Potent degradation of neuronal mirnas induced by highly complementary targets

Transcription

1 EMBO reports - Peer Review Process File - EMBOR Manuscript EMBOR Potent degradation of neuronal mirnas induced by highly complementary targets Manuel de la Mata, Dimos Gaidatzis, Mirela Vitanescu, Michael B. Stadler, Corinna Wentzel, Peter Scheiffele, Witold Filipowicz and Helge Großhans Corresponding author: Helge Großhans and Witold Filipowicz, Friedrich Miescher Institute for Biomedical Research Review timeline: Transfer date: 08 January 2015 Editorial Decision: 09 January 2015 Revision received: 21 January 2015 Accepted: 26 January 2015 Transaction Report: (Note: With the exception of the correction of typographical or spelling errors that could be a source of ambiguity, letters and reports are not edited. The original formatting of letters and referee reports may not be reflected in this compilation.) Transfer Note: Please note that this manuscript was originally submitted to the EMBO Journal where it was peer-reviewed. It was then transferred to EMBO reports with the original referees comments and the authors response attached. (Please see below) Editor: Esther Schnapp Transfer original referees comments and authors' response 08 January 2015 European Molecular Biology Organization 1

2 Referee #1: In this interesting, and nicely written manuscript, de la Mata et al. report on the characterization of mirna degradation induced by targets. Basically, the authors show here the molecular requirements for an efficient degradation of mirnas in primary neurons. It has been previously reported that the stability of some mature mirnas could be very quickly jeopardized in response to certain stimuli, such as dark to light transition in the retinal neurons. Others have reported that under some conditions, highly complementary targets, either artificial in the form of synthetic oligonucleotides, or natural as expressed by some viruses, could trigger decay of the targeting mirna. This mechanism, which the authors refer to as target RNA- directed mirna degradation (TDMD), appears to require 3' addition of nucleotides on the mirna (tailing), coupled to its 3' to 5' exonucleolytic degradation (trimming). In this report, the authors confirm that indeed TDMD is functional in primary neurons (and surprisingly enough, only in primary neurons with their approach) and that it depends both on the degree of complementarity of the target RNA and the bound mirna, and on the respective levels of the two molecules. Thus, they show that transducing a construct consisting of the coding sequence of GFP and 4 binding sites with a central bulge for mir- 132 results in an efficient decrease in mir- 132 levels. Using deep- sequencing, they then provide evidence that tailing and trimming of the mirna occurs, and that tailing appears to be initiated within Ago2. Disrupting the pairing at the 3' of the mirna, increasing the size of the bulge, or using regular target site results in the loss of TDMD. Only mirnas expressed under a certain absolute level can be targets of TDMD, and there is an inverse correlation between the targeting efficiency of the mirna and TDMD. In other words, when the mirna is expressed at sufficiently high levels, it will resume its regulatory activity, and will not be degraded. We thank the reviewer for commending on the qualities of our manuscript and the underlying work. At the same time, we would like to point out that his/her summary of the work, and the ensuing comments, reflect mostly our results with the 4x target. By contrast, they do not consider the differences found between the 4x and 1x targets. For instance, the referee concludes that only mirnas expressed under a certain absolute level can be targets of TDMD, but we show (Fig. 4) that even a highly expressed mirna such as mir- 124 (or overexpressed mir- 132) undergoes efficient TDMD with a 1x target. Hence, despite a high targeting efficiency of this mirna, it will not resume regulatory activity. There is thus no general inverse correlation between the targeting efficiency of the mirna and TDMD. Major comments The experiments are state of the art, are very well performed and described and they do back up the conclusions of the authors. However, the manuscript falls short in providing an explanation as to how TDMD occurs naturally in primary neurons. In its present form, it is merely a confirmation that indeed target mediated decay can work, but only using artificial targets. We respectfully disagree with the reviewer's claim that our manuscript is merely a confirmation of previous work. We believe this misunderstanding to result from the fact that our data showing that we can uncouple TDMD and mrna decay escaped the referee's attention (see comment above). By revealing that mrna and mirna decay are not linked; that the former but not the latter depends on target site cooperativity; and that TDMD exhibits an unforeseeably high efficiency in neurons, likely due to a multiple turnover activity, our work provides novel mechanistic insight into the thus- far poorly understood pathway of mirna decay. Consistent with a recommendation by referee #3, we will now highlight these important findings better in 2

3 the abstract. 1- It would have therefore been nice to identify at least one naturally expressed target in primary neurons that could explain the rapid decay of a specific mirna. This might indeed prove difficult, but it would add weight to the manuscript. 2- Similarly, there is no identification of a putative factor involved in the tailing or in the degradation of the mirna. Although we agree with the referee that both of these would make very exciting findings, identification of endogenous targets or contributing factors will indeed prove difficult particularly in primary neurons where limiting amounts of starting material compromise the feasibility of high throughput experiments. Our current efforts are aimed in this direction, but this will be a long- term project, beyond the scope of the current work. 3- The observation that TDMD seems to be more efficient in primary neurons is an interesting one, but maybe this is due to the fact that these are primary cells. It would have been nice to show whether TDMD is also functional in primary cells of another origin. We would like to point out that we already tested TDMD in primary cells of another origin, specifically murine embryonic fibroblasts (MEFs). In this primary cell type TDMD was again inefficient compared to primary neurons (Figure 5A). In addition to these major concerns, there are also more minor concerns that arose during the review of this manuscript Minor comments 1- In figure 2, the authors analyzed by deep- sequencing the effect of either a bulged site (inducing TDMD) or of a bulged site but with a mutated seed- match on mir- 132 accumulation and modification both in total RNA and in Ago2 IP. It would have been interesting to also perform this analysis with a target site presenting a perfect seed- match, but a mutated 3' site. In other words, how are mirnas that are not affected by TDMD affected and incorporated into Ago2. This is partly done in Figure E1, but not from Ago2 IP. Figure E1 shows that target sites with a perfect seed- match and a mutated 3' site neither induce TDMD nor change the tailing/trimming patterns when looking at total RNA. This implies that the effect of this target on the Ago2- associated mirnas will be negligible, especially considering that a significant fraction of mirnas are associated to FLAG/HA- Ago2 in our conditions (ca. 30%) and that there is a tight correlation in the effect of the extensively complementary target on input and Ago2- associated mirnas (Figure 2). Hence, the experiment proposed by the referee would aim at looking for a very small change in tailing or other type of modification induced by the seed- match target on the Ago2- associated mirnas. We respectfully suggest that this would not rule out, nor contribute significantly to understanding of, the causality of tailing for TDMD. 2- Although the authors did a fine job in showing that mirna decay seems to be initiated within Ago2, there is no mention of the fate of the target RNA. Is it more efficiently unloaded from Ago2 when there is high level of complementarity? Or is the tailing of the mirna occurring on Ago2 still 3

4 bound to its target? We provide a new figure (Figure E8) showing the fate of the WT vs. mut target on Ago2 after induction. The figure shows that as the mirna is depleted from Ago2 (Figure E8C), the WT target becomes unloaded from Ago2 relative to the mut target (Figure E8B). Thus, mirna decay and target unloading are well correlated in time. This further points at a direct effect of the target acting in a ternary complex with the mirna. 3- In Figure 4, the authors show that the level of the mirna is important for TDMD. Surprisingly, they also show that in some cases, a unique highly complementary binding site is more efficient to induce TDMD. At the same time, they also record the effect of the mirna binding site on the target by measuring mrna decay. However, they do not provide evidence that the effect on the mrna is linked to the effect at the protein level. Could it be that a binding site, which does not induce mrna decay, but induces TDMD, results in translational inhibition anyway? We provide a new figure (Figure E9) showing images of GFP fluorescence in neurons expressing the 1X, 4X and mut target against mir- 132 upon increasing amounts of transduced pri- mir- 132 (corresponding to Figure 4C- F). The pictures show a clear drop in GFP intensity for the 4X target but not for the 1X or mut targets upon increase in mir- 132 abundance. This indicates that a single binding site does not induce translational inhibition on the 1X target under these conditions, probably due to the high TDMD, which prevails over canonical mirna inhibition. (Note the the PDF file for Figure E9 consists on two pages: (A) GFP channel for the targets only and (B) merged GFP (targets) and Cherry(pri- mir- 132; the expression vector of which also encodes Cherry).) 4

5 Referee #2: In the manuscript "Potent degradation of neuronal mirnas induced by highly complementary targets" de la Mata et al. characterize the mechanism of target RNA- directed mirna decay in rodent primary neurons. Using lentiviral- mirna reporter transduction of primary hippocampal cultures, the authors attempt to qualitatively and quantitatively dissect target RNA- directed mirna decay. While the manuscript is generally interesting, much of the presented data regarding the molecular details of target RNA- directed mirna decay recapitulates previously reported findings in a different cellular context. Furthermore, a major problem arises from the quantitative studies on the comparison of TDMD and mirna- directed mrna turnover, where TDMD- efficiencies are determined using taxman assays (see below). Finally, comparison of TDMD- efficiency in different cellular contexts is interesting but requires more thorough quantitative studies on the molecular details of the processes (see below). We thank the referee for expressing his/her interest in our work. We also find his/her concerns about Taqman assays conceptually well founded, but, as will become evident from the detailed responses below, have indeed taken great pains to ensure that we are not examining experimental artifacts, by cross- validating Taqman assays, Northern blotting, and sequencing, all of which yield comparable results. Major points - Based on high- throughput sequencing of small RNAs derived from TDMD- reporter- transduced neuronal cultures the authors conclude, that "target- induced decay is highly correlated with tailing and trimming of targeted mirnas" (Figure 2). The presentation of the underlying data is quiet confusing. Perhaps the separate display of genome- matching, tailed, and trimmed mir- 132 species would facilitate to follow the authors' reasoning (e.g. divided into classes described in the test)? We could include additional labels to further differentiate what is genome- matching, tailed and trimmed species. However we do not think that displaying them separately (e.g. in different panels) would be helpful as it would hinder a comparison across all species. At any rate, we hope that our response to the referee's following comment will clarify the situation. Irrespectively, the authors should provide a quantitative measure for the proposed correlation together with statistical tests in order to support the conclusion. We have done this and quantified the association between the change in isoform expression and tailing/trimming by calculating the spearman rank correlation between the isoform rank in figure 2A and the corresponding isoform length. Reflecting the clear patterns visible from the figure, the correlation was 0.85 (p- value < 1.3e- 51), thus providing strong additional support for our conclusion. - Given the highly- quantitative aspect of the work it is unclear why the authors do not comment on an obviously strong differential effect of bulge geometry on TDMD: While a symmetric 3/3 mismatch bulge results in more than 25- fold reduction in mir- 132 levels, the asymmetric 4/3 bulge only causes a ~10- fold decrease, which is similar if not weaker to the effect observed with a symmetric 5/5 bulge. We thank the reviewer for this insightful comment, and we will comment on the differential effect of bulge geometry on TDMD. 5

6 - The authors propose that the relative abundance of target RNA and mirna dictate the regulatory outcome regarding TDMD or mrna decay, respectively. Furthermore, the presence of multiple sites is required for efficient mrna decay but not TDMD. It is unclear if the authors have determined the relative expression of mirna and mrna for each of the reporters, to exclude secondary effects on mrna abundance that may mask the outcome. Also, the primary expression data (not only fold change in abundance) may be useful to estimate the consistency in overall expression levels. We are in full agreement with the referee that absolute quantification is important to rule out potential confounding effects. In fact, we believe it to be a particular strength of our manuscript that we provide such data extensively - to our knowledge a first in the field. Thus, in Figure 4, we determine both the relative and absolute expression of mirna and mrna for each of the reporters. This is expressed as fold changes (WT vs mut) in Figure 4D and as the absolute number of molecules degraded (WT vs mut) in Figure 4E- F. Both ways of visualizing the data yield the same conclusion: multiple binding sites are required for efficient mrna decay whereas a single site is enough for efficient TDMD. Hence, we can exclude confounding effects such as those considered by the reviewer. If deemed relevant by the reviewer and/or editor, we can provide a table with the raw data for all mirna and mrna absolute expression values and the calculations involved. - For the absolute quantification of mirna levels, the authors rely on qpcr using the Taqman assay, which quantifies changes in abundance base on a single 3 isoform. This is a severe problem when assessing the relative decay of mirnas and target mrnas, because tailed and trimmed mirna species escape detection. Since these species contribute significantly to the overall concentration of the particular mirna the Taqman assay may severely overestimate TDMD efficiencies (i.e. underestimate the absolute concentration of mirnas inside the cell, which - under TDMD conditions - contain a much higher 3 end heterogeneity that is not detected by the Taqman assay, as shown in Fig. 2). We provide a new figure (Figure E10) showing that the quantification of mir- 132 TDMD with different techniques yields similar results. The Taqman assay overestimates TDMD efficiency by less than 10% compared to deep sequencing quantification of all mir- 132 isoforms (Figure E10A). This confirms that qrt- PCR detects mainly the mature mir- 132 isoform, which is the one most highly affected by the target at steady- state levels. The sum of all other mir- 132 isoforms are reduced to a lower extent by the 4X target, but, because they are overall less abundant, a similar overall TDMD is detected when quantifying all mir- 132 isoforms together. Similarly, quantification of TDMD by either Taqman assay or northern blot results in virtually identical results (Figure E10B). This shows that mature mir- 132 is essentially the only isoform detected by Northern blot (Figure E1), and also the one predominantly affected by the target. The fact that Northern blotting reveals a single band is explained by the fact that the most abundant trimmed and tailed species is 2 log2 units (i.e. 4- fold) lower than the mature form (Figure 2C). Under TDMD conditions (i.e. in the presence of a 4X WT target), despite the sum of tailed and trimmed species contributing significantly to the overall concentration of this particular mirna (Figure E10), they are individually too low to be detected as single bands. - Comparison of TDMD efficiencies in different cell types encounters the same problem described in the previous point. TDMD may be overestimated and may not be comparable between different cells, because a mere change in modification efficiency (not necessarily correlated to mirna turnover) would already interfere with detection and therefore wrongly assigned to mirna turnover. 6

7 We respectfully disagree with the referee's conclusion. Even if we overestimated TDMD in other cell types, this would in the worst case bolster, not undermine our conclusions. This is because we already detect very low levels of TDMD in these other cell types, and conclude that it is rather inefficient in non- primary neuronal cells. Hence, if we were overestimating it, the real magnitude of TDMD would be even lower. Nevertheless we provide a new figure (Figure E11) showing a Northern blot for mir- 132 in HEK cells expressing pri- mir- 132 in the presence of a 4X WT or mut target. The figure shows that mir- 132 is also detected as a single band and rules out, at least in this cell type, that targets might affect mainly their modification without changing the mature mirna levels. 7

8 Referee #3: In the manuscript "Potent degradation of neuronal mirnas induced by highly complementary targets", the authors describe a mirna degradation pathway induced by highly complementary mrna targets in primary neuronal cells. They name this phenomenon target RNA- directed mirna degradation or TDMD. Through a series of carefully designed experiments, the authors conclude that TDMD has several distinct features compared to mirna- induced target degradation. The following specific suggestions need to be addressed before publication: We thank this reviewer for his positive evaluation and constructive criticism of our work. Major points: 1. The manuscript lacks a general summary of the base- pairing rules that dictatestdmd. The authors should try to generalize their conclusions to other mirnas. The conclusion should summarize not only the results for mir- 132, but also that of other mirnas. For example, how many mismatches can be tolerated in TDMD? Is ANY kind of base- pairing interaction sufficient to induce TDMD (i.e. making compensatory mutations in the mirna and target mrna)? Is there a minimal ΔG requirement? We can add a paragraph in the discussion summarizing the pairing rules of to our analysis. If deemed necessary by the editor, we would also be happy to dissect pairing rules further for other mirnas by generating and testing a few more target constructs. 2. A major concern of the study is the quantification of mir- 132 levels using the Q- PCR assay throughout. The Methods section indicates that the Taqman assay used will only quantify levels of the mature mir- 132 sequence. However, as the authors show in Figure 2, there are many other mir- 132 isoforms present in cells. Some of these isoforms are expressed at levels similar to mir They are loaded into the Argonaute proteins, and are functional. The mature mir- 132 represents a portion of the total mir- 132 molecules (trimmed + tailed + mature mir- 132). Importantly, levels of these isoforms change dramatically upon expression of the targets as shown in Figure 2. It is crucial to know levels of the major mir- 132 isoforms in Figures 3, 4 and 5. The authors should use deep- sequencing, Q- PCR (different Taqman primers) or high- resolution Northern blots to measure levels of all major isoforms of mir Furthermore, although the authors show a mir- 132 Northern blot in Figure E1A, it is not clear why mir- 132 only appears as a single band. In Figure 2 and Figure E1E, there are several mir- 132 isoforms that are as abundant as mature mir The referee s observation that some isoforms are expressed at levels similar to mature mir- 132 is inaccurate. As explained above, the most abundant trimmed and tailed species are 2 log2 units (i.e. 4- fold) lower than the mature form (Figure 2C). Nevertheless (and as indicated in our response to Referee #2's concerns and repeated here) we provide a new figure (Figure E10) showing that the quantification of mir- 132 TDMD with different techniques detects similar effects. The Taqman assay overestimates TDMD efficiency by less than 10% compared to deep sequencing quantification of all mir- 132 isoforms (Figure E10A). This confirms that qrt- PCR detects mainly the mature mir- 132 isoform, which is the most highly affected one by the target at steady state levels. The sum of all other mir- 132 isoforms are reduced to a lower extent by the 4X target, but, because they are overall less abundant, a similar overall TDMD is detected when quantifying all mir- 132 isoforms together. Similarly, quantification of TDMD by either Taqman assay or northern blot result in essentially identical results (Figure E10B). This shows that mature mir- 132 is essentially the only isoform detected by Northern 8

9 blot (Figure E1), and also the one predominantly affected by the target. The fact that Northern blot detects a single band is explained by the fact that the most abundant trimmed and tailed species is 2 log2 units (i.e. 4- fold) lower than the mature form (Figure 2C). Under TDMD conditions (i.e. in the presence of a 4X WT target), the sum of tailed and trimmed species contribute significantly to the (reduced) overall concentration of this particular mirna (Figure E10). However, individually their levels are too low to be detected as single bands. 3. In the legend of Figure 2, the authors state that they omit a set of 34 mir- 132 isoforms that show expression differences between cells expressing wildtype and mutant targets. We do not understand why these isoforms were omitted because in Figure 2 there are many mir- 132 isoforms that show expression differences between cells expressing wildtype and mutant targets at time point 0. The authors must include all the experimental data, or they need to elaborate the biological and scientific reasons that justify omitting them. The expression of the inducible target is indeed leaky since it can be detected even at time point 0 (Figure E8). This explains why some mir- 132 isoforms show expression differences between cells expressing WT and mut targets at time point 0. The 34 omitted isoforms are already higher at time 0 for the WT target and do not change significantly after further induction of the target, meaning that their induction is already saturated throughout the time course. Therefore they do not bring insight into the dynamics of the system and were consequently filtered out. However they are not detrimental to our conclusions and can be included in the heatmap at the editor's discretion. 4. Readers would benefit more from the manuscript if the authors could summarize or discuss the practical information concerning degradation in neuronal cells: a. How many target mrna molecules (1X target site) are required to degrade a relatively abundant mirnas (e.g. mir124) at a certain time point (e.g. 48hr). b. Is it possible to combine two single target sites for different mirnas on the same mrna molecule to achieve double mirna knock- down? This is an excellent suggestion, and we will be happy to calculate such numbers. Moreover, the fact that 4x targets can be active for TDMD suggests that double mirna knock- down using a double 1x- target should be feasible, unless very abundant mirnas are targeted. If desired and deemed relevant by the editor, we could generate some double knock- down constructs to test this (e.g. mir- 132+miR- 212, mir- 132+miR- 128, etc). 5. The results presented in Figure 4 are important and interesting and should be summarized in the abstract. We thank the reviewer for this excellent suggestion and will revise the abstract as suggested. Minor points: 1. The authors use equal MOI of lentiviruses to achieve equal expression of exogenous genes throughout the manuscript. However, in most cases, there is no control of expression for different constructs (WT vs mut, highly complementary targets vs mimics of endogenous targets, etc.). The authors should quantitate the expression of the selection marker (such as GFP) as a control. We have measured GFP mrnas across several experiments and find them to correlate well with MOI. We will state this in the Materials and Methods and/or can include data in the extended section if preferred by the reviewer. 9

10 2. It is not clear that the mirna depicted in Figure 1A is mir The authors should also show base- pairing interactions between mir- 124 and the target. In Figure 1A, the length of the 3'UTR and the overall length of the mirna target sequences should be labeled. In addition, the authors should explain what the "N's" in Figure 1A are. We will make the suggested changes. Concerning the "N's", all four sites in the 4x targets contain different loops to reduce repetitiveness; this will now be explained in the figure legend. 3. Please use U instead of T in all the figures that show RNAs (Figures 1A, 3A, 3C, 3E etc.). We will make the requested corrections. 4. Figure E1A and E1B are important experiments. The authors should consider moving them to main figures. We will move these panels to the main figures. 5. Figure 2 should be relabeled so that it is clear to readers that each row represents a different mir- 132 isoform. Also, how many fold is the FLAG/HA- Ago2 overexpressed relative to the endogenous Ago2? We will relabel the figure accordingly. We will quantify the overexpression of FLAG/HA- Ago2 over endogenous Ago2 and provide the relevant numbers. 6. Page 7, last paragraph. Please clarify "we examined the small RNA sequencing data for isoforms of mir- 132": from what samples? We will clarify this sentence by including in rat hippocampal neurons transduced with the targets described below. 7. The authors claim that TDMD and target degradation are two antagonistic processes because relative TDMD effects seem to decrease with increasing mirna abundance. This is not true, because the authors point out later in the manuscript that the absolute mirna degradation amount (or absolute TDMD) is still higher with high mirna abundance, following the same trend as target degradation. Although the authors clearly show that TDMD and target degradation have distinct features, the word "antagonistic" should be changed perhaps into "independent". We will use the term independent. 8. The first sentence of the second paragraph on page 4 is misleading because not all the mirnas degraded by viruses (cited in the references) are anti- viral. We will change the sentence to: Certain viruses deploy TDMD to degrade host mirnas, some of which have antiviral activities. 9. In Figure 3E, the authors should remove "=seed" and " seed" labels because they are confusing. Similarly for the corresponding supplementary figure. We will remove the labels. 10

11 10. In Figure 4B, the authors should show base- pairing interactions between each of the mirnas and their targets. We would like to refrain from including such a schematic in the main figure, because it would make it too cluttered. However, they follow the same architecture as in shown Figure 1 for mir- 132, which we will indicate. If needed, we could further add a supplementary figure with all base- pairing interactions in detail. 11. On page 12, instead of saying "substoichiometrically" and "superstoichiometrically", please simply state the numeric fold change. For example, the mir- 124 level is about 3- fold more than the target mrna level. We will rephrase this sentence. 11

12 A Absolute 4X target mrna levels B 4x vs. mut target enrichment C Relative mir-132 levels on Ago2 Figure E8 12

13 A GFP channel 1:27 1:9 1:3 1x WT 4x WT 4x mut endo +pri mir

14 B GFP+Cherry merge 1:27 1:9 1:3 1x WT 4x WT 4x mut endo +pri-mir-132 Figure E9 14

15 Quantification of TDMD by different techniques A qrt-pcr Deep sequencing mature mir-132 All mir-132 isoforms mir-132 levels other isoforms mature mir mir-132 % reduction mut 94.5% WT mut 90.8% WT mut 86.2% WT B qrt-pcr Northern blot mir-132 levels mut WT mut WT mir-132 % reduction 81.5% 80% Figure E10: Quantification of mir-132 TDMD by different techniques detects similar effects. (A) Experiment comparing qrt-pcr (Taqman assay) vs. small RNA deep sequencing. qrt-pcr detects 94.5% reduction in mir-132 levels in total RNA from neurons infected for 6 days with a WT 4x target relative to a mut target. Small RNA deep sequencing detects a similar reduction in mature mir-132 levels (90.8%), and a slightly lower reduction in the level of all mir-132 isoforms (86.2%). This suggests that qrt-pcr detects mainly the mature mir-132 isoform which is the one mostly affected by the target at steady state levels. Other mir-132 isoforms are also reduced though to a lower extent, and, by being overall less abundant at steady state, they do not affect the overall change when quantifying all the isoforms together. (B) Experiment comparing qrt-pcr (Taqman assay) vs. Northern blot. Both techniques detect similar reduction in mature mir-132 levels in total RNA from neurons infected for 6 days with a WT 4x target relative to a mut target. This confirms that mature mir-132 is is the most abundant isoform, and also the one predominantly affected by that target. 15

16 HEK-293T cells - Northern blot Non-Transf. 4X WT mut mir132 U6 Figure E11 16

17 EMBO reports - Peer Review Process File - EMBOR st Editorial Decision 09 January 2015 Thank you for the transfer of your manuscript and point-by-point response to our journal. The former referee 2 has seen your response now and supports publication of the study by EMBO reports after minor revisions. Given that also referee 3 is supportive, we can offer to publish your manuscript. I suggest that you proceed as you propose in your point-by-point response. Regarding additional data, please include raw data and calculations used to obtain absolute expression values for mirnas and mrnas, as referee 2 suggests. These could be presented as source data that we will link to the respective figure legend/s. While the base-pairing rules that dictate TDMD should be summarized and discussed, pairing rules for other mirnas do not need to be tested. The generation of RNA targets with single binding sites for 2 different mirnas is also not required for publication of the study here. It will further be sufficient to state the correlation between GFP mrna and MOI and to justify in the text why you omit certain mir-132 isoforms from the analysis. A supplementary figure with all base-pairing interactions between the mirnas and their targets is also not essential. Importantly, all important findings should be mentioned in the abstract. Please send us the revised, final manuscript file as soon as possible. Given that the manuscript has 5 main figures, we can publish it as a short report with a longer text. However, please shorten the text as much as possible, especially the discussion is rather long. You can combine the results and discussion section, which may help to eliminate some redundancy that is inevitable when discussing the same experiments twice. Commonly used materials and methods can further be moved to the supplementary information, however, please note that materials and methods essential for the understanding of the experiments described in the main text must remain in the main manuscript file. Please keep the accession numbers in the main manuscript text. It would also be good to reduce the number of supplementary figures to below 10 (e.g. 7). Some of them may be combined, or added to the main figure/s. If you prefer, you could also include 1 or 2 more figures in the main manuscript file. But also in this case the manuscript text should be shortened. I look forward to seeing a revised version of your manuscript as soon as possible. Please let me know if you have questions or comments regarding the revision. 1st Revision - authors' response 21 January 2015 Referee #1: In this interesting, and nicely written manuscript, de la Mata et al. report on the characterization of mirna degradation induced by targets. Basically, the authors show here the molecular requirements for an efficient degradation of mirnas in primary neurons. It has been previously reported that the stability of some mature mirnas could be very quickly jeopardized in response to certain stimuli, such as dark to light transition in the retinal neurons. Others have reported that under some conditions, highly complementary targets, either artificial in the form of synthetic oligonucleotides, or natural as expressed by some viruses, could trigger decay of the targeting mirna. This mechanism, which the authors refer to as target RNA- directed mirna degradation (TDMD), appears to require 3' addition of nucleotides on the mirna (tailing), coupled to its 3' to 5' exonucleolytic degradation (trimming). In this report, the authors confirm that indeed TDMD is functional in primary neurons (and surprisingly enough, only in primary neurons with their approach) and that it depends both on the degree of complementarity of the target RNA and the bound mirna, and on the respective levels of the two molecules. Thus, they show that transducing a construct consisting of the coding sequence of GFP and 4 binding sites with a central bulge for mir- 132 results in an efficient decrease in mir- 132 levels. Using deep- sequencing, they European Molecular Biology Organization 17

18 EMBO reports - Peer Review Process File - EMBOR then provide evidence that tailing and trimming of the mirna occurs, and that tailing appears to be initiated within Ago2. Disrupting the pairing at the 3' of the mirna, increasing the size of the bulge, or using regular target site results in the loss of TDMD. Only mirnas expressed under a certain absolute level can be targets of TDMD, and there is an inverse correlation between the targeting efficiency of the mirna and TDMD. In other words, when the mirna is expressed at sufficiently high levels, it will resume its regulatory activity, and will not be degraded. We thank the reviewer for commending on the qualities of our manuscript and the underlying work. At the same time, we would like to point out that his/her summary of the work, and the ensuing comments, reflect mostly our results with the 4x target. By contrast, they do not consider the differences found between the 4x and 1x targets. For instance, the referee concludes that only mirnas expressed under a certain absolute level can be targets of TDMD, but we show (Fig. 4) that even a highly expressed mirna such as mir- 124 (or overexpressed mir- 132) undergoes efficient TDMD with a 1x target. Hence, despite a high targeting efficiency of this mirna, it will not resume regulatory activity. There is thus no general inverse correlation between the targeting efficiency of the mirna and TDMD. Major comments The experiments are state of the art, are very well performed and described and they do back up the conclusions of the authors. However, the manuscript falls short in providing an explanation as to how TDMD occurs naturally in primary neurons. In its present form, it is merely a confirmation that indeed target mediated decay can work, but only using artificial targets. We respectfully disagree with the reviewer's claim that our manuscript is merely a confirmation of previous work. We believe this misunderstanding to result from the fact that our data showing that we can uncouple TDMD and mrna decay escaped the referee's attention (see comment above). By revealing that mrna and mirna decay are not linked; that the former but not the latter depends on target site cooperativity; and that TDMD exhibits an unforeseeably high efficiency in neurons, likely due to a multiple turnover activity, our work provides novel mechanistic insight into the thus- far poorly understood pathway of mirna decay. Consistent with a recommendation by referee #3, we have now highlighted these important findings better in the abstract. 1- It would have therefore been nice to identify at least one naturally expressed target in primary neurons that could explain the rapid decay of a specific mirna. This might indeed prove difficult, but it would add weight to the manuscript. 2- Similarly, there is no identification of a putative factor involved in the tailing or in the degradation of the mirna. Although we agree with the referee that both of these would make very exciting findings, identification of endogenous targets or contributing factors will indeed prove difficult particularly in primary neurons where limiting amounts of starting material compromise the feasibility of high throughput experiments. Our current efforts are aimed in this direction, but this will be a long- term project, beyond the scope of the current work. 3- The observation that TDMD seems to be more efficient in primary neurons is an interesting one, but maybe this is due to the fact that these are primary cells. It would have been nice to show whether TDMD is also functional in primary cells of another origin. We would like to point out that we already tested TDMD in primary cells of another origin, specifically murine embryonic fibroblasts (MEFs). In this primary cell type TDMD was again inefficient compared to primary neurons (Figure 5A). European Molecular Biology Organization 18

19 EMBO reports - Peer Review Process File - EMBOR In addition to these major concerns, there are also more minor concerns that arose during the review of this manuscript Minor comments 1- In figure 2, the authors analyzed by deep- sequencing the effect of either a bulged site (inducing TDMD) or of a bulged site but with a mutated seed- match on mir- 132 accumulation and modification both in total RNA and in Ago2 IP. It would have been interesting to also perform this analysis with a target site presenting a perfect seed- match, but a mutated 3' site. In other words, how are mirnas that are not affected by TDMD affected and incorporated into Ago2. This is partly done in Figure E1, but not from Ago2 IP. Figure E1 shows that target sites with a perfect seed- match and a mutated 3' site neither induce TDMD nor change the tailing/trimming patterns when looking at total RNA. This implies that the effect of this target on the Ago2- associated mirnas will be negligible, especially considering that a significant fraction of mirnas are associated to FLAG/HA- Ago2 in our conditions (ca. 30%) and that there is a tight correlation in the effect of the extensively complementary target on input and Ago2- associated mirnas (Figure 2). Hence, the experiment proposed by the referee would aim at looking for a very small change in tailing or other types of modification induced by the seed- match target on the Ago2- associated mirnas. We respectfully suggest that this would not rule out, nor contribute significantly to understanding of, the causality of tailing for TDMD. 2- Although the authors did a fine job in showing that mirna decay seems to be initiated within Ago2, there is no mention of the fate of the target RNA. Is it more efficiently unloaded from Ago2 when there is high level of complementarity? Or is the tailing of the mirna occurring on Ago2 still bound to its target? We provide a new figure (new Figure E2E- G) showing the fate of the WT vs. mut target on Ago2 after induction. The figure shows that as the mirna is depleted from Ago2 (Figure E2F), the WT target becomes unloaded from Ago2 relative to the mut target (Figure E2G). Thus, mirna decay and target unloading are well correlated in time. This further points at a direct effect of the target acting in a ternary complex with the mirna on Ago. 3- In Figure 4, the authors show that the level of the mirna is important for TDMD. Surprisingly, they also show that in some cases, a unique highly complementary binding site is more efficient to induce TDMD. At the same time, they also record the effect of the mirna binding site on the target by measuring mrna decay. However, they do not provide evidence that the effect on the mrna is linked to the effect at the protein level. Could it be that a binding site, which does not induce mrna decay, but induces TDMD, results in translational inhibition anyway? We provide a new figure (new Figure E4) showing images of GFP fluorescence in neurons expressing the 1X, 4X and mut target against mir- 132 upon increasing amounts of transduced pri- mir- 132 (corresponding to Figure 4C- F). The pictures show a clear drop in GFP intensity for the 4X target but not for the 1X or mut targets upon increase in mir- 132 abundance. This indicates that a single binding site does not induce translational inhibition on the 1X target under these conditions, probably due to the high TDMD, which prevails over canonical mirna inhibition. Referee #2: In the manuscript "Potent degradation of neuronal mirnas induced by highly complementary targets" de la Mata et al. characterize the mechanism of target RNA- directed mirna decay in rodent primary neurons. Using lentiviral- mirna reporter European Molecular Biology Organization 19

20 EMBO reports - Peer Review Process File - EMBOR transduction of primary hippocampal cultures, the authors attempt to qualitatively and quantitatively dissect target RNA- directed mirna decay. While the manuscript is generally interesting, much of the presented data regarding the molecular details of target RNA- directed mirna decay recapitulates previously reported findings in a different cellular context. Furthermore, a major problem arises from the quantitative studies on the comparison of TDMD and mirna- directed mrna turnover, where TDMD- efficiencies are determined using taxman assays (see below). Finally, comparison of TDMD- efficiency in different cellular contexts is interesting but requires more thorough quantitative studies on the molecular details of the processes (see below). We thank the referee for expressing his/her interest in our work. We also find his/her concerns about Taqman assays conceptually well founded, but, as will become evident from the detailed responses below, have indeed taken great pains to ensure that we are not examining experimental artifacts, by cross- validating Taqman assays, Northern blotting, and sequencing, all of which yield comparable results. Major points - Based on high- throughput sequencing of small RNAs derived from TDMD- reporter- transduced neuronal cultures the authors conclude, that "target- induced decay is highly correlated with tailing and trimming of targeted mirnas" (Figure 2). The presentation of the underlying data is quiet confusing. Perhaps the separate display of genome- matching, tailed, and trimmed mir- 132 species would facilitate to follow the authors' reasoning (e.g. divided into classes described in the test)? Irrespectively, the authors should provide a quantitative measure for the proposed correlation together with statistical tests in order to support the conclusion. We indicate in the Figure trimmed as well as the genome matching mature mir- 132 species. We consider the proposed separate display of genome- matching, tailed and trimmed species problematic because it would hinder a comparison across all species. However, as proposed by the referee, we have quantified the association between the change in isoform expression and tailing/trimming by calculating the Spearman rank correlation between the isoform rank in Figure 2A and the corresponding isoform length. Reflecting the clear patterns visible from the figure, the correlation was 0.85 (p- value < 1.3e- 51), thus providing strong additional support for our conclusion. - Given the highly- quantitative aspect of the work it is unclear why the authors do not comment on an obviously strong differential effect of bulge geometry on TDMD: While a symmetric 3/3 mismatch bulge results in more than 25- fold reduction in mir- 132 levels, the asymmetric 4/3 bulge only causes a ~10- fold decrease, which is similar if not weaker to the effect observed with a symmetric 5/5 bulge. We thank the reviewer for this insightful comment. However, we feel that additional data points would be needed to establish a common theme of efficiencies for symmetric vs. asymmetric loops. We do not have these currently. - The authors propose that the relative abundance of target RNA and mirna dictate the regulatory outcome regarding TDMD or mrna decay, respectively. Furthermore, the presence of multiple sites is required for efficient mrna decay but not TDMD. It is unclear if the authors have determined the relative expression of mirna and mrna for each of the reporters, to exclude secondary effects on mrna abundance that may mask the outcome. Also, the primary expression data (not only fold change in abundance) may be useful to estimate the consistency in overall expression levels. European Molecular Biology Organization 20

21 EMBO reports - Peer Review Process File - EMBOR We are in full agreement with the referee that absolute quantification is important to rule out potential confounding effects. In fact, we believe it to be a particular strength of our manuscript that we provide such data extensively - to our knowledge a first in the field. Thus, in Figure 4, we determine both the relative and absolute expression of mirna and mrna for each of the reporters. This is expressed as fold changes (WT vs mut) in Figure 4D and as the absolute number of molecules degraded (WT vs mut) in Figure 4E- F. Both ways of visualizing the data yield the same conclusion: multiple binding sites are required for efficient mrna decay whereas a single site is enough for efficient TDMD. Hence, we can exclude confounding effects such as those considered by the reviewer. We provide a table with the raw data for all mirna and mrna absolute expression values and the calculations involved, which will be linked to the relevant figures in the published manuscript. - For the absolute quantification of mirna levels, the authors rely on qpcr using the Taqman assay, which quantifies changes in abundance base on a single 3 isoform. This is a severe problem when assessing the relative decay of mirnas and target mrnas, because tailed and trimmed mirna species escape detection. Since these species contribute significantly to the overall concentration of the particular mirna the Taqman assay may severely overestimate TDMD efficiencies (i.e. underestimate the absolute concentration of mirnas inside the cell, which - under TDMD conditions - contain a much higher 3 end heterogeneity that is not detected by the Taqman assay, as shown in Fig. 2). We provide a new figure (Figure E1D- E) showing that the quantification of mir- 132 TDMD with different techniques yields similar results. The Taqman assay overestimates TDMD efficiency by less than 10% compared to deep sequencing quantification of all mir- 132 isoforms (Figure E1D). This confirms that qrt- PCR detects mainly the mature mir- 132 isoform, which is the one most highly affected by the target at steady- state levels. The sum of all other mir- 132 isoforms are reduced to a lower extent by the 4X target, but, because they are overall less abundant, a similar overall TDMD is detected when quantifying all mir- 132 isoforms together. Similarly, quantification of TDMD by either Taqman assay or northern blot results in virtually identical results (Figure E1E). This shows that mature mir- 132 is essentially the only isoform detected by Northern blot (Figure 1C), and also the one predominantly affected by the target. The fact that Northern blotting reveals a single band is explained by the fact that the most abundant trimmed and tailed species is 2 log2 units (i.e. 4- fold) lower than the mature form (Figure 2C). Under TDMD conditions (i.e. in the presence of a 4X WT target), despite the sum of tailed and trimmed species contributing significantly to the overall concentration of this particular mirna (Figure E1), they are individually too low to be detected as single bands. - Comparison of TDMD efficiencies in different cell types encounters the same problem described in the previous point. TDMD may be overestimated and may not be comparable between different cells, because a mere change in modification efficiency (not necessarily correlated to mirna turnover) would already interfere with detection and therefore wrongly assigned to mirna turnover. We respectfully disagree with the referee's conclusion. Even if we overestimated TDMD in other cell types, this would in the worst case bolster, not undermine our conclusions. This is because we already detect very low levels of TDMD in these other cell types, and conclude that it is rather inefficient in non- primary neuronal cells. Hence, if we were overestimating it, the real magnitude of TDMD would be even lower. Nevertheless we provide a new figure (Figure E6C) showing a Northern blot for mir- 132 in HEK cells expressing pri- mir- 132 in the presence of a 4X WT or mut target. The figure shows that mir- 132 is also detected as a single band and rules out, at least in this cell type, that targets might affect mainly their modification without changing the mature mirna levels. European Molecular Biology Organization 21