Retinoic Acid Receptor γ Regulates B and T Lymphopoiesis via Nestin-Expressing Cells in the Bone Marrow and Thymic Microenvironments

Vitamin A is known to play an important role in regulating immune functions. Vitamin A deficiency in pregnant woman is associated with increased levels of infant morbidity and mortality (1). In addition, vitamin A supplementation has been reported to reduce mortality rates associated with measles and diarrheal infection (2, 3). Importantly, these improvements were linked to the immune-modulatory functions of vitamin A. Vitamin A supplementation in measles-infected children resulted in increased lymphocyte numbers and IgG Ab levels (4). Increased numbers of CD4 naive T cells have been reported after vitamin A supplementation in children (5). However, vitamin A supplementation in adults resulted in modest effects on lymphopoiesis (6). Furthermore, vitamin A diet restriction studies in rodents displayed an expansion of peripheral blood (PB) granulocytes and a decrease in PB B lymphocytes (7, 8).

We have previously reported that all-trans retinoic acid (ATRA), a naturally occurring vitamin A derivative, exerts pleiotropic effects on different immature hematopoietic subpopulations. ATRA aided in the in vitro maintenance of hematopoietic stem cell (HSC) potential when added to purified HSC-containing cultures, whereas ATRA enhanced the in vitro differentiation of myeloid progenitor cells in culture (9, 10). These varied effects of ATRA might be caused by the differential expression of RA receptors (RARs) among different immature hematopoietic subpopulations and by distinct effects of these RARs on hematopoiesis (11). In support of this, RARγ activation and deletion displayed reciprocal effects on HSCs, with activation enhancing HSC self-renewal and RARγ null mice (Rarγ^−/−) displaying reduced functional HSC content (11). In contrast, mice bearing RARα deletion (Rarα^−/−) displayed normal HSC number and function.

Adult Rarγ^−/− mice also developed a myeloproliferative-like disorder with increased granulocytes and granulocyte progenitors in the PB, bone marrow (BM), and spleen (12). Furthermore, BM erythropoiesis and B lymphopoiesis were suppressed in Rarγ^−/− mice. Interestingly, BM transplantation studies revealed that some of the deregulated hematopoiesis in Rarγ^−/− mice, including the elevated granulocytes and reduced BM B lymphopoiesis, was a consequence of Rarγ deletion in the BM microenvironment (BME), occurring extrinsically to the hematopoietic system (12).

HSCs reside in a specialized BME, which provides factors and signals necessary for HSC maintenance. Several cell types that form an integral part of the HSC BME have been identified, including cells of the osteoblast lineage, endothelial cells, and perivascular cells (13–17). In addition to HSC maintenance, populations of maturing hematopoietic cells have been revealed to develop within specialized BMEs, such as B lymphopoiesis (18). In B lymphopoiesis, CXCL12 is essential for the generation of the early-stage B cell precursor, the PreproB cell (19). Most PreproB cells were found in close proximity to CXCL12^hi reticular cells (20). As the B cells differentiated, the ProB cells migrated away from the CXCL12^hi reticular cells and were found to be associated with IL-7–expressing cells (20). Furthermore, Il7^−/− mice displayed a blockade at the ProB to PreB cell transition during B lymphopoiesis (21). In addition, osteoblast lineage-specific deletion of G protein α subunit resulted in impaired B lymphopoiesis at the ProB and PreB stages (22). These studies state the importance of BMEs in regulating B cell production. However, the cell type(s) constituting the specialized BME that aids in the regulation of PreB cells is largely unknown.

Similar to B cell development in distinct BMEs, T lymphopoiesis is supported by specialized microenvironment cells in the thymus. Early thymic progenitor cells that seed the thymus undergo a well-regulated T cell specification process coordinated by distinct thymic microenvironments (TMEs), which provide different factors necessary for T cell maturation (23). TMEs can be anatomically compartmentalized into two distinct cortical and medullary regions, containing epithelial cells, endothelial cells of the vasculature, and other stromal cells of mesenchymal origin. In addition to producing factors for T cell differentiation, cortical thymic epithelial cells (TECs) (cTECs) and medullary TECs (mTECs) and hematopoietic-derived dendritic cells ensure the development of self-restricted and self-tolerant mature T cells (24–26).

Furthermore, neural crest (NC)–derived thymic mesenchyme has been shown to have essential roles in regulating the embryonic stage of thymic development. Ablation of NC mesenchyme in the chick embryo resulted in the development of an abnormal thymus (27). NC mesenchyme has been suggested to regulate thymopoiesis by producing extracellular matrix necessary for T cell differentiation (28, 29). In addition, NC mesenchyme is involved in the development of TECs, thereby contributing to the development of different functional TMEs (30, 31). However, less is understood about the roles of NC mesenchyme in regulating thymic function postnatally (32, 33).

Retinoids and RARs can regulate different stages of hematopoietic differentiation. Furthermore, we have previously shown that the RARγ null microenvironment contributes to deregulated hematopoiesis; however, it was unclear which BMEs were causing these effects (12). Therefore, we wanted to characterize the microenvironmental cell types that regulate hematopoiesis via RARγ signaling. Recent studies using BME cell type–specific Cre transgenic mouse strains and cell lineage-tracing technologies based on reporter mouse strains have aided our understanding of the contribution of specific BME cell types toward hematopoiesis (34). In this study, we used such strains to conditionally delete RARγ in two distinct microenvironment cell types, osteoprogenitors (using Osterix [Osx]-Cre) and nestin (Nes)-expressing cells (using Nes-Cre). Osx is a zinc finger-containing transcription factor expressed specifically from the osteoprogenitor stage during osteoblastic differentiation (35). Nes is a type IV intermediate filament identified in the neural stem cell (36). Nes-expressing cells have been identified in the BME and share properties of BM mesenchymal stem cells (17).

In this study, we show that conditional deletion of Rarγ using OsxCre transgenic mice (37) crossed to Rarγ conditional knockout mice (38) (Osx-GFP:Cre⁺Rarγ^Δ/Δ mice) resulted in normal hematopoiesis, including HSC numbers and functions. Similar to Osx-GFP:Cre⁺Rarγ^Δ/Δ mice, targeted deletion of RARγ in Nes-expressing cells had no impact on HSC numbers or function. However, NesCre⁺:Rarγ^Δ/Δ mice had significantly increased PB granulocytes and significantly reduced numbers of PB B cells and CD4⁺ T cells, accompanied by significant reductions in BM PreB cells and altered thymic T lymphocyte cell development [significantly reduced thymus weights and reduced proportions of immature double-negative (DN) 1 and DN4 T cell precursors accompanied by significantly reduced numbers of TCRβ^low CD8 immature single-positive (ISP) cells and double-positive (DP; CD4⁺CD8⁺) T cells]. Moreover, a 10-d ATRA treatment of wild-type mice significantly increased BM PreB lymphocytes and increased the proportions of DN4 T lymphocytes.

Using NesCre transgenic mice crossed to Rosa26 enhanced yellow fluorescent protein (R26eYFP) reporter mice, we have also identified Nes-expressing cells in the TME. These cells share properties with NC-derived stromal cells, expressing cerebellar degeneration-related Ag 1 (CDR1) and platelet-derived growth factor receptor α (PDGFRα) and lacking expression of epithelial cell adhesion molecule (EpCAM) markers. Furthermore, these Nes-expressing cells express transcripts for Cxcl12 and stem cell factor (Scf).

Our studies reveal that Nes-expressing cells are an important BME contributor to the regulation of PreB lymphocytes. We have also identified Nes-expressing cells in the thymus and show that they contribute to the TME that is important in supporting the production of DN1, DN4, TCRβ^low CD8 ISP, and DP (CD4⁺CD8⁺) T lymphocytes.

The Osx-GFP:Cre, Nes:Cre, Rarγ^fl/fl, R26eYFP^ki/ki, and Rarγ^−/− mice have previously been described, and all were backcrossed onto the C57BL/6 background (37–41). The Rarγ strains were the kind gift of Pierre Chambon. Germline Cre recombinase activation and Rarγ germline deletion were identified by enhanced yellow fluorescent protein (eYFP) expression (in the Nes:Cre,Rarγ^Δ/ΔR26eYFP^ki/ki and Nes:Cre,Rarγ^+l+R26eYFP^ki/ki lines). All transgenic mice used in the experiments were male mice of 9–13 wk of age. B6.SJLPtprca (CD45.1⁺) and C57BL/6 (CD45.2⁺) male mice used in transplantation studies were at 8 wk of age at the time of transplantation and were purchased from Animal Resource Centre (Perth, WA, Australia). Short-term ATRA treatment studies were performed on C57BL/6 male mice at 8 wk of age at the start of treatment. The St. Vincent’s Hospital Animal Ethics Committee approved all animal experiments.

Mice were orally gavage-fed for 10 consecutive days with ATRA (5 mg/kg/d; Sigma-Aldrich, St. Louis, MO) or DMSO dissolved in peanut oil (Sigma-Aldrich), administered using sterile plastic feeding tubes 18 gauge × 38 mm (Instech Solomon, Plymouth Meeting, PA). Hematopoietic parameters in the mice were analyzed the day after the final dosage.

PBs obtained by retro-orbital bleeds were hemolyzed to remove RBCs, and the bones were flushed and crushed to obtain BM. Thymus and spleen were crushed to prepare single-cell suspensions in PBS with 2% FBS. PB parameters, BM, thymus, and spleen cellularity were analyzed using a hematological analyzer (Sysmex KX-21N; Sysmex, Kobe, Japan). Cells were stained with fluorescence or biotin-conjugated Abs against murine Ter119, CD71, B220, CD11b, Gr1, CD2, CD3, CD4, CD5, CD8, Sca-1, c-Kit, CD34, FLT3, FcγR (CD16/32), CD19, CD43, CD44, CD25, TCRβ (from eBioscience), IgM, CD105, and CD150 (from BioLegend). The lineage mixture for depletion of lineage-positive cells for the hematopoietic stem and progenitor cell analysis, and for the DN1-DN4 thymic cell analysis consisted of biotinylated Abs against CD2, CD3, CD4, CD5, CD8, CD11b, Gr1, B220, and Ter119. Biotinylated Abs were detected by secondary staining using streptavidin-conjugated Qdot-605 (Life Technologies). Cells were analyzed using a BD LSRII Fortessa (BD Biosciences). Analysis was performed using FlowJo software version 8.8.6 (Tree Star).

TME cells were isolated as previously described (42). In brief, the thymic lobes were minced in cold PBS, the fragments were allowed to settle on ice for 4 to 5 min, and the supernatant was discarded. Thymic stromal fragments were digested in 5 ml digestion medium containing collagenase type 1 (3 mg/ml; Worthington Biochemical) and DNase (1 mg/ml; Roche Diagnostics) at 37°C for 30–45 min. Fully digested thymic stromal cell suspensions were further enriched by centrifuging to 400 × g. The pellet fraction was washed and suspended in appropriate volumes of PBS/2% FBS/1% DNase for staining. Cells were stained with fluorescence-conjugated Abs against murine CD45, CD31, CDR1 (6C3 clone), EpCAM (G8.8 clone) (all from BioLegend), Ter119, B220, CD11b, Gr1, CD4, CD8 (all from eBioscience), and PDGFRα or IgG2a (from BD Biosciences). The lineage mixture for depletion of lineage-positive hematopoietic cells consisted of allophycocyanin-eFluor 780–conjugated Abs against CD4, CD8, CD45, Gr-1, B220, and Ter119, and the endothelial cells were also depleted using CD31 PECy7. Cells were sorted using a BD FACSAria Cell Sorter (BD Biosciences) or analyzed using a BD LSRII Fortessa (BD Biosciences). Analysis was performed using FlowJo software version 8.8.6 (Tree Star).

All transplant recipients were lethally irradiated (10 Gy, two split doses given 3 h apart) using a Gammacell 40 Exactor (Best Theratronics, Ottawa, ON, Canada) before transplantation. For the limiting dilution assay, cohorts of lethally irradiated CD45.1⁺ C57B6.SJL-Ptprca recipient mice (five mice per group) were injected with 10,000, 25,000, 50,000, and 200,000 BM cells obtained from either Nes:Rarγ^Δ/ΔR26eYFP or Osx:Rarγ^Δ/Δ mice or their wild-type littermates, mixed with 2 × 10⁵ Ptprca/C57BL/6 (CD45.2⁺/CD45.1⁺) congenic BM cells. Mice were assessed for donor cell PB reconstitution in lymphoid and myeloid lineages at 6 mo posttransplantation. HSC frequency was calculated and plotted using extreme limiting dilution analysis software (43).

RNA extraction and in-column DNase treatment were performed using the RNeasy Micro kit (Qiagen). cDNA were prepared using an AffinityScript quantitative PCR (qPCR) cDNA synthesis kit with random primers (Agilent Technologies). qPCR was performed using SensiMix II Probe Kit (Bioline) on the MX3000P Multiplex Quantitative PCR system (Stratagene). Primer sequences used for qPCR are in Supplemental Table I. The relative gene expression was calculated by 2^−ΔCT method using β₂-microglobulin as the reference gene.

Fresh frozen thymus samples were sectioned at 5 μm using a rotary RM 2125RTS microtome (Leica Microsystems, Wetzlar, Germany). The slides were fixed in ice-cold 100% acetone, rinsed three times in TBS, and blocked using 10% normal donkey serum in TBS. The slides were incubated for 1 h with chicken anti-GFP primary Ab or chicken anti-IgY control Ab. After washing three times in TBS, the slides were then incubated with donkey anti-chicken Alexa Fluor 488 for 1 h. The slides were then washed three times with TBS, incubated with DAPI for 1 min, washed with TBS, and mounted for viewing.

Unless otherwise described, statistics were performed using the unpaired two-tailed Student t test. Statistical analysis was performed using Prism 6 software (GraphPad), with p < 0.05 considered significant. For limiting dilution experiments, the data were analyzed with Poisson statistics to the single-hit model using extreme limiting dilution analysis software (43). The nonresponding values in log fraction were converted to the percentage of negative mice (absence of multilineage engraftment) by using the formula: percentage of negative mice = e anti-log fraction of nonresponding values.

To identify the BME cell type(s) contributing to the deregulated hematopoiesis observed upon global Rarγ deletion, we deleted Rarγ in osteoblast lineage cells using Osx-GFP:Cre mice (Cre recombinase fused to GFP expressed under the control of Osx promoter) (37). We also generated mice lacking Rarγ in Nes-expressing cells using Nes:Cre mice (Cre recombinase expressed under the control of the rat Nes promoter) (39). The transgenic Cre mice were crossed to conditional Rarγ^fl/fl mice (38). The majority of the NesCre⁺:Rarγ^fl/fl mice displayed germline Cre recombinase activation, which has been a reported unwanted side effect of this strain (34). In this study, we have therefore crossed the NesCre⁺;Rarγ^fl/fl strain to R26eYFP reporter mice (41) to readily screen and exclude mice bearing germline Cre recombinase activation.

The resultant mice (Osx-GFP:Cre⁺Rarγ^fl/fl, designated as Osx:Rarγ^Δ/Δ, and Nes:Cre⁺Rarγ^fl/flR26eYFP, designated as Nes:Rarγ^Δ/ΔR26eYFP) had no obvious physical phenotype compared with their control mice (Osx-GFP:Cre⁺Rarγ^+/+, designated as Osx:Rarγ^+/+, or Nes:Cre⁺Rarγ^+/+R26eYFP, designated as Nes:Rarγ^+/+R26eYFP).

We analyzed PB parameters in 9- to 12-wk-old male Nes:Rarγ^Δ/ΔR26eYFP and Osx:Rarγ^Δ/Δ mice and compared against their respective controls (Nes:Rarγ^+/+R26eYFP and Osx:Rarγ^+/+). Nes:Rarγ^Δ/ΔR26eYFP mice had similar PB leukocyte, erythrocyte, and platelet counts to their age-matched controls (Fig. 1A, Supplemental Fig. 1A, 1B). There were significant reductions in PB B220⁺IgM⁺ B cells and CD4⁺ T cells in the Nes:Rarγ^Δ/ΔR26eYFP mice in comparison with the control (Fig. 1B, 1C). The numbers of CD8⁺ T cells numbers were reduced in the Nes:Rarγ^Δ/ΔR26eYFP mice by 20%, although this was not significant (Fig. 1C). In contrast with the reductions in PB lymphoid cells, there were significantly increased PB granulocytes (CD11b⁺Gr1⁺) in the Nes:Rarγ^Δ/ΔR26eYFP mice (Fig. 1D). In contrast, we observed no alterations in any of the PB parameters analyzed in the Osx:Rarγ^Δ/Δ mice (Fig. 1E–H, Supplemental Fig. 1C, 1D). These data demonstrate that ablation of Rarγ in Nes-expressing cells results in significant reductions in the PB B and CD4⁺ T lymphocytes.

We next determined the impact of Rarγ deletion in either osteoblast lineage- or Nes-expressing cells on their ability to support BM hematopoiesis. There were no changes in BM leukocyte counts in Nes:Rarγ^Δ/ΔR26eYFP or Osx:Rarγ^Δ/Δ mice compared with their wild-type controls (Table I). Furthermore, both Nes:Rarγ^Δ/ΔR26eYFP and Osx:Rarγ^Δ/Δ mice showed normal levels of mature granulocytes (CD11b⁺Gr1^hi), immature granulocytes (CD11b⁺Gr1^lo), and erythroid cells (Ter119⁺) (Table I). We subdivided BM erythroid cells in the stages of erythroid differentiation based on their expression of CD71 and Ter119 and found no significant alterations in these cells in either the Nes:Rarγ^Δ/ΔR26eYFP or the Osx:Rarγ^Δ/Δ mice (Table I) (44). Furthermore, analysis of the spleen revealed no alterations in the weight and cellularity in either Nes:Rarγ^Δ/ΔR26eYFP or Osx:Rarγ^Δ/Δ mice (Table II), although there was a trend to reduced erythrocytes in the spleen of Osx:Rarγ^Δ/Δ mice compared with their wild-type controls (p = 0.08; Table II).

In contrast, in accordance with the reduced B lymphocytes observed in the PB, Nes:Rarγ^Δ/ΔR26eYFP mice had significantly reduced BM B lymphopoiesis. Nes:Rarγ^Δ/ΔR26eYFP mice displayed a 26% reduction in recirculating (B220^+highIgM⁺) (45) and significant reductions in BM-derived B220^+lowIgM⁺ and immature (B220⁺IgM⁻) B lymphocytes in comparison with the control (Fig. 1I, 1J). We further subdivided the immature B cells into stages of B cell differentiation based on CD19 and CD43 expression (Fig. 1I). Nes:Rarγ^Δ/ΔR26eYFP mice displayed a significant reduction in PreB (B220⁺IgM⁻CD19⁺CD43⁻) cells with no alteration in the ProB (B220⁺IgM⁻CD19⁺CD43⁺) or PreproB (B220⁺IgM⁻CD19⁻CD43⁺) stages of B lymphopoiesis (Fig. 1K).

We have previously reported that 8-wk-old Rarγ global knockout mice (Rarγ^−/−) displayed a significant reduction in both mature and immature B cells (12). In this study, we discovered a significant block at the ProB and PreB stages of B cell development in Rarγ^−/− mice (Fig. 1L). In contrast, we observed normal B cell development in the Osx:Rarγ^Δ/Δ mice in comparison with the control (Fig. 1M, Supplemental Fig. 1E). These data demonstrate that Rarγ deletion in Nes-expressing cells results in altered BM PreB lymphopoiesis, which may account for the reduced PreB cells observed in the global Rarγ^−/− mice (12).

We have previously reported that Rarγ^−/− mice have alterations in hematopoietic stem and progenitor cell numbers and function (11). Nes:Rarγ^Δ/ΔR26eYFP and Osx:Rarγ^Δ/Δ mice displayed similar levels of myeloid progenitor-containing [lineage⁻c-kit⁺Sca1⁻ (LKS⁻)] cells to their wild-type controls (Table III). No alterations were seen in the myeloid progenitor subfractions: granulocyte macrophage progenitors (LKS⁻CD34⁺CD16/32⁺) or common myeloid progenitors (LKS⁻CD34⁺CD16/32⁻) in either Nes:Rarγ^Δ/ΔR26eYFP or Osx:Rarγ^Δ/Δ mice (Fig. 2A, 2B). There were trends to increased megakaryocyte erythroid progenitors (MEPs; LKS⁻CD34⁻CD16/32⁻) in Osx:Rarγ^Δ/Δ mice (Fig. 2B), but no changes in MEPs in Nes:Rarγ^Δ/ΔR26eYFP mice compared with their controls (Fig. 2A). Furthermore, HSC-containing [lineage⁻c-kit⁺Sca1⁺ (LKS⁺)] cell numbers were not altered in either Nes:Rarγ^Δ/ΔR26eYFP or Osx:Rarγ^Δ/Δ mice (Table III). When the LKS⁺ cells were fractionated into long-term HSCs (LKS⁺CD34⁻Flt3⁻), short-term HSCs (LKS⁺CD34⁺Flt3⁻), and multipotent progenitors (LKS⁺CD34⁺Flt3⁺), there were also no alterations in the numbers of these immature HSC subpopulations in either Nes:Rarγ^Δ/ΔR26eYFP or Osx:Rarγ^Δ/Δ mice (Fig. 2C, 2D).

We then determined the in vivo repopulating potential of the BM cells from the Nes:Rarγ^Δ/ΔR26eYFP and Osx:Rarγ^Δ/Δ mice. We assessed the numbers of competitive repopulating units in the BM of either Nes:Rarγ^Δ/ΔR26eYFP or Osx:Rarγ^Δ/Δ mice by limiting dilution assays (46). We detected no significant differences in the HSC frequencies in either Nes:Rarγ^Δ/ΔR26eYFP or Osx:Rarγ^Δ/Δ mice when compared with their wild-type littermate controls (Fig. 2E, 2F). Taken together, these data suggest that Rarγ signaling in Nes-expressing cells and Osx-expressing osteoprogenitors does not contribute to HSC or myeloid progenitor cell maintenance.

There were no differences in the animal weights of 9- to 12-wk-old male Nes:Rarγ^Δ/ΔR26eYFP mice (Fig. 3A). Interestingly, however, 9- to 12-wk-old Nes:Rarγ^Δ/ΔR26eYFP mice displayed a 30% reduction in thymus weight (p = 0.005) (Fig. 3B). This was accompanied by significant reductions in the total cellularity of the thymus in the Nes:Rarγ^Δ/ΔR26eYFP mice (Fig. 3C). We observed significantly reduced numbers of CD4 single-positive (SP) cells (CD4⁺CD8⁻) and DP (CD4⁺CD8⁺) thymic T lymphocytes in the thymus (Fig. 3D, Supplemental Fig. 2A, 2B), but no significant alterations in the CD8 SP (CD4⁻CD8⁺) numbers in the thymus (p = 0.08) (Fig. 3D, Supplemental Fig. 2A, 2B). We next examined the numbers of TCRβ⁺ (high) CD4 and CD8 T cells and TCRβ^low CD8 ISP (Fig. 3E, Supplemental Fig. 2C, 2D). There were significant reductions of TCRβ^low CD8 ISP T cells in the thymus of Nes:Rarγ^Δ/ΔR26eYFP mice (Fig. 3E). In contrast, although there were trends to reductions, the numbers of TCRβ⁺ single CD4⁺ (p = 0.08) and TCRβ⁺ single CD8⁺ cells (p = 0.09) were not significantly altered in the thymus of Nes:Rarγ^Δ/ΔR26eYFP mice compared with their wild-type controls (Fig. 3E).

We analyzed the earlier stages of T cell differentiation based on CD44 and CD25 expression within the immature hematopoietic lineage-negative (CD4⁻8⁻; DN) fraction of the thymus. The Nes:Rarγ^Δ/ΔR26eYFP mice displayed no alterations in the proportions of DN1 to DN2 stages of T cell differentiation (Fig. 3F, 3G). Interestingly, however, there was a significant increase in the proportions of DN3 (Lin-CD44⁻CD25⁺) accompanied by a significant decrease in the percentages of DN4 (Lin⁻CD44⁻CD25⁻) T cell precursors in the Nes:Rarγ^Δ/ΔR26eYFP mice (Fig. 3F, 3G). When adjusting for cellularity differences between the two genotypes, there were significantly reduced numbers of both DN1 (Lin-CD44⁺CD25⁻) and DN4 T cell precursors (Fig. 3H). In contrast, we observed normal thymic size and T cell numbers in Osx:Rarγ^Δ/Δ mice (Table IV). Collectively, these data suggest that altering Rarγ in Nes-expressing cells might play a role in regulating thymic size and the production of immature DN4 T cells, TCRβ^low CD8 ISP T cells, and DP T cells in the thymus.

Nes-expressing cells have previously been reported in the BME (17). The occurrence of thymic atrophy together with the significant reductions in DN1 and DN4 immature T cells accompanied by significant reductions in the TCRβ^low CD8 ISP T cells and DP T cells in Nes:Rarγ^Δ/ΔR26eYFP mice prompted us to look for Nes-expressing cell in the TME. We used a lineage tracing approach by using the wild-type Nes:Rarγ^+/+R26eYFP mice, wherein all cells derived from Nes-expressing cells can be readily identified based on eYFP expression.

Investigation of the presence of Nes-targeted cells in whole thymus sections revealed eYFP⁺ cells in both the medullary and the cortical regions of the thymus, in perivascular regions (Fig. 4A, 4B). Analysis of the collagenase-digested TME cells revealed an absence of eYFP expression in the hematopoietic (lineage and CD45⁺) and endothelial (CD31⁺) thymic fractions of Nes:Rarγ^+/+R26eYFP mice (data not shown). We next separated lin⁻CD45⁻CD31⁻ TME cells, which constitute 1.70 ± 0.62% (mean ± SD) of the total thymic cells, based on EpCAM and CDR1 expression, and analyzed them for eYFP expression (47). Based on this analysis, we identified the four previously described distinct TME cell populations: 1) EpCAM⁺ CDR1^−ve cells (CDR1⁻ mTECs; 5.43 ± 2.16% of the TME cells), 2) EpCAM⁺ CDR1⁺ cTECs (CDR1⁺ cTECs; 6.23 ± 1.14% of the TME cells), 3) CDR1⁺EpCAM^−ve cells (CDR1⁺ stroma; 15.54 ± 4.38% of the TME cells), and 4) EpCAM^−ve CDR1^−ve cells (72.82 ± 5.74% of the TME cells) (Fig. 4C). Intriguingly, when analyzed for eYFP expression, we found eYFP expression restricted to ∼28% of the CDR1⁺ stroma (Fig. 4D, right panel). We next examined PDGFRα expression within the four different Lin⁻CD45⁻CD31⁻ TME cells, subdividing the CDR1⁺ stroma into eYFP⁺ and eYFP^−ve cells for these analyses. Only the eYFP⁺ CDR1⁺ stroma showed significant expression of PDGFRα above the IgG2a isotype control (Fig. 4E, right panel).

We next determined whether any of the TME cells were altered in the Nes:Rarγ^Δ/ΔR26eYFP mice. There were significant reductions in the proportions of CDR1⁺ cTECs (Fig. 4H), but no changes in the proportions of any of the other TME cells, including eYFP⁺ CDR1⁺ stroma (Fig. 4F, 4G, 4I, 4J), in the Nes:Rarγ^Δ/ΔR26eYFP mice compared with Nes:Rarγ^+/+R26eYFP mice.

To confirm that Nes expression was restricted to the CDR1⁺ stroma, we performed mRNA expression analyses on FACS-purified TECs and CDR1⁺ stroma from wild-type C57BL/6 mice (which did not express NesCre or R26eYFP). Nes mRNA was expressed at 24-fold abundance in the CDR1⁺ stroma compared with TECs (p = 0.0001; Fig. 5A). Moreover, mRNA levels of Rar subtypes, with the exception of Rarα1, were significantly higher in the CDR1⁺ stroma compared with TECs (Fig. 5B). Notably, among Rar subtypes, the Rarγ1 isoform showed the highest expression in the CDR1⁺ stroma compared with the TECs (Fig. 5B).

The reduction of DN1 and DN4 cells in the Nes:Rarγ^Δ/ΔR26eYFP mice was indicative of a CXCL12 defect, because CXCL12 is required for homing of immature T cell progenitors to the thymus (48), in addition to β selection and differentiation beyond the DN3 stage (49, 50). To determine whether the CDR1⁺ stroma expressed Cxcl12, we separated CDR1⁺ stromal cells based on eYFP expression and performed gene expression analysis of Cxcl12 and other T cell growth factors. Cxcl12 and Scf mRNA levels were found to be higher in the CDR1⁺ stroma, regardless of eYFP expression, compared with TECs (Fig. 5C, 5D). Furthermore, in accordance with previous reports, the expression of IL-7 (Il-7), Delta-like ligand 1 (Dll1), and Dll4 was largely restricted to the TECs (Fig. 5E–G) (51, 52). Taken together, we have identified an Nes-expressing CDR1⁺ stromal cell subpopulation within the thymus, which expresses PDGFRα and high levels of transcripts for the Rars and the T cell–supportive factors, Cxcl12 and Scf.

Nes-expressing cells have previously been reported in the BME (17), and studies using Nes-GFP mice have shown there are at least two populations of BME Nes-expressing cells, based on the intensity of their expression of Nes (53). Furthermore, the cells expressing the highest levels of Nes were found in perivascular locations (53). The majority of the Nes-GFP cells were found to coexpress PDGFRα (54). Interestingly, cells expressing CDR1 (identified as 6C3⁺) have also been identified in murine BME (55), but no one has determined the coexpression of PDGFRα and CDR1 on Nes-expressing murine BME cells. We examined the expression of PDGFRα and CDR1 in Nes-targeted eYFP⁺ cells obtained from collagenase-digested bone and BM cells from Nes:Rarγ^+/+R26eYFP mice. Four different populations could be identified based on expression of PDGFRα and/or CDR1 or no expression of PDGFRα or CDR1 (Supplemental Fig. 3).

Finally, we sought to determine whether gain of function of RARs could contribute to accelerated T and B lymphopoiesis. Adult wild-type C57BL/6 mice were gavage-fed with the pan RAR agonist, ATRA, for 10 d at a daily dose of 5 mg/kg and analyzed 1 d after the final dose was administered. Short-term ATRA treatment showed no alterations in PB leukocyte levels and BM cellularity when compared with the control (DMSO)-treated group (Fig. 6A, 6B). Given that the microenvironments in the BM and thymus caused the reductions in BM B lymphopoiesis and thymic T lymphopoiesis, respectively, in Nes:Rarγ^Δ/ΔR26eYFP mice, we focused our analysis on these lymphoid organs.

There were no obvious changes in the proportions of mature (B220⁺IgM⁺) or immature (B220⁺IgM⁻) B cells in the BM during this short time course of treatment (Fig. 6C). However, there were significant reductions in the proportions of ProB cells, accompanied by significantly increased proportions of PreB cells in the B220⁺IgM⁻ BM cells of ATRA-treated mice compared with DMSO-treated control mice (Fig. 6D). Analysis of the thymus revealed no significant alterations in thymus weight or cellularity after short-term ATRA treatment (Fig. 6E, 6F). In addition, the frequencies of CD4 SP, CD8 SP T cells, and DP T cells in the thymus were not altered after short-term ATRA treatment (Fig. 6G). In contrast, within the immature developing T cell subpopulations, we observed a significant reduction in the proportions of DN3 T cells accompanied by an increase (p = 0.06) in DN4 T cells after ATRA treatment (Fig. 6H). These data collectively show that short-term ATRA treatment accelerates in vivo BM PreB cell differentiation and the DN3-to-DN4 transition of developing thymocytes in the thymus, the opposite of that observed when Rarγ was deleted in Nes-expressing cells.

Loss of Rarγ in mice leads to the development of perturbed hematopoiesis, which has been shown to result, in part, from an altered hematopoietic cell microenvironment (12). To further characterize the microenvironment cell-type–specific contribution of Rarγ signaling toward deregulated hematopoiesis, we deleted Rarγ separately in Nes-expressing cells and Osx-expressing osteoprogenitor cells. We identified that when Rarγ was deleted in Nes-expressing cells, the mice developed perturbed PB, BM, and thymic lymphoid production, but, with the exception of increased PB granulocytes, no defects in other hematopoietic cell types were observed in Nes:Rarγ^Δ/ΔR26eYFP mice. In contrast, when Rarγ was deleted in Osx-expressing cells, there was no significant impact on hematopoiesis.

The increased granulocytes observed in Nes:Rarγ^Δ/ΔR26eYFP mice is interesting, because this is the most striking phenotype observed in Rarγ^−/− mice (12). However, the latter mice exhibit a much higher elevation in granulocytes, and this was also accompanied by significant increases in BM and spleen granulocyte populations, which were not observed in the Nes:Rarγ^Δ/ΔR26eYFP mice. It is therefore likely that other BME cell types contribute to the increased granulocytes observed in Rarγ^−/− mice.

Nes- and osteoprogenitor-specific deletion of Rarγ displayed no significant alterations in HSC and myeloid cell populations, nor did these mice exhibit extramedullary hematopoiesis, which was observed in the spleen in Rarγ^−/− mice (12). There was a trend toward increased numbers of LKS⁻ myeloid progenitors, in particular, MEPs, in the Osx:Rarγ^Δ/Δ mice; however, aside from a trend to reduced splenic erythrocytes in these mice, the mature erythrocytes and platelets in these mice were not altered, hence we cannot explain these trends to increased MEPs. The HSC defects observed upon global deletion of Rarγ are, at least in part, likely to be caused by the intrinsic loss of Rarγ (11). We cannot rule out, however, that loss of Rarγ in microenvironment cell type(s) other than Nes and osteoprogenitor cells contributes toward HSC maintenance.

Our studies do, however, reveal that Nes-expressing cells in the BM and TMEs have important roles in regulating lymphopoiesis. We have previously reported significant reductions in mature and immature B lymphocytes in the BM of Rarγ^−/− mice (12). In this article, we now show that this is due to significant reductions in the ProB and PreB stages of B cell development. Our data also reveal that Nes cell-targeted deletion of Rarγ resulted in significantly reduced BM-derived mature and immature B cells, occurring because of reduced numbers of PreB lymphocytes. Moreover, we also demonstrated that short-term in vivo pan RAR activation by ATRA treatment resulted in altered BM B lymphopoiesis, strikingly resulting in reduced proportions of ProB cells and increased proportions of PreB cells.

Our findings are in accordance with previous reports demonstrating a reduction in BM PreB cells upon induction of vitamin A deficiency and an increase in BM PreB cells in ATRA-treated mice (8, 56). Our study underpins the role of Rarγ in vitamin A–mediated regulation of BM B lymphopoiesis. Importantly, we show that the changes in PreB lymphocytes in vitamin A deficiency, and likely in ATRA-treated mice, are due to alterations in Nes-expressing BME cells, which, in turn, are regulated by Rarγ.

BM-derived Nes⁺ cells have been shown to produce cytokines important for B lymphopoiesis including Cxcl12 and Scf (17). It is yet to be determined whether Rarγ ablation leads to alterations in such cytokine production by BM Nes⁺ cells, because the germline deletion issues we have encountered have prevented us from obtaining sufficient numbers of mice to perform these studies. Importantly, however, we have uncovered a novel extrinsic role of Rarγ in regulating B lymphopoiesis through the Nes⁺ microenvironment cells, and future studies may reveal how PreB cells are regulated.

When Rarγ was deleted in Nes-expressing cells, we also observed significant reductions in thymic size, accompanied by significant reductions in the numbers of T cell precursors at the DN1 and DN4 stages of differentiation. This was accompanied by increased proportions (but not numbers) of DN3 T cells in the thymus. Furthermore, short-term pan RAR activation resulted in an increase in the proportions of DN4 T cell precursors, accompanied by significantly reduced proportions of DN3 T cell progenitors. We have also characterized an Nes-expressing cell that comprises part of the TME and likely plays a role in regulating DN4 T cell precursors. These cells are a subpopulation (∼30%) of the CDR1⁺EpCAM^−ve thymic stromal cells and express PDGFRα.

Interestingly, a recent study has shown that RA signaling regulates TEC proliferation in vitro (57). Furthermore, the thymic mesenchyme was identified as the major source of RA in the embryonic thymus (57). Moreover, a subset of CDR1⁺ thymic mesenchyme maintained the RA generation potential in the adult thymus (57).

Lineage tracing studies have previously reported that CDR1⁺EpCAM^−ve cells are NC mesenchymal cells (33). These cells were identified as α-smooth muscle actin–expressing pericytes found in close association with the thymic vasculature (33). Furthermore, PDGFRα-expressing NC mesenchyme has been shown to regulate TECs and the availability of intrathymic niches, thereby indirectly controlling T cell development (31, 32). Interestingly, the CDR1⁺EpCAM^−veeYFP⁺ thymic stromal cells express PDGFRα, whereas the other stromal cells, including the CDR1⁺EpCAM^−veeYFP^−ve thymic stromal cells, do not express PDGFRα. In further support, the Nes-targeted cells identified in our studies reside in perivascular regions (Fig. 4A, 4B), similar to that described for NC mesenchyme. Interestingly, the validation of another NesCre transgenic line using the R26LacZ reporter did detect some Nes-targeted cells in the thymus in perivascular regions; however, it was not clear whether these cells were endothelial cells or whether they were cells of another nature (58). Our studies have revealed that these cells do not express the endothelial cell marker, CD31, and that they are instead a subpopulation of thymic stromal cells.

The CDR1⁺EpCAM^−ve thymic stromal cells express high levels of Rar mRNA, particularly Rarγ. In addition, CDR1⁺EpCAM^−ve thymic stromal cells express high levels of Cxcl12 and Scf mRNAs compared with TECs. Both of these cytokines have been shown to have important roles in thymopoiesis, which may explain the immature lymphoid defects observed in Nes:Rarγ^Δ/ΔR26eYFP mice (35, 59, 60). It is currently not clear why there is a reduction in the proportions of DN1 T cell precursors. This could reflect reduced homing of early lymphoid progenitors from the BM to the thymus, because CXCL12 has been shown to be important to this process (48, 61). Homing properties and the numbers of early thymic progenitors were not determined but would be of interest in the future. However, our observation of a reduction of DN4 T cells upon Nes cell-specific deletion of Rarγ fits with the importance of CXCL12 in DN3-to-DN4 transition (49, 50). Because of significant germline deletion issues, we have not been able to obtain sufficient mice to determine whether loss of Rarγ in Nes-expressing CDR1⁺EpCAM⁻ thymic stromal cells results in deregulated expression of Cxcl12 or Scf, but these studies would also be of interest.

Our observation that the proportions of cTECs were significantly downregulated further supports previous studies showing that PDGFRα-expressing mesenchyme regulates the proliferation of epithelial cells, and hence thymic size (31). Interestingly, a separate study using fetal thymus organ cultures showed that RA signaling regulated the proliferation and differentiation of cTECs via mesenchymal cells (57). Our data suggest that RARγ signaling in Nes-positive NC mesenchyme cells is a key regulator of cTECs and thymic size.

Our data revise the previously suggested roles of TECs and NC-Mesenchyme in T cell development. Previously, it was thought that TECs were the sole source of Scf and CXCL12, which are critical for T cell development (61, 62). However, our data imply that the NC mesenchyme express much higher levels of Scf and Cxcl12 than TECs, whereas TECs express the Notch ligand Dll4, which is critical for T lymphopoiesis (52, 63). This would explain why stripping the mesenchyme from the embryonic thymus has a deleterious effect on T cell production (64), because removal of the mesenchyme would also deplete the major source of Scf and CXCL12 (64).

Taken together, our data demonstrate that RARγ signaling in Nes-expressing cells in the BM and TMEs is an important mediator of PreB and DN4 T lymphopoiesis, respectively. We further reveal that the Nes-expressing cells in the TME are a subpopulation of the NC mesenchyme and are important sources of Scf and CXCL12, which have also previously been shown to be expressed by Nes-expressing cells in the BME (17). Our gain-of-function studies in ATRA-treated wild-type mice have the opposite phenotype of the changes observed in BM PreB and thymic DN4 T lymphocytes in Nes:Rarγ^Δ/ΔR26eYFP mice. This also suggests that vitamin A predominantly regulates lymphopoiesis via Nes-expressing microenvironment cells, and that the immune defects observed in patients with vitamin A deficiency are likely to be a result of altered Nes-expressing lymphoid microenvironments.

We thank Pierre Chambon for generously providing the Rarγ strains used in this study. We also thank David Izon, Monique Smeets, Carl Walkley, Juan-Carlos Zuniga-Pflucker, and Daniel Gray for excellent discussions and comments on the manuscript. We thank Tanja Jovic for excellent technical assistance and the SVH Bioresources Centre for care of experimental animals.

The authors have no financial conflicts of interest.