Main

T lymphocytes, unlike other hematopoietic cells, differentiate in the thymus, a specialized organ that provides a unique environment for the acquisition of immunological competence. Analysis of the phenotype and stage of differentiation of the thymus-settling progenitors (TSPs) is important for characterization of the thymus-independent stages of T cell development and for devising strategies for cell-replacement therapy in lymphopenic patients. Although many studies have attempted to identify TSPs, reports over time have been contradictory. Depending on the experimental system, TSPs have been defined as T cell– and natural killer (NK) cell–committed cells, common lymphoid progenitors (CLPs) or lymphoid-primed multipotent progenitors (LMPPs). Integrating these conflicting data has proven challenging1.

The identification of the CLP2,3,4,5 in the bone marrow (BM) led to the proposal that TSPs are derived from a lymphoid cell–committed population. That was disputed by the observation that mice deficient in the transcription factor Ikaros, which lack CLPs in the BM, still have early thymic progenitors (ETPs), the most immature thymocytes derived from TSPs, which suggests that hematopoietic progenitor cells more undifferentiated than CLPs colonize the thymus6. Evidence that ETPs from adult mice generate myeloid cells, as well as B cells, in culture has further strengthened the view that ETPs are not derived from CLPs7,8. A fate-map analysis, however, found almost no myeloid cells derived from cells expressing CD127 (the interleukin 7 receptor α-chain)9, which indicates that the myeloid potential observed in ETPs does not contribute substantially to the myeloid lineage in vivo. Moreover, under particular culture conditions that include stromal cells that express Delta-like 4 (DL4), a ligand for the Notch family of signaling receptors7, even CLPs can generate myeloid cells9, which suggests that the results noted above7,8 could be explained by the artificial induction, in vitro, of nonphysiological differentiation pathways.

Analysis of thymocytes obtained from mouse embryos at embryonic days 11–12 (E11–E12), before the thymus is structurally well defined, has shown that TSPs are engaged in the T cell–NK cell pathways, which suggests that commitment occurs before contact with the thymic epithelium10,11,12. Consistent with that view, several laboratories have characterized progenitors of T cells–NK cells in the embryo and have found that cells with similar restricted potential are present in fetal blood (FB)13,14 and fetal liver (FL)15,16,17,18. However, in newborn mice, ETPs display robust potential to develop into T cells, B cells and myeloid cells but no potential to develop into megakaryocytes or erythrocytes and thus functionally resemble LMPPs19.

Thymocyte differentiation starts with CD4−CD8− double-negative (DN) cells, which also lack expression of the invariant signaling protein CD3. At this stage progenitor cells expand their populations, rearrange the δ-, γ- and β-chains of their T cell antigen receptors (TCRs)(the DN2–DN3 stage) and undergo pre-TCR selection (the DN3–DN4 stage). The most immature DN stage (DN1) is characterized by expression of the activation and memory marker CD44 and no expression of the T cell–activation marker CD25. The combination of expression of the stem cell factor receptor CD117 (c-Kit)6,20 and B cell–differentiation marker CD24 (HSA)21,22 can be used to further subdivide the DN1 compartment in five subsets in which DN1a and DN1b correspond to ETPs, while the other subsets have low T cell potential. Under physiological conditions, persistent thymic input is required for continuous T cell production23,24, but analysis of TSPs in the adult thymus is difficult because of the low number of cells that reach the thymus every day (estimated to be less than ten cells)25.

In this work we analyzed embryonic thymus with a new phenotypic characterization that combines expression of CD127, CD24 and the CLP marker CD135 (the receptor tyrosine kinase Flt3) and applied it to ETPs from E12 to E18. We found that TSPs were DN1 CD117+CD135hiCD24lo cells. During the time period of E12–E18, there was bimodal variation in the absolute number and differentiation potential of TSPs. In the first wave, between E12 and E15, TSPs in fetal thymic organ culture (FTOC) rapidly differentiated to generate embryonic γδ T cells that expressed γ-chain variable region 3 (Vγ3) of the TCR and mature γδ and αβ T cells. The TSPs had a phenotype similar to that of CLPs, lacked any detectable myeloid potential and lost B cell potential before they entered the thymus. In vivo, TSPs from that first wave rapidly generated T cells that maintained a small amount of reconstitution for up to 8 weeks after transfer. TSPs from a later wave (starting at E16) lacked the potential to develop into Vγ3 γδ T cells, extensively expanded their populations before maturation and thus gave rise to large numbers of CD4+CD8+ double-positive thymocytes. TSPs from the second wave phenotypically resembled LMPPs, lost the potential to develop into B cells or myeloid cells only in the thymus and generated large numbers of T cells in vivo. Consistent with their higher proliferative capacity, TSPs from the second wave had higher expression of cyclin D1 and lower expression of the cyclin-dependent kinase inhibitors p21 (encoded by Cdkn1a) and p57 (encoded by Cdkn1c) than did those from the first wave. We propose a model that reconciles the divergent views that accommodate the existing data. Similar to the avian thymus26,27, the mouse thymus is colonized by two successive waves of TSPs that represent different stages of hematopoietic development that follow distinct kinetics of differentiation.

Results

CD135 and CD127 identify new DN1a subsets

As noted above, CD24 and CD117 are used to characterize five DN1 subsets in lineage marker–negative (Lin−) thymocytes from adult mice (DN1a–DN1e)21. To study the first ETPs at the onset of thymopoiesis, we analyzed DN1 subsets at E12–E13 with a similar set of markers and compared those with cells from newborn thymus. The frequency of DN1a (Lin−CD117+CD44+CD25−CD24lo) cells was maximal at E12 and decreased by E13 (Fig. 1a), which suggested that the DN1a subset may contain the earliest thymic progenitor cells, in the absence of more differentiated subsets. We next analyzed whether DN1a cells had the phenotype of CLPs (CD135hiCD127+) or LMPPs (CD135hiCD127−)6,19,28. We assessed the expression of CD135 and CD127 on DN1a and DN1b cells at E12 and E13. The frequency of CD135hiCD127+ cells was always higher in the DN1a compartment than in the DN1b compartment and was highest at E12 (Fig. 1a; gating strategy and positive and negative controls, Supplementary Fig. 1). The finding that they had their highest frequency at E12 suggested that they represented the earliest ETPs.

Figure 1: DN1 compartment from the thymus at E12 and E13.
figure 1

(a) Flow cytometry of thymocytes obtained from mice at E12 and E13 and from newborn mice, then stained for CD3, CD8, CD4, NK1.1, CD44, CD25, CD24 and CD117. Outlined areas (left) indicate DN1 (Lin−CD44+CD25−) cells subcategorized as CD24loCD117+ (DN1a) cells (bottom region) and CD24+CD117+ (DN1b) cells (top region) and analyzed further (right); numbers adjacent to outlined areas (middle and right) indicate percent CD135hiCD127+ cells (top region) or CD135loCD127+ cells (bottom region) among DN1a cells (middle) and DN1b cells (right), with gates set by comparison with DN2 cells (negative internal control for CD135 and positive internal control for CD127; Supplementary Fig. 1a–c). (b) Flow cytometry analyzing the B cell, T cell, NK cell, DC and myeloid cell potential of single progenitor cells obtained and sorted as in a and cultured on a monolayer of OP9-DL4 stroma (T cell potential (n = 96 cells) and NK cell potential (n = 48 cells)) or OP9 stroma (B cell potential (n = 96 cells) and myeloid cell potential (n = 96 cells)) or in stroma-free culture (DC potential (n = 48 cells)). FL LMPPs serve as a positive control (far right). ND, no positive wells detected. Data are representative of six (E12), twenty-five (E13) or four (newborn) independent experiments (a) or three independent experiments (b).

Full size image

DN1a CD135hiCD127+ ETPs lack B cell and myeloid potential

We assessed the potential of DN1a CD135hiCD127+ cells to develop into T cells, B cells and myeloid cells at E13 because there is a higher cell yield at E13 than at E12. Single-cell analysis of DN1a CD135hiCD127+, DN1a CD135lo and DN1b subsets showed that they had high T cell potential (as many as 50% of the cells developed into T cells) but no detectable B cell or myeloid potential (Fig. 1b). We were unable to find B cell or myeloid progeny of any of those three subsets in three independent experiments in which we analyzed 96 single cells in each experiment. Consequently, we concluded that the frequency of progenitors of B cells or myeloid cells at those stages was below 1 in 288. In contrast, we found NK cell and dendritic cell (DC) potential of 1 in 3 and 1 in 5, respectively (Fig. 1b), consistent with published reports showing that the NK cell and DC potential is lost only in the thymic DN2 stage29.

To establish lineage relationships among the three subsets and to identify the most undifferentiated progenitor cells, we used two complementary approaches: short-term culture and FTOC. We analyzed sorted DN1 subsets cultured for 22 h on OP9 stromal cells expressing DL4 (OP9-DL4 cells) (Fig. 2a–c). We found that 49% of DN1a CD135hiCD127+ cells lost CD135 expression, although they remained in the DN1a compartment (Fig. 2a), and 33% of DN1a CD135lo cells upregulated CD24 expression and thus acquired the phenotype of DN1b cells (Fig. 2b), while most sorted DN1b cells remained CD24hi (Fig. 2c). After 48 h of FTOC of E12 cells, differentiation progressed to the DN3 stage (Fig. 2d) and we no longer detected DN1a CD135hiCD127+ cells, the main progenitor compartment at the onset of the culture. After culture, the majority of ETPs (55%) were DN1b cells (Fig. 2d).

Figure 2: DN1a CD135hiCD127+ cells are the most immature progenitors of T cells.
figure 2

(a–c) Flow cytometry of purified DN1a CD135hiCD127+ thymocytes (a), DN1a CD135lo thymocytes (b) and DN1b thymocytes (c) at E13, analyzed immediately after purification (Before culture; top) or after culture for 22 h on OP9-DL4 stroma (After culture; bottom); similar results were obtained at E12, E15 and E18. Numbers adjacent to outlined areas indicate percent cells in each throughout; letters adjacent indicate subset (a, DN1a; b, DN1b). (d) Flow cytometry of cells from thymic lobes at E12, analyzed immediately after isolation (top) or cultured for 48 h in FTOC (bottom), sorting cells into compartments DN1–DN3 (left) and the DN1 subsets (middle) and assessing CD135 expression in DN1a cells (right); similar results were obtained for cultured thymus at E13. (e) Flow cytometry of cells from FB collected from embryos at E13, analyzing CD24 expression (bottom right) in Lin−α4β7−CD127+CD117+Sca-1loCD135hi cells. (f) Flow cytometry analyzing the B cell, T cell and myeloid cell potential (as in Fig. 1) of FL CLPs (Lin−α4β7−CD117+Sca-1loCD127+CD135hi), FB CRLPs (Lin−α4β7−CD117+Sca-1loCD127+CD135hi), FB LMCPs (Lin−CD117+Sca-1loCD127−CD135hi) and FB MCPs (Lin−CD117+Sca-1loCD127−CD135lo); n = 96 cells (CLPs and CRLPs) or n = 48 cells (LMCPs and MCPs). ND, no positive wells detected (less than 1 positive well among 96 total wells). Data are from one experiment representative of three (a–d), thirteen (e) or two (f) independent experiments.

Full size image

Thus, DN1a CD135hiCD127+ cells rapidly differentiated into DN1b cells by sequentially downregulating CD135 expression and upregulating CD24 expression. DN1a CD135lo cells progressed to the DN1b stage of differentiation. Therefore, in the thymus at E13, DN1a CD135hiCD127+ cells were the most immature ETPs.

CD135hiCD127+ progenitor cells are present in the circulation

In FL at E13, most CLPs are CD135hi, and acquisition of expression of the integrin α4β7 coincides with the loss of B cell and T cell potential and progression into the NK cell and innate lymphoid cell lineages16,17,30. The most immature thymic progenitor cells express CD117, CD127 and CD135, a phenotype similar to that of the CD135hi CLPs present in FL. Because FL cells can reach the thymus only via the circulation, we analyzed FB at E13 for the presence of progenitor cells with a phenotype and differentiation potential similar to that in FL (Fig. 2e). CD135hiα4β7− CLPs could be further subcategorized by CD24 expression (Supplementary Fig. 2a). Among Lin−Sca-1+c-Kit+ (LSK) cells, the level of CD135 expression distinguishes LMPPs (LSK CD135hi cells), multipotent progenitors (LSK CD135int cells) and a population with substantial enrichment for hematopoietic stem cells (LSK CD135− cells). Populations with similar phenotypes were also present in FB (Fig. 2e); we called these circulating counterparts 'lympho-myeloid circulating progenitors' (LMCPs: LSK CD135hi cells), 'multipotent circulating progenitors' (MCPs: LSK CD135int cells), 'circulating hematopoietic stem cells' (LSK CD135− cells) and 'circulating restricted lymphoid progenitors' (CRLPs: Lin−α4β7−CD135hiCD117+Sca-1loCD127+ cells).

We investigated the differentiation potential of progenitor cells in FL and FB (Fig. 2f). FL CLPs had potent T cell and B cell potential but no myeloid potential (Fig. 2f). In contrast, CRLPs had similarly high T cell potential (1 in 2) but less capacity to give rise to B cells (1 in 47) (Fig. 2f). As expected, LMCPs and MCPs had T cell, B cell and myeloid cell potential (Fig. 2f). These results indicated that CRLPs were biased toward the T cell lineage.

Our results showed that circulating CRLPs were lymphoid progenitor cells with high T cell potential and low B cell potential. The thymus develops from the third and fourth pharyngeal pouches and is structurally defined as an organ only at E12. Before that stage, hematopoietic progenitor cells in the vicinity of the anlage are not in contact with the thymic epithelium31,32. To investigate whether the CRLPs are detectable in the blood before the onset of thymopoiesis, we analyzed progenitor cells from FB at E11 (Supplementary Fig. 2b). We found lymphoid progenitor cells in FB at E11 with a phenotype, frequency and differentiation potential (Supplementary Fig. 2b,c) similar to those of cells in FB at E13 (Fig. 2f).

T cell–biased CD24lo CRLPs express thymus-tropic molecules

The chemokine receptors CCR9 and CCR7 are involved in the seeding of progenitor cells in the thymus12,33. The proportion of CLPs and CD24lo CRLPs that expressed CCR9 was higher than that of CD24+ progenitor cells (Fig. 3a, top and middle). Moreover, CCR9+ cells were also more frequent among CD24lo CRLPs than among FL CD24lo CLPs (Fig. 3a, top and middle). Notably, only a subset of DN1a CD135hiCD127+ cells in the thymus expressed CCR9 (Fig. 3a, bottom), which suggested that DN1a CD135hiCD127+ cells downregulated CCR9 expression when they differentiated into DN1a CD135lo cells.

Figure 3: Functional characterization of CRLPs at E13.
figure 3

(a) Expression of CD117 and CD24 in Lin−α4β7−CD117+Sca-1loCD127+CD135hi cells (left column), and CCR9 expression in CD24lo CLPs and CRLPs (middle column) and CD24+ CLPs and CRLPs (right column), from FL (top) and FB (middle) and thymocytes (T; bottom) at E13. Bottom right, CCR9 expression in DN1a CD135hiCD127+ and DN1a CD135lo thymocytes at E13 (sorted at left); number below bracketed line indicates percent CCR9+ cells. (b) Flow cytometry of progenitor cells sorted from FB at E13, cultured for 3 d on OP9-DL4 stroma and stained for CD44 and CD25 to assess the DN compartments of cultures of MCPs (n = 200), LMCPs (n = 700), CD24+ CRLPs (n = 200) and CD24lo CRLPs (n = 150) after elimination of Lin+ (CD4+CD8+CD3+) cells. Data are from one experiment representative of three experiments.

Full size image

Quantitative RT-PCR analysis confirmed those results and showed more expression of CCR9 in CD24lo cells than in CD24+ cells in FL and FB (Supplementary Fig. 3a). In the thymus, CCR9 expression was downregulated after the DN1a CD135hiCD127+ stage (Supplementary Fig. 3b). That was also true for DN1a thymocytes at later stages, such as E17 (Supplementary Fig. 3c). There was also high expression of CCR7 in most ETPs at both stages (Supplementary Fig. 3d).

To compare the progression of progenitor cells from FB along the T cell pathway, we did short-term culture of cells on OP9-DL4 stroma. MCPs, LMCPs and CD24+ CRLPs differentiated to the DN1 stage, but only a few cells reached the DN2 stage after the 3 d of culture (Fig. 3b). In contrast, 14% of CD24lo CRLPs differentiated to the DN3 stage during the same period. Thus, CD24lo CRLPs were the most advanced progenitor cells along the T cell pathway found in the blood.

Activation of the Notch pathway upregulates Notch-dependent transcripts34 and leads to T cell commitment and, as a consequence, the downregulation of B cell–related genes, such as Ebf1 (which encodes the B cell–specification factor EBF1) and Pax5 (which encodes the transcription factor Pax5). CD24lo CRLPs and DN1a CD135hiCD127+ cells had significantly higher expression of Hes1 (a transcript that is a target of Notch) and lower expression of Ebf and Pax5 than did FL CLPs or CD24+ CRLPs (Supplementary Fig. 4), which suggested activation of the Notch pathway before colonization of the thymus. These results were consistent with the lack of B cell potential of CD24lo CRLPs and further emphasized the similarities between CD24lo CRLPs and DN1a CD135hiCD127+ cells. We concluded that at E11–E13, the thymus was colonized by CRLPs biased to generate T cells.

Thymic colonization by two progenitor waves

As shown above, TSP populations found at the onset of thymopoiesis (E12–E13) were composed of lymphoid compartment–restricted progenitor cells with the phenotype of CLPs and a T cell bias. In contrast, the most immature thymic progenitor cells in newborn mice have been shown to have B cell and myeloid potential19. To determine whether different cell types colonize the thymus at different stages of development, we analyzed TSPs throughout gestation.

We detected DN1a and DN1b (ETP) subsets at all gestational days, whereas we first detected DN1c and DN1e subsets only at E14 and detected the DN1d subset at E17 (Fig. 4a). The frequency and number of DN1a CD135hiCD127+ cells decreased from E12 to E15 and then increased again after E16 (170 at E12, 30 at E15, 148 at E16, 294 at E17 and 617 at E18 (cells per thymus); Fig. 4a,b), which indicated that thymic progenitor cells colonized the thymus in two sequential waves. Notably, CD135 expression in DN1a cells from newborn mice was lower than that observed in cells at E13 (Supplementary Fig. 1b). To investigate potential differences in TSPs involved in the two waves, we assessed the ability of DN1a CD135hi cells from E13 to E18 to differentiate into T cells, B cells and myeloid cells. As noted above for cells at E13 (Fig. 1), we detected no B cells or myeloid cells in cultures of DN1 subsets at E15 (refs. 10,11,12) (Supplementary Fig. 5). However, we detected B cell and myeloid potential, as well as T cell, NK cell and DC potential, in DN1a CD135hi cells at E16–E18 (Fig. 4c, left and middle, and Supplementary Fig. 5b). Furthermore, limiting-dilution analysis showed that the frequency of B cell and myeloid cell progeny, although lower than that of LSK cells in FL (1 in 3; Supplementary Fig. 5), averaged 1 in 25 (Fig. 4c, right).

Figure 4: Thymic colonization occurs in two successive waves during embryonic development.
figure 4

(a) Flow cytometry of cells from thymic lobes at E13–E18, stained for lineage markers, CD44, CD25, CD24, CD135 and CD127 and sorted into DN1 subsets with gating on Lin−CD44+CD25− cells (left), followed by analysis of the expression of CD135 and CD127 by DN1a cells (middle) or DN1b cells (left). (b) Frequency (left) and absolute number (right) of DN1a CD135hi cells from thymic lobes at E13–E18. *P = 0.005, E12 versus E15, and **P = 0.001, E15 versus E16 (Student's t-test). (c) B cell potential (left) and myeloid cell potential (middle) of DN1a CD135hi cells at E13, E16, E17 and E18, cultured on OP9 stroma with the cytokines Flt3L, IL-3, IL-7, GM-CSF and M-CSF. Right, limiting-dilution assay of the B cell and myeloid cell potential of DN1a CD135hi cells at E18, for cultures with 10, 20 and 50 cells per well with 48 replicates each (n = 3,840 sorted DN1aCD135hi cells). Data are representative of twenty-five (E13), ten (E15), seven (E16) or five (E18) independent experiments (a) or three independent experiments (b,c; mean and s.e.m. in b).

Full size image

The multilineage differentiation potential found in TSPs after E15 prompted us to measure the expression of CD127 mRNA and Sca-1 protein in DN1a CD135hi cells. Quantitative RT-PCR analysis showed lower expression of CD127 transcripts in DN1a CD135hi cells after E16 (Supplementary Fig. 6a). Consistent with that result, CD127 expression (detected by antibody staining) was also lower in DN1a cells at E17 and E18 (Supplementary Fig. 6b) than in all other populations analyzed. In contrast, expression of Sca-1 protein was higher at E18, with levels similar to those found in LMPPs (Supplementary Fig. 6c). The B cell and myeloid cell potential of DN1a CD135hi cells at E16–E18, together with the low CD127 expression and high Sca-1 expression, indicated that in contrast to result obtained at E11–E15, TSPs at E16–E18 resembled LMPPs. That conclusion was further supported by the analysis of circulating cells (Fig. 5a). While we found CRLPs and LMCPs in FB at E13 (Fig. 2), CRLPs were almost completely absent from FB at E18 (Fig. 5a).

Figure 5: TSPs from the two waves have different potential and kinetics of differentiation.
figure 5

(a) Flow cytometry of FB and FL progenitor cells at E18 (as in Fig. 3). (b) Quantification of CD45.1+ cells expressing Vγ3 and Vδ1 (Vγ3+Vδ1+ T cells), the δγ TCR (δγ T cells), CD4 and CD8 (DP cells) and the αβ TCR (αβ T cells) among cells from irradiated CD45.2+ thymic lobes at E14 colonized by purified CD45.1+ DN1a thymocytes (n = 500 cells) at E13 and E18 and cultured for 12 d, followed by dissociation of the thymic lobes, presented as total donor cells recovered from each lobe. Each symbol represents one colonized FTOC; small horizontal lines indicate the mean (and s.e.m.). NS, not significant; *P = 0.02 and **P = 0.0001 (Student's t-test). (c) Proliferation of DN1 CD117+ thymocytes at E14 and E18, assessed by staining with Ki67 and DAPI and presented as the frequency of cells in S-G2-M (Ki67+DAPI+) and G1 (Ki67+DAPI−). (d) Reconstitution of blood in Cd3−/− CD45.2+ mice (n = 8) engrafted with two CD45.2+ thymic lobes colonized by 1 × 103 CD45.1+CD45.2+ TSPs at E13 and two CD45.2+ thymic lobes colonized by 1 × 103 CD45.1+ TSPs at E18, followed by weekly blood collection and staining of blood for CD45.1, CD45.2, CD3, CD4, CD8, TCRβ and TCRδ, presented as the frequency of T cells from DN1a cells at E13 or E18 among CD3+ cells at days 17 and 31 after transplantation. *P = 0.003 and **P = 0.001 (Student's t-test). (e) Ratio of T cells derived from DN1a cells at E18 to those from DN1a cells at E13 on days 17, 24 and 31 after transplantation as in d. Data are representative of three independent experiments (a,c) or are from one experiment representative of two experiments (b) or one experiment (n = 8 recipient mice) representative of three experiments with experimental design variations (d,e).

Full size image

To determine whether we could detect TSPs in the blood before the onset of FL hematopoiesis, we analyzed embryos at E10. Intraembryonic hematopoiesis starts at E10 (the 30-somite stage), with detection in the liver of the first progenitor cells derived from the aorta-gonad-mesonephros region35. Before that stage, hematopoietic activity is highest in the yolk sac, which generates cells with limited differentiation potential. We detected CD135hiCD24lo CRLPs in FB at E10.5 (Supplementary Fig. 6d) but not at E10 (Supplementary Fig. 6e). This indicated that TSPs were generated only in the FL and did not come from the yolk sac.

TSPs from each wave have different properties

A hallmark of embryonic T cell development is the ability to generate invariant Vγ3+ dendritic epithelial T cells36,37. To assess the capacity of DN1a cells from the first and second waves to give rise to that cell type, we colonized thymic lobes from mice at E14 with purified DN1a cells from mice at E13 and E18 (E13 DN1a cells and E18 DN1a cells, respectively). Although both populations generated a similar number of γδ T cells, only E13 DN1a cells generated dendritic epithelial T cells (Fig. 5b, left); the few Vγ3+ cells present in lobes colonized by E18 DN1a cells did not express the characteristic Vδ1 chain (Fig. 5b). More notably, while E13 and E18 DN1a progenitor cells were equally able to generate αβ T cells, they did so with different kinetics. At day 12 of culture, cultures initiated with E13 DN1a cells generated a larger number of mature αβ T cells and tenfold fewer DP cells than did cultures initiated with E18 DN1a cells (Fig. 5b, middle right and far right), which indicated that E13 DN1a cells differentiated more rapidly. We concluded that the two waves of TSPs differed not only in phenotype and differentiation potential but also in the kinetics of their maturation into CD3+ T cells and their ability to generate embryonic subsets of T cells. The fast differentiation and low number of DP cells generated in FTOC suggested that DN1a cells from the first wave might be impaired in progression through the cell cycle. Analysis with the proliferation marker Ki67 in combination with the DNA-intercalating dye DAPI allowed us to distinguish cells in the G0, G1 and S-G2-M phases of the cell cycle. Less than 7% of E14 DN1a cells but more than 30% of E18 DN1a cells were in S-G2-M (Fig. 5c). In contrast, nearly 90% of E13 cells were in G1 (Fig. 5c), which suggested that they were partially arrested in the cell cycle.

E18 TSPs generate more T cells than do E13 TSPs

To assess the repopulation capacity of TSPs in vivo, we irradiated thymic lobes at E15 and colonized them with TSPs from E13 (E13 TSPs) or E18 (E18 TSPs), then grafted two lobes of each under the kidney capsule of CD3-deficient (Cd3−/−) mice. Blood samples obtained 17 d after transplantation showed that the progeny of E13 TSPs differentiated before the progeny of TSPs at E18 did and constituted the majority of peripheral CD3+ cells at this time point (Fig. 5d, left). Two weeks later, the frequencies were reversed, and E18 TSPs contributed the majority of the peripheral T cell pool (Fig. 5d, right). The ratio of the progeny of E18 TSPs to that of E13 TSPs varied between days 17 and 31 after transplantation from less than onefold to more than sixfold (Fig. 5e). That was consistent with the results obtained by FTOC indicating faster differentiation but poorer proliferation of E13 TSPs than of E18 TSPs (Fig. 5b). Notably, T cells derived from E13 TSPs averaged 2% of total lymphocytes throughout the experiment, while those derived from E18 TSPs rapidly increased in number and stabilized to an average of 12% of total lymphocytes after day 31 (Fig. 6a). That result indicated that E18 TSPs generated at least fivefold more CD3+ cells than did E13 TSPs (Fig. 6a).

Figure 6: The slow differentiation and high proliferation properties of TSPs at E18 are reflected in their transcriptional profiles.
figure 6

(a) Frequency, in recipient blood, of T cells derived from E13 or E18 progenitor cells (from the transplantation experiment in Fig. 5c,d; time after transplantation, horizontal axes), calculated as a fraction of total cells in the lymphocyte gate. NS, not significant; *P = 0.0005 (analysis of variance). (b) Heat map of a selected set of genes (from the gene-ontology analysis and the database of the Immunological Genome Project) with different expression in cells at E13 versus those at E18. Products encoded (as grouped along right margin) are involved in cell-cycle regulation (purple), T cell differentiation (orange), cytokine and chemokine signaling (pink), regulation of the actin cytoskeleton, transendothelial migration and tight junctions (gray) and sensing of cytosolic DNA and RIG-like receptor signaling (blue). (c) Quantitative RT-PCR analysis of cDNA from TSPs at E13 and E18 and DN1a thymocytes at postnatal day 10 (P10), confirming the transcriptome analysis of cyclin-dependent kinase inhibitors (Cdkn1a and Cdkn1c), cyclin D1 (Ccnd1), regulators of the Notch pathway (Hes1, Hey1 and Dtx1), granzyme B (Gzmb) and gelsolin (Gsn); results are presented in arbitrary units (AU) relative to expression of the control gene Hprt or Gapdh. *P = 0.025, **P = 0.01, ***P = 0.001, †P = 0.0006, ‡P = 0.0003 and §P = 0.0001 (Student's t-test). Data are representative of three experiments with three biological replicates (a) or are from one experiment with three biological replicates (b) or one experiment representative of three independent experiments (c; mean and s.e.m.).

Full size image

We did genome-wide transcriptome analysis of both TSP populations (E13 and E18 TSPs). We generated a heat map of the differences between E13 TSPs and E18 TSPs most represented in the transcription profile, after gene-ontology analysis (Fig. 6b). Consistent with the defective progression through the cell cycle, E13 TSPs had higher expression of Cdkn1a and Cdkn1c (which encode cyclin-dependent kinase inhibitors) and Gadd45g (which encodes a signaling molecule), together with lower expression of Ccnd1 (which encodes cyclin D1) (Fig. 6b,c). Additionally, higher expression of genes encoding the cytokine-response inhibitor SOCS2 and the tumor suppressors Hic1 and Perp by E13 TSPs might also account for the defective cell-cycle progression (Fig. 6b,c). Consistent with their rapid differentiation, E13 TSPs had higher expression of gene encoding of CD3γ, granzymes A and B, pre-TCRα and the RNA-binding protein KHDC1A (Fig. 6b). E13 TSPs also had higher expression of Nrarp, Hey1 and Deltex1 (whose products regulate different aspects of the Notch signaling pathway) than did E18 TSPs (Fig. 6b,c), which could also be interpreted as a consequence of faster developmental progression. Genes encoding molecules involved in cytokine and chemokine signaling were among the most represented genes among those with different expression in E13 TSPs versus E18 TSPs (Fig. 6b), which suggested that TSPs from each wave were responsive to different signals. We also found substantial representation of transcripts encoding molecules involved in regulation of the actin cytoskeleton, transendothelial migration and tight junction formation in E13 TSPs, with greater representation of genes encoding molecules related to the sensing of cytosolic DNA and helicase RIG-I–like signaling pathways and histidine metabolism in E18 TSPs (Fig. 6b). We concluded that TSPs from the first wave generated fewer T cells than did those from the second wave because of impaired cell-cycle progression, reflected by lower expression of cyclin D1 and higher expression of cyclin kinase inhibitors.

Discussion

Here we found colonization of the mouse thymus by successive waves of progenitor cells composed of hematopoietic cells that differed in their developmental stage and function. Our results provide an explanation for apparently conflicting results in the literature by showing that the first colonization wave was composed of T cell–biased CLP-like cells, whereas cells from the second wave resembled LMPPs, with B cell and myeloid cell potential. The discontinuous migration of progenitor cells into the thymus has been described in birds, in which three waves of thymic immigrants have been defined26,27. Even though such a phenomenon has been suggested to also occur in mice38, experiments similar to those done with the avian model were not possible in mammals because embryonic development does not progress normally after surgical manipulation. However, we used a new phenotyping strategy (DN1a CD135hiCD127+ cells) to identify, in the unmanipulated embryo, a small ETP subset with properties expected in TSPs. We counted TSPs throughout embryonic development and found that TSP numbers varied in a bimodal manner: the first wave spanned E11–E15, and the second wave occurred after E16. A difference in their phenotype and potential of cells from the two waves indicated that TSPs switched from being T cell–biased progenitor cells to LMPPs in the first wave and second wave, respectively.

More notable were the different biological properties of progenitor cells from each wave. Contrary to TSPs of the first wave, E18 TSPs (representative of the second wave) lost the ability to generate embryonic populations of γδ T cells (Vγ3+Vδ1+ dendritic epithelial T cells) and thus resembled postnatal ETPs in both phenotype and function36,37. The first wave of TSPs not only generated mature αβ T cells faster but also generated fivefold fewer mature T cells than did TSPs from the second wave. Faster differentiation of embryonic thymocytes has been proposed on the basis of transcriptional profile analysis of ETPs from embryos at E15 and ETPs from adults39. The low number of T cells from the first wave could be partially explained by slow progression through the cell cycle due to higher expression of the cyclin kinase inhibitors p21 (Cdkn1a) and p57 (Cdkn1c) that arrest cells in G1 and of Hic1 and Perp, a direct target of the tumor suppressor 53 (data not shown), and lower expression of cyclin D1. The higher expression of SOCS2 and differences in the regulation of cytokine receptors also suggested that E13 TSPs and E18 TSPs respond differently to cytokine signals. All of these factors would allow fast differentiation and thus provide a small but sizable number of functional αβ T lymphocytes soon after birth. We found among E12 thymocytes considerable enrichment for DN1a CD135hiCD127+ cells with several distinct properties: they had the ability to generate T cells with high efficiency; they had the unique ability to irreversibly generate all other DN1 subsets and to rapidly differentiate, such that they were no longer detected after 22 h in culture; and they had the lowest expression of transcripts of the Notch target Hes1 of all the DN1 subsets. Together these properties indicated that DN1a CD135hiCD127+ cells were the most immature ETPs.

Although several thousand ETPs can be found in the adult thymus, the number of TSPs that enter the thymus every day is probably as low as ten cells25, consistent with the proposal that most ETPs from adults are not TSPs21,28,40,41. In contrast, at E12–E13, the most immature ETP compartment (DN1a CD135hiCD127+ cells) is totally renewed in less than 22 h. This indicates that at any given time point, fetal ETPs correspond to cells that seeded the thymus during the previous 24 h.

CCR9 and CCR7 are required for the homing of progenitor cells to the thymus because in mutant embryos lacking both CCR9 and CCR7, ETPs are almost completely absent42,43,44,45. Consistent with that, CCR9 was expressed by 35% of DN1a CD135hiCD127+ TSPs at E13 and was not detectable in DN1a CD135lo cells, while CCR7 was present in most DN1a cells. We also found cells with phenotype and differentiation potential similar to those of the most immature ETPs in FB, where they showed considerable enrichment for CCR9+ cells. Moreover, they differentiated more rapidly along the T cell pathway than did other CD127+ subsets or LMPPs12,42. Thus, CD24lo CRLP populations with a phenotype similar to that of thymic DN1a CD135hiCD127+ cells included TSPs that rapidly differentiated after colonizing the thymus, thereby losing CCR9 expression. Other receptors linked to homing to the thymus, such as CXCR4 and CD44, were not expressed differently by these subsets (data not shown).

As shown before12,13,14,15,16,17,46,47, we found that the first ETPs, isolated between E12 and E15, were devoid of myeloid potential and, more notably, were devoid of B cell potential, which indicated either that TSPs entered the thymus already biased toward the T cell–developmental pathway or that DN1a CD135hiCD127+ cells at E12–E15 lost B cell potential soon after reaching the thymus. In favor of the former possibility, several reports have found progenitor cells engaged in the T cell–differentiation pathway in FB and FL13,14,15,16,17,46. Notably, cells with restricted T cell–NK cell potential have also been found in the FL15 and FB12,47 of nude (nu/nu) mice, which indicates that the first stages of lineage restriction are pre-thymic. Consistent with that, we found circulating cells with TSP phenotype and differentiation potential at E10.5 and E11, before the onset of thymopoiesis, but not at earlier stages (E10); this suggested that TSPs did not originate before the FL functions as a hematopoietic organ. Consistent with that, lymphoid compartment–restricted cells have not been found in early hematopoietic sites other than FL48. We found higher expression of Hes1 in CRLPs than in their FL counterparts, which may be consistent with activation of the Notch pathway during or after exit from the FL. However, E18 TSPs that maintained B lineage potential had moderately higher expression of Hes1 than did E13 TSPs. We are therefore tempted to speculate that the loss of B cell and myeloid cell potential by E13 TSPs was Notch independent and Hes1 could be induced in these cells independently of the Notch pathway (by Wnt or Shh), as suggested before49,50. Challenging the possibility that the first wave of DN1a CD135hiCD127+ cells lost B cell potential only after reaching the thymus is the observation that the E12 thymus is not fully mature and that the expression of Notch ligands in the thymic epithelium substantially increased after E15 (data not shown), precisely the time at which we first detected B cell and myeloid cell potential in DN1a thymocytes. We conclude that the TSPs between E12 and E15 have restricted differentiation potential before colonizing the thymus, while cells from the second wave, after E16, retain B cell and myeloid cell potential. It thus appears that the first wave of thymus-colonizing progenitor cells was evolutionarily selected to rapidly give rise to mature αβ T cells and to generate innate γδ T cells.

Methods

Mice.

C57BL/6 mice (Charles River) and Cd3−/− C57BL/6 mice (a gift from A. Freitas) were used between 6 and 8 weeks of age. Mice were allowed to mate overnight. The next day, mice with a vaginal plug were separated from other mice and were considered to be at E0. All experiments were according to the Pasteur Institute ethic charter approved by French Agriculture ministry and to the EU guidelines.

Cell suspensions.

All cells were suspended in Hanks' balanced-salt solution (HBSS) supplemented with 1% FCS (Gibco). Fetal liver and thymus were micro dissected under a binocular magnifying lens. Organs were rinsed with Hanks' balanced salt solution plus 1% FCS for removal of contaminating blood cells. Then, thymus and fetal liver cells were passed through a 26-gauge needle of a 1-ml syringe and were filtered. Adult thymi were dissected, and single-cell suspensions were obtained through the use of a nylon mesh. Blood from embryos was obtained for 20 to 30 min in Hanks' balanced salt solution supplemented with 1% of FCS without calcium and magnesium. Blood cells were resuspended in a 70% (vol/vol) solution of Percoll (GE Healthcare BioSciences) topped by a 40% (vol/vol) solution of Percoll. After 30 min of centrifugation at 3,000 r.p.m. for depletion of red blood cells, cells were collected at the 40%-70% interface.

Flow cytometry and cell sorting.

Cell suspensions were stained with antibodies from BD Biosciences, eBioscience, Biolegend, R&D Systems and Invitrogen (Supplementary Table 1). Cells were incubated for 15–30 min at 4 °C in the dark. Biotinylated antibodies were detected by incubation for 5–10 min at 4 °C in the dark with streptavidin coupled to Pacific blue, phycoerythrin-indotricarbocyanine or Qdot605. FL and thymus of newborn and adult mice were depleted of Lin+ cells by magnetic cell separation (Miltenyi Biotec). Each antibody was pre-titrated for the optimal dilution (Supplementary Table 1). Antibodies to lineage markers included anti-CD19, anti-Gr1, anti-Ter119, anti-NK1.1, anti-CD11c, anti-CD3, anti-CD4 and anti-CD8 (all identified in Supplementary Table 1). Cell cycle was analyzed by fixing of cells after staining with antibodies to surface markers (identified in Supplementary Table 1) with the eBioscience fixation kit and staining with Ki67. DAPI (4,6-diamidino-2-phenylindole) was added 7 min before analysis. CANTO I, CANTO II and LSRII from BD Bioscience were used for flow cytometry analysis. A MoFlo (Dako Cytomation) and FACSAria III (BD Bioscience) were used for cell sorting.

Frequency assay.

Single cells were sorted into 96-well plates containing a monolayer of OP9 stroma (for analysis of the differentiation of myeloid cells and B cells) or OP9-DL4 stroma (for analysis of the differentiation of T cells), with complete medium OPTI-MEM plus 10% FCS, penicillin (50 units/ml), streptomycin (50 μg/ml) and β-mercaptoethanol (50 μM) supplemented with a saturating amount of the following cytokines: interleukin 7 (IL-7) and Flt3 ligand, for the differentiation of T cells and B cells; and macrophage colony-stimulating factor, IL-3, c-Kit ligand and granulocyte-macrophage colony-stimulating factor, for myeloid differentiation. Cytokines were obtained from the supernatant of myeloma cell lines (provided by F. Melchers) transfected with cDNA encoding those cytokines.

For DC assays, single cells were directly sorted into wells of 96-well pates and were supplemented with the same cytokines used for myeloid differentiation; for the detection of NK cells, IL-2 was added to cultures of OP9-DL4 stromal cells.

After 9 to 14 d, wells with colonies were stained with anti-CD19, anti-CD3, anti-CD4, anti-CD8, anti-CD11b, anti-CD11c and anti-NK1.1 (all identified in Supplementary Table 1) and were analyzed by flow cytometry as described above. CD19+ cells were considered B cells, CD8+CD4+CD3+ cells were considered T cells, Gr-1+CD11b+ cells were considered myeloid cells, CD3−NK1.1+ cells were considered NK cells and CD11c+CD11b+CD19−CD3− cells were considered DCs.

The OP9 and OP9-DL4 cell lines were regularly assessed for contamination by mycoplasma and were consistently found to be negative for contamination.

Cell frequencies, determined with ELDA software ('extreme limiting-dilution analysis') from the Walter and Eliza Hall Institute Bioinformatics Division, are presented as the number of positive wells and the number of total tested wells.

E12 FTOC.

After dissection, E12 thymi were placed on filters (Millipore) in Petri dishes with 3 ml of complete medium. After 48 h of culture, thymi were gently shredded and analyzed by flow cytometry.

Reconstitution FTOC.

Irradiated thymic lobes (3,000 rads) from CD45.2+ mouse embryos at E14 or E15 were colonized for 48 h, on a hanging drop, in Terasaki plates by 500 DN1a cells from CD45.1+ embryos at E13 or E18, in complete medium. After colonization, E14 thymi were cultured for 12 d on a filter floating on 3 ml of complete medium. After culture, thymi were dissociated and analyzed by flow cytometry.

In vivo reconstitution.

Irradiated thymic lobes (3,000 rads) from CD45.2+ mouse embryos at E15 were colonized for 48 h by 1,000 DN1a CD135hi cells from CD45.1+CD45.2+ mouse embryos at E13 or DN1a CD135hi cells from CD45.1+ mouse embryos at E18, in complete medium. Two thymi colonized with E13 TSPs and two thymi colonized with E18 TSPs were placed under the kidney capsule of each Cd3−/− mouse. Blood was obtained from the retro-orbital plexus at regular intervals.

Microarrays and statistical analysis.

1 × 104 to 3 × 104 sorted TSPs isolated from the thymus at E13 and E18 were analyzed by microarray. RNA was extracted with an RNeasy Micro kit (Quiagen). Agilent Whole Mouse Genome Oligo Microarrays were done by Miltenyi Biotech after quality control of the RNA with an Agilent 2100 Bioanalyzer, linear T7-based amplification of RNA and labeling with indocarbocyanine with the Agilent Low Input Quick Amp Labeling Kit (Agilent Technologies).

For analysis of microarray data, Agilent Mouse 8x60k Array data sets were analyzed by GenoSplice technology. Data were analyzed with R software and with the LIMMA (linear models for microarray data) software package from the Walter and Eliza Hall Institute Bioinformatics Division. Data were normalized by quantile normalization. Background correction was made by the 'normexp' method. Control probes and probes with low expression were removed, then replicate spots were averaged. An unpaired Student's t-test was used for comparison of gene intensities in the biological replicates. Genes were considered to be significantly regulated with a difference in expression of ≥1.5-fold and a P value of ≤0.05.

For hierarchical clustering, the distance from the gene signal in a given sample to the corresponding average in all the samples was calculated for each regulated gene. Corresponding values were displayed and grouped into clusters with the Java tool MeV4.6.2 (multiple-experiment viewer) from The Institute of Genome Research, by Euclidean distance and average linkage clustering.

For pathway and gene-ontology analyses, significant (genes whose expression was significantly different in the populations assessed) pathways of the Kyoto encyclopedia of genes and genomes (accession code 22130871) and gene-ontology terms were retrieved from the DAVID bioinformatics database (Database for Annotation, Visualization and Integrated Discovery; accession code 19131956). All regulated genes and upregulated and downregulated gene lists were used to retrieve significant pathways and gene-ontology terms. Results corresponded to union of these three analyses.

Data were analyzed with an unpaired Student's t-test (Figs. 4b,5b,d and 6c and Supplementary Figs. 3a,b, 4 and 6a) or by analysis of variance (Fig. 6a). A P value (two-tailed Student's t-test) of <0.05 was considered significant; a P value of >0.05 was considered not significant.

Relative quantitative RT-PCR.

After extraction of mRNA with an RNeasy Micro kit (Quiagen), mRNA was reverse-transcribed into cDNA with a PrimeScript RT reagent kit (Takara Bio), followed by quantitative PCR with TaqMan primers and TaqMan Universal Master Mix from Applied Biosystem on ABI Prism 7300 from Applied Biosystem. Primers for the following genes were used (Applied Biosystem assay identifiers in parentheses): Hprt (Mm00446968_m1), Ebf1 (Mm01288946_m1), Pax5 (Mm00435501_m1), Notch1 (Mm00435245_m1), Hes1 (Mm00468601_m1), Ccr9 (Mm02620030_m1), Il7ra (Mm00434295_m1), Rag2 (Mm00501300_m1), Cdkn1c (Mm01272135_g1) and Dtx1 (Mm00492297_m1). For each PCR, relative expression is determined using HPRT expression as a housekeeping gene. Alternatively, Universal Master Mix SybrGreen from Applied Biosystem was used as follows: Ccnd1 forward, 5′-GCGTACCCTGACACCAATCTC-3′, and reverse, 5′-CTCCTCTTCGCACTTCTGCTC-3′; Gzmb forward, 5′-CCACTCTCGACCCTACATGG-3′, and reverse, 5′-GGCCCCCAAAGTGACATTTATT-3′; Cdkn1a forward, 5′-ATCACCAGGATTGGACATGG-3′, and reverse, 5′-CGGTGTCAGAGTCTAGGGGA-3′; Hey1 forward, 5′-GCGCGGACGAGAATGGAAA-3′, and reverse, 5′-TCAGGTGATCCACAGTCATCTG-3′; Gapdh forward, 5′-AGGTCGGTGTGAACGGATTTG-3′, and reverse, 5′-TGTAGACCATGTAGTTGAGGTCA-3′; and Gsn (gelsolin) forward, 5′-ACCTTCTCCGGCTACTTCAAG-3′, and reverse, 5′-CAGAGCCACACCACTGATAGA-3′.

Accession codes.

GEO: microarray data, GSE50910.