Although stem cell properties have been characterized in many tissues (1
), we are only beginning to understand how stem cell function is regulated at the molecular level. Gene expression profiles have been described for uncultured hematopoietic stem cells and cultured central nervous system neurospheres (2
), but not for prospectively identified, uncultured neural stem cells. Because stem cell properties change in culture (9
), the gene expression profile of uncultured neural stem cells might better reflect their properties in vivo.
Molecular links between stem cell function and disease are of particular interest. Many diseases involve defects in neural development and may be caused by mutations that impair neural stem cell function. One potential example is Hirschsprung disease, a relatively common (1 in 5000 births) gut motility defect caused by a failure to form enteric nervous system ganglia in the hindgut. This can lead to fatal distention of the gut (megacolon). Although a number of the mutations that cause Hirschsprung disease have been identified (12
), the ways in which these mutations affect neural development have been controversial, and it is unknown whether they affect gut neural crest stem cell (NCSC) function.
Gut NCSCs are self-renewing and multipotent, give rise to diverse types of neurons and glia in vivo, and persist in the gut throughout adult life (13
). Uncultured gut NCSCs can be isolated by flow cytometry by selecting freshly dissociated fetal gut cells that express the highest levels of p75 (the neurotrophin receptor) and α4
). These p75+
cells represent only 1 to 2% of cells in the E14.5 (embryonic day 14.5) rat gut (14
). Of the single p75+
cells that were added to culture, 60 ± 9% survived to form colonies, and 80 ± 7% of these colonies contained neurons (peripherin), glia [glial fibrillary acidic protein (GFAP)], and myofibroblasts [smooth muscle actin (SMA)]. These colonies typically contained 1 × 105
to 2 × 105
cells after 14 days of culture. These colonies are characteristic of NCSCs (13
We compared the gene expression profiles of gut NCSCs and whole-fetus RNA using oligonucleotide arrays (26,379 probe sets). Three independent 10,000-cell aliquots of freshly isolated, uncultured gut NCSCs were sorted by flow cytometry. Target RNA was independently extracted from the NCSCs and from three E14.5 fetuses, amplified through two rounds of in vitro transcription, and hybridized to each set of arrays.
The reproducibility of sample isolation and amplification was high. The variability among gut NCSC samples (mean ± SD: R2
= 0.975 ± 0.004) and among whole-fetus samples (R2
= 0.981 ± 0.003) was comparable to what would be expected from chipto-chip variation (R2
= 0.973 for the same sample on different chips). In contrast, the correlation coefficient between whole-fetus and gut NCSC samples was R2
= 0.855 ± 0.006. Arrays probed with whole-fetus or gut NCSC RNA contained 13,189 (50.0%) or 12,424 (47.1%) probe sets, respectively, at which transcript expression was detected. Genes corresponding to 475 probe sets were expressed at higher levels (by a factor of > 3; P
< 0.05) in gut NCSCs, and 970 probe sets were expressed at higher levels in whole-fetus RNA ( and tables S1 and S2
Table 1 Known genes that were more highly expressed in gut NCSCs relative to whole-fetus RNA by a factor of >5 [only expressed sequence tags that were highly similar (HS) to known genes were listed]. Ret, Sox10, Gfra1, and EDNRB have been linked toHirschsprung (more ...)
To assess the accuracy of the microarray results, we compared the expression of a subset of genes by quantitative (real-time) reverse transcription polymerase chain reaction (qRT-PCR). The same trends in expression levels were observed by microarray analysis and qRT-PCR in 20 of 21 cases (). Also, genes that encoded cell surface proteins and that appeared to be expressed by NCSCs by microarray analysis were also expressed at the protein level by flow cytometry (). The only exception was α1
integrin (CD49A), for which low-intensity signals were apparent by microarray analysis but which was undetectable by flow cytometry (18
). Overall, the results from microarray analysis, qRT-PCR, and flow cytometry were consistent.
Table 2 Comparison of the expression of selected genes in E14.5 gut NCSCs and whole fetuses by microarray analysis, qRT-PCR, and flow cytometry. Microarray intensities <100 were similar to background and were set to 100 for purposes of calculating ratios. (more ...)
Genes that have been linked to Hirsch-sprung disease were frequently expressed at higher levels in gut NCSCs. Of the 10 known genes that were most highly expressed in gut NCSCs relative to whole-fetus RNA, mutations in four of these genes have been linked to Hirschsprung disease: Ret
, and endothelin receptor type B
) ( and ).
To ensure that these genes were expressed in NCSCs rather than contaminating restricted neural progenitors or differentiated cells, we used qRT-PCR to compare their expression in E14.5 gut p75+
NCSCs with E14.5, E19.5, or postnatal day 4 (P4) gut cells that expressed moderate levels of p75med
that are enriched for restricted progenitors and more differentiated cells (fig. S1
, and EDNRB
were all expressed at significantly higher levels in NCSCs (P < 0.01). Most of the other 17 genes tested were also expressed at significantly different levels in NCSCs as compared with p75med
gut cells. Thus, there are significant differences in gene expression between gut NCSCs and restricted neural progenitors/differentiated cells.
The genes that were up-regulated in gut NCSCs relative to whole fetal RNA were not necessarily NCSC-specific. Whereas some of these genes (Ret
, and EDNRB
) were expressed at lower levels by p75med
gut cells, other genes (D
) were expressed at comparable or higher levels by p75med
cells (fig. S1
). Nonetheless, Ret
, and EDNRB
were all expressed at high levels by gut NCSCs, which raised the possibility that mutations in these genes cause severe defects in enteric nervous system development by impairing the function of gut NCSCs.
Mutations in GDNF
, its receptor Ret
, or its coreceptor Gfra-1
all lead to Hirschsprung disease in humans and aganglionic megacolon in mice (19
). GDNF promotes the survival, proliferation, and migration of mixed populations of neural crest cells in culture (27
). However, Ret protein was reported to be expressed by restricted gut neural crest progenitors but not by migrating trunk NCSCs (31
). These data raise the question of whether GDNF and Ret regulate gut NCSC function.
To analyze Ret receptor expression, we stained live gut NCSCs from the stomach and intestines with an antibody to Ret (). Virtually all gut NCSCs expressed Ret protein on the cell surface. In contrast, other populations of migrating and postmigratory trunk NCSCs failed to express Ret (18
). To study the function of Ret, we cultured E13.5 to E14.5 rat guts in collagen gels supplemented with GDNF (10 ng/ml). In the presence of GDNF, large numbers of cells migrated into the collagen gel (). Cells also migrated in the general direction of beads soaked in GDNF (). This is consistent with reports that GDNF is expressed in the gut in advance of migrating neural crest cells and is chemoattractive for neural crest cells in culture (29
Fig. 1 Flow-cytometric analysis of Ret, and CD29 (β1 integrin) expression by E14.5 gut p75+α4+ NCSCs and E14.5 gut p75−α4− epithelial progenitors from the same dissociated guts. As summarized in , the gut NCSCs (more ...)
Fig. 2 GDNF signaling promotes gut NCSC migration and is required for the migration of NCSCs into the intestines. [(A) to(E)] In nine independent experiments, E13.5 toE14.5 rat guts (*) were dissected and cultured in collagen gels. In the absence of GDNF (A (more ...)
To test whether GDNF promoted the migration of NCSCs (a small minority of gut neural crest cells), we extracted the migrated cells from the gel and cultured them at clonal density (13
). In five independent experiments, 2.5 ± 1.2% of migrating cells formed multilineage NCSC colonies. More than 13 times as many NCSCs could be extracted from collagen gels supplemented with GDNF as from control cultures (). This increase appears to be entirely explained by a promotion of migration, as GDNF did not affect the survival (), proliferation (), or differentiation of NCSCs into neurons and glia () under these culture conditions. Consistent with previous reports (27
), GDNF did appear to promote the proliferation and/or survival of restricted neural crest progenitors under the same conditions (fig. S2
To test whether NCSCs fail to migrate in vivo in the absence of GDNF signaling, we examined NCSC migration in the guts of Ret
-deficient mice. Few neural crest cells migrate beyond the esophagus in Ret−/−
), but the NCSCs in these mice have not been studied. In the esophagus of E13.5 mice, we found a factor of 4 reduction in the frequency of Ret−/−
NCSCs (), although this difference was not statistically significant because one of the Ret−/−
mice had normal numbers of NCSCs in the esophagus. The proliferation and differentiation of these Ret−/−
NCSCs in culture were indistinguishable from Ret+/+
NCSCs (fig. S3
), suggesting that there was no intrinsic defect in their stem cell potential. In contrast, in the stomach and intestine we found a factor of 20 reduction in the frequency of NCSCs in Ret−/−
mice (). A failure of Ret−/−
NCSCs to migrate beyond the esophagus is sufficient to explain the absence of enteric ganglia in the distal stomach and intestines of Ret−/−
Because GDNF did not affect NCSC survival or proliferation in culture, the precipitous reduction in NCSC frequency in the stomach and intestine is likely caused primarily by a defect in migration. However, Ret signaling may also be required for the survival or proliferation of NCSCs before E12.5 in the esophagus or before their entry into the esophagus (32
). Most neural crest cells that colonize the gut are Ret-dependent and derive from the vagal neural crest, whereas a minority of neural crest cells that colonize the esophagus are Ret-independent and derive from the trunk neural crest (33
). One possibility is that NCSCs are depleted from the esophagus and virtually absent from the stomach and intestine because only trunk-derived NCSCs are able to migrate into the foregut of Ret−/−
mice. This would suggest that Ret signaling is required not only for the migration of NCSCs within the gut but also for the migration of most vagal-derived NCSCs into the esophagus. Irrespective of the precise fate of Ret−/−
vagal-derived NCSCs, these data demonstrate that Ret
is required for the colonization of the gut by NCSCs.
It is likely that loss-of-function mutations in Gfra-1
, and Sox10
also lead to Hirschsprung disease by impairing gut NCSC function. Sox10 has recently been shown to regulate the multipotency of NCSCs (34
The mutations responsible for about one-half of Hirschsprung cases have not yet been identified (35
). Given that mutations in 4 of the 10 most up-regulated genes in gut NCSCs have already been shown to cause Hirschsprung disease, the remaining genes that are highly up-regulated in gut NCSCs represent a resource of candidates that could also cause or modify the risk of Hirschsprung disease when mutated.
Two studies recently identified subsets of genes that were up-regulated in three stem cell populations, relative to other cells, and concluded that the genes they identified were indicative of “stemness” or the “molecular signature of stem cells” (6
). Only one gene, α6 integrin
, was up-regulated in gut NCSCs (NCSC/fetus = 4.6, P
< 0.0001) and was present on both of these previously published lists (table S3
). α6 integrin−/−
mice develop to birth but then die neonatally as a result of severe blistering in the skin and other epithelia (36
). Keratinocyte stem cells and spermatogonial stem cells also express α6
). It will be interesting to determine whether α6
integrin is necessary for stem cell function in multiple tissues.
Our results demonstrate the value of combining the analysis of stem cell phenotype and function with microarray analysis and reverse genetics. The results we obtained by microarray analysis were consistently confirmed by qRT-PCR (), flow cytometry (), and functional analysis (). We believe this combination of approaches will provide critical insights into the cellular and molecular mechanisms underlying diseases.