Chipping away at stem cells.

U recently, conventional wisdom has held that the lineage potential of stem cells was restricted to the tissue of origin. That is, stem cells derived from tissue A could only give rise to differentiated cells of tissue A and not to differentiated cells of tissues B, C, or D. This notion has been challenged since by a plethora of reports suggesting that stem cells have an inherent plasticity that allows them to respond to extrinsic signals present in the transplanted environment (1–3). In these studies, trans-differentiation or reprogramming has been substantiated mostly by cell morphology andyor the expression of antigenic proteins specific to the transplanted tissue environment and not to the tissue of origin. Although more rigorous criteria of functional activity and sustainable multilineage engraftment should be applied to future reports of trans-differentiation (4), if true, the phenomenon of trans-differentiation suggests that most if not all stem cells share an intrinsic genetic program that is not present in nonstem cells. Put another way, one hypothesis is that stem cells originating from different tissues may share a common genetic program responsible for maintaining them in an undifferentiated proliferative state. Once placed in a novel environment, the plastic nature of stem cells would allow them to respond to local cues and activate the appropriate differentiation pathway. Whether stemspecific genes exist or not is at present difficult to test given that the existence of stem cells itself is difficult to prove in many organ systems. In the case of neural stem cells (NSCs) of the murine central nervous system (CNS), the lack of known surface markers has hindered the prospective identification of these cells (i.e., for direct isolation of NSCs from fresh tissues). Currently, NSCs of the CNS are identified retrospectively. The existence of stem cells is inferred by the analysis of their differentiated progeny, using a complex series of cloning and differentiation assays. If there is indeed a universal stem cell gene-expression profile, the process of maintaining the ‘‘stem’’ state is in all likelihood highly complex, requiring the interactions and contributions of many different cellular genes in a spatial and temporal order. Thus, unraveling this gene puzzle is no easy task. In this issue of PNAS, Terskikh et al. (5) address the question of stem-specific genes by building on two previous studies that have examined separately the geneexpression profiles of hematopoietic stem cells (HSCs; ref. 6) and NSCs (7). In these earlier studies, cDNA derived from cells with minimal to no functional ‘‘stem activity’’ was subtracted from cDNA derived from cells enriched for stem activity. This approach was necessary to remove normal housekeeping genes along with transcripts unrelated to stem cell biology. The results are two cDNA libraries that are enriched selectively for transcripts believed to be specific to fetal HSCs (6) and postnatal day 0 (P-0) NSCs (7). Subjected to high-throughput sequencing, biochemical analysis, and verification [i.e., quantitative reverse transcription (RT)PCR, Northern and in situ hybridization], the subtracted cDNA libraries eventually were used to produce fetal HSC (6) and NSC (7) cDNA arrays. By using an extension of this strategy, Terskikh et al. ask whether there are adult HSC-enriched transcripts that are also expressed in mouse NSCs (5). In all three studies, cDNA chips were used to examine gene-expression profiles of HSC andyor NSC (5–7). To interpret more accurately the differences or similarities between two expression profiles, it is best to start with purified or homogeneous populations of all of the specific cell types under study. For HSCs and NSCs, obtaining a homogeneous population is especially critical, as they are present at very low frequencies. It has been estimated that in the fetal liver (the site where fully functional HSCs are first found) and the bone marrow (the site where HSCs are located throughout adulthood), HSCs are present at a frequency of 1 in 104 or 105 cells (8). Similarly, NSCs are estimated to comprise only 3–4% of neurospheres (murine CNS cells tend to proliferate as neurospheres or balls of cells in culture; ref. 9). The rarity of these cells and the possibility that stem-specific transcripts may be present at low abundance signify that in the absence of homogeneity, the majority of the gene expression data would be from nonstem cells. The problem of cell and tissue heterogeneity is thus one of the most formidable issues confronting stem cell biologists. In the hematopoietic system, the problem has been largely solved as stem cells are identified and purified prospectively through a combination of negative and positive selections. In day 14 fetal liver and bone marrow, HSCs with the phenotype A A41Sca1Kit1Linnegy low and Sca1Kit1Thy1.1lowLinnegylow, respectively, are isolated by using f luorescenceactivated cell sorting (FACS; refs. 5, 6, and 10). Earlier studies have shown that when these two respective hematopoietic cell populations are transplanted back into a lethally irradiated host whose endogenous hematopoietic system has been disabled, the cells are able to reconstitute the normal hematopoietic system functionally (10). Thus, in vivo transplantation studies have established these phenotypic cell populations as HSCs (i.e., they are able to self-renew and give rise to mature blood cells). In the present study by Terskikh et al. (5), bone marrow cells are sorted into two cell populations: the Sca1Kit1Thy1.1lowLinnegylow population (HSC) and the non-HSC remaining fraction of the bone marrow (BM). cDNA prepared from the BM cell population was subtracted from HSC cDNA, thus selectively enriching for transcripts specific to adult HSCs (5). It is noteworthy that this same strategy was used earlier to generate a gene-expression profile of HSCs derived from fetal day-14 liver (6). Interestingly, a comparison of the expression profiles of fetal (available in the Princeton Stem Cell