|Home | About | Journals | Submit | Contact Us | Français|
Myeloproliferative syndromes (MPS) are largely considered to be intrinsic to hematopoietic cells. Here, we demonstrate that mice null for retinoic acid receptor gamma (RARγ), have an MPS that was induced solely by the RARγ deficient microenvironment. Eight week old RARγ-/- mice had significantly increased granulocyte/macrophage progenitors in bone marrow (BM), with elevated peripheral blood, BM and spleen granulocytes. The MPS phenotype continued for the lifespan of the mice and was more pronounced over time. Unexpectedly, transplant studies revealed this disease was not intrinsic to the hematopoietic cells. Wildtype BM transplanted into mice with an RARγ null microenvironment rapidly developed the MPS. Significantly elevated TNFα in RARγ-/- mice contributed to the MPS, but transplantation of TNFα-/- BM did not prevent the microenvironment-induced MPS. These data show that loss of RARγ results in a non-hematopoietic cell intrinsic MPS, revealing the capability of the microenvironment to be the sole cause of hematopoietic disorders.
Studies of the roles of the hematopoietic microenvironment have markedly increased in recent years, primarily due to the identification of the bone-forming osteoblast as being a critical component of the hematopoietic stem cell (HSC) niche, the place where HSCs reside and are primarily regulated (Calvi et al., 2003; Zhang et al., 2003). In addition to this important role in regulating HSC self-renewal and differentiation, the bone marrow (BM) microenvironment has been proposed to consist of various other niches, first termed hematopoietic inductive microenvironments (Trentin, 1971), which are areas of the BM that are highly specialized for the development of different maturing hematopoietic cell types. This concept has been supported by the recent identification of specific niches for B lymphocytes (Tokoyoda et al., 2004) and megakaryocytes (Avecilla et al., 2004) in the BM.
The hematopoietic microenvironment is not only supportive for the development of hematopoietic cells, but is also known as a highly preferred site for the metastasis of certain cancer cell types, including breast, prostate, and melanoma cancers. A recent report demonstrated that the cytokine receptor activator of NF-κB ligand (RANKL), a member of the tumor necrosis factor (TNF) family of cytokines, and which is expressed by osteoblasts, recruited epithelial and melanoma cancer cells expressing RANK to the BM, where they subsequently lodged and formed secondary tumors (Jones et al., 2006). A similar chemoattractant role for non-hematopoietic cancer cells has been described for the chemokine, CXCL12 (Muller et al., 2001), which is also expressed by osteoblasts and is known to be an important mediator of the homing of HSCs to the BM (Lapidot et al., 2005). Despite these recent advances, however, virtually nothing is known about the potential involvement of the BM microenvironment in the initiation of different diseases of hematopoietic origin, or the underlying factors that influence the development of these disorders.
Myeloproliferative syndromes (MPS) are a heterogeneous subclass of nonlymphoid hematopoietic neoplasms for which, with the exception of chronic myeloid leukemia (CML), hypereosinophilic syndrome, and juvenile myelomonocytic leukemia, the causes remain largely unknown (Van Etten and Shannon, 2004). To date, MPS have been considered to be of hematopoietic origin (Kogan et al., 2002), and the involvement of Bcr-Abl in CML (Van Etten and Shannon, 2004) together with various studies utilizing mouse models support this hypothesis (Araki et al., 2004; Le et al., 2004; Passegue et al., 2004; Wernig et al., 2006; Yan et al., 1994). Despite this, the molecular basis of a significant subset of other MPS, such as myelofibrosis and myelodysplastic syndrome (MDS), remains unknown. Likewise, it is unclear as to whether the hematopoietic (BM) microenvironment may play an active part in promoting and/or supporting the development of MPS.
Retinoic acid receptors (RARs) have been associated with many different diseases, including cancer, and retinoid-based therapies are increasingly being utilized to treat such disorders (Altucci and Gronemeyer, 2001). There are three RAR subtypes: RARα, RARβ and RARγ, all of which are highly conserved amongst species. The vitamin A derivative, all-trans retinoic acid (ATRA), is the naturally occurring ligand for all RARs.
The RARs are nuclear hormone receptors that act as ligand-dependent transcriptional regulators: in their liganded state they activate transcription, whereas in the non-liganded form they repress transcription of their target genes (Minucci and Pelicci, 1999). RARs have numerous direct target genes which have retinoic acid response elements in their promoter region (Balmer and Blomhoff, 2002; Balmer and Blomhoff, 2005).
The generation of RAR-specific mutant mice has allowed delineation of the different roles of the RARs in organogenesis. We have recently shown that RARγ, but not RARα, null mice have 3-fold reduced numbers of hematopoietic stem cells (HSCs) (Purton et al., 2006). Here we show that loss of RARγ also results in a myeloproliferative-like disease. This MPS was not intrinsic to the hematopoietic cell, but was induced by the RARγ null microenvironment. These novel findings reveal that the microenvironment can play a dominant role in both inducing and supporting the progression of hematopoietic diseases such as MPS and potentially leukemias.
We have observed that RARα and RARγ, but not RARβ isoforms are widely expressed in immature and maturing hematopoietic cell types ((Purton et al., 2006) and unpublished data). Furthermore, the natural ligand for RARs, ATRA, has potent effects on enhancing HSC self-renewal and promoting differentiation of more mature granulocyte/macrophage progenitors, which we have previously demonstrated are likely due to the different actions of RARα and RARγ on hematopoietic cells (Purton et al., 1999; Purton et al., 2000; Purton et al., 2006). RARα-/- mice do not have any observable hematopoietic defects in vivo (Kastner et al., 2001; Purton et al., 2006), whereas RARγ-/- mice have significantly reduced numbers of HSCs accompanied by increased numbers of immature progenitor cells in their BM (Purton et al., 2006). To determine if loss of RARγ also affected the production of mature hematopoietic cells, we examined the cellularity and hematopoietic composition of peripheral blood (PB), BM, spleen and thymus preparations obtained from RARγ null, heterozygous and wildtype mice.
Eight week old RARγ-/- mice had significantly elevated PB and BM leukocytes compared to their wildtype littermates (Figure 1A). The elevated leukocyte levels were comparable to those achieved by oncoretroviral overexpression-induced models of MPD in C57Bl/6 background strains (Wernig et al., 2006). Spleen leukocyte numbers were also elevated in RARγ-/- mice, but this increase was not significant compared to the wildtype mice. Eight week old RARγ-/- mice also presented with splenomegaly, with spleen weights significantly increased (1.5 to 3-fold) compared to their wildtype littermates. In contrast, the thymic cellularity was significantly reduced (25%) in RARγ-/- mice compared to their wildtype littermates (Figure 1A). RARγ-/- mice have a growth deficiency (Lohnes et al., 1993), and at 8 weeks of age were approximately 20% smaller than their wildtype littermates (Table S1). When normalized to body size the thymocyte cellularity was similar to that of wildtype mice, however, there were significant increases in BM and spleen cellularity in RARγ-/- mice.
We used immunophenotypical analysis to further explore the effects of loss of RARγ on the production of mature hematopoietic cell lineages. There were marked increases in the numbers of CD11b+ Gr-1+ granulocytes in the PB (Figure 1B and Table S2), BM (Figure 1C and Table S2) and spleen (Figure 1D and Table S2) of RARγ-/- mice compared to their wildtype littermates. These increases occurred in both the immature and mature granulocyte compartments in the BM (Walkley et al., 2002) (Figure 1C and Table S2).
There were also significant reductions in the numbers of B lymphocyte subsets and erythrocytes in the BM of the RARγ-/- mice (Tables S1 and S2), however the numbers of these cells were unaltered in the PB and spleen (Table S1). In contrast to RARγ-/- mice, RAR +/- mice did not have altered numbers of mature hematopoietic cells (Figure 1 and Table S1).
We further focused on investigating the mechanisms behind the increased granulopoiesis in the RARγ-/- mice. Both the immature (CD11b+ Gr-1 dim) and mature (CD11b+ Gr-1 bright) granulocyte subsets were significantly elevated in the BM, suggesting that the increase in granulocytes was arising from an immature progenitor population (Walkley et al., 2002). Furthermore, we have recently shown that RARγ-/- mice have increased numbers of immature colony-forming cells (CFU-GEMM), which likely arise from accelerated differentiation of HSCs into progenitor cells (Purton et al., 2006). We therefore investigated the numbers of more committed granulocyte progenitors in the BM of RARγ-/- mice.
The numbers of day 7 GM-CSF, SCF+G-CSF and G-CSF-responsive colony-forming cells (CFCs) were significantly increased in RARγ-/- BM compared to that of wildtype littermates (Figures 2A and 2B). The numbers of more mature day 3 cluster-forming cells were also significantly increased in these mutants compared to wildtype mice (Figure 2C). The PB and spleen of RARγ-/- mice also had increased numbers of mature myeloid CFCs (CFU-GM) (Figures 2C and D), but the numbers of immature CFCs (CFU-GEMM) and immunophenotypic HSC/progenitor cells in these organs were not significantly different to RARγ+/+ mice (Table S3). These data demonstrated that there were increased numbers of committed myeloid progenitors in the BM, spleen and PB of RARγ-/- mice. In contrast, the lack of CFU-GEMM and HSC/progenitor cells in spleen and PB suggested that RARγ-/- mice were not presenting with an HSC mobilization phenotype.
To determine if the increased numbers of progenitors could be due to altered cytokine sensitivity, we investigated the numbers of BM CFCs formed in submaximal and supramaximal concentrations of G-CSF (Walkley et al., 2002). Both RARγ-/- and RARγ+/+ BM cells showed a similar cytokine response, hence it did not appear that the progenitor cells had altered sensitivity to cytokines (Figure 2F). A similar result was also observed in response to IL-3 (unpublished data).
It was possible that the increased numbers of CFCs were due to increased survival of these progenitors. To assess this, we established CFC assays and delayed cytokine addition to 24 and 48 hours. The numbers of CFCs that formed were then assessed at 7 days post-initiation. There were no differences in the survival of progenitors stimulated with G-CSF (Figure 2G), suggesting that the myeloid expansion was not a secondary response due to altered cell death of granulocyte progenitors.
Finally, we assessed the frequencies of common myeloid progenitors (CMP), granulocyte/macrophage progenitors (GMP) and megakaryocyte/erythroid progenitors (MEP) in RARγ-/- and RARγ+/+ BM (Akashi et al., 2000). The frequency of GMPs was significantly increased (1.8-fold) in RARγ-/- BM compared to that of their wildtype littermates (Figure 2H). Given that RARγ-/- mice also had significantly increased BM leukocyte cellularity compared to RARγ+/+ mice (Figure 1A), this resulted in an overall significant 2.1-fold increase in absolute numbers of GMPs in RARγ null BM compared to wildtype BM.
The elevated granulopoiesis in RARγ-/- mice could be compensatory if these cells had impaired functional capacity, hence we tested several of their functional properties. The oxidative bursts response was similar in BM samples of either genotype (Figure S1). Furthermore, recruitment of cells after intradermal injection of Zymosan A (which elicits an acute inflammation response) was similar in RARγ-/- mice compared to their wildtype littermates (Figure S1), and there was no difference between the morphology of the cell types that were recruited to the ear in response to Zymosan A (data not shown). Finally, there was no difference in the percentage of BM granulocytes undergoing apoptosis when assessed by annexin V staining (Figure S1). These data suggested that RARγ-/- granulocytes had similar functional potential to those of RARγ wildtype granulocytes.
We have recently reported that small numbers of RARγ-/- mice survive to approximately 12 months of age (Purton et al., 2006). To determine if the increased granulocyte phenotype persisted for the lifespan of RARγ-/- mice we investigated the cellularity and hematopoietic composition of different organs obtained from 12 month old RARγ-/- and RARγ+/+ mice.
The average PB leukocyte counts of the older RARγ-/- mice were dramatically increased compared to their wildtype littermates (RARγ+/+= 10.54 ± 0.51; RARγ-/-= 29.6 ± 3.32 × 103/μl blood, P<0.005). This was accompanied by significantly increased numbers of granulocytes in the PB, spleen and BM (Figure 3A-C). Platelets were also significantly elevated in the PB of these older knockout mice (RARγ+/+= 1480 ± 46.7; RARγ-/-= 1888 ± 66.7 × 103/μl blood, P<0.005). In contrast, erythrocyte numbers were significantly reduced in the PB of these aged mutant mice (RARγ+/+= 9.77 ± 0.32; RARγ-/-= 8.87 ± 0.15 × 106/μl blood, P<0.05).
Analysis of immature progenitors in these mice revealed that RARγ-/- mice had significantly increased numbers of CFU-GEMM and CFU-GM in their PB and spleen (Figure 3D, 3F, 3G and 3I), accompanied by strikingly elevated numbers of lineage-restricted granulocyte/macrophage CFCs in their bone marrow (Figure 3E and H). Significantly reduced numbers of mature B cells and erythrocytes were also observed in the bone marrow of the 12 month old RARγ-/- mice (Figure 3J).
The frequencies of the lineage-negative, c-kit-positive, Sca-1-negative (LKS-) progenitor cells were significantly increased in BM and spleens of 12 month old RARγ-/- mice (Table S4). The HSC-containing lineage-negative, c-kit-positive, Sca-1+ (LKS+) cells were also markedly elevated in the spleens of these mice (Table S4).
Histological analysis of different organs of 9-12 month old mice revealed that the bone marrow of the RARγ-/- mice was extremely hypercellular compared to their wildtype littermates (Figures 4A and 4B). The trabecular bones were virtually absent, and the cortical bones were dramatically thinner in these older RARγ-/- mice (Figure 4B). Bone marrow cells of the RARγ-/- mice were predominantly developing myeloid cells, with some megakaryocytes also obvious (Figures 4B and 4D). Reduced B lymphocyte foci were also evident in the spleens of the RARγ-/- mice compared to their wildtype littermates (Figure S2). Myeloperoxidase staining revealed that extramedullary hematopoiesis was occurring in the liver of the RARγ-/- mice, however we did not observe significant numbers of hematopoietic cells in their kidneys compared to their wildtype littermates (Figure S2). Strikingly, there were large foci of immature and maturing hematopoietic cells (including granulocytes, monocytes, megakaryocytes and erythrocytes) developing in adipose tissue in these older of the RARγ-/- mice, but not in their wildtype littermates (Figures 4E-H). During their lifespan, however, none of the animals developed leukemia or lymphoma.
Given the significantly increased myeloid compartment in the mice with tissue infiltration, together with lack of evidence of malignant transformation, the phenotype of the RARγ-/- mice best corresponds with that of a myeloproliferative-like disease (MPD-like, or MPS) (Kogan et al., 2002).
Myeloproliferative-like diseases are thought to arise from hematopoietic cells (Kogan et al., 2002). In a previous study we investigated the HSC frequency in the RARγ-/- BM (Purton et al., 2006). During the 6 months of monitoring recipient mice post-transplant we did not observe a MPS in wildtype mice that were transplanted with BM from the RARγ-/- mice. However, these transplants were performed with competing BM from wildtype congenic mice, which may have masked or prevented the occurrence of the myeloproliferation. Therefore we repeated the transplants using whole BM without competing cells and compared the hematopoietic phenotypes to that of 8 week old RARγ null mice (Figures 5A and 5B).
In contrast to the MPS observed in RARγ-/- animals, at 8 weeks post-transplant all wildtype recipients of either RARγ+/+ or RARγ-/- BM had similar PB leukocyte and granulocyte counts (Figures 5C and 5D). The PB cellularity did not increase during the 6 months the mice were monitored post-transplant (at 6 months post-transplant, leukocytes: RARγ+/+= 13.72 ± 0.93; RARγ-/-= 11.5 ± 0.61 × 103/μl blood; granulocytes: RARγ+/+= 0.91 ± 0.17; RARγ-/-= 0.80 ± 0.12 × 103/μl blood). Bone marrow and spleen cellularity and lineage contribution were also comparable between the recipients of the two genotypes (data not shown). Complete reconstitution by RARγ+/+ or RARγ-/- cells was confirmed by immunophenotypical analysis at each time point of analysis, hence these results were not due to an inability of the RARγ-/- BM cells to engraft in the congenic recipients. These data therefore demonstrate that the myeloproliferation did not occur when RARγ-/- hematopoietic cells were supported by a wildtype microenvironment.
To determine if the microenvironment of the RARγ null mice was inducing the MPS we performed reciprocal transplants. Congenic wildtype cells were transplanted into lethally irradiated RARγ+/+ or RARγ-/- recipient mice. By 5 weeks post-transplant, RARγ-/- mice transplanted with wildtype cells had significantly elevated PB leukocytes and granulocytes (Figures 5E and 5F, Table S5). The increases in leukocytes and granulocytes were even more profound compared to 8 week old non-transplanted RARγ-/- mice (Figures 5A and 5B, Table S2). Immunophenotypical analysis confirmed the hematopoietic cells were of wildtype origin rather than endogenous recovery of RARγ null hematopoietic cells.
In addition to the markedly elevated granulocytes in these transplant recipients, BM B lymphopoiesis and erythropoiesis were also significantly suppressed when the wildtype congenic BM was transplanted into the RARγ-/- mice (Figure 5G, Table S5). This was also more profound compared to the B lymphocyte and erythrocyte phenotype observed in the BM of 8 week old RARγ-/- mutants (Table S1 and S2). Hence, the myeloproliferative-like disease observed in RARγ-/- mice was not intrinsic to the hematopoietic cells, but was induced by the RARγ deficient microenvironment.
We have previously reported that 8 week old RARγ-/- mice have three-fold reduced numbers of HSCs accompanied by increased numbers of progenitors, including day 12 colony-forming unit-spleen (CFU-S) (Purton et al., 2006). Given that the trabecular bone has been described as being a key component of the HSC niche (Calvi et al., 2003; Zhang et al., 2003) and that trabecular bone was virtually absent in 12 month old RARγ-/- mice (Figure 4B), we wished to determine whether the microenvironment of 8 week old RARγ-/- mice was impaired in its ability to support HSCs.
Histological sections of undecalcified tibiae revealed significantly reduced trabecular bone in 8 week old RARγ-/- mice. Histomorphometric quantitation of the tibiae sections from RARγ-/- mice demonstrated significantly fewer and more dispersed trabeculae, decreased trabecular volume but normal trabecular thickness in tibiae obtained from RARγ-/- compared to wildtype mice (LEP and NA Sims, manuscript in preparation). This resulted in an overall 1.6-fold reduction in the number of trabeculae in these mice. Despite this, there were similar percentages of osteoblasts per bone surface in 8 week old RARγ-/- mice compared to wildtype mice. In contrast, the reduction in trabecular bone was accompanied by 1.6-fold increased numbers of osteoclasts per bone surface, indicating increased osteoclastogenesis was the predominant cause of the osteopenia in 8 week old RARγ-/- mice (LEP and NA Sims, manuscript in preparation).
To assess whether this reduction in trabecular bone impaired HSC self-renewal, we transplanted lethally irradiated RARγ+/+ or RARγ-/- recipient mice with wildtype congenic BM. At 8 weeks post-transplant (a similar time of exposure to the niche as that of the 8 week old mutants used in our previous studies (Purton et al., 2006)) we assessed the numbers of CFCs, CFU-S and HSCs (using the limiting dilution assay) in the transplanted mice as previously described (Purton et al., 2006; Szilvassy et al., 1990; Walkley et al., 2005).
The numbers of CFU-GM were significantly increased in PB and spleen, but not BM of RARγ-/- recipient mice (Figure S3A-C). In contrast, there were no differences in the numbers of CFU-GEMM, LKS+, CFU-S or HSCs when wildtype BM was exposed to an RARγ-/- microenvironment compared to an RARγ+/+ niche (Figure S3D-H). In vitro co-culture assays revealed that RARγ-/- stromal cells increased the production of maturing wildtype BM cells during 14 days of culture (Figure S3I). Collectively, these data suggest that the RARγ-/- microenvironment increases the proliferation and production of relatively mature hematopoietic cells, but does not affect the numbers of immature progenitors and HSCs.
In order to further elucidate the mechanisms behind the myeloproliferation in the RARγ-/- mice, we examined the expression of different inflammatory mediators, myeloid cytokines and JunB, which is a putative target of retinoic acid (Balmer and Blomhoff, 2002), and induces an MPD (Passegue et al., 2004), in the hematopoietic organs of the RARγ mutants.
The expression of TNFα was significantly increased in all three hematopoietic organs in RARγ-/- mice compared to their wildtype littermates (Figure 6A). In contrast, expression of other pro-inflammatory cytokines, IL-2 and IL-6 were not markedly altered in the organs (Table S6). The expression of IL-4 was significantly reduced in the BM and slightly but significantly elevated in the spleen of RARγ-/- mice (Table S6).
The expression of JunB was not altered in the BM (Table S6). Neither of the two major myeloid cytokines, GM-CSF or G-CSF were elevated in any organ (Table S6). In contrast, G-CSF expression was significantly reduced in the thymus of RARγ-/- mice compared to their wildtype littermates (Table S6).
Given that we had not observed altered numbers of T lymphocyte subsets in RARγ-/- mice (Table S1), we further investigated the expression of TNFα in purified populations of T lymphocytes. TNFα was predominantly expressed by thymocytes expressing CD4 and/or CD8 (Table S6), consistent with a previous report investigating the synthesis of TNFα by T lymphocyte populations (Giroir et al., 1992).
TNFα is a pro-inflammatory cytokine that is known to reduce the numbers of B lymphocytes in the BM and preferentially promote granulopoiesis via reductions in CXCL12 expression in BM (Ueda et al., 2005; Ueda et al., 2004). CXCL12 was unaltered in RARγ-/- BM and spleen, and was elevated in the thymus (Table S6). Hence, the reduced B lymphopoiesis observed in RARγ-/- BM (Table S1) did not appear to be a result of reduced CXCL12 expression that normally occurs in the BM during inflammation.
To determine the contribution of deregulated TNFα signaling to the occurrence of the MPS we transplanted RARγ+/+ or RARγ-/- mice with TNFα-/- BM. PCR genotyping of BM obtained at the time of analysis confirmed full engraftment of the transplanted mice with TNFα-/- BM. Furthermore, Q-RT-PCR studies on BM, spleen and thymus harvested from these mice confirmed that the levels of expression of TNFα in RARγ-/- recipients were reduced to that of the wildtype hosts transplanted with TNFα-/- BM (data not shown).
Transplantation of TNFα-/- BM significantly reduced the MPS in the RARα-/- recipient mice (Figures 6 and S4). When compared to the myeloproliferation observed in RARγ-/- mice transplanted with wildtype BM, in those transplanted with TNFα-/- BM the fold-increases in PB leukocytes were reduced by approximately 50% (Figure 6B). The numbers of BM leukocytes were not altered (data not shown), whereas the spleen leukocyte counts were restored to the levels of the wildtype transplant recipients (Figure 6F).
While still elevated above that of wildtype recipients, the fold-increases in granulocytes in all organs were reduced by approximately 50% in BM and spleen, and were 3-fold reduced in PB of RARγ-/- mice transplanted with TNFα-/- BM compared to those transplanted with wildtype BM (Figures 6C, 6E, 6G, S4).
The numbers of B220+/IgM+ cells remained significantly lower in the BM of RARγ-/- recipients transplanted with TNFα-/- BM, however, the numbers of immature B220+/IgM- cells in these recipients were restored to that of wildtype recipients (Figure 6H). Finally, BM erythropoiesis in RARγ-/- recipients transplanted with TNFα-/- BM was restored to the levels observed in wildtype recipients (Figure 6H). These data indicate that deregulated TNFα production significantly contributes to, but is not the sole cause of, the myeloproliferative-like disease that is induced by the microenvironment of RARγ null mice.
Approximately 1/3rd of the RARγ-/- recipients had PB leukocyte counts above 48 × 106 cells/ml (twice that normally achieved by oncoretroviral MPD in a C57BL/6 background) at 8 weeks post-transplant, and had to be euthanized due to poor condition. These mice also had severe anemia, reflected by significantly reduced PB erythrocyte counts and hemoglobin content (data not shown).
PB smears of these recipient mice revealed highly elevated numbers of circulating immature myeloid cells and numerous abnormal erythrocytes in the RARγ-/- recipients compared to the wildtype recipients (Figures 7A and 7B). FACS analysis confirmed the cells in both recipient types were of wildtype donor origin (Figures 7C and 7E), and showed that the majority (74.9 ± 1.89%, n = 3) of the PB leukocytes in the RARγ-/- recipients were immature myeloid cells co-expressing intermediate levels of Gr-1 and CD11b (Figure 7F). In contrast the same donor cells transplanted into a wildtype microenvironment had PB content of approximately 15.1 ± 1.55% cells (n = 4) expressing higher levels of Gr-1 and CD11b (Figure 7D).
To determine if leukemic transformation of the cells had occurred, we transplanted whole spleen cells consisting of 2.5 × 106 leukocytes (containing 25% immature granulocytes by FACS) from one such primary RARγ-/- recipient (Figures 7B, 7E and 7F) into lethally irradiated wildtype and RARγ-/- secondary recipient mice and monitored their recovery during 8 weeks post-transplant.
At 8 weeks post-transplant the average CD45.1+ donor reconstitution was similar between groups (approximately 80%, data not shown). Surprisingly, the MPD-like phenotype was highly dependent upon the secondary recipient genotype, demonstrating that the hematopoietic cells had not undergone leukemic transformation during the primary transplant. When injected into wildtype secondary recipients, the percentage of donor-derived granulocytes at 8 weeks post-transplant reverted to 26.6 ± 6.36% (n = 5), with many mature granulocytes evident (Figures 7G and 7H). In marked contrast, when injected into RARγ-/- recipients the splenic leukocytes reconstituted an average of 78.87 ± 9.31% immature granulocytes in the PB of these recipients (Figures 7I and 7J, n = 5, P<0.003 wildtype vs RARγ-/- recipients). These data therefore revealed that an RARγ null microenvironment was absolutely required to maintain the MPS.
Our studies show that loss of one of the major receptors for vitamin A, RARγ, results in a microenvironment-induced MPS. The myeloproliferation likely arises from the GMPs in the bone marrow and progresses markedly with age. Profound loss of trabecular bone in 12 month old mice correlated with marked increases in the numbers of immature CFCs detectable in PB and spleen and extramedullary myelopoiesis in other organs including liver and adipose tissue. The microenvironment-induced myeloproliferation partially involves elevated TNF expression, as transplantation of TNFα-/- BM into RARγ-/- mice markedly ameliorated the MPD-like phenotype. However, we did not observe any evidence of progression to leukemia during the lifespan of these mice. Furthermore, the RARγ-/- microenvironment was absolutely required to sustain the MPS, as the phenotype reverted when either BM (Figures 5C and 5D) or spleen cells (Figures 7G and 7H) from RARγ-/- mice were transplanted into wildtype recipients.
An interesting observation in the older RARγ-/- mice was the hematopoietic support potential of the adipose tissue in these mice. Adipocytes are present in bone marrow, albeit in low numbers in mice. Several previous reports have suggested that adipocytes and adipose tissue are capable of supporting hematopoiesis, including myelopoiesis (Corre et al., 2004; Cousin et al., 2003; Hangoc et al., 1993). Leptin, a hormone produced by adipocytes, has been shown to augment myeloid colony formation (Bennett et al., 1996). In addition, the leptin receptor has been shown to be expressed by immature myeloid cells (Gainsford et al., 1996). Adiponectin, another protein expressed by adipocytes, has also been linked to the regulation of immature myeloid cells, however, it has been described as being a negative regulator of these progenitors (Yokota et al., 2000). Interestingly, TNFα has been shown to increase leptin and reduce adiponectin expression in adipose tissue (Vettor et al., 2005), hence it is possible that increased TNFα signaling directly enhanced the hematopoietic support potential of adipose tissue in the older RARγ null mice. Regardless of the mechanism, it appears that under conditions of extreme hematopoietic stress, adipose tissue can serve as an additional organ capable of supporting extramedullary hematopoiesis.
A previous study also showed that vitamin A deficiency (VAD) in mice resulted in myeloproliferation (Kuwata et al., 2000). The VAD mice did not show external signs of infections, and sera screening and cytokine analysis of these mice did not reveal any infectious cause of the myeloproliferation (Kuwata et al., 2000). This MPD-like phenotype took approximately 14 weeks to become evident and was more profound than that we have observed in non-transplanted RARγ-/- mice (Figure 1) but less severe than we have observed in approximately 1/3rd of RARγ-/- recipients transplanted with wildtype congenic BM (Figures 6B, 6E and 6F). When the VAD mice were fed a vitamin A supplemented diet, their granulocyte phenotype was reversed (Kuwata et al., 2000).
Although there are similarites in the studies performed by Kuwata et al (Kuwata et al., 2000) to ours, there are also distinct differences. The numbers of B220+ cells were unaltered in the BM of VAD mice compared to those a fed vitamin A-supplemented diet. In contrast, we observed a dramatic reduction in the numbers of B lymphocytes in mice that had an RARγ null microenvironment. Furthermore, unlike our findings, the numbers of colonies that formed from BM from VAD mice were not significantly increased compared to BM from mice fed the supplemented diet. Instead, the increased granulocytes were found to result, at least in part, to impaired apoptosis in the granulocytes of VAD mice (Kuwata et al., 2000). We did not observe altered granulocyte apoptosis in our studies (Figure S1).
A similar elevated granulocyte phenotype to the one observed in RARγ-/- mice was also observed when mice were fed a pan-RAR antagonist for 10 days (Walkley et al., 2002). Like that of RARγ-/- mice, this increase was accompanied by increased numbers of granulocyte progenitors and was not due to altered survival properties of granulocytes or their progenitors.
Vitamin A activates both the RARs and the retinoid X receptors (RXRs), which form heterodimers with both RARs and other members of the steroid hormone super family to activate transcription. It is therefore likely that some of the defects observed in VAD mice, including impaired apoptosis of granulocytes were due to loss of RXR function. Our studies show, however, that the increased numbers of granulocytes observed in VAD mice were likely due, at least in part, to loss of RARγ function, and were not intrinsic to the hematopoietic cell.
TNFα is classically known as a pro-inflammatory cytokine, which raises the possibility that the myeloproliferation observed in RARγ null mice results from an inflammatory process. However, the markedly elevated TNFα expression we observed was not accompanied by increases in other cytokines normally associated with inflammation (IL-2, IL-6 or IL-4 (Table S6)), nor was the expression of IFNγ deregulated (data not shown). Furthermore, there was no evidence of infection in RARγ-/- mice or in the VAD mice studied by Kuwata et al (Kuwata et al., 2000), and the myeloproliferation also occurred in RARγ-/- mice that were administered antibiotics from the time of weaning to analysis.
Myeloproliferative syndromes have, to date, been considered to be hematopoietic cell intrinsic. Studies utilizing either transgenic mice such as JunB (Passegue et al., 2004), Ptpn11 (Araki et al., 2004) or Nf1 (Le et al., 2004), or oncoretroviral overexpression of acquired somatic mutations such as Jak2V617F (Wernig et al., 2006) have demonstrated that some MPS are intrinsic to the hematopoietic cell. However, given that the BM microenvironment is known to be an important regulator of hematopoiesis it is not unreasonable to speculate that the microenvironment may play a larger role in the development of MPS than previously thought. Our data and that of Walkley et al (accompanying paper) support this possibility.
Indeed, another recent study has provided additional support to the potential role of the microenvironment in regulating MDS. Conditional germline deletion of the inhibitor of NF-κB (IκBα) resulted in hypergranulopoiesis that was not cell autonomous (Rupec et al., 2005). Loss of IκBα also resulted in increased numbers of dysplastic hematopoietic cells (erythrocytes, megakaryocytes and granulocytes), resulting in MDS with progression to secondary acute myeloid leukemia (Rupec et al., 2005). The mice died between 6 and 7 days after birth. Additional studies in this report showed that Notch1 was upregulated in IκBαΔ/Δ granulocytes, and the Notch1 ligand, Jagged1, was shown to be upregulated in IκBαΔ/Δ hepatocytes (Rupec et al., 2005). Inhibition of Jagged1 was able to abolish the hypergranulopoiesis that occurred when wildtype BM cells were co-cultured with IκBαΔ/Δ hepatocytes (Rupec et al., 2005). It is unclear, however, whether the non-hematopoietic microenvironment in the IκBαΔ/Δ mice was the sole contributing factor to the myelodysplastic syndrome and subsequent progression to leukemia observed in IκBαΔ/Δ mice, or whether, like that demonstrated in the accompanying paper by Walkley et al, loss of IκBα was also required in hematopoietic cells to exert these effects.
Retinoic acid receptors regulate many different gene products and aberrant expression or function of RARs have been identified in many different diseased states. It is likely that there are multiple contributing factors to the microenvironment-induced MPS observed in RARγ-/- mice. Increased production of TNFα is clearly involved, as transplantation of TNFα-/- BM into the RARγ-/- mice partially abrogated the myeloproliferative disease, restored BM erythropoiesis to normal and reduced the severity of the BM B lymphocyte defects observed in RARγ-/- mice (Figures 6 and S4). There are numerous links with aberrant expression of TNFα and human disease, including MDS (Flores-Figueroa et al., 2002) and cancer (Szlosarek et al., 2006). Here we provide the first demonstration that an abnormal microenvironment created by RARγ deficiency can be the sole causative mechanism for MPS. These data provide further support for evaluating the bone marrow microenvironment as a source of hematologic disease and potential target for future therapies for such diseases.
RARγ (Lohnes et al., 1993) mutant mice were backcrossed 10 generations onto the C57BL/6J background for these studies. For transplant studies, donor mice were either B6.SJL-PtprcaPep3b/BoyJ (herein referred to as Ptprca; CD45.1+), obtained from Animal Resources Centre, Perth, WA, Australia, or TNFα-/- mice (Korner et al., 1997) (generously provided by Dr Mark Smyth, Peter MacCallum Cancer Centre, Melbourne, Vic., Australia). Recipient mice were either CD45.1+ Ptprca mice or CD45.2+ RARγ mutant mice. Except when otherwise described, all mice were used between 8 and 10 weeks of age. All experiments performed were approved by the Peter MacCallum Cancer Centre animal experimentation ethics committee and were conducted in strict compliance to the regulatory standards of the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes.
Peripheral blood, BM, spleen and thymic cell content were measured on a Sysmex K1000 Auto analyzer (Sysmex).
Bone marrow, spleen, thymus and hemolyzed PB cells were stained with antibodies to B lymphocytes (B220-PE and IgM-FITC), T lymphocytes (CD4-FITC and CD8a-PE), granulocytes (CD11b-PE and Gr-1-FITC) and erythrocytes (Ter119-PE) as previously described (Walkley et al., 2002). Samples were analysed on a FACSCAN (Becton Dickinson).
Whole mice were presented for necropsy fixed in Bouin’s fixative. Tissues from all major organ systems were collected, processed, sectioned, and stained with hematoxylin and eosin by routine methods. Tissue sections with spleen, liver, kidney, adipose, and BM were immunostained with myeloperoxidase (MPO; Abcam, Cambridge, MA) to identify myeloid cells. Briefly, sections were deparaffinized and dehydrated followed by antigen retrieval in Reveal Citra buffer (Biocare Medical, Walnut Creek, CA) using a pressurized antigen-decloaking chamber (Biocare). Sections were blocked with normal serum followed by an avidin block then incubated overnight at 4°C with a rabbit polyclonal MPO (pre-diluted). The sections were then blocked with biotin and incubated with biotinylated goat anti-rabbit (1:200) for 35 minutes at room temperature. Lastly sections were treated with streptavidin-peroxidase (Biogenix, San Ramon, CA), Romulan AEC chromagen (as per manufacturer’s instructions), and counterstained with Richard-Allan hematoxylin.
Recipient mice were lethally irradiated with a total dose of 10 Gy, given in two equal fractions 3 hours apart, delivered by 2 opposing 137Cs sources (Gammacell 40, Atomic Energy of Canada) on the day of transplant. Each mouse was transplanted with 5 × 106 BM leukocytes obtained from either RARγ mutant (CD45.2+), Ptprca (CD45.1+) or TNFα-/- (CD45.2+) mice. The hematopoietic parameters of the transplanted mice were assessed between 5-8 weeks post-transplant. Donor cell engraftment 5-8 weeks post-BMT was confirmed by either FACS analysis after staining hemolyzed PB cells with CD45.1-PE and CD45.2-FITC or by PCR-based assays on BM obtained from the transplanted mice (for TNFα-/- donor BM cell transplants).
To assess potential leukemic transformation of the transplanted cells, spleen cells consisting of 2.5 × 106 leukocytes obtained from a RARγ-/- recipient were transplanted into new lethally irradiated RARγ-/- or wildtype recipients. PB cellularity, donor cell engraftment and hematopoietic cell content were analyzed 8 weeks post-BMT as described above.
Colony-forming cell assays were performed using BM, spleen and hemolzyed PB cells as previously described (Walkley et al., 2002). Progenitor cell survival and cytokine responsiveness were measured as previously described (Walkley et al., 2002). The numbers of common myeloid progenitors (CMPs), granulocyte/macrophage progenitors (GMPs) and megakaryocte/erythroid progenitors (MEPs) were assessed as previously described (Walkley et al., 2005).
The results are expressed as the mean ± standard error of the mean (SEM) for n given samples. Data were analysed using the two-tailed Student’s t-test, with any P value less than or equal to 0.05 being considered significant.
We thank D Raikowski, P Della Pelle and B Martin for excellent technical assistance, NA Sims for generously sharing unpublished data, M Smyth for providing mice and Peter MacCallum Cancer Centre (PMCC) Animal Facility Staff for care of experimental mice. We also thank S Orkin, DA Williams, K Shannon, DG Gilliland and M Mielcarek for comments on the manuscript and useful suggestions. This work was supported in part by grants from the Cancer Council of Victoria (L.E.P.), the National Health and Medical Research Council (L.E.P.), National Institutes of Health (NIH) DK84551 (D.T.S.) and NIH DK71773 (L.E.P). C.R.W. was a recipient of an Australian Postgraduate Award and is a Special Fellow of the Leukemia and Lymphoma Society.
The authors declare that they have no conflicting interests pertaining to this research.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.