TECHNOLOGY DEVELOPMENT

Transcription

1 TECHNOLOGY DEVELOPMENT Optimized Flow Cytometric Analysis of Central Nervous System Tissue Reveals Novel Functional Relationships Among Cells Expressing CD133, CD15, and CD24 DAVID M. PANCHISION, a HUI-LING CHEN, a FRANCESCA PISTOLLATO, a DANIELA PAPINI, b HSIAO-TZU NI, c TERESA S. HAWLEY d a Center for Neuroscience Research, Children s Research Institute, Children s National Medical Center, Washington, DC, USA; b Stem Cell Core Facility, Children s Research Institute, Children s National Medical Center, Washington, DC, USA; c R&D Systems Inc., Minneapolis, Minnesota, USA; d Flow Cytometry Core Facility, George Washington University Medical Center, Washington, DC, USA Key Words. Brain Brain tumors Flow cytometry Cell viability Stem cells Progenitor cells Disclosure of potential conflicts of interest is found at the end of this article. ABSTRACT Although flow cytometry is useful for studying neural lineage relationships, the method of dissociation can potentially bias cell analysis. We compared dissociation methods on viability and antigen recognition of mouse central nervous system (CNS) tissue and human CNS tumor tissue. Although nonenzymatic dissociation yielded poor viability, papain, purified trypsin replacement (TrypLE), and two purified collagenase/neutral protease cocktails (Liberase-1 or Accutase) each efficiently dissociated fetal tissue and postnatal tissue. Mouse cells dissociated with Liberase-1 were titrated with antibodies identifying distinct CNS precursor subtypes, including CD133, CD15, CD24, A2B5, and PSA-NCAM. Of the enzymes tested, papain most aggressively reduced antigenicity for mouse and human CD24. On human CNS tumor cells, CD133 expression remained highest after Liberase-1 and was lowest after papain or Accutase treatment; Liberase-1 digestion allowed magnetic sorting for CD133 without the need for an antigen re-expression recovery period. We conclude that Liberase-1 and TrypLE provide the best balance of dissociation efficiency, viability, and antigen retention. One implication of this comparison was confirmed by dissociating E13.5 mouse cortical cells and performing prospective isolation and clonal analysis on the basis of CD133/CD24 or CD15/CD24 expression. Highest fetal expression of CD133 or CD15 occurred in a CD24 hi population that was enriched in neuronal progenitors. Multipotent cells expressed CD133 and CD15 at lower levels than did these neuronal progenitors. We conclude that CD133 and CD15 can be used similarly as selectable markers, but CD24 coexpression helps to distinguish fetal mouse multipotent stem cells from neuronal progenitors and postmitotic neurons. This particular discrimination is not possible after papain treatment. STEM CELLS 2007;25: INTRODUCTION The analysis of neuroepithelial precursor cells is indispensable for studying the mechanisms controlling proliferation, mitotic arrest, and lineage commitment in the nervous system [1]. Neural stem cells are a subclass of precursor cells that exhibit (a) extensive self-renewal and (b) multipotency, the ability to generate neurons and glia. Clonal analysis is the standard method for measuring these defining characteristics. This can be performed in vitro by plating cells at limiting dilution, or both in vitro and in grafting studies by genetically marking individual cells within the larger population, such that all cells in a cluster can be identified as the clonal progeny of a single original cell [2 4]. Multipotency is measured by quantifying the variety of differentiated derivatives generated from an individual clone [2]. Although the isolation and manipulation of neural precursors has the potential to introduce artifacts in their behavior, the method is clearly valuable for understanding the basic biology of these cells. Furthermore, the ability to efficiently manipulate neural precursors offers therapeutic tools for either replacing cells that are lost because of neural degeneration [5] or targeting dysfunctional cells in central nervous system (CNS) tumors [6]. Cell-surface antigen-based selection has allowed prospective isolation of subclasses of CNS and neural crest cells [7 11]. This method of enrichment is so named because the selectable property of the cell closely predicts cell function. Recent studies indicate that human CNS precursor cells expressing high levels of the surface antigen CD133 (CD133 /hi ), with little or no CD24 (CD24 /lo ), had the highest frequency of initiating clones as measured by neurosphere formation [12, 13]. There is also evidence that these antigens are useful for characterizing similar subpopulations from rodent CNS [9, 14 17]. High CD24 expression has been used to identify transit-amplifying cells [18] as well as differentiated neurons [19], and CD24 is required for terminal differentiation of neuronal progenitors [20]. The antigen CD15, also known as LeX or stage-specific embryonic antigen 1 (SSEA1), has also been identified as a positive select- Correspondence: David M. Panchision, Ph.D., Children s National Medical Center, Center for Neuroscience Research, 5th Floor, Suite 5340, 111 Michigan Avenue, N.W., Washington, DC 20010, USA. Telephone: ; Fax: ; dpanchision@ cnmcresearch.org Received April 28, 2006; accepted for publication February 20, 2007; first published online in STEM CELLS EXPRESS March 1, AlphaMed Press /2007/$30.00/0 doi: /stemcells STEM CELLS 2007;25:

2 Panchision, Chen, Pistollato et al able marker for rodent multipotent neural stem cells [21 24]. The polysialylated or embryonic form of NCAM (PSA-NCAM, E-NCAM) has been used to identify neuronally committed progenitors or immature neurons [7, 25 29], whereas A2B5 has been used to prospectively isolate bipotent glial-restricted precursors [7, 10, 27, 30, 31]; both cell types appear to be limited in self-renewal potential. A critical step in flow cytometric analysis of the nervous system is the dissociation of tissue into single cells [32]. Many methods of cell dissociation have been successfully used on neural tissue, including nonenzymatic trituration [14, 33, 34], papain [10, 11, 21, 24, 35 37], trypsin [6, 12, 30, 38, 39], and collagenase with either dispase [14, 27, 40] or other neutral proteases [41]. Nonenzymatic trituration has the disadvantage of killing large percentages of cells and leaving undissociated tissue, whereas crude enzymatic preparations can cleave important surface antigens. This hinders antibody titration and reproducible flow cytometric analysis. Proprietary reagents such as recombinant trypsin-like replacement (TrypLE [Invitrogen, Carlsbad, CA, or collagenase/neutral protease (Liberase Blendzymes [Roche Diagnostics, Basel, Switzerland, or Accutase [Innovative Cell Technologies, San Diego, innovativecelltech.com]) solutions have been introduced to eliminate undefined factors and reduce lot-to-lot variability. However, the comparative effect of these different dissociation methods on the analysis of important selectable neural antigens is poorly understood. Here, we compared a number of these dissociation methods on neural tissue for cell viability, dissociation efficiency, and retention of CD133, CD15, and CD24 antigenicity. We found that papain, Liberase-1, TrypLE, and Accutase efficiently dissociate mouse CNS tissue and human CNS tumors into single cells. Liberase-1 and TrypLE provide the best balance of dissociation efficiency, viability, and antigen retention. Using Liberase-1 dissociation, we characterize novel patterns of selectable marker coexpression in mouse fetal cortex; lateral, medial, and caudal ganglionic eminence; and postnatal subventricular zone. We also show that CD133 and CD15 expression closely parallel each other in mouse brain cells and can be used similarly for prospective isolation. However, the highest mean levels of CD133 and CD15 are expressed in neuronal progenitors rather than multipotent stem cells of the fetal mouse forebrain. Selection using CD24 coexpression enhances the discrimination of these populations, but this type of selection is limited by the manner of dissociation. MATERIALS AND METHODS Tissue Dissociation Animal tissue was acquired using an approved protocol in accordance with Institutional Animal Care and Use Committee guidelines. Human pediatric tumor tissue was acquired using an approved protocol in accordance with institutional review board guidelines. Tissues included embryonic day 10.5 (E10.5) mouse cervical spinal cord; 13.5 mouse lateral, medial, and caudal ganglionic eminence (LGE, MGE, and CGE, respectively) and cortex; postnatal day 2 (P2) lateral ventricle wall; and human CNS tumor tissue acquired directly from surgery. Human tumor tissue was gently minced with a scalpel, whereas mouse tissue was left undisturbed before digestion. Initial comparisons were performed with E13.5 mouse LGE, where each LGE pair was resuspended in 1 ml of 1 Hanks buffered saline solution (HBSS; Ca 2 /Mg 2 -free, plus Hepes and 1.55 g/l glucose, without bicarbonate, ph 7.2; Invitrogen) containing 200 units/ml DNase I (Roche) and 1 mm MgCl 2 ; the exception was one LGE pair that was resuspended in growth medium (described later herein), one that was resuspended in purified collagenase/neutral protease cocktail in phosphate-buffered saline (PBS)/ EDTA (Accutase; Innovative Cell Technologies; used neat, one lot tested) and another LGE pair that was resuspended in a purified trypsin-like replacement in Dulbecco s PBS/EDTA (TrypLE Select; Invitrogen; used neat, two lots tested) along with DNase/MgCl 2.To the remaining tubes, we added different concentrations of an highperformance liquid chromatography-purified collagenase/dispase cocktail (Liberase Blendzyme1; Roche; four lots tested, seven vials total) or papain (12 units/ml; Worthington, Lakewood, NJ, two lots tested) that was preactivated as directed in 1.1 mm EDTA, mm mercaptoethanol, and 5.5 mm cysteine-hcl for 30 minutes before addition. The absence of bicarbonate in the HBSS allowed dissections and incubations to be performed in a room atmosphere. Samples were placed in a 37 C rotator oven during digestion. In some experiments, incubation times were varied for each enzyme preparation (highest concentration only for Liberase-1). Samples were then spun at 200g for 5 minutes, resuspended in fresh HBSS/ DNase/MgCl 2 without enzyme, and triturated with three rounds of eight passes through an ART 1000E pipette tip (Molecular Bioproducts, Inc., San Diego, a fourth round of trituration was performed with a flamed Pasteur pipette. After each round of trituration, the tissue was allowed to settle for 5 minutes, and the top 800 l of suspended cells were transferred into a new tube to avoid further disturbance. The combined cell suspensions for each group were spun down again and resuspended in medium appropriate for further use. On the basis of initial comparisons, the standard protocol for tissue digestion for multiple labeling and sorting became 30 minutes (fetal), 60 minutes (neonatal), or 90 minutes (human tumor) incubation with 200 g/ml Liberase-1, corresponding to 0.62 Wünsch unit (WU)/ml collagenase and 66.7 units/ml dispase. The final step of trituration with a flamed Pasteur pipette was eliminated because it introduced variability in the procedure. In some instances, human tumor cells were dissociated after acquisition and grafted into nonobese diabetic/severe combined immunodeficiency (NOD/SCID) mice as previously described [6], using an approved animal protocol. After 4 12 weeks to allow tumor growth, tumor tissue was dissected and enzymatically dissociated before further analysis. Flow Cytometry For flow cytometry, cells were resuspended in flow cytometry buffer, consisting of 1 HBSS, ph 7.2, containing 1.55 g/l glucose and 0.1% fraction V of bovine serum albumin (BSA; Sigma-Aldrich, St. Louis, Cells were counted and diluted to a density of 10 6 cells per milliliter of buffer; all mouse analysis was performed with 100- l aliquots containing cells. For viability analysis, we added 300 l of 1.33 Annexin buffer (diluted from 10 stock) followed by 2 l of Annexin-V- APC (BD Biosciences, San Diego, and incubated at room temperature for 15 minutes; for the final 5 minutes of this incubation, we also added 7-amino-actinomycin-D (7-AAD, Invitrogen) to a final concentration of 50 ng/ml before analysis. Higher concentrations of either dye substantially shifted the fluorescence of the entire population and interfered with detection of other fluorochromes. For mouse surface marker analysis, we used antibodies against CD24 (phycoerythrin [PE]-conjugated rat IgG 2b ; BD Biosciences), CD15 (fluorescein isothiocyanate [FITC]-conjugated mouse IgM; BD Biosciences), CD133 (biotinylated-rat-igg 2b ; R&D Systems Inc., Minneapolis, PSA-NCAM (mouse IgM; Chemicon, Temecula, CA, com), A2B5 (mouse IgM; Chemicon), BMPRIA (goat IgG; R&D Systems), and BMPRIB (mouse IgG 2a ; R&D Systems). Intracellular analysis of BMP activation was performed with a Phospho-Smad1/ 5/8 antibody (rabbit IgG; Cell Signaling Technology, Beverly, MA, For human tumor analysis, we used antibodies against human CD133 (clone AC141-PE, as directed; Miltenyi Biotec, Bergisch Gladbach, Germany, miltenyibiotec.com) and human CD24 (FITC-conjugated mouse IgG 2a ; 1:25 dilution; BD Biosciences). Antibodies were titrated over

3 1562 CNS Cell Dissociation and Flow Cytometry Figure 1. Flow cytometric analysis of mouse fetal central nervous system cell viability. E13.5 mouse lateral ganglionic eminence was dissociated with HBSS alone (A), 400 g/ml Liberase-1 in HBSS (B), 12 units/ml preactivated papain in HBSS (C), TrypLE purified trypsin replacement (D), or Accutase (E). The top panel shows physical characteristics of entire unfiltered population measured by object size (forward scatter) and granularity (side scatter). A polygon delimits the main population of intact cells, whereas events outside of main population consist of pyknotic cells and cellular debris; number indicates main population as a percentage of total events. The bottom panel shows viability of cells within the main population; high Annexin-V staining indicates early apoptotic cells, whereas 7-AAD staining indicates late apoptotic and necrotic cells. Numbers indicate percentages of main population events within each quadrant for this experiment only; see Table 1 and supplemental online Table 1 for complete results. Abbreviations: 7-AAD, 7-amino-actinomycin-D; HBSS, Hanks buffered saline solution. a semi-log scale to determine appropriate dilution and incubated on ice for 30 minutes. Cells were washed in buffer, and then secondary fluorescent-conjugated antibody (if needed) was added at the appropriate dilution and incubated on ice for 15 minutes. Cells were washed once and resuspended in buffer for viability dye staining and analysis. Cells were analyzed on an FACSCalibur flow cytometer (BD Biosciences). Background fluorescence was measured using unlabeled cells and cells labeled with isotype control or secondary antibody alone; these set gating parameters between positive and negative cell populations. Cell aggregates and small debris were excluded from analysis or isolation on the basis of side scatter (measuring cell granularity) and forward scatter (measuring cell size); dead cells were excluded from analysis on the basis of viability dye fluorescence. Fluorescent intensities for cells in the population were point-plotted on two-axis graphs or histogram using CellQuest software (BD Biosciences). Cell Sorting Cells were filtered through a 70- m nylon mesh before final centrifugation, although filtration was not required after fetal tissue dissociation with Liberase-1, then resuspended in flow cytometry buffer. All sorts were performed on cells dissociated with 200 g/ml Liberase-1 (0.62 WU/ml collagenase and 66.7 units/ml dispase). Mouse cells were sorted on fluorescence-activated cell sorter (FACS), either a FACSAria (BD Biosciences) or an Influx (Cytopeia, Seattle, WA, sorter. Single viable cells were gated on the basis of Annexin-V exclusion and pulse width, and then physically sorted into collection tubes for limiting dilution plating. Postsort purity analysis was performed on aliquots from each sort group. FACSDiva (BD Biosciences) or FlowJo ( software was used for analysis. Human tumor cells were sorted by magnetic cell separation (MACS; Miltenyi Biotec) using both selection and detection antibodies against human CD133 (clone AC133/1-microbead and clone AC141-PE, as directed; Miltenyi Biotec). Cell Culture and Clonal Analysis Dissociated mouse cells were assayed for viability by trypan blue exclusion, plated onto fibronectin-coated tissue culture dishes, and cultured in 5% O 2,5%CO 2 in the presence of 20 ng/ml basic fibroblast growth factor (bfgf; R&D Systems) either acutely or added daily as previously described [2, 42]. For acute SMAD activation experiments on passaged cells, BMP2 (R&D Systems) was added at 10 ng/ml immediately after dissociation for 15 minutes; cells were then fixed in fresh 4% paraformaldehyde and permeabilized before staining for flow cytometry. For clonal analysis, sorted mouse cells were plated at 10 cells per mm 2 on coverslips, expanded for 4 days, and differentiated for 6 days as previously described [2, 42]. Immunofluorescence was performed using primary antibodies against MAP2a b (mouse, 1:500, Sigma-Aldrich), glial fibrillary acidic protein (GFAP; rabbit, 1:800; Dako, Fort Collins, CO, and O4 (mouse, 1:100, Sigma-Aldrich), followed by fluorescent secondary antibodies (Alexa dyes; Invitrogen). Cells were counterstained with 4-6-diamidino-2-phenylindole (DAPI) to measure total cell number. Staining was visualized by epifluorescence (BX60 upright microscope; Olympus, Tokyo, and images were compiled for figures using Photoshop 7.0 (Adobe, San Jose, CA, RESULTS Comparison of Mouse CNS Tissue Dissociation Methods We compared several methods of dissociation using nonenzymatic buffers or increasing incubation times of four enzyme preparations: papain, purified trypsin (TrypLE), and two purified collagenase/neutral protease cocktails (Liberase-1 or Accutase). Initial comparisons were done with E13.5 forebrain tissue and extended to E10.5 mouse cervical spinal cord, which contains substantial numbers of proliferative precursors [24]. To assess viability, we incubated the dissociated cells with Annexin-V to identify early apoptotic cells and 7-AAD to identify late apoptotic and necrotic cells (Fig. 1; Table 1). By flow cytometric analysis, we found that the poorest results came from dissociating the cells with growth medium or HBSS alone. In contrast, each of the four enzyme preparations greatly improved dissociation and viability, both in terms of percentage of the main population (intact cells) and the total population (all detected events, including debris). Cell aggregates, as measured by high forward scatter, were common in nonenzyme dissociated cells but minimal in the enzyme-treated preparations. Pulsewidth analysis of Liberase-1-treated tissue also verified a high efficiency of single-cell dissociation (supplemental online Fig. 1A, 1B).

4 Panchision, Chen, Pistollato et al Table 1. Viability assays after dissociation Tissue Dissociation Minutes Trypan blue Main 7AAD Total 7AAD Main Annex V Total Annexin V % SEM No. % SEM No. % SEM No. % SEM No. % SEM No. Mu E10.5 Lib Papain TrypLE Accutase Mu E13.5 DFNF HBSS Lib Lib Lib Lib Lib Lib Papain Papain Papain b TrypLE TrypLE TrypLE Accutase Accutase Accutase Mu P2 Lib n/d n/d HuTu a Lib Papain TrypLE b Accutase HuTu c Lib n/d n/d n/d n/d HuTu c CDB n/d n/d n/d n/d Numbers indicate mean percentage of cells excluding dyes SEM. Trypan blue staining identifies nonviable cells in initial hemocytometer reading; 7-AAD staining identifies late apoptotic and necrotic cells, whereas Annexin-V staining identifies early apoptotic as well as late apoptotic cells. Cytometric gating is as follows: % Main, percentage of intact cells as measured by forward/side scatter characteristics (main population); % Total, percentage of all detected events, including pyknotic cells, debris and statistical noise (total population). Viability of most cultured human tumor cells was quantified by trypan blue exclusion only. a Performed on high-grade glioma. b Two additional papain-treated mouse samples and one additional TrypLE-treated human tumor sample contained substantial aggregates of free-dna and could not be reliably counted. c Performed on passaged cells only. Abbreviations; 7-AAD, 7-amino-actinomycin-D; CDB, cell dissociation buffer; DFNF, DMEM/F12 with N2 supplements and bfgf; HBSS, Hank s Buffered Saline Solution; Mu, mouse; HuTu, human central nervous system tumor; Lib, Liberase-1; n/d, not determined. Although papain and TrypLE were usually highly effective in dissociating tissue, they gave more variable results. On two occasions with papain and three occasions with TrypLE, we saw free-dna aggregation even in the presence of DNase I, indicating substantial cell lysis. We could not definitively determine the cause of this variability. Two additional papain and one TrypLE sample could not be properly assayed by hemocytometer because of DNA aggregation and so were discarded. Accutase and Liberase-1 never generated free DNA aggregates. Flow cytometric viability agreed with trypan blue exclusion counts shortly after dissociation (Table 1). Viability of cells isolated with Liberase-1 stayed consistent after sequential primary and secondary antibody labeling, washing and up to 8 hours on ice postdissociation, which are typical durations when cell sorting (supplemental online Table 1). Because CNS tissue dissociation is typically followed by in vitro culturing, we plated cells in monolayer culture to determine how plating efficiency was affected by dissociation method (supplemental online Fig. 1C 1F). Nearly all cells adhered within 1 hour of plating, although papain and TrypLEtreated cells remained phase bright longer, suggesting weaker adhesion. Cells were counted 16 hours after plating to determine viability (supplemental online Fig. 1E) and the number of clusters of three or more cells (unlikely to result from cell division; supplemental online Fig. 1F). Treatment for 30 minutes with Liberase-1 or minutes with papain yielded the highest cell viability after plating. Although the majority (64%) of surviving HBSS-dissociated cells remained in clusters, all four enzyme treatments strongly reduced the number of clusters. Of these, Accutase left the most cell aggregates after the digestion and trituration process, even after 60 minutes of treatment. Although both Liberase-1 and Accutase are cocktails of collagenase and neutral protease, it is possible that differences in protease type or proprietary formulation could account for the different dissociation efficiencies. On the basis of these results, we conclude that all four enzymatic preparations yield efficient single-cell dissociation and viability compared with HBSS alone, with a rank order of papain Liberase-1 TrypLE Accutase HBSS. Titration of Antibodies to Mouse CNS Cell Surface Antigens For assaying surface marker expression, we first used 30 minutes of Liberase-1 treatment to acutely isolate cells from E13.5 cortex and postnatal day 2 (P2) mouse lateral ventricle wall, which is enriched in multipotent stem cells and neuronal progenitors [43]. For each marker, we titrated the concentration of the antibody to optimally distinguish positive from negative cell

5 1564 CNS Cell Dissociation and Flow Cytometry Figure 2. Titration of antibodies for CD133, CD15, CD24, A2B5, and PSA-NCAM against mouse central nervous system cells. E13.5 mouse cortex (A C) or postnatal day 2 mouse lateral ventricle tissue (D, E) was dissociated with 200 g/ml Liberase-1 in Hanks buffered saline solution. Cells numbering in 100 l of flow cytometry buffer were incubated with serial dilutions of antibodies against CD24 (A), CD15 (B), CD133 (C), PSA-NCAM (D), or A2B5 (E), followed by secondary antibodies where appropriate, then analyzed by flow cytometry. Each column shows isotype control or secondary-only antibody (nonspecific staining) at top followed by increasing concentrations of each antibody. Exact antibody concentration is listed when known; fold dilution is listed for PSA-NCAM and A2B5. Frequency histogram for each antibody concentration is shown by black filled distribution; nonspecific staining is overlaid in a gray line for comparison. Concentrations used for subsequent experiments (per 100 l of suspension): 0.01 g of CD24; 0.6 g of CD15; 1.0 g of CD133; 1/300,000 PSA-NCAM; 1/30,000 A2B5. Abbreviation: K, 1,000. populations. We identified distinct populations of mouse cells on the basis of the expression of CD133, CD15, CD24, A2B5, and PSA-NCAM (Fig. 2). The importance of high viability to proper titration was underscored by the observation that dead cells preferentially bound anti-cd133, thereby saturating out the specific binding capacity of the antibody (Fig. 3A). Tissue dissociation with Liberase-1 solved this problem by increasing cell viability (Fig. 3B). Although all four enzymes preserved equivalent antibody recognition of CD133 and CD15 (not shown), papain completely eliminated staining for CD24, whereas Liberase-1, TrypLE, and Accutase preserved similar CD24 staining (Fig. 3C 3F). Additionally, staining for BMP receptors IA and IB was lost after Liberase-1 or papain digestion (Fig. 3G 3J); staining was not tested after TrypLE or Accutase treatment. Loss of BMP receptors was verified by the reduction of SMAD activation (measured by phospo-smad1, -5, and -8 recognition) in cultured precursor cells after treatment with BMP2, which is known to efficiently induce neural crest-like fates in these cells [3, 44, 45]. Comparison of Human CNS Tumor Dissociation Methods We also tested the ability of these enzymes to dissociate human neural tissue. A solid high-grade glioma was divided with a scalpel and dissociated with Liberase-1, papain, TrypLE, or Accutase (Fig. 4A 4O). Viability was high for all four treatment groups (Table 1). Analysis of CD24 expression (Fig. 4F 4J) showed that TrypLE and Accutase yielded the highest number of CD24 cells (approximately 90%), whereas Liberase-1 yielded less (44%) and papain yielded greatly reduced CD24 numbers (14%). All four methods yielded CD133 cells (Fig. 4K 4O), although the highest percentage came after digestion with Liberase-1 (62%) and the lowest came from papain (37%) and Accutase (35%). Liberase-1 effectively dissociated a wide variety of resected tumors (supplemental online Table 2) with minimal undissociated tissue, except when tumors contained necrotic tissue. In these cases, the necrotic portions failed to digest and thus did not affect the viability counts. Additionally, we were able to more effectively passage cultured tumor cells with Liberase-1 compared with Cell Dissociation Buffer, an HBSS-based buffer containing a calcium chelator (Table 1). Because CD133 expression is predictive of tumor reinitiating activity in some CNS tumors [6, 46, 47], we used Liberase-1 to dissociate cells from a freshly resected pediatric atypical teratoid/rhabdoid tumor, and then immediately incubated cells with anti-cd133 conjugated to magnetic beads. A second PEconjugated anti-cd133 antibody was also used to label the same cells for postsort purity analysis by flow cytometry. CNS tumor cells could be easily distinguished from red blood cells on the basis of forward scatter (Fig. 4P, consistent with hemocytometer analysis) and showed an enrichment of CD133 cells from the total population after magnetic sort-

6 Panchision, Chen, Pistollato et al Figure 3. Effect of dissociation method on antigen recognition. (A, B): Nonenzymatic dissociation interferes with discrimination of CD133 cells. Mouse fetal cortex was dissociated with HBSS alone (A) or with 200 g/ml Liberase-1 (B), followed by labeling with anti-cd133. Viability analysis with 7-AAD showed that CD133 antibody preferentially bound dead (7-AAD ) cells, thus limiting availability for binding live cells. Increased cell viability after Liberase-1 dissociation of tissue allowed a distinct CD133 population to be identified. Numbers indicate percentage of CD133-stained cells that are living and dead. (C F): CD24 antigenicity is lost after papain treatment but preserved equally after Liberase-1, TrypLE, or Accutase treatment. CD133 and CD15 antigenicity is preserved equally after all four methods (not shown). Numbers show percentage of CD24 hi cells. (G J): Liberase-1 and papain treatment reduce or eliminate BMP responsiveness of central nervous system precursors. Expanded cultures from E13.5 cortex were passaged with HBSS and then incubated for 15 minutes in suspension with HBSS alone (G J), 200 g/ml Liberase-1 (G I), or 12 units/ml papain (J). Cells were then treated with 10 ng/ml BMP2 for 15 minutes before labeling with an antibody against phospho-smad-1/5/8. Cells passaged with HBSS alone (gray line distribution) show a shifted peak and shoulder indicating BMPR-IA and BMPR-IB staining and a separate peak indicating SMAD phosphorylation, whereas cells treated with Liberase-1 or papain (dark filled distribution) show lower BMP receptor immunoreactivity and/or BMP responsiveness as measured by reduced SMAD phosphorylation. Abbreviations: 7-AAD, 7-amino-actinomycin-D; HBSS, Hanks buffered saline solution. ing (Fig. 4Q 4T). This shows that Liberase-1 permits efficient and immediate separation of cell subtypes from human CNS tissue without the need for a recovery period for CD133 re-expression [6]. CD133 and CD15 Coexpress in a Similar Manner in Mouse Forebrain Previous studies have used high expression of CD133 or CD15 to enrich for multipotent mouse stem cells [17, 21, 23], but it is not known whether these markers identify the same cell type. Unlike human CNS cells [12, 13], mouse CNS cells have not been assayed for CD133/CD15 coexpression, or for coexpression with other selectable markers such as CD24, a marker of the neuronal lineage [19, 20]. To address this, we dissociated E13.5 LGE and P2 subventricular zone (SVZ), and then performed triple labeling with CD24, CD133, and either CD15, A2B5, or PSA-NCAM (Fig. 5; supplemental online Fig. 2). From E13.5 LGE, we found one distinct population (approximately two thirds of total) that expressed high levels of CD24 and no/low levels of CD15 or CD133 (i.e., CD15 lo/ CD133 lo/ CD24 hi ). A second distinct population centered on the high expression of CD15 and/or CD133 with low expression of CD24 (i.e., CD15 hi CD133 hi CD24 lo ). Within this second population, some cells expressed lower levels of CD15/CD133 but were not negative. Comparison of CD15 and CD133 expression indicated that most cells expressed similar levels of both antigens (Fig. 5H; supplemental online Fig. 3A 3E), suggesting that they identify similar cell types. However, there was a small but distinct CD15 hi CD133 lo CD24 hi population in both E13.5 LGE (Fig. 5F, 5I) and P2 SVZ (supplemental online Fig. 2F, 2I) that appeared repeatedly. We also triple labeled cells with CD133, CD24, and either A2B5, a glial progenitor marker [7, 10, 27, 30, 31], or PSA-NCAM, a neuronal progenitor and neuron marker [7, 25 29]. Triple labeling with CD133, CD24, and A2B5 showed that A2B5 cells (approximately 20% in both tissues) were predominantly CD133 hi (Fig. 5J 5L; supplemental online Fig. 3F 3J). Nearly all PSA-NCAM cells were CD133 lo and CD24 hi (Fig. 5M 5O; supplemental online Fig. 3K 3O). Analysis of CD24 against increasing concentrations of CD133 and PSA-NCAM (not shown) showed that this distribution remained the same. We then compared coexpression of CD133 and CD24 from other mouse E13.5 forebrain regions, including the cortex, MGE, and CGE (Fig. 5P 5R). The LGE, MGE, and CGE are known to generate distinct interneuron derivatives and varying contributions of oligodendrocytes [48 50]. We found that the general expression profile of CD133/CD24 expression was similar in each region, suggesting that it is conserved throughout the developing brain. However, the percentages of cells within each quadrant were different. We found that cortex, which is approaching its peak neuronal birth rate at this age [51, 52], contained the largest proportion of CD133 hi CD24 lo and CD133 hi CD24 hi cells (Fig. 5P). In contrast, the ganglionic eminences, which generate neurons on an accelerated schedule relative to the cortex [53, 54], had comparatively few of these cells but much larger numbers of CD133 CD24 hi cells (Fig. 5F,

7 1566 CNS Cell Dissociation and Flow Cytometry Figure 4. Dissociation, CD24 and CD133 labeling, analysis, and sorting of human tumor cells. (A O): Solid high-grade glioma recovered from nonobese diabetic/severe combined immunodeficiency mouse graft was divided four ways and dissociated with different enzymes before antibody labeling. (A): Light scatter profile after Liberase-1 digestion shows that whole cells (main) makes up about 60% of total events. Background fluorescence (B) and viability staining of main population cells with Annexin-V and 7-AAD (C) after Liberase-1; see Table 1 for other enzymes. (D O): Profile of viable cells based on main, Annexin-V gating. (D, E): Background fluorescence in detection channels for CD24 and CD133; same cells with different overlaid gates. (F J): CD24 antigenicity was retained at highest levels after TrypLE or Accutase digestion, and was reduced by half after Liberase-1 digestion and nearly eliminated after papain digestion; (F) shows isotype control. (K O): CD133 antigenicity was retained with all four dissociation methods but was highest after Liberase-1 digestion; (K) shows isotype control. (P T): Surgically resected pediatric atypical teratoid/rhabdoid tumor was dissociated with 200 g/ml Liberase-1 for 60 minutes and immediately incubated with anti-cd133 (AC133/1-microbeads plus AC133/2-PE) for magnetic selection. Light scatter profile (P) shows that whole central nervous system (CNS) cells (main, partially off scale) make up 4.1% of total events; concentrated low side scatter events are putative erythrocytes (93% of viable cells by trypan blue exclusion). (Q T): Profile of viable CNS cells based on main, Annexin-V gating. (Q): Nonspecific staining. (R): Before selection, CD133 cells make up 7.1% (polygon) of all intact CNS tumor cells. (S): After magnetic selection, unbound (flow-through) cells contain 4.7% CD133 cells. (T): Bound and eluted fraction is 97.5% CD133 cells. No recovery step to permit CD133 re-expression is necessary. Differences in physical profile in (A) and (P) are due in part to forward scatter voltage settings and side scatter scale (log in [A], linear in [P[). Abbreviation: 7-AAD, 7-amino-actinomycin-D. 5Q, 5R). Interestingly, the LGE, MGE, and cortex all contained approximately 20% CD133 lo CD24 lo cells. CD133/CD24 and CD15/CD24 Coexpression Similarly Distinguish Fetal Mouse Forebrain Stem Cells from Neuronal Progenitors and Neurons In E13.5 LGE, the most highly CD133-/CD15-expressing cells often expressed high levels of CD24 (Fig. 5F, 5I). This was even more pronounced in E13.5 cortex (Figs. 5P, 6A), which has a high rate of new neuronal birth at this stage [51]. This expression pattern is surprising because highest expression of CD15 alone [21, 23], or lowest expression of CD24 [9], has been used to enrich for stem cells. In contrast, enrichment of postnatal cerebellum stem cells with CD133 showed many CD133 cells expressing other lineage markers [17]. To investigate differences in potency among these cells, we freshly dissociated E13.5 mouse cortex with Liberase-1, prospectively isolated cells based on CD133 (or CD15) versus CD24 expression (Fig. 6B 6D; supplemental online Fig. 4), and performed in vitro clonal analysis. Cell division in the presence of bfgf could easily be visualized within the first 16 hours because of the mirroring behavior of daughter cells (Fig. 6E), which allowed extrapolation of how many initially plated cells survived and divided (Fig. 6F 6G). Notably, CD133 /lo CD24 hi cells did not divide; these cells had extensive fine filopodia that distinguished them from the short process-bearing dividing cells. Cells sorted on the basis of CD15/CD24 expression showed similar plating and clonal efficiencies (supplemental online Fig. 4G, 4H) to those seen after CD133/CD24 sorting, further supporting the contention that CD133 and CD15 expression predicts similar mouse cell populations. After differentiation by bfgf withdrawal, we performed quadruple staining of cells for MAP2a b (neurons), GFAP (astrocytes), O4 (oligodendrocytes), and DAPI (total nuclei) and classified clones into seven types (Fig. 6H 6N). The CD133 /lo CD24 hi fraction that did not divide yielded only single complex neurons (referred to as type 1). The CD133 hi CD24 hi fraction generated predominantly neuron-only clones of 4 16 cells with long processes (type 2) or cells with short bipolar morphologies (type 3). CD133 lo CD24 lo and CD133 hi CD24 lo cells generated mostly large three-fate clones with faint glial staining (type 4) or strong glial staining (type 5), indicating that they are multipotent. Additionally, we found three-fate clones

8 Panchision, Chen, Pistollato et al Figure 5. Coexpression of CD24, CD133, CD15, PSA-NCAMs, and A2B5 on fetal mouse forebrain cells. (A O): E13.5 mouse lateral ganglionic eminence was freshly dissociated with 0.2 mg/ml of Liberase-1. Light scatter profile (A) shows that whole cells (main) make up 90% of total events. (B): 7-amino-actinomycin-D (7-AAD) staining (shown without main gate) shows that live cells are 98% of main population, 88% of total events. (C O): Profile of live cells based on main, 7-AAD gating. (C): Example of isotype control (for CD24) on twodimensional plot. (D O): Frequency histogram for CD133 (D) and CD24 (E), and dot plot of coexpression (F); CD15 histogram (G) and CD15 coexpression with CD133 (H) and CD24 (I); A2B5 histogram (J) and A2B5 coexpression with CD133 (K) and CD24 (L); PSA-NCAM histogram (M) and PSA-NCAM coexpression with CD133 (N) and CD24 (O). Green overlay indicates controls for nonspecific staining. CD133 and CD15 are coexpressed similarly among cortical cells, with highest expression in cells also expressing high levels of CD24 and A2B5. PSA-NCAM is expressed only in cells expressing high levels of CD24 and is rarely coexpressed with CD133. (P R): Separate analysis of CD133 and CD24 staining in E13.5 mouse cortex (P) medial ganglionic eminence (Q) and caudal ganglionic eminence (R); profile of live cells based on main 7-AAD gating. Profile is similar when gating is performed on 7-AAD or Annexin- V, cells. that contained nearly all astro-oligo fates (type 6) or neuronoligo fates (type 7). Although both CD24 lo quadrants generated three-fate clones, CD133 lo CD24 lo and CD15 lo CD24 lo cells generated a higher proportion of type 4 clones, whereas CD133 hi CD24 lo cells generated higher numbers of type 5 and 7 clones. Similar results were found after CD15/CD24 selection, although type 7 clones also were generated from the CD15 hi CD24 hi population (Fig. 6O). On the basis of these results, we conclude that CD133/CD24 and CD15/CD24 expression can be used in a similar manner to predict mouse CNS precursor potency. Interestingly, the cell fraction enriched in clone-forming neuronal progenitors expresses higher levels of CD133 and CD15 than does the cell fraction enriched in multipotent stem cells. CD24 expression distinguished mouse fetal stem cells from postmitotic neurons and their progenitors; however, treatments that degrade CD24 expression (such as papain; Fig. 3) prevent such functional characterization. DISCUSSION Methods of neural tissue dissociation vary widely and lead to differences in dissociation efficiency and viability as measured by flow cytometry [32, 35]. We find that papain, or purified preparations of trypsin replacement (TrypLE) or collagenase/ dispase/neutral protease (Liberase-1, Accutase), each yield efficient dissociation of mouse CNS tissue and human CNS tumors into single cells that remain viable for long periods of time before culture. Liberase-1 or TrypLE work best on mouse and

9 1568 CNS Cell Dissociation and Flow Cytometry Figure 6. Prospective isolation and clonal analysis of fetal mouse cortical cells. (A): Colors denote sort gates overlaid on CD133/CD24 expression profile from E13.5 mouse cortex after Annexin-V (nonviable) cells were gated out; (B D): example of purity analysis showing cell enrichment after gating (rectangles) for CD133 lo CD24 lo (B), CD133 hi CD24 lo (C), and CD133 hi CD24 hi (D); each plot shows mean relative fluorescent intensity of CD24 (x-mean) and CD133 (y-mean) for enriched population, along with percentage of cells within gate area before (top) and after (bottom) sorting. (E): Example of two-cell clone, 16 hours postclonal density plating. (F, G): Cell survival (F) and percentage of cells dividing (G) at 16 hours postplating. (H M): examples of clone types generated after sorting, 3 days of expansion, 6 days of differentiation, and staining for MAP2a b (neurons), GFAP (astrocytes), O4 (oligodendrocytes) and DAPI (all nuclei). Type 1 is a single neuron and is not shown. (N O): Quantitation of each clone type generated from CD133-/CD24-sorted (N) or CD15-/ CD24-sorted (O) cells. Mean SEM, n 3 for (F, G, N, O). Abbreviations: AO, astrocyte-oligodendrocyte clone; DAPI, 4-6-diamidino-2-phenylindole; GFAP, glial fibrillary acidic protein; N, neuron-only clone; NAO, neuron-astrocyte-oligodendrocyte clone; NO, neuron-oligodendrocyte clone. Scale bars 50 (D) and 100 m (C F). human tissue based on the balance of dissociation efficiency, viability, and preservation of tested antigens. Papain most aggressively degrades CD24 antigenicity, whereas Accutase preserves antigenicity but is least efficient in dissociation; however, we could not rule out that even longer Accutase incubation times might yield more complete digestion. Using Liberase-1 dissociation, we identify novel patterns of selectable marker expression in mouse fetal cortex, LGE, MGE, CGE, and postnatal SVZ. We also show that CD133 and CD15 expression closely parallel each other in mouse brain cells and are roughly interchangeable for prospective isolation, unlike that reported in human brain [41]. Interestingly, a neuronal progenitor population expresses higher levels of CD15 and CD133 than multipotent stem cells of the fetal mouse forebrain. Coexpression with CD24 enhances the discrimination of these two populations, indicating that Liberase-1, TrypLE, and Accutase have an advantage over papain for this type of analysis in mouse. In human CNS tissue studies, cruder preparations of trypsin can cleave important cell surface antigens such as CD133, requiring a recovery period for antigen re-expression [6]. We find that all four tested enzymatic methods efficiently dissociate human tumor tissue while preserving CD133, although Liberase-1 preserves antigenicity the best. Consistent with studies in various tissues [35, 55 58], there were differences in the aggressiveness of each enzyme preparation. First, papain and TrypLE sometimes generated free DNA during dissociation, suggesting cell lysis. Second, mouse and human CD24 staining was greatly reduced after papain treatment, whereas BMPR-IA and BMPR-IB were both rendered antigenically and functionally inactive after either Liberase-1 or papain digestion (TrypLE and Accutase were not tested). However, this last effect may be a useful property in mouse and human embryonic stem (ES) cell applications, where levels of BMP signaling are critical determinants of pluripotency [59, 60]. Suspended aggregate (neurosphere) cultures are often difficult to dissociate into single-cell suspensions without killing a substantial proportion of the cells. We found that viability and efficiency of passaging tumor cultures, in monolayer or in suspended aggregates, worked well with Liberase-1 (Table 1). TrypLE also worked well for this purpose, whereas papain yielded poor viability (not shown). Mouse cell passaging from fibronectin-coated dishes was not facilitated by Liberase-1 when compared with nonenzymatic passaging (not shown). We did not compare the effectiveness of these enzymes in passaging mouse neurospheres. The high viability of cell dissociation also allowed us to perform multiantigen analysis with high confidence that results are representative of the cell types in the tissue. Our results show that CD133 and CD15 can be used almost interchangeably along

10 Panchision, Chen, Pistollato et al with CD24 to distinguish mouse multipotent stem cells from neuronal progenitors and postmitotic neurons. Interestingly, we find that the cells expressing the highest levels of CD133 and CD15 are more likely to be neuronal progenitors than multipotent stem cells. In contrast, we find that P2 SVZ has fewer CD133 hi CD15 hi CD24 hi cells, which may be why selection on the basis of CD15 alone works for enriching multipotent stem cells from postnatal SVZ [21]. We did not test this by comparative prospective isolation and clonal analysis using these markers to determine the profile of neuronal progenitors from P2 SVZ cells. Although neuronal progenitors (type A cells) in the postnatal SVZ express CD24 [18], they may express lower levels of CD133 or CD15 than do neuronal progenitors in the E13.5 cortex. Alternatively, these may be a distinct subset of neuronal progenitors from PSA-NCAM neuronal progenitors [7, 25, 28], which we show are nearly all CD24 hi and express little or no CD133/CD15. In our hands, these cells showed almost no proliferation in either bfgf (Fig. 6) or epidermal growth factor (EGF) (not shown); however, it is quite possible that our stringent medium does not support the survival of these cells. Additionally, a CD133 lo CD24 hi neuronal progenitor may be better responsive to other mitogens such as Sonic Hedgehog [17]. Our results show that multipotent fetal cells are enriched in the CD24 lo fraction, similar to results in adult SVZ [9], but that there is heterogeneity in the resulting clone types. Three-fate clones enriched in oligodendrocytes (types 6 7) were most frequently generated from CD15 hi CD24 lo and CD133 hi CD24 lo cells. Because we see that many cells in this region also express the glial-committed progenitor marker A2B5 (Fig. 5K) [7, 10, 27, 61], this suggests that CD15 hi A2B5 hi selection will enrich for type 6 or 7 clone-forming cells. We found another interesting heterogeneity within the CD24 lo population: Cells expressing lower levels of CD15 and CD133 preferentially generated threefate clones with simple bipolar morphologies and weak staining for neuronal and glial markers. In contrast, cells expressing higher levels of CD15 and CD133 more frequently generated three-fate clones with extensive processes and stronger cell-type marker staining. Although we have yet to determine lineage relationships between these precursors, one intriguing possibility is that CD133 lo CD15 lo CD24 lo cells are a more primitive multipotent cell population than CD133 hi CD15 hi CD24 lo cells. Consistent with this idea, the majority of neurospheres from postnatal SVZ comes from CD24 lo/ transit-amplifying type C cells rather than type B stem cells; expansion in EGF promotes multipotency in these type C cells [18]. Likewise, fetal cortical CD133 hi CD15 hi CD24 lo cells may retain potency that they do not exhibit in vivo, whereas the CD133 lo CD15 lo CD24 lo population may be stem cells both in vitro and in vivo. It is also possible that bfgf and EGF may differentially expand or instruct [62 65] these cell subgroups. Our future studies will test these possibilities. In summary, we show that the flow cytometric analysis of the important neural antigens CD133, CD15, and CD24 are affected by the manner of dissociation. Although nonenzymatic dissociation does not digest antigens, this method is not useful for flow cytometric analysis since poor viability prevents proper titration of antibodies like CD133. We show that CD24 antigenicity, which is retained during Liberase-1, TrypLE, or Accutase treatment but lost during papain treatment, provides a useful discrimination between mouse fetal multipotent stem cells, neuronal-committed progenitors and neurons when paired with CD133 or CD15. These results reinforce the importance of comparing multiple methods of dissociation when characterizing novel selectable markers in future neural lineage studies. ACKNOWLEDGMENTS This study was supported by funds from the NIH Mental Retardation and Developmental Disabilities Research Center Grant P30HD40677, the Frank and Nancy Parsons Fund, the Georgia Derrico and Rod Porter Fund for the Children s National Medical Center (CNMC)/University of Padova Sister Program, the CNMC Research Advisory Council, and two CNMC Board of Visitor Grants. We thank William King and Bhargavi Rajan for flow cytometry support; Dr. Brian Rood and Hui-Zhen Zhang for assistance in acquiring human tumors; and Dr. Vittorio Gallo and Dr. Giuseppe Basso for advice and support. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST The authors indicate no potential conflicts of interest. REFERENCES 1 Panchision DM, McKay RD. The control of neural stem cells by morphogenic signals. Curr Opin Genet Dev 2002;12: Johe KK, Hazel TG, Muller T, Dugich-Djordjevic MM, McKay RD. Single factors direct the differentiation of stem cells from the fetal and adult central nervous system. Genes Dev 1996;10: Rajan P, Panchision DM, Newell LF, McKay RD. BMPs signal alternately through a SMAD or FRAP-STAT pathway to regulate fate choice in CNS stem cells. J Cell Biol 2003;161: Dupin E, Ziller C, Le Douarin NM. The avian embryo as a model in developmental studies: Chimeras and in vitro clonal analysis. Curr Top Dev Biol 1998;36: Lindvall O, Kokaia Z, Martinez-Serrano A. Stem cell therapy for human neurodegenerative disorders-how to make it work. Nat Med 2004; 10(suppl):S Singh SK, Hawkins C, Clarke ID et al. Identification of human brain tumour initiating cells. Nature 2004;432: Mujtaba T, Piper DR, Kalyani A et al. Lineage-restricted neural precursors can be isolated from both the mouse neural tube and cultured ES cells. Dev Biol 1999;214: Morrison SJ, White PM, Zock C et al. Prospective identification, isolation by flow cytometry, and in vivo self-renewal of multipotent mammalian neural crest stem cells. Cell 1999;96: Rietze RL, Valcanis H, Brooker GF et al. Purification of a pluripotent neural stem cell from the adult mouse brain. Nature 2001;412: Maric D, Maric I, Chang YH et al. Prospective cell sorting of embryonic rat neural stem cells and neuronal and glial progenitors reveals selective effects of basic fibroblast growth factor and epidermal growth factor on self-renewal and differentiation. J Neurosci 2003;23: Belachew S, Chittajallu R, Aguirre AA et al. Postnatal NG2 proteoglycan-expressing progenitor cells are intrinsically multipotent and generate functional neurons. J Cell Biol 2003;161: Uchida N, Buck DW, He D et al. Direct isolation of human central nervous system stem cells. Proc Natl Acad Sci USA 2000;97: Barraud P, Stott S, Mollgard K et al. In vitro characterization of a human neural progenitor cell coexpressing SSEA4 and CD133. J Neurosci Res 2007;85: Murayama A, Matsuzaki Y, Kawaguchi A et al. Flow cytometric analysis of neural stem cells in the developing and adult mouse brain. J Neurosci Res 2002;69: Fargeas CA, Florek M, Huttner WB et al. Characterization of prominin-2, a new member of the prominin family of pentaspan membrane glycoproteins. J Biol Chem 2003;278: Weigmann A, Corbeil D, Hellwig A et al. Prominin, a novel microvillispecific polytopic membrane protein of the apical surface of epithelial cells, is targeted to plasmalemmal protrusions of non-epithelial cells. Proc Natl Acad Sci U S A 1997;94: Lee A, Kessler JD, Read TA et al. Isolation of neural stem cells from the postnatal cerebellum. Nat Neurosci 2005;8: Doetsch F, Petreanu L, Caille I et al. EGF converts transit-amplifying