P. putida WCS358r or its DAPG-GMM: secondary PCR products after subtraction with DNA extracted from rhizosphere samples treated with P. putida WCS358r or WCS358r::phl. A: lane 1: from WCS358r as tester, and WCS358r::phl as driver. Lane 2: unsubtracted DNA corresponding to lane 1. B: lane 3: from WCS358r::phl as tester, and WCS358r as driver. Lane 4: unsubtracted DNA corresponding to lane 3. M = 100 bp DNA markers. correspond to those that were obtained by subtraction from the spiked E. coli DNA (lanes 5, 7). No enrichment took place in the unsubtracted samples (lane 2). When an identical rhizosphere sample was spiked with phage DNA and used as both tester and driver (lane 3), the phage DNA was subtracted. Since in the tests the bacteriophage DNA fragments constituted the only differences between the tester and driver, the remaining faint bands (1 and 2) and the prominent lower bands (3, 4 and 5) must represent fragments that apparently were not removed by subtractive hybridization. These fragments also constitute the main bands in unsubtracted samples (lanes 2, 4). These experiments demonstrate that unique sequences in one DNA pool can be detected. However, not all sequences shared by both DNA pools could be effectively subtracted. I d e n t i f i c a t i o n o f t r e a t m e n t - s p e c i f i c D N A s e q u e n c e s Since the microarray analysis of samples from 2002 did not lead to the identification of DAPG-sensitive bacteria, further analysis of the rhizosphere samples treated with WCS358r or WCS358r::phl from this year was performed with SSH to detect differences such as had been observed by DGGE in the previous study (Chapter 3). In a first experiment, the sample treated with WCS358r was used as the tester, and the sample treated with the DAPGproducer WCS358r::phl was used as the driver. Results are shown in Fig. 5A. There were clear differences between subtracted (lane 1) and unsubtracted PCR products (lane 2). This indicates that enrichment of specific sequences took place as a result of the subtraction procedure.