Expression and function of CD105 during the onset of hematopoiesis from Flk1⁺ precursors

During vertebrate embryogenesis, the onset of hematopoiesis and vasculogenesis occurs in the extra-embryonic yolk sac with the formation of blood islands from aggregates of mesodermal precursors. Cells within these clusters differentiate into primitive erythrocytes while those at the periphery differentiate into endothelial cells. The temporal and spatial coupling in the appearance of hematopoietic and endothelial cells led to the hypothesis that these lineages are derived from a common progenitor.1 Recent studies have shown that the fms-like receptor tyrosine kinase Flk1 (also known as vascular endothelial growth factor receptor–2), which is expressed on subsets of mesoderm,2 is critical for the normal development of both hematopoietic and endothelial lineages. Flk1-deficient (flk1^−/−) mice die in utero at embryonic day 8.5 (E8.5) owing to defects in blood island and vasculature formation.3 Further investigations revealed that flk1^−/− embryonic stem (ES) cells failed to contribute to hematopoietic or endothelial cells in chimeric mice, although residual hematopoietic and endothelial activities were observed during differentiation in vitro.4-6 Moreover, in vitro clonal assays provided direct evidence that hematopoietic and endothelial cell lineages are derived from a common precursor2,7 that expresses Flk1.2 Thus, Flk1 expression may serve to identify the hemangioblast, a putative bipotent progenitor for the hematopoietic and endothelial lineages. The molecular events contributing to hematopoietic and endothelial development immediately following the Flk1⁺ stage remain to be elucidated.

Members of the transforming growth factor (TGF)–β superfamily exert multiple effects during development, including roles in mesoderm patterning and hematopoietic differentiation.8 CD105 (endoglin) is an accessory receptor for several members of the TGF-β superfamily.9,10 CD105 was first identified on human leukemic cells of the pre-B phenotype11 and was then shown to be transiently expressed on subsets of normal human hematopoietic lineages, such as proerythroblasts, macrophages, and fetal marrow early B cells.12-15 It is expressed on all types of vessels and is implicated in endothelial cell function.16CD105-deficient (eng^−/−) mice exhibit normal vasculogenesis but die in utero of impaired vascular and cardiac development, suggesting that CD105 is critical for angiogenesis.17-19 A role for CD105 in hematopoiesis is suggested by its expression on hematopoietic subsets and its potential involvement in signaling by members of the TGF-β superfamily. However, hematopoietic development beyond yolk sac erythropoiesis was not assessed in eng^−/− mice owing to embryonic lethality at E10.0 to E10.5.

Differentiation of ES cells in vitro provides a powerful model system to study hematopoietic development.20 The progression of events appears to parallel that of the developing embryo, and various hematopoietic lineages can be generated, including erythroid and myeloid cells.20-23 In particular, ES cells differentiated on the macrophage colony-stimulating factor–deficient bone marrow stromal cell line, OP9, are also able to generate lymphocytes, allowing for the characterization of myeloerythroid and lymphoid development within the same system.24-26 Therefore, we used this approach to elucidate the expression and function of CD105 during lymphohematopoietic development. We demonstrate that during ES cell differentiation in vitro, CD105 is coexpressed on Flk1⁺precursors with hematopoietic potential and, furthermore, that expression is maintained at intermediate levels on the earliest detectable CD45⁺ cells. These data suggest that CD105 may be a useful marker to further investigate early hematopoietic development from Flk1⁺ precursors. In addition, our findings suggest that CD105 plays an important functional role in hematopoietic differentiation from Flk1⁺ mesodermal cells, as we observed severely diminished myeloerythropoiesis in the absence of CD105. However, CD105 does not appear to play a prominent role in lymphopoiesis. Thus, our data implicate CD105 as a lineage-specific regulatory molecule during the onset of hematopoiesis from Flk1⁺ precursors.

OP9 cells were originally obtained from Dr T. Nakano (Osaka University, Japan). Experiments in Figures 1 to 3 were performed with R1 ES cells which were obtained from Dr G. Caruana (Mt Sinai Hospital, Toronto, ON, Canada). The eng^+/− ES cells (clone 4A-36) were previously generated17 by gene-targeting the parental wildtype 129/Ola-derived E14 ES cell lines, deleting 609 base pairs (bp), including eng exon 1 and its initiation codon, and leaving the endoglin promoter intact. These G418-gancyclovir–resistant clones were able to give germline transmission in mice.17 G418-resistant ES cells that had randomly integrated the gene-targeting construct were used as control (eng^+/+) cells, and results with these cells were similar to those obtained with E14 wild-type cells. Theeng^−/− ES cells were derived in vitro fromeng^+/− clone 4A-36 by selection with high-concentration G418 (2.4 mg/mL).27 Screening of resistant colonies for eng^−/− was performed by multiplex polymerase chain reaction (PCR) as previously described.17 Two independently derivedeng^−/− clones were used for ES-cell–differentiation experiments.

ES-cell/OP9–differentiation cocultures were performed as previously described.25 26 Briefly, 10⁴ ES cells were seeded onto OP9 monolayers in 6-well plates, or 5 × 10⁴ES cells were seeded onto OP9 monolayers in 10-cm dishes. After 5 or 6 days of coculture, cells were harvested and made into single-cell suspensions by vigorous pipetting and 0.25% trypsin treatment. Cells were then washed and reseeded onto new OP9 cell monolayers, and media were changed or cells were passaged (without trypsin) every 3 to 5 days.

Flk1⁺ cells were sorted on day 5 or 6, reseeded onto new OP9 cell monolayers, and harvested on various days for analysis by flow cytometry and/or reverse-transcriptase (RT)–PCR analyses. For Flk1 subset analysis shown in Figure 3C, 2.5 × 10⁴ cells were sorted on day 5, reseeded onto new OP9 cell monolayers, and analyzed on various days. The Flk1⁺CD105⁻CD31⁺ subset (Figure3B) represented a minor population that was not usually detectable and therefore not included in the progenitor analysis shown in Figure 3C. However, hematopoietic potential from this population was lower than that observed for the Flk1⁺CD105⁻CD31⁻ subset.

Preparations of samples and flow cytometry analysis were performed as previously described.28 Postsort analyses showed a purity of at least 95%. Conjugated antibodies for flow cytometry were purchased from Pharmingen (San Diego, CA), except for CD105 (clone MJ7/18), which was purchased from Pharmingen in purified form and biotinylated. For analysis, live cells were gated on the basis of forward and side scatter and lack of propidium iodide uptake.

A 1:4 serial dilution from 1.92 × 10⁴ to 3 × 10² sorted Flk1⁺ cells fromeng^+/+ and eng^−/−ES-cell/OP9 cocultures (day 5) were reseeded onto new OP9 cell monolayers in 24-well plates, with 6 replicate wells (n = 6) for each dilution. All the cells from each well were harvested and analyzed individually by flow cytometry at day 14 for cell surface expression of CD45 and TER119 (data not shown). The presence of CD45⁻TER119⁺ erythrocytes was scored and the progenitor frequency was determined by the method of maximum likelihood applied to the Poisson model.29 The χ² analyses of CD45⁻TER119⁺ erythrocytes indicated satisfactory conformance of the experimental data to the Poisson model: χ$[3]2$ = 1.598(eng^+/+) and χ$[3]2$ = 3.332(eng^−/−). Cocultures were observed under an inverted microscope and analyzed by flow cytometry at limiting dilution to determine approximate clone size.

Total RNA was isolated by means of the Trizol RNA isolation protocol (Gibco BRL, Gaithersburg, MD). Complementary DNA (cDNA) was prepared from each RNA sample with the use of random hexamer primers and the cDNA Cycle Kit (Invitrogen, San Diego, CA). All PCR reactions were performed with the same cDNA batches as shown for β-actin, and all PCR products corresponded to the expected molecular sizes. Gene-specific primers used for PCR are as follows (5′ → 3′): β-actin (500 bp) F-GAT GAC GAT ATC GCT GCG CTG, R-GTA CGA CCA GAG GCA TAC AGG; CD105 (640 bp) F-GGT GTT CCT GGT CCT CGT TT, R-CAA AGG AGG TGA CAA TGC TGG; T βR-II (858 bp) F-TCT TCT ACT GCT ACC GTG TCC A, R-CGT AAT CCT TCA CTT CTC CCA; ALK1 (662 bp) F-GAA CAC GGC TCC CTC TAT GA, R-ACT TTG GGC TTC TCT GGA TTG; brachyury (423 bp) F-TAC TCT TTC TTG CTG GAC TT, R-ATC TTT GTG GTC GTT TCT TT; β-globin (439 bp) F- CAC AAC CCC AGA AAC AGA CA, R-CCA CTC CAG CCA CCA CCT TC; ζ-globin (445 bp) F-ATG TGG GAG AAG ATG GCT GCT, R-CAA TAA AGG GGA GGA GAG GGA; tie-2 (728 bp) F-GTC CTT CCT ACC TGC TAC TT, R-TTC CAC TGT TTA CTT CAA TG. PCR reactions were performed as follows: 94°C for 2 minutes; 35 cycles of 30 seconds at 94°C, 30 seconds at optimal annealing temperatures (54°C to 59°C), and 60 seconds at 72°C; and a final extension at 72°C for 6 minutes. Products were separated by agarose gel electrophoresis on a 1.0% gel and visualized by ethidium bromide staining; reverse photo images are shown. Total RNA from OP9 cells and E13 fetal liver were used as controls for all primer sets.

ES cells were differentiated on OP9 cells (ES-cell/OP9 coculture) as previously described,25,26 and the generation of multiple hematopoietic cell lineages was analyzed by flow cytometry (Figure 1). A small fraction of cells expressing the receptor tyrosine phosphatase CD45 (leukocyte common antigen), which is present on all hematopoietic cells except mature erythrocytes,30,31 is detectable as early as day 5 of coculture.26 Some variation is observed in the temporal kinetics of the different hematopoietic lineages in independent ES-cell/OP9 cocultures. Typically, erythropoiesis in ES-cell/OP9 cocultures peaks between day 12 and day 14, and myelopoiesis peaks between day 12 and day 16. Although significant lymphoid populations can be observed by day 14, lymphopoiesis generally peaks after day 16, when myeloerythropoiesis subsides; thus, after day 19, the cocultures usually consist primarily of lymphocytes. Analysis by flow cytometry (representative of day-14 to day-16 cocultures) identified 4 distinct populations, labeled a-d (Figure 1A), as determined by cell surface expression of CD45 and CD24 (heat-stable antigen).32,33Figure 1B shows further analysis of these populations with lineage-specific markers as follows: CD45⁻CD24⁺ cells (Figure 1A, a) corresponded to TER119⁺ erythrocytes34 (Figure 1B, a); CD45^intCD24⁺ cells (Figure 1A, b) corresponded to CD45R (B220)⁺ B lymphocytes,35,36 which also coexpressed CD19 (Figure 1B, b; and data not shown); CD45⁺CD24⁺ cells (Figure 1A, c) corresponded to CD11b⁺ (Mac-1) myeloid cells37 (Figure 1B, c); and CD45^hi cells (Figure 1A, d) corresponded to NK lymphocytes characteristically lacking CD24 expression (Figure 1A, d) and expressing NK cell markers DX5 and CD90 (Thy1) (Figure 1B, d).28 38 

To study events during the onset of hematopoiesis, we sought to characterize the hematopoietic potential from Flk1⁺precursors. A transient wave of Flk1 expression was observed during the in vitro differentiation of ES cells into embryoid bodies (EBs).39 The majority of cells with hematopoietic potential were shown to be Flk1⁺ during the early stages of differentiation and Flk1⁻ at later stages.39ES cells differentiated on collagen IV–coated plates were able to give rise to Flk1⁺ hemangioblasts, some of which expressed the vascular endothelial cadherin, CD144.2 We assessed the temporal appearance of Flk1⁺ precursors during ES-cell/OP9 coculture and determined that Flk1 expression peaked between day 4 and day 6 (Figure 2), with subsets of cells coexpressing CD144. We further determined that hematopoietic potential was predominantly contained within the Flk1⁺ fractions at day 5 and day 6 (Figure 3 and data not shown). These findings are consistent with previous reports in the EB differentiation system,39 demonstrating the transient nature of Flk1⁺ expression by a population of cells containing the earliest hematopoietic precursors.

In order to further define the population of Flk1⁺precursors with hematopoietic potential, we characterized the Flk1⁺ subset on the basis of the expression of CD105 and CD31 (platelet endothelial cell adhesion molecule 1), which have been reported to be present on subsets of hematopoietic cells, including early progenitors.12,15,40-42 Flow cytometric analysis of day-5 ES-cell/OP9 cocultures revealed that CD105 and CD31 expression subdivided the Flk1⁺ fraction into discrete populations (Figure 3B). OP9 cells did not express any of the markers indicated in Figure 3 (data not shown). An equal number of cells (2.5 × 10⁴) from each of these subsets was isolated by flow cytometric cell sorting at day 5, reseeded onto OP9 cells, and analyzed by flow cytometry for hematopoietic activity on various days, with the initial seeding of ES cells designated as day 0. Analysis for the surface expression of CD45 (Figure 3C) and TER119 (data not shown) revealed that hematopoietic potential was largely contained within the Flk1⁺ fractions (Figure 3C, bottom 3 rows), compared with residual levels within the Flk1⁻ fraction (Figure 3C, top row). Notably, CD105⁺ subsets accounted for the majority of hematopoietic potential within the Flk1⁺ fraction (Figure3C, rows 2 and 3), compared with CD105⁻ subsets (Figure3C, row 4 and data not shown; see “Materials and methods”). In contrast, similar levels of hematopoietic activity were observed in CD31⁺ and CD31⁻ cocultures (Figure 3C, rows 2 and 3). It was previously suggested by Kabrun et al39 that Flk1 expression defines early hematopoietic precursors that could represent the onset of embryonic hematopoiesis. Our data support this notion. In addition, our findings that CD105⁺ cells accounted for the majority of hematopoietic potential within Flk1⁺ fractions suggest that early hematopoietic precursors coexpress Flk1 and CD105. Moreover, induction of CD105 expression was observed in cells that had been sorted CD105⁻ (Figure 3C, fourth row, days 6 and 8), and at day 6 a population of cells expressed CD105, prior to the detection of CD45⁺hematopoietic cells at day 8. Furthermore, CD45⁺ cells did not coexpress Flk1 (data not shown), and the majority of CD45⁺ cells at day 6 and day 8 were CD31⁻(data not shown), but intermediate levels of CD105 expression were maintained on emerging CD45⁺ cells (Figure 3C, day 8). We cannot exclude the possibility that some CD45⁺ cells observed at day 11 were generated directly from CD105⁻precursors. Nonetheless, taken together, our data suggest that CD105 should serve as a useful marker to further dissect events during the progression of developmental stages from Flk1⁺CD45⁻ to Flk1⁻CD45⁺ cells. Expression of CD105 on emerging hematopoietic cells appeared to be transient, as coexpression on some CD45⁺ cells was diminished after day 8 (Figure 3C). Interestingly, expansion of CD45⁺CD105⁻ cells by day 11 (Figure 3C) corresponds to the approximate time when lineage-specific differentiation is observed in ES-cell/OP9 cocultures. Thus, the developmentally regulated expression of CD105 on CD45⁺ cells may serve to identify the earliest hematopoietic cells, and further suggests that CD105 may play an important role during the onset of hematopoiesis.

The expression of CD105 during ES-cell/OP9 coculture was confirmed by RT-PCR (Figure 4). Consistent with the flow cytometric analysis, CD105 transcripts were present at day 5 of coculture and diminished by day 12. In addition, we observed the expression of type II TGF-β receptor (TβR-II) transcripts, indicating the presence of this TGF-β receptor that can interact with CD105.10 CD105 and TβR-II messenger RNAs were not expressed in OP9 cells but were present at low levels in E15 fetal liver cells (Figure 4). Since CD105 and ALK1 (a type-I TGF-β receptor) are mutated in hereditary hemorrhagic telangiectasia type I and type 2, respectively,17 43 expression of ALK1 was examined. Figure 4 shows that ALK1 was expressed in OP9 cells and in ES-cell/OP9 cocultures at days 5, 8, and 12, but not in E15 fetal liver cells. Control RT-PCRs were performed in the absence of cDNA templates (dH₂O), and with the use of β-actin primers.

To determine whether CD105 is functionally important for hematopoietic development, we assessed the differentiation potential ofeng^−/− ES cells in vitro. Theeng^−/− ES cells were generated from heterozygous eng^+/− ES cells (clone 4A-36; Bourdeau et al17) following selection in high-concentration G418, and confirmed by multiplex PCR as previously described17 (Figure 5A). Figure 5B shows that eng^−/− andeng^+/+ ES cells were comparable in their ability to differentiate into Flk1⁺ precursors after 6 days of ES-cell/OP9 coculture. An equal number of Flk1⁺CD45⁻ cells (7 × 10³) fromeng^−/− and eng^+/+cocultures were sorted and reseeded onto new OP9 cells for flow cytometric and RT-PCR analyses on various days. Figure 5C shows the results of RT-PCR analysis from Flk1⁺CD45⁻cells directly sorted at day 6, or after coculture for an additional 3 days (coculture day 9). Analysis ofeng^−/−, eng^+/−, andeng^+/+ cocultures revealed similar expression levels of brachyury, a mesoderm-specific transcription factor,44,45 and tie-2, a receptor tyrosine kinase associated with endothelial cell differentiation and reportedly expressed in fetal liver hematopoietic stem cells (HSCs).46-49 Thus, expression analysis of Flk1,tie-2, and brachyury suggests that the early differentiation potential of eng^−/− ES cells is normal. In contrast, hematopoietic differentiation appears to be impaired (Figure 5D). Flow cytometric analysis at day 9 revealed that CD45⁺ hematopoietic cells were severely diminished ineng^−/− as compared witheng^+/+ cocultures (Figure 5D); this was observed in 4 independent experiments. However, the presence of a small fraction of CD105⁻CD45⁺ cells ineng^−/− cocultures indicated that CD105 function, albeit important, was not absolutely required for hematopoietic commitment and further differentiation. Although we previously determined that hematopoietic activity is predominantly contained in Flk1⁺ fractions from day-5 and day-6 ES-cell/OP9 cocultures (Figure 3 and data not shown), we addressed the possibility that hematopoietic potential could be shifted to the Flk1⁻ fraction in the absence of CD105. Consistent with previous observations, the Flk1⁻ sorted fraction contained minimal hematopoietic activity in eng^−/−cocultures (data not shown).

To assess the hematopoietic precursor potential ofeng^−/− ES cells, Flk1⁺ cells were sorted from day-5 or day-6 ES-cell/OP9 cocultures, reseeded onto OP9 cells, and analyzed by flow cytometry on various days for cell surface expression of erythroid and myeloid lineage markers. Figure6A shows that erythroid and myeloid cells were efficiently generated from sorted Flk1⁺ precursors derived from control (eng^+/+) cocultures. In contrast, Flk1⁺ precursors fromeng^−/− cocultures exhibited severely diminished myeloerythroid potential (Figure 6A). Althougheng^−/− ES cells could differentiate into erythroid (CD45⁻TER119⁺) and myeloid (CD45⁺CD11b⁺) cells, erythropoiesis was diminished by approximately 15-fold (Figure 6B; day 12), and myelopoiesis by 5- to 8-fold (Figure 6B; days 9 and 12) ineng^−/− cocultures, as compared witheng^+/+ cocultures. Similar results were observed with the use of 2 different eng^−/− ES cell clones (Figure 6B-C), and in 4 independent experiments. Theeng^+/− ES cells were not impaired in their ability to generate erythroid and myeloid cells (Figure 6C). The elevated cell numbers for eng^+/− ES cells in Figure 6C are not deemed to be significant as this was not consistently observed. Erythropoiesis and myelopoiesis ineng^−/− cocultures, albeit at much reduced levels, followed the same time course and duration as foreng^+/+ and eng^+/−control cocultures (Figure 6C). This suggests that the temporal kinetics of hematopoietic differentiation were not altered by the targeted deletion of the eng gene.

Our findings indicated that sorted Flk1⁺ precursors from day-5 and day-6 ES-cell/OP9 cocultures typically gave rise to TER119⁺ erythrocytes by day 12 (Figure 6). Nakano et al22 previously reported that 2 waves of erythropoiesis are observed in ES-cell/OP9 cocultures, similar to erythropoiesis during mouse ontogeny, with the first transient wave generating primitive erythrocytes expressing ζ-globin at day 6, and the second wave generating definitive erythrocytes expressing adult β-globin beginning at day 10. Consistent with the observations of Nakano et al,22,ζ-globin transcripts were not detected in any day 12 cocultures derived from sorted Flk1⁺precursors (Figure 6D). To determine whether definitive erythropoiesis was affected by the absence of CD105, we examined the expression of β-globin from day 12 cocultures. RT-PCR analysis clearly indicated that β-globin expression was severely diminished ineng^−/− cocultures, as compared witheng^+/+ cocultures (Figure 6D). This finding, taken together with our flow cytometric analysis, demonstrates that definitive erythropoiesis is impaired in the absence of CD105. Two groups observed significant levels of erythrocytes in the yolk sac, suggesting that primitive erythropoiesis occurred efficiently.17,18 However, another group19reported severe anemia in eng^−/− yolk sacs. As Flk1⁺ precursors are isolated from day-5 and day-6 cocultures, a comparison of primitive erythropoiesis ineng^−/− and eng^+/+ES-cell/OP9 cocultures would be difficult to interpret owing to the fact that primitive erythropoiesis occurs concomitantly with, or soon after, the reseeding of Flk1⁺ cells onto OP9 cells. Thus, the extent to which primitive erythropoiesis is also dependent on CD105 function remains unclear.

Potential differences in erythroid progenitor frequency and colony size were determined by limiting dilution analysis ineng^−/− and eng^+/+cocultures. Day-5 sorted Flk1⁺ cells were titrated by serial dilution, reseeded onto OP9 cells, and analyzed by flow cytometry at day 14. Progenitor frequency was estimated by the statistical method of maximum likelihood29 (applied to the Poisson model) from the analysis of individual cocultures that were scored for the presence of CD45⁻TER119⁺erythrocytes. From this analysis, erythroid progenitor frequency from Flk1⁺ precursors was estimated to be approximately 16-fold lower in eng^−/− cocultures, with a frequency of 1 in 7843 (95% confidence limits [CLs], [3711-16 579]) as compared with eng^+/+, which had a frequency of 1 in 463 (95% CLs, [205-1043]). This difference was statistically significant (P < .025) and is consistent with the data (Figure 6) showing severe erythropoietic defects ineng^−/− cocultures. Flow cytometric analysis of positive cocultures at limiting dilution and examination under a microscope revealed no obvious differences in colony size between colonies from eng^−/− andeng^+/+ cocultures.

We previously reported that efficient lymphopoiesis occurs in ES-cell/OP9 cocultures (Figure 1).26 However, B lymphocytes (CD45^intCD19⁺) were not consistently generated from eng^−/− ES cell–derived Flk1⁺ precursors. In these experiments, low numbers of sorted Flk1⁺ cells (7 to 8 × 10³) were seeded per well onto OP9 cells. At this number of input cells, even Flk1⁺ precursors derived from control ES cells failed to give rise to B cells in a consistent manner. Therefore, we considered that B lymphopoiesis might be inefficient owing to low progenitor frequency in the Flk1⁺ subset. Thus, we performed separate experiments in which 8 × 10⁴ sorted Flk1⁺ cells were seeded per well. This approach revealed that B lymphopoiesis (CD45^intCD19⁺CD11b⁻) (Figure7) and NK lymphopoiesis (CD45^hiCD19⁻CD11b⁻) (Figure 7) were similar in eng^−/− and controleng^+/− cocultures compared with severe defects observed in myeloerythropoiesis (CD45⁺CD19⁻CD11b⁺) (Figure 7; and data not shown) that were still evident ineng^−/− cocultures, as in previous experiments (Figure 6). However, data from Figure 7 and limiting dilution analysis indicate that a possible mild defect may be exhibited in lymphopoiesis from eng^−/− ES cells (Figure 7, day 19; S.K.C. et al, unpublished observations, November 2000). The extent to which lymphopoiesis may be dependent on CD105 function remains to be determined. Nonetheless, these data suggest that lymphoid and nonlymphoid hematopoiesis may be distinguishable on the basis of their developmental requirement for CD105 in that myeloerythropoiesis is strongly dependent on CD105 function.

Early hematopoietic and endothelial precursors, which may include hemangioblasts, have been reported to express Flk1.2,7,39,50 Our results show that CD105 is coexpressed on Flk1⁺ early hematopoietic precursors, and CD105 expression can be induced on Flk1⁺CD105⁻cells. The Flk1⁺CD105⁺ population of early hematopoietic precursors may also include committed endothelial precursors and hemangiogenic cells. Indeed, CD105 has been reported to be expressed on cells of both hematopoietic and endothelial lineages.12-16 Nonetheless, we show that CD105 expression is maintained on CD45⁺ cells that have committed to the hematopoietic lineage. Furthermore, our data revealed that CD105 is not required for the differentiation of CD45⁺ cells but rather appears to play an important role in early myeloerythropoietic progenitors, suggesting that CD105 plays a role in hematopoietic development after specification to the hematopoietic lineage. The initial reports that examined endothelial development in CD105-deficient mice showed that while vasculogenesis appeared normal, subsequent stages in angiogenesis appeared to be defective.17-19 Taken together, these data support the notion that CD105 functions after differentiation is specified to either the hematopoietic or the endothelial lineage. Moreover, Flk1⁺ mesodermal precursors were efficiently generated fromeng^−/− ES cells. This finding further suggests that CD105 could play a role in hematopoietic development following the Flk1⁺ stage. This is supported by the more severe and earlier defects in Flk1-deficient mice compared with CD105-deficient mice.3,4 17-19 However, it remains to be determined whether CD105 function is dependent on Flk1 signaling in a common differentiation pathway.

The most striking phenotype we observed in CD105-deficient ES cells was the profound reduction in myeloid and erythroid cells, which suggests that the survival, self-renewal, or proliferation of a common myeloerythroid progenitor may be strongly dependent on CD105 function. Normally, the microenvironment created by OP9 stromal cells allows for the efficient differentiation of erythroid, myeloid, and lymphoid lineages.26 Lymphopoiesis did not appear to be significantly altered in eng^−/− cells. However, in the absence of CD105, it appears that either inhibitory cues were not being sufficiently antagonized, or stimulatory cues were not being sufficiently amplified during myeloerythropoiesis. In addition, the balance between differentiation and self-renewal signals may have been dysregulated. Our observations are consistent with a role for CD105 in regulating differentiation at early stages of myeloerythropoiesis, rather than later stages. First, CD105 expression was highest at day 8 and was down-regulated by day 12 in normal ES cells, suggesting that defects due to the absence of CD105 would have a direct impact on early myeloid and erythroid progenitors. Thus, in the absence of CD105, it appears that a smaller pool of myeloerythroid precursors is generated, which is, however, able to differentiate normally. This is supported by the observations thateng^−/− cells exhibited a lower progenitor frequency but displayed kinetics and colony size similar to control cells.

The absence of CD105 appears to dampen early hematopoietic differentiation from ES cells, but other factors probably determine the extent of this effect. CD105 is an accessory receptor for members of the TGF-β superfamily and can bind TGF-β1, TGF-β3, activin-A, bone morphogenetic protein (BMP)–2, and BMP-7 in complex with their cognate receptors.9,10 When present in the receptor complex, CD105 can modulate cellular responses to TGF-β1 and is capable of acting as an antagonist of inhibitory and stimulatory signals.51-53 However, it can also potentiate effects of TGF-β1 and is therefore best described as a regulatory component of the receptor complex.54,55 The phenotype ofeng^−/− embryos is reminiscent of that observed for TGF-β1^−/− and TGF-β receptor II^−/−embryos56,57 and therefore suggests an important role for CD105 in conjunction with this growth factor and its receptor complex in angiogenesis and hematopoiesis. In general, TGF-β1 exerts a negative control on the cell cycle of primitive murine hematopoietic cells and shows a preferential growth inhibitory effect on early progenitors.56,58,59 For example, TGF-β1 was shown to inhibit the expression of stem cell factor and its receptor CD117 (c-kit)60 and was also shown to be a major regulator of erythropoiesis, inhibiting early stages but stimulating later stages.58 Interestingly, Pierelli et al40reported that CD105 is expressed on primitive HSCs and suggested that autocrine TGF-β1 helps to maintain the resting state and self-renewal capacity of these cells. Other members of the TGF-β superfamily, such as activin-A, BMP-2, BMP-4, and BMP-7, have also been implicated as important regulators of mesodermal specification to the hematopoietic fate, or of early hematopoietic differentiation.58,61-64Mechanistically, TGF-β signals are transduced by Smad proteins8 that can synergize with components of the Wnt signaling cascade65,66 which have also been shown to regulate hematopoietic differentiation.67-70 Thus, although the precise mechanism remains to be determined, it is likely that CD105 functions as a regulator of microenvironmental cues delivered through TGF-β cytokines.

Our studies have identified a potential role for CD105 during the onset of hematopoiesis from Flk1⁺ precursors. With the multiple effects exerted by members of the TGF-β superfamily throughout development,8 71 and the potential ability of CD105 to regulate responses to several of these factors, it will be important to determine which pathways are regulated in specific lineages during hematopoietic development.

We thank Dr Norman Iscove for helpful discussion and Cheryl Smith for technical assistance with cell sorting. We would also like to thank Dr Daniel J. Dumont for advice regarding the derivation of CD105-deficient ES cells.