Fetal and Adult Hematopoietic Stem Cells Give Rise to Distinct T Cell Lineages in Humans

See allHide authors and affiliations

Science  17 Dec 2010:
Vol. 330, Issue 6011, pp. 1695-1699
DOI: 10.1126/science.1196509

This article has a correction. Please see:


Although the mammalian immune system is generally thought to develop in a linear fashion, findings in avian and murine species argue instead for the developmentally ordered appearance (or “layering”) of distinct hematopoietic stem cells (HSCs) that give rise to distinct lymphocyte lineages at different stages of development. Here we provide evidence of an analogous layered immune system in humans. Our results suggest that fetal and adult T cells are distinct populations that arise from different populations of HSCs that are present at different stages of development. We also provide evidence that the fetal T cell lineage is biased toward immune tolerance. These observations offer a mechanistic explanation for the tolerogenic properties of the developing fetus and for variable degrees of immune responsiveness at birth.

The developing fetal immune system is generally believed to induce immune tolerance after exposure to foreign antigens (1). In the mouse, this tolerogenic tendency has been attributed to the absence of a “mature” adaptive immune system before birth (2). In the human, however, fetal exposure to foreign antigens, notably maternal alloantigens, can lead to the generation of immune tolerance (35), even though the immune system develops at a substantially earlier developmental stage (1, 6). Recently, we reported that tolerance induction in the human fetus is in part mediated by an abundant population of fetal regulatory T cells (Tregs) (7), a cell population containing a significantly greater percentage (~15%) of total peripheral CD4+ T cells in the developing human fetus than is found in healthy infants and adults (~5%) (8). Unlike adult T cells, we also observed that fetal T cells exhibit enhanced proliferation after exposure to alloantigens and are poised to become Tregs upon stimulation, a process dependent on transforming growth factor–β (TGF-β) (5). Given these observations, we hypothesized that the human fetal T cell compartment may not simply be an “immature” version of the adult T cell compartment, but instead one derived from a separate lineage of T cells that is poised to deliver a tolerogenic response to all antigens encountered in utero.

Although the human fetal T cell compartment begins to develop at approximately 10 gestational weeks (g.w.) (9), much of what we know about it arises from studies of cord blood obtained at birth. A few reports indicate that the majority of fetal T cells found at mid-gestation (~16 to 24 g.w.) have a surface phenotype comparable to that of conventional naïve T cells found in neonates and in adults (1012). To characterize such cells more completely, we analyzed 18- to 22-g.w. fetal CD4+ T cells obtained from mesenteric lymph nodes (mLNs) for the expression of a panel of known surface antigens specific for naïve CD4+ T cells in adults and found that many have a phenotype (CD45RA+CCR7+CD95CD25) similar to that of naïve adult CD4+ T cells (Fig. 1A). Next, we assessed the proliferative response of sort-purified fetal and adult naïve T cells to stimulation with allogeneic peripheral blood mononuclear cells (PBMCs) in a primary mixed lymphocyte reaction (MLR). Fetal naïve CD4+ T cells were much more highly responsive to stimulation with allogeneic cells: After 6 days of stimulation, more than 50% had divided as compared to only ~10% of adult naïve CD4+ T cells. Activated fetal naïve CD4+ T cells were also more likely to adopt a Treg fate, as measured by up-regulation of Foxp3 and CD25 (Fig. 1B). Although Foxp3 and CD25 can be transiently expressed by some T cells as a consequence of activation, our previous results indicate that activated fetal T cells exhibit sustained expression of Foxp3 and, unlike adult T cells, are prone to up-regulate Foxp3 even as a result of “spontaneous” activation in tissue culture (5, 7). We have also demonstrated that fetal Tregs are capable of suppressing both proliferation and cytokine production, suggesting that their function is similar to that of adult Tregs (5, 7).

Fig. 1

Fetal naïve CD4+ T cells display functional differences as compared to adult naïve CD4+ T cells. (A) Flow cytometric analysis of the phenotype of naïve CD4+ T cells isolated from fetal mLNs (18 to 22 g.w.) and adult PBMCs (25 to 35 years old). Top panels depict initial gating on CD3CD4+ T cells, showing CD45RA versus CCR7 staining, and bottom panels depict initial gating on CD45RA+CCR7+ cells (hightlighted in black in top panels), showing CD95 versus CD25 staining. Cells considered to be “naïve” CD4+ T cells were CD45RA+CCR7+CD25CD95 (highlighted in black in bottom panels). Data are representative of at least three independent donors. (B) Sorted naïve CD4+ T cells were either left unstimulated or cultured for 6 days in the presence of irradiated allogeneic PBMCs. Proliferation of fetal and adult naïve CD4+ T cells was measured by carboxyfluorescein diacetate succinimidyl ester (CFSE) dilution and analyzed by flow cytometry. Tregs were phenotypically identified by up-regulation of CD25 and Foxp3.

These findings suggested that fetal and adult naïve CD4+ T cells may be phenotypically similar but functionally distinct. To better define the differences between the two populations, we performed gene expression analysis by microarray on sorted populations of fetal and adult naïve CD4+ T cells (fig. S1) and found substantial differences (Fig. 2, A and B, and table S1). A parallel analysis revealed that many of the same differences that were identified for fetal and adult naïve CD4+ T cells were similar to those found for fetal and adult Tregs (Fig. 2, A to C, and table S2). We also noted a substantial overlap in genes that were abundant in both fetal and adult Tregs, when compared to fetal or adult naïve T cells, that was consistent with Treg genes identified in previous studies (13, 14) (Fig. 2D and tables S3 and S4). To understand the relative contributions of age and functional phenotype, principal components analysis (PCA) of the different CD4+ T cells was performed. Age (adult versus fetal origin) was the largest PC, accounting for 36.1% of the observed variance, whereas a second PC (functional subtype, naïve versus Treg) accounted for 25.9% (fig. S2). The consistent observation of differential gene expression in fetal and adult T cells (both naïve and Treg) suggests that they represent different lymphocyte populations. As is individually highlighted in each scatterplot (Fig. 2, B and C), this point is illustrated by selected genes that were found to be highly expressed within fetal or adult T cell populations (fig. S3).

Fig. 2

Fetal and adult naïve CD4+ T cells show significant differences in gene signature. (A) Unbiased cluster analysis on the basis of gene expression for fetal (mLN) and adult (PBMC) naïve CD4+ T cells and CD4+CD25+ Tregs. (B to D) Scatterplots of pairwise global gene expression comparisons comparing (B) fetal naïve CD4+ T cells and adult naïve CD4+ T cells, (C) fetal and adult CD4+CD25+ Tregs, and (D) CD4+CD25+ Tregs and naïve CD4+ T cells (both fetal and adult). Gene expression values are plotted on a log scale. Genes that were differentially expressed between groups (determined using a 5% false discovery rate and a fold change ≥2) are indicated in red and blue. Specific genes that were differentially expressed based on age or subset are highlighted in green.

We hypothesized that adult and fetal human T cells are different because they are derived from distinct populations of multilineage hematopoietic stem cells (HSCs). Thus, previous work in avian and murine models has demonstrated that fetal and adult HSCs differ in phenotype (15), tissue localization (16), functional properties (for example, regarding turnover) (15, 17), and developmental potential (1820). For example, different populations of HSCs seed the murine and avian thymus during fetal development (20, 21), giving rise to successive waves of thymocytes. In the mouse, the earliest to be found are dendritic epidermal T cells (DETCs), identified by their expression of a specific γδ T cell receptor (TCR) (Vγ3Vδ1) (18, 20). These cells are replaced by a second wave of thymocytes with a diverse array of both TCRγδ and TCRαβ cells, which are then replaced by a third wave of thymocytes at around the time of birth, in which αβ T cells are the primary population.

In similar fashion, human fetal αβ CD4+ T cells may differ from their adult counterparts because they are derived from distinct populations of HSCs. During the course of fetal development in humans, the location of the HSC pool switches from the aorta gonad mesonephros region to the fetal liver and, finally, to the fetal bone marrow (BM), where the majority of HSCs are thought to reside throughout adulthood (22). The first wave of thymocytes is derived primarily from HSCs that are present in the fetal liver, whereas thymocytes present at later stages of fetal development and in adults are likely to develop predominantly from BM-derived HSCs (22). Thus, besides intrinsic differences between fetal and adult HSC populations, there are also differences in tissue distribution that could contribute to different developmental potential (23).

To determine whether distinct HSC populations give rise to distinct lineages of T cells in the human fetus, we isolated total CD34+ hematopoietic stem/progenitor cells (HSPCs) (including primitive CD34+CD38 HSCs and CD34+CD38+ progenitors with high proliferative potential) from fetal liver (at 18 to 22 g.w.), fetal BM (18 to 22 g.w.), and adult BM (19 to 43 years old) (fig. S4). We used the SCID-hu Thy/Liv mouse model of thymopoiesis to allow these populations of HSPCs to mature into single positive (SP) thymocytes in vivo. SCID-hu Thy/Liv mice were generated by engraftment of fetal thymus and liver from an HLA-A2 donor into the immunodeficient C.B17 scid/scid mouse (24). The animals were irradiated and then injected with 500,000 HSPCs isolated from HLA-A2–positive fetal liver, fetal BM, adult BM, or vehicle only (n = 5 mice per group) (fig. S4). After 7 to 8 weeks, mature thymocytes were harvested from the Thy/Liv implant. At this time, we were able to identify donor-derived (HLA-A2+) mature CD3+ thymocytes [CD3+CD4+CD8+ (DP), CD3+CD4-CD8+ (SP8), and CD3+CD4+CD8 (SP4)] in each of the groups receiving CD34+ HSPCs but not in animals receiving vehicle alone (figs. S4 and S5).

We asked whether fetal and adult HSPCs might differ in their capacity to generate Tregs in the thymus, because fetal T cells appear to be predisposed to generating Foxp3+ Tregs upon stimulation in vitro, and there is an abundance of Tregs in the peripheral tissues of mid-gestation fetuses (7, 8). Both fetal liver and fetal BM-derived HSPCs were found to generate populations of CD25+Foxp3+ SP4 thymocytes, whereas adult BM-derived HSPCs generated a significantly lower frequency of CD25+Foxp3+ SP4 thymocytes (Fig. 3, A to C, and fig. S6). Thus, it appears that fetal HSPCs display a heightened capacity to generate Tregs during thymic maturation.

Fig. 3

Fetal and adult HSPCs give rise to mature SP4 thymocytes that show similar functional differences to fetal and adult peripheral naïve CD4+ T cells. (A) Representative flow cytometry graphs showing thymocyte populations isolated from SCID-hu Thy/Liv implants after transplantation of donor (HLA-A2+) CD34+ HSPCs isolated from fetal liver, fetal BM, or adult BM. Panels on the left show initial gating on CD4 versus CD8 to select for SP4 thymocytes, and panels on the right depict subsequent gating on donor-derived (HLA-A2+) CD3+ mature SP4 thymocytes. (B) Flow cytometric analysis of the frequencies of CD25+Foxp3+ Tregs dervied from fetal liver, fetal BM, or adult BM hematopoietic progenitors. (C) Quantification of CD25+Foxp3+ Treg frequencies observed in three to five different thymic implants from each group. (D) Flow cytometric analysis of proliferation and Foxp3 expression by SP4 thymocytes derived from fetal liver, fetal BM, or adult BM HSPCs in response to stimulation with irradiated allogeneic PBMCs (7-day stimulation). (E) Quantification of the frequencies of CD25+Foxp3+ Tregs after stimulation with allogeneic PBMCs. Data show at least three separate implants for each group. Data in bar graphs represent mean ±SD, and statistical significance was measured by Student’s t test.

The initial observation that prompted our comparison between fetal and adult T cells was the elevated proliferation and differentiation of fetal T cells into Tregs after stimulation. To determine whether SP4 thymocytes derived from fetal HSPCs were functionally similar to SP4 thymocytes derived from adult HSPCs, we stimulated whole thymocyte populations in vitro with allogeneic antigen-presenting cells in a primary MLR (at a 1:3 ratio) and thereafter monitored proliferation and Foxp3 expression. SP4 thymocytes derived from fetal HSPCs behaved similarly to peripheral fetal naïve CD4+ T cells, exhibiting robust proliferative responses and a significantly greater expansion of Foxp3+ cells after stimulation as compared to SP4 thymocytes derived from adult HSPCs (Fig. 3, D and E, compare to Fig. 1B). Thus, we conclude that the functional differences observed between fetal and adult naïve T cells in the periphery can be, in part, accounted for by differences in the progenitor populations from which they arise.

We next sought to determine whether the gene signature of T cells derived from fetal HSPCs was different than that of T cells derived from adult HSPCs. To make this comparison, HLA-A2+ SP4 thymocytes were sorted from SCID-hu Thy/Liv implants injected with fetal liver, fetal BM, or adult BM HSPCs, and microarray analysis was performed using the same platform previously used to test peripheral T cells. This analysis revealed that fetal BM- and fetal liver-derived HSPCs gave rise to SP4 thymocytes that were indistinguishable from one another, indicating that the fetal HSPCs in each of these tissues have similar developmental potential (Fig. 4, A and B, and table S5). In contrast, the transcriptional profile of SP4 thymocytes matured from adult BM-derived HSPCs was vastly different (Fig. 4, A and B, and table S5). Thus, we find that fetal HSPCs and adult HSPCs give rise to mature SP4 thymocytes with profoundly different functional properties and gene expression patterns. These differences are likely to reflect developmental changes in progenitor populations, rather than differences related to compartmentalization of stem cell subsets.

Fig. 4

Fetal and adult HSPCs give rise to SP4 thymocytes with distinct gene signatures. (A) Unbiased cluster analysis based on gene expression for SP4 thymocytes from fetal BM, fetal liver, and adult BM hematopoietic progenitors.(B) Total number of genes found to be significantly different, comparing SP4 thymocytes derived from fetal and adult progenitor populations. (C) PCA showing clear separations between SP4 thymocyte populations and peripheral T cells (PC1) as well as differences between fetal and adult SP4 and peripheral naïve CD4+ T cell populations (PC2). (D) Scatterplot depicting genes that are differentially expressed between fetal and adult SP4 thymocytes and between fetal and adult peripheral naïve CD4+ T cells. (E) Scatterplot depicting genes that are differentially expressed between adult and fetal naïve T cells and between adult naïve and adult Tregs. For both (D) and (E), differential expression was determined using a false discovery rate of ≤5% and a fold change ≥2. The significance of overlap between groups of genes was determined using a chi-squared test for independence.

Finally, we sought to determine whether the differences observed in the gene expression patterns of peripheral fetal and adult naïve CD4+ T cells could be explained by differences in the fetal and adult stem cell populations. We performed PCA on both peripheral T cells and HSPC-derived thymocytes to identify the sources of variation in gene expression and their relative magnitude. In this analysis, developmental stage (thymocytes versus mature T cells) was found to be the most prominent source of variance (PC1, 70.4% of total variance), followed by age (PC2, fetal versus adult, 6.3%) (Fig. 4C). We observed a significant overlap (P = 1.0 × 10−8) between genes that were abundant in both fetal naïve and fetal HSPC-derived SP4 cells (253 probe sets) and genes that were abundant in adult naïve and adult HSPC-derived SP4 cells (79 probe sets) (Fig. 4D). Because fetal naïve cells displayed a propensity to differentiate into Tregs as compared to adult naïve T cells, we performed an additional correlation to determine whether fetal and adult naïve T cells shared a common gene signature with Tregs. We identified a significant number of genes that were highly expressed in both adult Tregs and fetal naïve T cells, but not in adult naïve T cells, some of which have been previously implicated as playing a specific role in determining Treg fate specification (such as IKZ2 and IKZ4) (Fig. 4E and fig. S3). This is also reflected in the PCA performed on all T cell subsets (Fig. 4C), which showed that fetal naïve cells are more closely related to adult Tregs than to adult naïve T cells along PC2. We can conclude that fetal T cells are a separate population of cells, distinct from those found in adults, and that this in part reflects differences in the populations of stem cells, which give rise to the T cell compartment across different stages of development.

Our results support the hypothesis that, as in mice and birds (21, 25), in humans the development of the immune system is layered. This hypothesis suggests that hematopoiesis occurs in distinct waves, including an early one that is fetal and a later one that is adult, with each generating distinct populations of cells that may coexist (layer) for a period of time (fig. S7). The propensity for fetal CD4+ T cells to adopt a Treg fate upon stimulation suggests that the initial wave of T cell differentiation may favor a population of T cells whose role is to promote tolerance to self (and potentially foreign) antigens encountered during development. Whether these cells are enriched in a specific Treg precursor, which has been suggested to exist in the adult thymus and periphery in mice (26, 27), remains to be determined. Although the focus of this study has been CD4+ T cells, the fetal peripheral lymphoid tissues contain a full array of hematopoietic cells that have not been extensively studied and may also be distinct from their adult counterparts. For example, although our findings are different from those made in the mouse, the results of our study do not rule out the existence of a comparable population of γδ DETCs in the human fetus. Our findings have broad implications for the understanding of human hematopoiesis and provide a framework for studying a range of biological processes, from infectious diseases to transplantation strategies and immune tolerance.

Supporting Online Material

Materials and Methods

Figs. S1 to S7


References and Notes

  1. We thank all of the patients who generously donated tissue that was used in these studies. We also thank S. J. Fisher and D. F. Nixon for valuable discussions; A. J. Butte, M. E. Moreno, and G. Kosikova for technical support; and the Gladstone Genomics Core for microarray analysis. Support for this work was provided by grants from NIH to J.M.M. (OD000329 and AI40312) and to K.W. and A. J. Butte (CA049605 and AG033314) and from the Harvey V. Berneking Living Trust. This project has been funded in part with federal funds from the National Institute of Allergy and Infectious Diseases, NIH, under contract no. N01-AI-70002. S.V. is supported by the Deborah Addicott Fellowship for Bioinformatics and Stem Cell Biology. J.M. is supported by the Swedish Research Council. T.D.B. received funding from American Pediatric Society, the American Academy of Pediatrics, and the March of Dimes as a fellow of the Pediatric Scientist Development Program. J.M.M. is a recipient of the NIH Director’s Pioneer Award Program, part of the NIH Roadmap for Medical Research, through grant DPI OD00329. The microarray data have been deposited in the Gene Expression Omnibus with accession code GSE25119.

Stay Connected to Science

Navigate This Article