|Home | About | Journals | Submit | Contact Us | Français|
Experimental models of myelin disorders can be treated by the transplantation of oligodendrocyte progenitor cells (OPCs) into the affected brain or spinal cord. OPCs express gangliosides recognized by MAb A2B5, but this marker also identifies lineage-restricted astrocytes and immature neurons. To establish a more efficient means of isolating myelinogenic OPCs, we asked if FACS could be used to sort PDGFα receptor+ cells from fetal human forebrain, based on expression of the PDGFRα epitope CD140a. CD140a+ isolates were maintained as mitotic bipotential progenitors that could be instructed to either oligodendrocyte or astrocyte fate. Transplanted CD140a+ cells were highly migratory, and rapidly and robustly myelinated the hypomyelinated shiverer mouse brain, more efficiently than did A2B5-sorted cells. Microarray analysis of CD140a+ cells revealed their differential expression of CD9, as well as of PTN-PTPRZ1, wnt, notch and BMP pathway components, indicating the dynamic interaction of self-renewal and fate-restricting pathways in these cells.
Disorders of myelin are among the most prevalent and diverse neurological disorders. They include conditions as diverse as the childhood hereditary leukodystrophies, cerebral palsy, subcortical stroke, multiple sclerosis, and vascular white matter disease, among others1–3. These conditions have in common the loss of central oligodendrocytes, the sole myelin-producing cell of the brain and spinal cord, and each requires oligodendrocytic replacement and remyelination as an antecedent to recovery. Myelinogenic oligodendrocyte progenitor cells pervade the parenchyma of both the developing and adult human brain4–6, and we have developed strategies for their isolation and transplantation, by which we have established the ability of these cells to efficiently remyelinate the dysmyelinated brain and spinal cord 7–9. However, we have hitherto isolated these cells on the basis of their expression of gangliosides recognized by monoclonal antibody (MAb) A2B5 5, 10–12, the targets of which are expressed by multiple phenotypes besides oligodendrocyte progenitors. In particular, developing neurons also express A2B5, necessitating their exclusion from A2B5-sorted isolates derived from fetal human brain tissue. We thus used PSA-NCAM-based depletion of neuronal progenitors, prior to A2B5-based immunoselection, to prepare enriched populations of OPCs from fetal human brain tissue, which proved able to myelinate and ultimately rescue the congenitally hypomyelinated shiverer mouse7, 9. Yet these studies also revealed that fetal A2B5-defined OPCs were still phenotypically heterogeneous, and included a large fraction of non-myelinogenic cells restricted to astrocytic fate. To more precisely enrich myelinogenic OPCs, we thus sought to identify better markers for their definitive isolation. To this end, we asked if genes already recognized as selectively expressed by adult human OPCs, which are highly biased to oligodendrocyte fate, might serve as markers by which to isolate oligodendrocyte-biased OPCs from the fetal brain as well. In this regard, we had previously found that human A2B5+ oligodendrocyte progenitors highly over-express mRNAs encoding both the NG2 chondroitin sulfate proteoglycan, and PDGFRA, the platelet derived growth factor-α receptor13, suggesting the utility of each as potentially more efficient at isolating oligodendrocyte-biased OPCs14, 15. While NG2 is also expressed by pericytes, which limited its utility as a selection marker, PDGFRα appears to be expressed specifically by OPCs in both rodents 16–18 and humans6, 13, 19, 20.
Yet despite the potential attraction of a PDGFαR-based immunoselection strategy, no one had ever enriched OPCs using that’s strategy. In part, this failure was due to the incompatibility of standard dissociation techniques with PDGFαR-based immunoselection. To address these issues, we thus paired the use of an antibody against CD140a, which recognizes a human PDGFRα-ectodomain, with a compatible papain-based dissociation procedure7, to isolate CD140a+/PDGFRα+ OPCs from the fetal human brain. We confirmed that these cells were able to give rise to both oligodendrocytes and astrocytes, in a BMP and serum-dependent fashion, and that their oligodendrocytic progeny could robustly myelinate the hypomyelinated shiverer mouse brain. Importantly, CD140a+ cells exhibited more rapid and efficient myelination than did their A2B5+/PSA-NCAM− counterparts. In addition, expression profiling confirmed that CD140a+ cells differentially express transcripts characteristic of oligodendroglial progenitors. These differentially expressed genes encoded surface antigens such as CD921, 22, thereby identifying alternative ectodomains by which to sort OPCs, as well as components of the wnt and notch pathways, which may be critical to the turnover and fate of fetal human OPCs.
In the human forebrain, OPCs are most actively generated during the second trimester. To assess the geographic distribution of CD140a/PDGFRα+ cells during that period, we immunolabeled sections of 22 week g.a. forebrain with monoclonal anti-CD140a IgG, which recognizes a PDGFRα ectodomain23. Immunostaining revealed that CD140a+ cells pervaded the forebrain, extending through the intermediate zone to the developing cortical mantle (Fig. 1A), with a morphology similar to that of NG2-defined adult OPCs24, and a distribution which anticipated that observed in the adult human forebrain4–6. Virtually all CD140a+ cells co-expressed the oligodendrocyte lineage transcription factor olig2 (Fig. 1B), while sox2 was co-expressed by a fraction (Fig. 1C). A significant proportion of CD140a+ cells expressed Ki67, a marker of actively cycling cells (Fig. 1D). In contrast, no co-localization of CD140a-immuno-reactivity was noted with either astrocytic glial fibrillary acidic protein (GFAP) or neuronal Hu protein. Together, these data suggest that CD140a+ cells comprise a pool of uncommitted OPCs that ubiquitously co-express olig2.
Flow cytometry revealed an abundant population of CD140a+/PDGFRα+ cells at 20–22 weeks gestation, by which point 4.8 ± 0.7% of cells within cortical dissociates, which included admixed intermediate zone and cortical mantle, were sorted as CD140a+ (n=13, standard error [SE]). (Fig. 2A/B). At earlier gestational ages however, relatively few CD140a+ cells were noted; appreciable numbers first began to appear in forebrain dissociates at 16 weeks g.a.. Over the period of 16–23 weeks, we noted a significant correlation between the incidence of cortical CD140a+ cells and gestational age (p<0.0001, r2=0.43, df = 26) (Fig. 2C). In contrast, the relative incidence of CD140a+ cells within the dissected ventricular and subventricular zones (VZ/SVZ) did not increase with gestational age (p=0.57, r2=0.04, df = 10) (Fig. 2D). The relatively constant proportion of CD140a+ cells in the VZ/SVZ, in contrast to the rapidly increasing proportion of CD140a+ cells in cortical dissociates (3.1 ± 0.4 vs. 1.2 ± 0.1%; unpaired t-test, p=0.012, df=37), suggested a monotonic colonization of the developing brain by newly generated VZ/SVZ-derived OPCs (Fig. 2E).
A2B5+/PSA-NCAM− cells derived from the fetal human brain comprise a phenotypically heterogeneous cell population. Although these cells include a myelinogenic fraction of bipotential OPCs7, 9, they also include a GFAP+ fraction that appear restricted to astrocyte phenotype. We thus asked if CD140a+/PDGFRα+ OPCs comprised a discrete fraction of the A2B5+/PSA-NCAM− population. Dissociated fetal cells were first depleted of PSA-NCAM+ cells by immunomagnetic sorting, and then subjected to two-color flow cytometry for A2B5 and CD140a (Fig. 3A–C). We found that CD140a+ cells were significantly more abundant in the A2B5+/PSA-NCAM− pool than in the unsorted population from which both derived (23.2 ± 3% in A2B5+/PSA-NCAM− vs. 10.9 ± 3% in the initial dissociate; t-test p=0.0324, n=4) (Fig. 3D).
These data indicated that CD140a+ cells comprised a subpopulation of A2B5+ OPCs. On that basis, we next asked whether the oligodendrocyte-competent fraction of glial progenitors was specifically enriched in the CD140a subpopulation. To this end, A2B5+/PSA-NCAM− and CD140a+ cells derived in parallel from the same tissue samples (n=3) were plated into base media supplemented with 1% platelet-depleted serum (PD-FBS) and triiodothyronine (T3; 30 ng/ml), to assess their potential to differentiate as oligodendrocytes (n=3 samples, each done in triplicate wells/phenotype). After 7 days, matched cultures were immunostained using MAb O4, which recognizes immature post-mitotic oligodendrocytes in humans5, 25, 26. We found that the proportion of O4+ oligodendrocytes was substantially greater in all CD140a+ fractions than in their CD140a− counterparts (n=3 sorts) (Fig. 3E).
Since only a fraction of A2B5+ cells generate oligodendrocytes7, and since only some A2B5+ cells co-express CD140a, we next asked if oligodendrocytes arose exclusively from the CD140a+ subpopulation. To this end, we used fluorescence-activated cell sorting (FACS) to separately isolate CD140a+ and CD140− cells. Immediately after sorting, the CD140a+ cells were found to be uniform in size, relatively small and phase bright. Within 24 hrs, most were observed to elaborate fine processes, initially as bipolar cells (Supplemental Fig. 1A). These cells uniformly expressed CD140a/PDGFRα (Supplemental Figs. 1B–C), and co-expressed the transcription factors olig2 and sox2, which are both expressed by uncommitted glial progenitors (Supplemental Figs. 1D–E). In contrast, the CD140a− fraction was largely devoid of olig2+ cells (Supplemental Fig. 1F). The vast majority of CD140a− cells exhibited a neuronal phenotype, expressing both βIII-tubulin and the neuronal mRNA binding protein Hu (Supplemental Fig. 1G). In addition, a subpopulation of cells in the CD140a− fraction also expressed astrocytic markers, including both GFAP and aquaporin 4 (Supplemental Fig. 1H).
To promote oligodendrocytic differentiation, the two populations were then each raised in T3-supplemented media. Within 1–2 days, most CD140a+ cells exhibited a bipolar progenitor morphology and expressed nuclear olig2. After 4 days, 36.6 ± 5% of CD140a-sorted cells expressed the oligodendrocytic sulfatide recognized by MAb O4 (n=13) (Fig. 4A), whereas <0.1% of CD140a− cells expressed O4-immunoreactivity at this point (n=7) (Fig. 4B). CD140a-derived O4+ cells matured as O1+ and MBP+ oligodendrocytes by 7 days in vitro (Fig. 4C). In contrast, at 7 days the vast majority of CD140a− cells were either βIIItubulin+ neurons (63 ± 8%, n=4 samples), or GFAP+ astrocytes (6 ± 2%, n=4) (Fig. 4D). A subset of A2B5+ cells (21 ± 1%, n=3), some of which co-expressed GFAP, was also noted in CD140a− cultures. Nonetheless, neither exposure to serum concentrations up to 10%, nor supplementation by 20 ng/ml IGF1, triggered the appearance of O4+ oligodendrocytes in CD140a− cultures. Overall, we noted a >200-fold increase in the percentage of O4+ cells arising from CD140a+ cells, relative to their CD140a− counterparts (37% CD140a+ vs. 0.14% CD140a−, p<0.0001, two-tailed t-test; df=18). Thus, O4+ oligodendrocytes were produced only by CD140a+ progenitor cells.
Since oligodendrocytic fate potential was restricted to CD140a+ cells, we next asked if CD140a+ cells could also be induced to astrocytic fate, using serum or BMP exposure, as previously noted in adult-derived OPCs13. We found that addition of either serum or BMP-4 after FACS induced rapid GFAP+ astrocytic differentiation. Within 7 days of post-sort exposure to 0.5% PD-FBS, cultures exhibited a >3.5-fold rise in GFAP+ astrocytes, from 8.8% to 30.7% (p=0.002, unpaired t-test, df=5). Similarly BMP-4 exposure (50 ng/ml) yielded a dose-dependent increase in astrocytes, from 8% to >40% at 4 days (p<0.01; F = 12.4; ANOVA with Dunnett’s post hoc analysis). In contrast, oligodendrocyte differentiation was inhibited by BMP-4; O4+ cells declined from >30% to <5% at 4 days (p<0.01; F = 12.7). As a result, the GFAP/O4 ratio in these cultures increased in response to BMP-4, from <0.5 to > 15.
Importantly, CD140a+ cells could be maintained as bipotential progenitors in vitro. PDGF-AA and FGF2 (each 20 ng/ml) inhibited oligodendrocytic differentiation, such that only 2.4 ± 1.2% and 4.8 ± 1.3 of PDGF-AA/FGF2-exposed CD140a+ cells developed O4 immuno-reactivity by days 4 and 7 (n=5); under these conditions, most cells remained as A2B5+ progenitors (75 ± 7%, n=6), accompanied by the gradual appearance of GFAP+ astrocytes (8.8± 1.9%, n=4). Importantly, FGF2 withdrawal and addition of T3 favored oligodendrocyte differentiation, increasing the incidence of O4+ cells to 13.7 ± 2.9 % (p=0.01, t-test; n=3). In addition, astrocyte lineage commitment could be blocked by noggin, a soluble competitive antagonist of the BMPs. When added to CD140a+ cells in the absence of growth factors for 4 days, 100 ng/ml noggin potently inhibited BMP4-induced astrocytic differentiation (p<0.05, Tukey’s post hoc test), and was also sufficient to reverse the inhibitory effect of BMP-4 on oligodendrocyte commitment (21.0 ± 3% O4+ cells in BMP-4 + noggin, vs. 4.3 ± 1% O4+ cells in BMP-4 alone; p<0.05). Interestingly, baseline levels of astrocytic and oligodendrocytic differentiation (8.1 ± 3% and 22.6 ± 4%, respectively) were unaffected by 100 ng/ml noggin in the absence of exogenous BMP (8.1 ± 4% GFAP+ and 21.0 ± 3% O4+).
We next asked if CD140a-sorted cells could differentiate as mature oligodendrocytes in vivo as well as in vitro, and whether their derived oligodendrocytes were competent to myelinate axons in vivo. To this end, we transplanted 10 neonatal myelin-deficient shiverershi/shi mice with 5 × 104 cells each of CD140a-sorted OPCs. The xenografts were delivered shortly after birth as intracallosal injections, as previously described7. Under these conditions, A2B5+/PSA-NCAM− sorted fetal human OPCs typically begin to generate myelin by 8 weeks after transplant, myelinating relatively large volumes of otherwise hypomyelinated white matter by 12 weeks7. To assess the relative competence of CD140a-sorted cells to effect myelin production and axonal ensheathment, the CD140a-engrafted shiverers were sacrificed at either 8 (n=7) or 12 (n=3) weeks of age, and their brains cryosectioned and immunostained for myelin basic protein (MBP), which is not otherwise expressed in shiverer. By 8 weeks, the CD140a-engrafted mice indeed exhibited significant and widespread forebrain myelination, in a distribution similar to but generally broader than that which we had observed using A2B5-sorted OPCs (Figs. 5 and and6A).6A). Quantification within the fimbria revealed that at 8 weeks after injection, a time-point at which myelinogenesis from implanted A2B5-sorted fetal human OPCs is just beginning (Windrem et al., 2004), 8 ± 3% of CD140-sorted cells had already differentiated as MBP+ oligodendrocytes (n=3, Figs. 5A, D). At that time-point, the majority of human cells were of oligodendrocytic lineage, such that 68 ± 14% of human nuclear antigen (hNA)+ cells co-expressed olig2, while <5% of human CD140a+ cells had differentiated as GFAP+ astrocytes (Fig. 5C, D). In addition, a third persisted as NG2+ OPCs (31 ± 14%; Fig. 5B, D); a fraction of these remained mitotically competent, as defined by Ki67 (8 ± 4%, n=3; Fig. 5D).
We next compared the performance at 12 weeks of CD140a+ cells, in terms of both their migration and myelination competence, to that of A2B5+/PSA-NCAM− cells whose in vivo myelination competence we had previously described at that time point. To this end, we assessed the efficiency of callosal axonal ensheathment by CD140a+ cells, using the same transplantation and assessment protocols, source gestational ages, sacrifice time-point and imaging metrics that we had previously described for callosal myelination by fetal A2B5+/PSA-NCAM− cells (Fig. 6). We found that at 12 weeks, the CD140a-engrafted shiverers exhibited ensheathment of 26.7 ± 2.5% of all callosal axons by human MBP-defined myelin (mean ± SE; n=3 mice) (Figs. 6B–E). Remarkably, in especially dense areas of donor cell engraftment, ensheathment efficiencies approaching 40% were often noted. These data contrasted sharply to the average axonal myelination efficiency by A2B5+/PSA-NCAM− cells at 12 weeks, of 10.1 ± 2.8% (n=6) (Windrem et al., 2004). Comparing these data sets, the CD140a+ cells myelinated with significantly and substantially higher efficiency at 12 weeks than did their A2B5+/PSA-NCAM- counterparts (Figs. 6E–F) (n=6; p <0.01 by Student’s t-test). These results suggested that CD140a+ cells exhibited both a higher efficiency of oligodendrocytic differentiation, and a more rapid initiation of myelinogenesis, than did fetal A2B5-defined OPCs; indeed, the CD140a+ cells myelinated with an efficiency and time course more similar to that which we had previously observed with adult brain-derived A2B5+ OPCs7. Unlike adult OPCs, however, fetal CD140a+/PDGFRα+ cells migrated widely, extending as broadly throughout the brain and brainstem as their fetal A2B5+/PSA-NCAM− counterparts, and with even greater penetration of both cortical and subcortical grey matter (Fig. 6A). Importantly, despite their efficient oligodendrocytic production and myelinogenesis, engrafted CD140a+ cells remained bipotential in vivo, in that they reliably generated astrocytes as well as oligodendrocytes in recipient brains (Fig. 6G–H). Together, these findings indicate that CD140a+ fetal human OPCs exhibit widespread migration and myelinogenesis in vivo, and do so more rapidly and efficiently than the larger pool of A2B5+/PSA-NCAM− OPCs.
To better define the differentiation state of CD140a+ human OPCs, we next assessed their expression profiles, relative to CD140a− cells. We first used Taqman low density arrays to achieve high throughput RT-qPCR of a panel of 48 marker genes, that we chose as potentially predictive of glial cell fate (Table 1). Using 6 fetal cortical samples of 20–22 weeks g.a., sorted on the basis of CD140a, we compared the marker gene expression patterns of the CD140a+ and CD140a− fractions (Table 1). PDGFRA mRNA was >500 fold higher in CD140a+ than CD140a− cells (paired t-test, p=0.013), validating both the stringency of our sorts, and the use of anti-CD140a as a means of isolating PDGFRα+ cells. We also noted significantly higher expression of OPC and oligodendrocyte-lineage markers. For instance, the CSPG4 proteoglycan NG2 was over-expressed almost 50-fold in CD140a+ cells, while the oligodendrocyte lineage transcription factors OLIG2 and SOX10 were expressed >150-fold more in CD140a+ than CD140− cells. In addition, the progenitor profile of CD140a+/PDGFRα+ cells was confirmed by their significant expression of SOX227, 28. In contrast, neither mature oligodendrocytic genes, such as MBP, PLP MOBP or MOG, nor those selective to astrocytes, such as GFAP or AQP4, were enriched in CD140a+ OPCs (Table 1). Interestingly, while we also noted over-expression of the microglial-selective transcripts CD68 and CD86, migrating human glial progenitors and microglia may share these antigens29.
Using Affymetrix U133+2 arrays, we next compared the transcription profiles of fetal human CD140a-sorted and depleted cells (see Methods). 408 genes were identified as both significantly over-expressed by CD140a+/PDGFRα+ cells (Supplemental Table 1). In accord with our qPCR data, the microarray analysis identified several OPC markers among genes highly expressed by CD140a-sorted cells (Supplemental Table 2). Gene ontology-based and Ingenuity-based pathway analysis identified nervous system development (GO:0007399; p=8.6 × 10−5), neurogenesis (p=1.6×10−8) and, interestingly, schizophrenia (p=5.0 × 10−12). Gene set enrichment analysis (GSEA) identified the WNT, NOTCH, and EGF signaling pathways as over-represented in CD140a-sorted cells (q = 1.7 ×10−3, 5.1 ×10−5, and 9.3 ×10−5, respectively). CD140a active signaled via WNT as indicated by over-expression of several known WNT target genes (p=0.011) (Supplemental Figure 2). In addition, GSEA of KEGG-defined pathways identified those involved in glycan synthesis as the most differentially regulated by CD140a+ cells (Supplemental Figs. 3A and 3B).
Among membrane proteins, the gene encoding CD9 was 7.4-fold enriched in CD140a+ cells. We thus asked if human OPCs might be further separable using CD9 ectodomain-directed FACS. Flow cytometry revealed that CD9+ cells comprised 2.75 ± 0.7% of all cells in the fetal intermediate zone and cortical plate (n=5), and were less common in the VZ (0.51 ± 0.1%). Two-color cytometry then revealed that the CD9+ pool partially overlapped with that of CD140a+ cells, such that half of CD140a+ OPCs expressed CD9 (49.6 ± 5%, n=6); as such, 2.6 ± 1.5% of the entire dissociated pool co-expressed CD140a and CD9 (Supplemental Figure 4). Indeed, used alone, CD9 based sorting yielded a highly significant, 14-fold enrichment in CD140a+/PDGFRα+ cells relative to their unsorted dissociates (t-test, p=0.00002). These data suggest that CD9 identifies an antigenically distinct subpopulation of OPCs in the fetal human brain, and suggest the utility of concurrent CD9 and CD140a-directed FACS for isolating a highly enriched, oligodendrocyte-competent fraction of human glial progenitor cells.
In this study, we developed a CD140a/PDGFRα-based strategy for the direct isolation of OPCs from the fetal human brain. These cells were bipotential for oligodendrocytes and astrocytes, could be instructed to generate one or the other in vitro, and proved robustly myelinogenic in vivo. Importantly, they were even more efficient at in vivo oligodendrocytic differentiation and myelination than were matched isolates of fetal A2B5+/PSA-NCAM−-sorted cells, which comprised the previous standard by which oligodendrocyte progenitors were isolated for therapeutic assessment. In addition, we found that CD140a+ cells expressed a transcriptional profile that manifested the importance of notch, wnt and EGF signaling to both their homeostatic turnover and differentiation. Indeed, their gene expression patterns suggest that fetal human OPCs actively signal to and modulate their local environment, by their robust and selective production of a host of extracellular matrix molecules and secreted ligands.
Our data suggest that in the human forebrain, this process may begin early in the second fetal trimester, with the appearance of the first forebrain OPCs. Indeed, little had previously been known about the ontogeny of OPCs in the human forebrain. PDGFRα and NG2-expressing oligodendrocyte progenitors have been described at 18 weeks g.a. in the developing telencephalon29, 30, yet we found that CD140a+ cells could be isolated from the fetal brain by 16 weeks g.a., and significantly populated the cortical mantle by 18 weeks g.a. Interestingly, previous studies have noted that O1-defined oligodendrocytes appear at 28–30 weeks g.a. in the human forebrain, coincident with the initial appearance of myelin30. Thus, the period spanning 18–30 weeks g.a. appears to comprise the principle period of glial progenitor expansion and oligodendrocytic specification in the developing human forebrain.
We previously noted that fetal human OPCs, when defined as A2B5+/PSA-NCAM−, differentiated as both oligodendrocytes and astrocytes upon neonatal xenograft, and could rescue the otherwise lethal shiverer phenotype7, 9. Yet in the present study, we found that not all A2B5+/PSA-NCAM− cells were CD140/PDGFRα+; rather, CD140a+ cells were enriched within the larger population of A2B5+/PSA-NCAM− cells. Importantly, only CD140a+ cells were capable of oligodendrocyte differentiation in vitro, indicating that CD140a+ cells comprised the entire fraction of A2B5+/PSA-NCAM− cells responsible for myelination upon xenograft. Thus, our data indicated that the A2B5+/PSA-NCAM− immunophenotype comprised a heterogeneous pool, that included both astrocyte-restricted CD140a− cells, and oligodendrocyte-competent CD140a+ progenitors.
The oligodendrocytic bias of fetal human CD140a+ cells was reflected in both their antigenic and gene expression profiles, which were highly enriched in known markers of oligodendroglial lineage (PDGFRA, NG2, ST8SIA1, NKX2.2, OLIG1, OLIG2, and SOX10). Importantly though, human fetal CD140a+/PDGFRα+ cells were not restricted to oligodendrocyte phenotype, but rather were bipotential for astrocytes and oligodendrocytes alike; since when stimulated in vitro with serum or BMP4, they readily exhibited astrocytic differentiation. As such, CD140a-sorted fetal human OPCs behave similarly to both their adult human and rodent counterparts, which proliferate in response to PDGF-AA and FGF212, 31, 32, and can develop as astrocytes in response to serum or BMPs13, 33, 34, yet differentiate as oligodendrocytes when raised in the absence of mitogens or BMPs. As such, CD140a+ cells seem both lineally and phenotypically analogous to bipotential O2A progenitors of the developing rodent brain 7, 33, 34.
The bipotentiality of CD140a+ OPCs is of great practical import: While our data suggest that these cells are superior to fetal A2B5-defined cells in terms of their myelinogenic potential, their differentiated phenotype still depends upon the environment into which they are introduced. Thus, whereas CD140a+ OPCs might be expected to efficiently differentiate as oligodendrocytes and thus effectively remyelinate congenitally hypomyelinated brain, such as in Pelizaeus-Merzbacher disease or cerebral palsy, their use in acute demyelinating disorders, such as in multiple sclerosis or post-ischemic demyelination, might prove to be limited by their astroglial differentiation in an injury environment35. Thus, the bilineage competence of these cells must be considered as appropriate clinical targets for their use are further defined.
Although human fetal CD140a+ cells appear bipotential for astrocytes and oligodendrocytes both in vitro and in vivo, we cannot exclude the possibility of discrete CD140a+ subpopulations already restricted to either one phenotype or the other. In particular, it is important to note that even in BMP-stimulated cultures of CD140a+ cells directed to astrocyte phenotype, some oligodendrocytes were noted to develop. This suggested that the CD140a+ pool, while including all OPCs with oligodendrocyte lineage competence, might also include a fraction restricted to oligodendrocyte phenotype. We were intrigued then to note the selective expression of mRNA encoding the pro-oligodendroglial tetraspanin CD922 by fetal OPCs. Flow cytometry revealed that CD9 was expressed by a discrete subset comprising almost half of all CD140a+ cells, from which the CD9+/CD140a+ fraction could be isolated (Supplemental Figure 4). Interestingly, in development, the pool of CD9-expressing OPCs declines as myelination proceeds22, suggesting the depletion of a discrete pool of oligodendrocyte-biased progenitors. As such, the CD9+ fraction of CD140a+ OPCs may define an intriguing pool for further assessment as a potential therapeutic vector.
Besides their potential use as transplantable cellular agents for remyelination, OPCs comprise clear targets for therapeutic mobilization and induced differentiation, whether as oligodendrocytes or astrocytes. To better assess the receptors expressed by these cells, and their cognate signaling pathways by which their mobilization and fate are determined, we assessed their patterns of differential gene expression, relative to CD140a- cells. We found that the transcriptional profile of fetal CD140a-defined OPCs largely resembled that of adult A2B5-defined OPCs13. Both phenotypes selectively over-expressed a number of chondroitin sulfate proteoglycans, as well as their cis-interacting partners. These included the heparin-binding ligand pleiotrophin (PTN), and its principle receptor, protein tyrosine phosphatase-β/ζ (PTPRZ1), a molecule important to both the turnover and fate determination of adult OPCs13. One target of PTRPZ1 is β-catenin36, which may regulate wnt signaling in oligodendroglia37. Indeed, this pathway may be regulated differently in fetal than in adult OPCs, as we found significantly higher expression of the wnt transcription factors TCF7L1 and TCF7L2 in fetal CD140a+ cells, than in their A2B5-sorted adult counterparts. These data suggest a possible role for the PTN-PTPRZ1 interaction in regulating wnt signals in fetal human OPCs, differentially so relative to its function in adult human OPCs.
In summary, we have defined a method for the direct, CD140a/PDGFRα-based isolation of a highly myelinogenic pool of OPCs from the fetal human brain, that comprises the entire population of cells capable of rapid oligodendrocyte differentiation. Like A2B5-defined oligodendrocyte progenitors, fetal CD140a+ cells robustly differentiate as myelinating oligodendrocytes when transplanted into the shiverer forebrain. Yet unlike the heterogenous population of A2B5-expressing progenitor cells, which includes some but not all oligodendrocyte-competent cells and is also admixed with lineage-restricted fibrous astrocytes, the CD140a-defined pool comprises an exclusive population of cells that uniformly retains oligodendrocytic lineage competence. Importantly, CD140a+ cells differentiate and myelinate more rapidly and efficiently than do fetal A2B5-sorted OPCs, suggesting the potential advantage of CD140a+ cells as cellular vectors for treating disorders of myelin. Indeed, given their relative homogeneity, broad migration competence and rapid myelinogenesis, our data suggest the superiority of CD140a+ cells, relative to previously studied immunophenotypes, as therapeutic reagents for adult as well as pediatric myelin disorders.
Fetal brain tissue was obtained from 38 cases (15 – 22 weeks gestational age). Samples were obtained from patients who consented to tissue use under protocols approved by the University of Rochester-Strong Memorial Hospital Research Subjects Review Board. Cortical tissue was dissected into ventricular and sub-ventricular zone and overlying cortical mantle and chilled on ice. Briefly, the minced samples were dissociated using papain and DNase as described 7, always within 2 h of extraction, and maintained overnight in DMEM/F12/N1 base media (composition below), with 20 ng/ml FGF-2 (Sigma Aldrich).
The day after dissociation, cells were prepared for either magnetic separation or FACS. PSA-NCAM-defined depletion of cells by magnetic separation was performed as described 9. CD140a is the same gene as PDGFR alpha in man and the CD140a monoclonal antibody therefore recognizes human PDGFRα 23. For CD140a or CD9 FACS (BD Pharmingen), cells were re-suspended in PBS with 2mM EDTA and 0.5% BSA and incubated with primary antibody. Isotype and fluorescence-minus one controls were used to set appropriate gates. Each batch of CD140a-PE IgG2a and IgG2a isotype-PE control antibodies (BD Pharmigen) were optimized using flow cytometry, typically over a concentration range of 1:15–1:75. Single cells were discriminated using pulse width and height measurements. After sorting, cells for expression analysis were directly frozen, while cells for in vitro assessment were maintained in a base medium composed of DMEM/F12 (Invitrogen) with HEPES (17.9m), insulin (5 ug/mL), L-glutamine (6.35 mM), D-glucose (46.4 mM), sodium pyruvate (1.5 mM), non-essential amino acids (0.2 mM,), penicillin/streptomycin (40 U/mL), progesterone (57 ng/ml), putrescine (92 μg/ml), selenite (0.065 μg/ml), and transferrin (5mg/mL). For transplantation, cells were maintained in this base medium supplemented with FGF-2 (20 ng/ml) for 1–3 days.
Cultures were stained for the early progenitor and oligodendrocyte markers, A2B5 and O4 respectively, as described 5, 13. Both O4 and A2B5 were localized on live cells that were then fixed with 4% paraformaldehyde. O4 supernatant (gift of R. Bansal and S. Pfeiffer, University of Connecticut) was used at a dilution of 1:100, and monoclonal antibody A2B5 supernatant (clone 105, American Type Culture Collection) was diluted in a 1:1 with DMEM/F12/N1, each was applied for 40 minutes at 4°C. Post-fixation, cultures were stained for astrocyte markers, GFAP (1:1000, Chemicon) and AQP4 (1:200, Chemicon), and for neuronal markers, βIII-tubulin (clone TuJ1, 1:1000, Covance) and HuD (clone 16A11, 10ng/ml, Invitrogen). Oligodendrocyte lineage cells were labeled using Olig2 (1:200, AbCam). Neural progenitor cells were labeled using anti-SOX2 antibody (1:500, R & D Systems). Secondary antibodies, Alexa-488, 594 and 647 conjugated goat-anti mouse IgM, rabbit and rat antibodies respectively were used at a dilution of 1:400 (Invitrogen, La Jolla, CA). Fixed cultures were counterstained with DAPI (10 ng/ml; Invitrogen).
The number of phenotypically labeled cells were counted in 10 randomly chosen fields from individual replicate samples (n=3 at each dosage level). Statistical significance was assessed by one-way repeated measures analysis of variance (ANOVA), followed by Dunnett’s multiple comparisons test (GraphPad Prism 5, p<0.05).
Transplanted cells were identified using antibody 1281 to human nuclei (1:400, Chemicon), monoclonal antibody 2029 (clone 9.2.27) to human chondroitin sulfate proteoglycan Ng2 (1:200, Chemicon), goat antibody to human Olig2 (1:200, R&D), rabbit antibody 5804 to GFAP (1:1000, Chemicon), rabbit antibody to Ki67 (1:200, LabVision), and rat antibody 7349 to MBP (1:25, Abcam), all as described 7–9, 12. Confocal imaging was done using an Olympus Fluoview mated to an IX70 inverted microscope, as described. Argon laser lines were used to achieve three-channel immunofluorescence detection of Alexa488-,568- and 647-tagged goat or donkey secondary antibodies (each 1:400, Invitrogen). NG2 immuno-fluorescence was obtained following incubation with biotinylated secondary antibody (1:200, Jackson), and avidin-Alexa488 (1:500, Invitrogen). Three sections at least 350 mm apart were counted (Stereo Investigator, MicroBright Field, Williston,VT), representing rostral, middle, and caudal regions of the fimbria and corpus callosum white matter engraftment for each animal.
Homozygous shiverer × rag2 immunodeficient mice were bred in our colony. Within 1–2 days of birth, pups were cryoanesthetized for cell delivery. Donor cells (0.5 × 105) in 1 μl of HBSS were injected through a pulled glass pipette and inserted through the skull into the presumptive corpus callosum. Transplants were directed to the corpus callosum at a depth of 1.0–1.2 mm, depending on the weight of the pup, which varied from 1.0 to 1.5 g. Pups were killed at 8 and 12 weeks thereafter. To prevent rejection of xenografts, pups were injected daily with an immunosuppressant FK-506 (5 mg/kg, Tecoland Inc) after reaching 2 weeks of age.
Six fetal samples were FACS sorted for CD140a/PDGFRα (21–22 weeks gestational age) and positive and negative fractions collected for molecular analysis. Total RNA was extracted using RNeasy (Qiagen, Chatsworth, CZ) and amplified using ribo-SPIA based whole transcriptome based-amplification (NuGen). The expression of 47 cell type-specific marker genes was assessed using a 48 gene low-density Taqman-based array (TLDA, Applied Biosystems). The complete list of primer probes is given in Supplemental Table 3. The relative abundance of transcript expression was calculated by DDCt analysis, and the expression data normalized to GAPDH. Genes whose expression was not detected in the more than half of the RNA samples were excluded. Statistical analysis was performed on log2-transformed data and p-values were corrected for multiple testing using false-discovery rate 38.
Extracted total RNA was amplified using 3′-biased ribo-SPIA (NuGen Ovation) and hybridized onto Affymetrix U133+2 arrays according to manufacturer’s instructions (Affymetrix). Raw CEL data was processed using RMA 39 and downstream analysis performed using Bioconductor and R 40. Initial quality control was performed and included measurements of RNA degradation and signal distribution. One sample was removed from further analysis after identification as an outlier following principle component and hierarchical clustering analysis. Genes defined as specifically expressed by PDGFRα+ cells were greater than 3 fold expressed and significant using a moderated t-test statistic with 5% false discovery rate cut-off (q<0.05, n=5; linear modeling empirical Bayes test statistic)41. The differentially expressed genes were further filtered to removed genes highly expressed by human microglial cells (CD11b-defined, n=3), relative to adult A2B5-sorted OPCs 13. Gene ontology and Ingenuity Pathway Analysis were performed on the list of PDGFRα specific genes. Gene ontology over-representation was performed using NIH DAVID42. Significance p-values calculated in Ingenuity were based on a right-tailed Fisher’s Exact test to identify over-represented functional/pathway annotations, that is annotations which have more differentially expressed genes than expected by chance. Individual networks of interesting genes were selected according to relevance score and composition of gene constituents. Gene set enrichment analysis was performed using the PGSEA package 43 and significance assessed by fitting the relative enrichment of individual pathways to a linear model and using a moderated t-test statistic to assign significance. The resulting p-values were corrected for multiple testing correction using the false discovery rate. The complete microarray dataset is available at GEO under the accession no. GSE29368.
Supplemental Figure 1. CD140a sorted cells express markers of OPCs
The phenotype of CD140a-sorted cells was assessed using immunocytochemistry at 24hrs post-sort. At this stage, CD140a+/PDGFRα+ sorted cells uniformly expressed CD140a/PDGFRα immunoreactivity (A & B). At high power, CD140a+ expressing cells (C) co-expressed transcription factor markers of oligodendrocyte lineage, Olig2 (D) and neural progenitors, Sox2 (E). CD140a−depleted cells did not express CD140a immunoreactivity (F). Most CD140a/PDGFRa-depleted cells were neuronal lineage expressing βIII-tubulin and HuD (G). A smaller subset of depleted cells consisted of GFAP/AQP4 double-labeled astrocytes (H).
Scale = 10 μm.
Supplemental Figure 2. WNT target genes in CD140a+ cells To investigate the functional significance of wnt signaling pathway relative enrichment in CD140a-sorted cells, GSEA was performed using annotated wnt target genes (http://www.stanford.edu/~rnusse/wntwindow.html). P-GSEA indicated a significant enrichment of wnt target genes in CD140a+ cells (p=0.011). Among those, 15 genes were significantly differentially expressed by CD140a+ and CD140a− cells (>3 fold change, 5% FDR). This figure shows the resulting heat map of those genes plotted for each fetal sample (n=5).The majority 12 of 15 were significantly greater expressed in CD140a-sorted cells.
Supplemental Figure 3. Gene set enrichment analysis of CD140a sorted profiles
The profiles of CD140a+ cells were analyzed by parametric gene set enrichment using the KEGG database of cell-cell signaling pathways. Significant pathways were selected on the basis of a moderated t-test statistic and 5% FDR cutoff. The results are illustrated in decreasing order of significance in a Heatmap, whereby each sorted cell sample is represented by a single cell; relative enrichment in the CD140a+ fraction, relative to CD140a− cells, is indicated by darker red color, and depletion by blue (A). The top two most significantly enriched pathways were involved in glycan biosynthesis. B, shows a heatmap of expression data for those significantly regulated genes annotated in the KEGG ‘Glycan structures biosynthesis 1 pathway’.
Supplemental Figure 4. CD140a/CD9 double cytometry
Fetal dissociates were incubated with CD140a and CD9 antibodies. One such example of CD140a/CD9 two-color cytometry is shown in panel A. Fluorescence-minus controls utilizing isotype control antibodies were used to set positive sort gates for each antigen. CD140a/CD9 cytometry was performed on six separate fetal samples (18–20wk gestational age). Approximately half of CD9+ cells co-expressed CD140a, whereas approximately one third of CD140a+ cells expressed CD9. This represented a highly significant overlap and would translate to a 14.4 ± 2.3 fold enrichment of CD9 positive cells following CD140a sorting (p=0.0002). These results are consistent with the 7.4 fold higher expression of CD9 mRNA in CD140a cells compared to the depleted population.
Supplemental Table 1. Differentially expressed genes in CD140a+ cells. The microarray profiles of CD140a-sorted cells were compared directly to their depleted controls. CD140a enriched genes were defined as those genes greater than 3 fold higher expressed in sorted cells and significant following paired t-tests using a moderated t-test statistic and 5% false discovery rate cut-off. Using these criteria 408 genes were identified as expressed by CD140a+ OPCs. The mean relative expression and corrected p-value (q-value) are shown for each gene. For those genes with multiple probe sets, data is presented from the most significant probe set.
Supplemental Table 2. OPC-specific and cell-cell signaling genes are highly expressed by fetal CD140a-defined cells. The microarray profiles of CD140a-sorted cells were compared directly to their depleted controls. CD140a enriched genes were defined as those genes greater than 3 fold higher expressed in sorted cells and significant following paired t-tests using a moderated t-test statistic and 5% false discovery rate cut-off. Using these criteria 408 genes were identified as expressed by CD140a+ OPCs. Selected genes on the basis of cell type expression or signaling pathway membership are shown. The complete list is available as Supplemental Table 1. The mean relative expression and FDR corrected p-value (q-value) are shown for each gene. For those genes with multiple probe sets, data is presented from only the most significant probe set.
Supplemental Table 3. Taqman gene expression assays used in low density array. Real-time PCR was performed in parallel using low density Taqman arrays (TLDAs) from Applied Biosystems. The arrays were designed with 48 genes, utilizing GAPDH as an internal control. The table lists the assay IDs for each gene and NCBI gene ID.
This work was supported by NINDS R01NS039559, the Adelson Medical Research Foundation, the Mathers Charitable Foundation, the National Multiple Sclerosis Society, and the New York State Stem Cell Research Board (NYSTEM). We would like to thank Xiao-Jie Li for technical support, and Michelle Zanche, Andrew Cardillo and the URMC Functional Genomics core for Affymetrix and Taqman array support.
Author contributionsFJS directed both the in vitro experiments and genomics analysis; CM planned and performed the in vitro studies with FJS, and conducted the assessment of OPC ontogeny; SS performed the transplants of A2B5+ and CD140+ cells into shiverer mice; TP assisted FJS and CM in the in vitro studies and FACS analysis; MSW directed the comparative assessment of myelination by A2B5 and CD140 cells in vivo; SAG co-designed the experiments, co-analyzed the data with FJS and MW, and wrote the paper together with the co-authors.