c-Kit+ Cells in Adult Salivary Glands do not Function as Tissue Stem Cells

A rare population of salivary gland cells isolated based on c-Kit immunoreactivity are thought to represent tissue stem cells since they exhibit the most robust proliferative and differentiation capacity ex vivo. Despite their high promise for cell-based therapies aimed at restoring salivary function, the precise location and in vivo function of c-Kit+ stem cells remain unclear. Here, by combining immunostaining with c-KitCreERT2-based genetic labeling and lineage tracing in the adult mouse salivary glands, we show that c-Kit is expressed in a relatively large and heterogeneous cell population that consists mostly of differentiated cells. Moreover, c-Kit does not mark ductal stem cells that are known to express cytokeratin K14. Tracking the fate of in vivo-labeled c-Kit+ or that of K14+ cells in spheroid cultures reveals a limited proliferative potential for c-Kit+ cells and identifies K14+ cells as the major source of salispheres in these cultures. Long-term in vivo lineage tracing studies indicate that although c-Kit marks at least two discrete ductal cell lineages, c-Kit+ cells do not contribute to the normal maintenance of any other cell lineages. Our results indicate that c-Kit is not a reliable marker for salivary gland stem cells, which has important implications for salivary gland regenerative therapies.

differentiation 18 . At birth, however, c-Kit expression is shifted from proacinar cells to ductal cells, where it continues to be expressed to adulthood 19 . In histological sections of adult mouse SMGs, c-Kit immunopositive cells are detected in various ductal compartments including the IDs, SDs, and EDs at much higher frequencies than those reported by flow cytometry analyses 8,9,20,21 . Although the broad tissue distribution of c-Kit + cells has hindered efforts to map the precise location of c-Kit + stem cells, several studies have suggested that these stem cells reside in the EDs 5,9,21 . Transcriptional profiling of the Lin − CD24 + c-Kit + Sca1 + cell fraction has revealed high levels of basal cell cytokeratins K14 and K5 in this population, implying that c-Kit + stem cells express K14/K5 9 . Furthermore, co-immunostaining of tissue sections for c-Kit and K5 has shown co-localization of these two markers in a small subset of ductal cells in the ID and the ED where c-Kit + and K14/K5 + cells are juxtaposed 21,22 . K14 has been identified as a marker of the lineage-restricted ductal stem cells that reside in the basal layer of the ED and at the ID/GD junction 20,22 . Long-term lineage tracing of K14 + stem cells, however, did not reveal any lineal relationship between c-Kit + and K14 + cells in the ID. Although similar studies have not been performed in the ED, it is clear that K14 + cells in the ED do not function as multi-lineage stem cells in the adult gland 20,22 . Recent long-term lineage tracing of several cell populations including ductal and acinar cells have indicated that homeostasis of the adult salivary gland does not rely on multipotent stem cells [22][23][24] . Despite the high promise of c-Kit as a marker for SG stem cells and its potential application for cell-based therapies, the in vivo location and function of c-Kit + stem cells remain unclear.
Here, we first verified the frequency and distribution pattern of c-Kit + cells in all major SGs of adult mice through genetic labeling and immunostaining. We then used an inducible genetic lineage-tracing approach to investigate the fate of in vivo-labeled c-Kit + cells in spheroid cultures and during homeostatic renewal of the adult SMG. Our analysis did not identify c-Kit as a marker for salivary gland stem cells.

c-Kit is broadly expressed in the major salivary glands of adult mice.
To clarify the discrepancy between c-Kit immunoreactivity of tissue and single cell suspensions of adult SMGs, we used a genetic approach to label c-Kit-expressing cells with a fluorescent protein and then compared the frequency and distribution pattern of the genetically-labeled cells with that of c-Kit-immunoreactive cells in the SMG. A well-characterized c-KitCre ERT2/+ mouse line with a tamoxifen (TAM)-inducible form of Cre recombinase driven by an endogenous c-Kit locus was crossed with R26R-tdTomato (TdT) reporter strain carrying a floxed stop codon between the ubiquitously expressed Rosa26 promoter and a gene encoding TdT, a variant of red fluorescent protein 25,26 . TAM was administered to bi-transgenic mice (8 wks of age, 3 females and 2 males) for four consecutive days to efficiently label c-Kit-expressing cells 25 and three days later, SGs were removed for immunofluorescent and flow cytometry analysis (Fig. 1B). Analysis of TdT-labeled cells in tissue sections or single cell suspensions of SMGs showed that TAM administration induced TdT expression in about 20% of total population of SMG cells ( Fig. 1C-E). The TDT-expressing cells were predominantly mapped to the salivary ducts (Fig. S1). To verify the cell specificity of c-Kit-driven Cre recombination of the R26R-TdT locus, histological sections and single cell suspensions of TdT-labeled SMGs were immunostained for c-Kit protein (Fig. 1F,G). Immunofluorescent microscopy revealed a strong concordance between TdT expression and c-Kit protein in tissue sections of SMGs, with a distribution pattern consistent with previous reports (Fig. 1F) 20,21 . TdT + c-Kit + cells were detected throughout the salivary ducts and were either organized into uniformly labeled cell clusters in the ID or distributed more sporadically in the larger salivary ducts (Fig. 1F). In contrast, flow cytometry analyses of single cell suspensions prepared from the contralateral gland using the same anti-c-Kit antibody (clone 2B8) showed no correlation between TdT labeling and c-Kit immunopositivity (Fig. 1G,H). The minimal immunoreactivity of c-Kit-expressing cells in single cell suspensions of SMGs was verified using a different anti-c-Kit antibody (clone Ack2), and it was consistent with previous flow cytometric analysis indicating a frequency for c-Kit immunopositive cells ranging from 0.058% 8 to 0.97% 13 in total population of SMG cells. Flow cytometry analysis of TdT-labeled SMG cells for expression of other parenchymal cell surface markers that are often used in combination with c-Kit to enrich for stem cells, including CD24, CD49f (pan-epithelial, Figs 1D and S2), and Sca1 (a marker expressed in salivary ducts, Fig. S2), showed that almost all TdT-expressing cells in SMGs are epithelial cells, and more than 50% express Sca1 (Fig. 1I,J). These data indicated that c-Kit is expressed in a relatively large and heterogeneous epithelial cell population in SMGs. Moreover, the minimal concordance between TdT expression and c-Kit-immunopositivity indicated that c-Kit epitopes are altered in single cell suspensions of SMGs.
Analysis of TdT-labeled cells in other major SGs of bi-transgenic mice revealed a striking difference in the frequency and distribution pattern of c-Kit-expressing parenchymal cells in parotid glands when compared to SMG and sublingual glands (Figs 2A,B and S3). The acinar to ductal ratio is comparable for the parotid and sublingual glands (9 acinar to 1 ductal) and much higher than that in the SMG (6 acinar to 4 ductal) 27 . Despite this, the number of TdT-labeled cells in parotid glands was significantly higher than the other major salivary glands ( Fig. 2A,B). Immunostaining for aquaporin 5 (Aqp5), a water channel expressed exclusively in acinar cells 28 , revealed widespread TdT expression in parotid acini (Fig. 2C,D). Immunostaining for c-Kit protein verified concurrent expression of TdT and c-Kit in acinar cells and appropriate localization of c-Kit on the surface of acinar and ID cells (Fig. 2E). Overall, the widespread expression of c-Kit in the differentiated compartments of SGs such as GDs, SDs, and parotid acini indicate that c-Kit is not a specific marker for SG stem cells.

c-Kit and K14 mark distinct cell populations.
Previous studies have suggested that c-Kit + stem cells co-express basal cytokeratins (K14/K5) and that they reside in EDs 9,21 . To more precisely map the position of c-Kit-expressing cells with respect to K14 + ductal cells, sections of TdT-labeled SMGs were immunostained for K14. In SGs, K14 marks ductal stem cells in EDs and, IDs, as well as myoepithelial cells 22,29 . Fluorescent microscopy analysis of the EDs and IDs showed no overlap between TdT and K14 expression (n = 480 K14 + ID cells and Scientific REPORTS | (2018) 8:14193 | DOI:10.1038/s41598-018-32557-1 n = 340 K14 + basal ED cells) (Fig. 3A,B arrowheads and Fig. S4). In the IDs, TdT was expressed in luminal cells juxtaposed to K14 + ductal stem cells (Fig. 3A), confirming our previous studies indicating that c-Kit and K14 mark two discrete cell lineages 22 . In the EDs, c-Kit-expressing cells (TdT + ) were scattered among the K14 + basal cells, however, they displayed a distinct columnar, flask-like morphology (Figs 3B and S4, arrows). The morphology and distribution pattern of c-Kit + cells in the EDs were comparable to that of c-Kit + cells in highly differentiated GDs and SDs (Figs 3C and S1). In these compartments identified by K19 expression, c-Kit-expressing cells were scattered among the principal ductal cells and displayed a unique flask-like morphology with narrow and long basolateral projections similar to what has been described for "Tuft" cells (Figs 3C and S4, arrows) 30 . Tuft cells are highly specialized chemosensory cells found in various epithelial tissues that play a key role in regulating innate immunity 31 . Overall, these findings indicate that K14 and c-Kit mark distinct cell populations, and they underscore the heterogeneity of cKit-expressing cells in salivary ducts. Contribution of c-Kit + and K14 + cells to salisphere formation in cultures of SMGs. The expression of c-Kit by differentiated cells in salivary ducts and parotid acini does not exclude the existence of a small subset of c-Kit + cells that may function as SG stem cells. Since the robust proliferative and differentiation capacity of a subset of SG cells in spheroid cultures has been attributed to c-Kit + cells 9,17,32 , we traced the fate of in vivo-labeled TdT-expressing cells in spheroid cultures established from total population of SMG cells. TAM was administered to cKitCre ERT2/+ : R26RTdT mice (8-10 wks of age) to label c-Kit + cell in vivo. Three days later, both SMGs were removed, one gland was processed for immunofluorescent analysis to quantify the efficiency of TdT-labeling in c-Kit-expressing cells, the other gland was dissociated into a single cell suspension to generate salispheres (Fig. 4A). Analysis of cultures from TAM-treated and non-treated littermates showed a comparable sphere-forming efficiency of about 2.5% ± 0.8% for spheres measuring ≥0.2 mm in diameter at 7 days post-seeding (Fig. S5A). Interestingly, despite a fairly high efficiency of in vivo labeling for c-Kit + cells at 72% ± 12% (Fig. 4B,C), over 95% of salispheres formed in culture did not express TdT (n = 1320 from 3 mice; Fig. 4B,C). In fact, although TdT-labeled cells accounted for about 20% of cells initially present in culture, the majority failed to grow to spheres larger than 0.1 mm in diameter (Fig. 4B,C). Moreover, upon passage, the proportion of TdT + spheres further declined to 0.3% ± 0.4% (n = 1560; Figs 4D and S5B). Failure of TdT-labeled cell in contributing to salispheres formed in cultures of SMGs suggested that a different cell population contributes to salispheres. Given that K14 marks an actively cycling stem cell population in adult SMGs 22 , we wondered if this population was the source of salispheres. A similar labeling strategy was applied to K14Cre TRE : R26RTdT mice to mark K14 + cells with TdT in vivo and assess their sphere-forming capacity in vitro (Fig. 1E-G). Analysis of spheroid cultures showed that the overall salisphere-forming efficiencies of SMG cells were comparable between the two mouse models (Fig. 4F). However, in spheroid cultures established from K14Cre TRE :R26RTdT mice, approximately half of the large and highly branched salispheres were uniformly labeled with TdT, suggesting they were descendants of a single K14 + TdT + cell (Fig. 4E). More importantly, the labeling efficiency of K14 + ductal stem cells within the tissue corresponded directly to the percentage of TdT + spheres in culture, indicating that K14 + stem cells were the major source of salisphere formation in culture (Fig. 4G). In addition, when these primary spheroid cultures were dissociated and re-plated, the proportion of TdT-labeled spheres remained unchanged, confirming the self-renewal ability of K14 + cells in culture (Fig. 4G). Overall, these data demonstrate that SG stem cells contribute to the formation of salispheres in culture, however, no subset of c-Kit + cell population is capable of extended growth in salisphere cultures.

c-Kit + cells do not contribute to the maintenance of other cell lineages during homeostasis. A
recent lineage tracing study has demonstrated that c-Kit + ID cells do not contribute to acini 24 , consistent with previous studies indicating that acini are maintained independent of ductal progenitors 23 . However, this study offered limited analysis of c-Kit + cells in other ductal compartments 24 . Therefore we used a similar lineage tracing approach to more quantitatively analyze the fate of all subsets of c-Kit + cells in adult SMGs. TAM was administered to 6-wk old c-KitCre ERT2/+ : R26RYFP mice for three consecutive days to pulse label c-Kit + cells with YFP, and then the number and position of RosaYFP-labeled cells were analyzed and compared over 6 months (Fig. 5A). Analysis of YFP-labeled cells at 4 days or following a 2-or 6-month chase period showed a similar pattern of distribution for YFP-labeled cells (Figs 5B and S6). YFP-labeled cells remained restricted to the ID cells and the  tuft-like cells in larger ducts during the entire chase period. Moreover, quantitative analysis of YFP-labeled cells showed no significant change in their number during the 6-month chase period (Figs 5B,C and S6). As expected, the descendants of Rosa-YFP cells in the ID cells did not expand into acini over the 6-month chase period (Fig. 5B arrowheads, n = 260 ID/acini junctions), confirming previous studies 23,24 . More importantly, however, even after a 6-month chase period, the YFP-labeled cells with flask-like/columnar morphology were detected in the EDs, SDs and GDs (Fig. 5D, arrows). Co-immunostaining of SMGs for YFP and K14 verified the persistence of YFP-labeled cells juxtaposed to the K14 + progenitors in the ED, suggesting no lineal relationships between c-Kit + and K14 + cells (Fig. 5D). The stable frequency of YFP + cells in salivary ducts indicated that c-Kit + cells were not replaced at the population level by unlabeled progenitor or stem cells. Overall, our lineage tracing experiments revealed the presence of at least two distinct cell lineages marked by c-Kit, one in the ID and the other in the larger ducts that are maintained independent of K14 + ductal stem cells. Moreover, a lack of contribution of c-Kit + cells to any other cell type during homeostasis is not consistent with the behavior of stem cells 33 .

Discussion
Tissue stem cells are defined as self-renewing cells that generate at least one type of differentiated descendant during tissue homeostasis [33][34][35] . The identification of c-Kit as a marker for SG stem cells was based on ex vivo characterization of a c-Kit-immunoreactive cell population isolated from the adult mouse SMGs 8,9,13,17 . Our genetic labeling and lineage tracing studies were aimed at providing direct evidence for the existence of c-Kit + stem cells in the adult gland. However, the broad expression of c-Kit in SGs, the lack of contribution of c-Kit + cells to any other differentiated cell lineages during tissue homeostasis, and their limited proliferative potential in spheroid cultures were all inconsistent with the existence of c-Kit + stem cells in adult SGs.
The minimal contribution of genetically-labeled cKit-expressing cells to salisphere formation revealed by our in vitro fate analysis did not validate the widely-accepted notion that c-Kit + stem cells are the major producers of salispheres in cultures of adult SMGs 9,10,17,32 . Previous studies have relied on c-Kit immunoreactivity to enrich for a rare cell population in single cell suspensions of adult SMGs that displayed a robust salisphere-forming capacity in culture 8,9,11,13 . However, the reported frequencies of c-Kit immunopositive cells in single cell suspensions were too low (ranging from 0.05% to 1%) and inconsistent with the broad c-Kit expression in ductal cells observed in histological sections of SMGs 8,9,13,20,21 . To overcome the potential problems stemming from unreliable c-Kit immunoreactivity, we used a transgenic mouse line in which endogenous activity of the c-Kit locus dictated expression of Cre and recombination of a Rosa locus resulted in the expression of a fluorescent protein, thus enabling purification, mapping, and lineage tracing of c-Kit cells without relying on c-Kit immunoreactivity 25,26 . The cell-type specificity of Cre-mediated recombination of R26R-TdT locus was confirmed by immunostaining for endogenous c-Kit protein in tissue sections, however, TdT expression was not in concordance with c-Kit immunoreactivity in single cell suspensions of SMGs. These data strongly suggest that the enzymatic tissue digestion of salivary glands can damage or alter externally exposed c-Kit epitopes as shown in other tissues 36 . Therefore, the results of studies using c-Kit immunoreactivity to isolate SG stem cells should be interpreted with caution as the isolated cell populations do not represent c-Kit + cells. It is worth noting that even when a combination of four surface markers (Lin − CD24 + cKit + Sca1 + ) was used to enrich for SG stem cells, the isolated stem cell population remained heterogeneous with a salisphere-forming efficiency of less than 3% 9 . Furthermore, transcriptome profiling of the isolated cell population indicated high levels of K5 and K14 transcripts, suggesting inclusion of K14 + cells in the sorted population 9 . In the absence of validation of c-Kit expression at single cell levels in the sorted cell population, it is not clear whether salisphere formation by this heterogeneous population was solely attributable to the c-Kit + cells or to contaminating K14 + ductal stem cells or other progenitor populations. Indeed, another study has shown that SMG cells enriched for high levels of pan epithelial cell surface markers CD24 and CD29 were highly efficient in salisphere formation, but further purification based on c-Kit immunoreactivity did not improve the proliferative capacity of this population 32 .
Our data clearly demonstrate that c-Kit and K14 mark two distinct cell populations in the adult SMGs. Previous studies have suggested that a population of K14 + c-Kit + multipotent stem cells resides in the basal layer of EDs 5,9,21 . However, the lack of co-localization of K14 protein (antibody) and TdT expression (c-Kit expressing cells), the distinct morphology of c-Kit + cells when compared to K14 + basal cells in the ED, the differential contribution of K14 + cells and c-Kit + cells to salispheres in culture (50% for K14 + cells vs. 2.5% for c-Kit + cells in primary spheres and 50% vs 0.3% in secondary cultures), and the long-term lineage tracing analysis of c-Kit cells in this study and that of K14 + cells in our previous studies showing a lack of lineal relationship between these cell populations 20,22 are all inconsistent with the existence of a c-Kit + K14 + cell population in the adult gland. Although a K14 + Kit + embryonic progenitor population has been identified during early development of SMGs, the lineage of K14 + and Kit + cells may diverge early during development. This is supported by a recent study indicating that c-Kit and K14 are inversely regulated in response to retinoic acid during development 37 . In the present study, we relied on the Cre-lox system in transgenic mice, the "gold standard" for addressing the problems of in vivo cellular origins 38 . The genetic labeling did not appear to have an adverse effect on the proliferative potential of c-Kit + cells in culture since neither tamoxifen administration nor expression of Cre and TdT altered the overall salisphere-forming efficiency of SMG cells. Furthermore, a similar approach using K14Cre TRE :TdT mice in which the efficiency of Cre recombination in ductal stem cells was about 50% (Fig. 4G) demonstrated uniform TdT expression in 50% of salispheres formed from total population of SMG cells indicating that K14 + ductal stem cells are the major contribute to salisphere formation in cultures. One caveat of our genetic labeling is that not every c-Kit + cell was labeled with TdT and therefore we cannot directly rule out the possibility of a bias against Cre activation in the presumptive multipotent stem cell population in EDs. However, the persistence of genetically labeled c-Kit + cells in EDs revealed by our long-term in vivo lineage tracing argues against this possibility. Given the relatively rapid turnover rate in EDs 20 , if there was a c-Kit + K14 + stem cell population in which c-KitCre locus was not active, then the YFP-labeled c-Kit + cells should have been gradually replaced by descendants of this unlabeled stem cell population, resulting in a gradual decline in the number of YFP-labeled cells during the 6-month chase period. Clearly, the persistence of YFP-labeled cells in EDs indicates that the lineage of c-Kit + cells is related neither to the K14 + stem cells, nor to the presumptive K14 + Kit + stem cells.
Our long-term lineage tracing of c-Kit + cells confirmed a recent lineage tracing analysis of c-Kit + ID cells demonstrating that c-Kit + ID cells do not contribute to acini in submandibular and sublingual glands 24 . Classical nuclear tracking studies have suggested that IDs harbor stem cells for both differentiated ductal and acinar cells 1,18,39 . We have recently identified the ductal stem cells in IDs that contribute to the maintenance of the K19 + differentiated cell lineage in GDs 22 . However, since c-Kit is uniformly expressed by all ID cells localized between the ductal stem cells and acini, the lack of expansion of lineage-traced cells into acini does not support these earlier studies, but rather is consistent with recent studies indicating that acini homeostasis relies entirely on the proliferative capacity of existing acinar cells 23 . Although neither populations of c-Kit + cells contribute to any other differentiated cell lineages in the adult SMGs, the persistence of genetically-labeled cells in both IDs and larger ducts reveals two distinct ductal cell lineages. The maintenance of these two lineages is independent of K14 + ductal stem cells and underscore the complexity of ductal cell lineages in adult SGs. In the larger ducts, c-Kit is expressed in luminal cells with a distinct barrel-shaped or flask-like morphology and distribution pattern similar to that described for highly specialized tuft cells 30 . Although c-Kit expression in tuft cells must be further investigated, c-Kit signaling may be involved in the regulation of various secretory, chemosensory, and immune regulatory functions attributed to tuft cells 30,31 . In addition to salivary ducts, c-Kit expression is most prominent in the acini of parotid glands. The basis for differential expression of c-Kit in acini between parotid and SMGs is not clear because in rodents, both glands are composed of mostly serous acini 3 . However, recent studies have indicated functional differences in ion transporter activity between the parotid and SMG acinar cells 27 . Clearly, the functional significance of c-Kit signaling in these highly specialized cell populations requires further investigation.
In summary, the results presented here indicate that c-Kit is not a reliable marker for SG stem cells and argues against the existence of a Kit + stem cell population in SGs. Furthermore, given the therapeutic potential of salisphere-derived cells in reversing irradiation-induced hyposalivation 9,10,17 , the identification of K14 + stem cells as an expandable source of SG cells in culture has important implications for cell-based therapies aimed at restoring SG function.

Methods
Animals. Transgenic c-Kit-Cre ERT2+/− mice in which a tamoxifen-inducible form of Cre recombinase is driven from an endogenous c-Kit locus have been described previously 25,26 . These mice were kept on C57Bl/6 background. Cre reporter strains R26R-tdTomato (stock# 007914) and R26R-YFP (Stock# 006148) were purchased from Jackson Laboratory (Bar Harbor, Maine). Mice were genotyped by polymerase chain reaction using mouse genomic DNA from tail biopsy specimens. Selective labeling in c-Kit + cells in adult mice was induced by intraperitoneal injection of tamoxifen (TAM, 1 mg/50 ul of peanut oil/mouse; Sigma Aldrich, St. Louis, MO) for 3 or 4 consecutive days as described by previous studies 26 . Transgenic K14-Cre TRE/+ mice with Cre expression regulated by doxycycline was used to induced selective labeling of K14 + cells as described previously 22 .
Adult male and female mice aged between 8-10 weeks were used in all experiments with the exception of long-term lineage tracing studies in which Tam was administered at 6 weeks of age. The care and experimental use of animals were approved by the Institutional Animal Care and Use Committee (IACUC protocol#287271) of the Stony Brook University, Stony Brook, New York, USA. Animals were maintained in the Stony Brook University-Health Science Center Division of Laboratory Animal Resources, an AAALAC-accredited center, and housed in groups of up to five mice per cage under maximum isolation with fresh food and water and regular cleaning. All animal experiments were performed in accordance with institutional guidelines set forth by the State University of New York.

SG Cell Preparation and Spheroid
Cultures. SMGs were harvested, the attached fat and connective tissues were carefully removed, and the glands were separated with one processed for histology and the other for cell preparation. For cell preparation, SMGs were dissociated by mechanical and enzymatic digestion (0.025% Collagenase, 0.05% hyaluronidase, 1U/ml dispase) to single cell suspensions as described previously 20 . Ten thousand SMG cells from this single cell suspension were re-suspended in 40 μl media, mixed with 60 μl Matrigel (BD Biosciences, San Jose, CA) on ice, seeded evenly around the periphery of a well of a 12-well plate (in triplicates) to ease counting of spheres. Plates were placed at 37 °C for 20 min to allow Matrigel to solidify. Cell-Matrigel mixture was covered with 1 ml of salisphere growing media containing epidermal growth factor (20 ng/ml), fibroblast growth factor-2 (20 ng/ml), N2 supplement, insulin (10 μg/ml), dexamethasone (1 μM), 5% fetal bovine serum and 10 μM ROCK inhibitor Y-27632 (Sigma) and half of the media was changed every 2-3 days 40 . Cultures were grown for 7-8 days before being analyzed by fluorescent phase-contrast microscopy. To establish secondary cultures, primary salispheres were first released from Matrigel by incubating in 2U/ml dispase (Life Technologies, Grand Island, NY) for 60 min at 37 °C and then trypsinized to obtain single cells. Four thousand cells were mixed in Matrigel and cultured as described above. Images were captured from the entire rim of each well and used to quantify the number, size, and percentage of labeled spheres in each culture.
Statistical Analyses. All statistics including power analysis were performed using SPW12 software (Systat software Inc., San Jose, CA). When comparing two groups, the numbers of samples sufficient to result in statistically significant differences was determined using standard power calculations with α = 0.05 and a power of 0.8. Differences in means were evaluated by un-paired 2-sided Student's t test when comparing two groups, and one way analysis of variance and the Tukey's HSD post hoc comparison when comparing several groups. Significance was set at p < 0.05.

Availability of Materials and Data
All materials, data and associated protocols promptly available to readers without undue qualifications in material transfer agreements.