BMP4 and Madh5 regulate the erythroid response to acute anemia

Laurie E. Lenox, John M. Perry and Robert F. Paulson


Acute anemia initiates a systemic response that results in the rapid mobilization and differentiation of erythroid progenitors in the adult spleen. The flexed-tail (f) mutant mice exhibit normal steady-state erythropoiesis but are unable to rapidly respond to acute erythropoietic stress. Here, we show that f/f mutant mice have a mutation in Madh5. Our analysis shows that BMP4/Madh5-dependent signaling, regulated by hypoxia, initiates the differentiation and expansion of erythroid progenitors in the spleen. These findings suggest a new model where stress erythroid progenitors, resident in the spleen, are poised to respond to changes in the microenvironment induced by acute anemia.


Erythropoiesis in the bone marrow is primarily homeostatic; however, the situation is dramatically different in the adult spleen in response to acute erythropoietic stress, where rapid, expansive erythropoiesis occurs. Previous work suggested a model where acute anemia leads to tissue hypoxia, which induces expression of erythropoietin (Epo) in the kidney. Increased levels of serum Epo mobilize cells from the bone marrow, which migrate to the spleen where they expand and differentiate.1,2 The spleen contains a unique microenvironment that can support expansive erythropoiesis.3 However, the signals that regulate the increase in splenic erythropoiesis in response to acute anemia are not clear.

The expansive erythropoiesis observed in the adult spleen is similar to fetal liver erythropoiesis during development.4 In both cases rapid erythroid development occurs. Similar to the spleen, fetal liver stromal cells are capable of supporting the expansion of erythroid progenitors.5 Because of these common features, it has been suggested that splenic and fetal liver erythropoiesis may be mechanistically similar. This link between the fetal liver and spleen is apparent in mice with a mutation at the flexed-tail (f) locus. During fetal development, f/f mutant embryos exhibit a severe microcytic, hypochromic anemia.6-8 The f/f fetal livers contain about 50% the normal number of erythroid progenitors9,10 and have a maturation defect, which results in the production of large numbers of siderocytes or erythrocytes that contain nonheme iron granules.8,11 Despite these defects, the anemia of f/f mice resolves about 2 weeks after birth. Adult f/f mice exhibit normal numbers of steady-state erythroid progenitors.12 However, they are unable to respond rapidly to acute erythropoietic stress. This defect is manifested by a delay in the expansion of erythroid progenitors in the spleen and a delay in the appearance of reticulocytes following phenylhydrazine (PHZ)–induced acute anemia.13 Despite the delayed response, adult f/f mice do not exhibit the maturation defect present during fetal liver erythropoiesis because no siderocytes are observed during the recovery from acute anemia.14 These observations demonstrate that the f gene product plays a key role in regulating the expansion and maturation of erythroid progenitors at times of great erythropoietic need.

Previous work has suggested that f/f mice have a mutation in sideroflexin 1 (sfxn1), a putative mitochondrial transporter, which is proposed to play a role in the transport of molecules required for heme biosynthesis.15 In this report, however, we show that f/f mice have a mutation in the Madh5 gene, which directly affects the ability of f/f mice to respond to acute anemia. Madh5 functions as a receptor-activated Smad downstream of the bone morphogenetic protein 2 (BMP2), 4, and 7 receptors.16,17 Previous work has implicated BMPs and in particular BMP4 in the development of mesodermal cells that will give rise to hematopoietic cells early in development.18 Our work shows that in response to acute anemia, BMP4 is rapidly induced in the spleen. BMP4 acts on an immature progenitor cell causing it to differentiate into an Epo-responsive stress erythroid progenitor. Cell-sorting experiments showed that BMP4-responsive cells exhibit the same cell surface phenotype as the bone marrow-derived megakaryocyte-erythroid progenitors (MEPs)19; however, only spleen MEPs respond to BMP4. These results demonstrate that these spleen progenitors exhibit properties that are distinct from bone marrow erythroid progenitors, suggesting that they represent a population of “stress erythroid progenitors” resident in the spleen whose function is to rapidly generate erythrocytes at times of great erythropoietic need.

Materials and methods


C57BL/6 and C57BL/6-f mice were obtained from the Jackson Laboratory (Bar Harbor, ME). Madh5+/– mice were obtained from Dr C. Deng (National Institutes of Health–National Institute of Diabetes and Digestive and Kidney Diseases, Bethesda, MD).20 All mice were approximately 6 to 8 weeks old, and controls were matched for age. Acute anemia was induced by injection of PHZ (Sigma, St Louis, MO) at a concentration of 100 mg/kg mouse in phosphate-buffered saline (PBS).

Colony assays for BFU-Es

Splenocytes, bone marrow, and peripheral blood cells were isolated from C57BL/6 control, f/+, and f/f mice. Then, 1 × 105/mL nucleated bone marrow and peripheral blood cells and 2 × 106/mL nucleated splenocytes were plated in methylcellulose media (StemCell Technologies, Vancouver, BC, Canada) containing 3 U/mL Epo with or without either 10 ng/mL interleukin 3 (IL-3; Sigma) or 0.15 to 15 ng/mL BMP4 (R&D Systems, Minneapolis, MN) where indicated. Erythroid burst-forming units (BFU-Es) were scored as described.21 For the BMP4 preincubation experiment, splenocytes and bone marrow cells from C57BL/6 and f/f mice were incubated for 24 hours in Iscove modified Dulbecco medium (IMDM) plus 5% fetal calf serum (FCS) with or without 15 ng/mL BMP4. Colony assays were then performed as indicated with or without 15 ng/mL BMP4 for each.

Characterization of the Epo sensitivity of the stress BFU-Es

Colony assays were performed as described (see “Colony assays for BFU-Es”) on bone marrow and spleen cells in the presence of 0.1, 0.3, 1, 3, and 10 U/mL Epo as indicated. Additionally, bone marrow cells were supplemented with 50 ng/mL stem cell factor (SCF), whereas splenocytes were supplemented with 15 ng/mL BMP4. Colonies were scored as described.21

Cloning of Madh5 mRNAs from f/f, f/+, and control mice

Total RNA was isolated from single-cell suspensions of spleen cells using the TriZol reagent (Invitrogen, Carlsbad, CA) according to the manufacturer's instructions. cDNA was generated and Madh5 cDNA was amplified using 5′-GGGGCCGAGCTGCTAAT-3′ and 5′-CTATGAAACAGAAGAAATGGGG-3′ primers.

Analysis of BMP4 expression

Total RNA isolated from bone marrow or spleen cells homogenized in TriZol (Invitrogen) was reverse transcribed into cDNA. Polymerase chain reaction (PCR) was performed using primers 5′-TGTGAGGAGTTTCCATCACG-3′ and 5′-TTATTCTTCTTCCTGGACCG-3′.

Staining of spleen sections with anti-BMP4 antibodies

Spleens were harvested on the indicated days after PHZ-induced anemia and fixed in zinc fixative, and paraffin-embedded tissue sections were cut. The expression of BMP4 was analyzed as described22 using anti-BMP4 antibody (Novocastra Laboratories/Vector Laboratories, Burlingame, CA). Slides were analyzed by confocal microscopy.

Cell staining and sorting

Bone marrow and spleen MEPs were sorted as previously described19 with the exception that fluorescein isothiocyanate (FITC)–conjugated anti-c-kit was used (and allophycocyanin [APC]–conjugated anti-c-kit and FITC-conjugated anti-CD34 eliminated) for spleen sorts after determining that FITC-conjugated anti-CD34 did not stain spleen cells. Cells were washed twice and sorted using a Coulter Elite ESP flow cytometer (Coulter Electronics, Hialeah, FL). Cells were plated in methylcellulose and scored as described (see “Colony assays for BFU-Es”).

Analysis of BMP4 signaling in W-20-17 osteoblast cells

The cDNAs coding the f/f truncated transcripts, as well as full length Madh5, were cloned into the MSCVneo retroviral construct. Recombinant virus was generated as previously described23 and used to infect W-20-17 cells (American Type Culture Collection, Manassas, VA). Pools of neomycin-responsive (neoR) colonies were plated and induction of alkaline phosphatase by BMP4 was measured as described.24


The flexed-tail mutant mice exhibit a delayed expansion of erythroid progenitors in the spleen in response to acute anemia.13 To identify the origin of the defect, we characterized the response of f/f, f/+, and control mice to acute anemia using a modified PHZ-induced hemolytic anemia protocol that rapidly induced severe anemia (hematocrit 30% treated versus 50% untreated) in 12 hours. Similar to previous findings, we observed that early erythroid progenitors (BFU-Es) in the bone marrow were not elevated in response to anemia and in fact gradually declined over time (Figure 1A). The greatest difference was observed in the spleen and peripheral blood. In the spleen, control mice exhibited an expansion of BFU-Es that peaked at 36 hours after anemia induction. Similarly, f/+ mice showed the greatest expansion at 36 hours but unlike control mice we also observed a significant expansion later at 4 and 6 days after anemia. In contrast, the expansion was significantly delayed in the f/f mice where it peaked at 4 days after anemia induction (Figure 1B). These results correlate with previous data showing that the f/f mice were delayed in the expansion of erythroblasts in the spleen following anemia induction.13 In the peripheral blood, however, we did not identify any BFU-Es potentially migrating from the bone marrow to the spleen at any of the time points in the f/f, f/+ or control mice (data not shown). These results suggest a new model where progenitor cells resident in the spleen mediate the response to acute anemia.

Figure 1.

Analysis of BFU-E expansion during the recovery from PHZ-induced acute hemolytic anemia. (A) Bone marrow and (B) spleen BFU-Es from C57BL/6-f/f (▪), C57BL/6-f/+ (▦), and C57BL/6-+/+ control mice (□) during the recovery from PHZ-induced acute anemia. Cells were plated in methylcellulose media containing Epo (3 U/mL) and IL-3 (10 ng/mL). (C) Bone marrow and (D) spleen BFU-Es from C57BL/6-f/f, C57BL/6-f/+, and C57BL/6-+/+ control mice during the recovery from PHZ-induced acute anemia. Cells were plated in methylcellulose media containing only Epo (3 U/mL). (E) Sensitivity of bone marrow and spleen BFU-Es to Epo. A total of 5 × 105 bone marrow or spleen cells were plated in methylcellulose media containing the indicated concentrations of Epo plus 50 ng/mL SCF (bone marrow, ▪) or 15 ng/mL BMP4 (spleen, □). The asterisk indicates P = .02. Error bars represent standard deviation.

Splenic erythroid progenitors that expand in response to acute anemia exhibit distinct properties

Bone marrow BFU-E colonies have a distinct morphology, develop in 7 days in culture, and require 2 signals to develop. The first signal is Epo and the second is a burst-promoting activity (BPA), which in vivo is SCF, but in vitro IL-3 or granulocyte-macrophage colony-stimulating factor (GM-CSF) can substitute. We observed that the BFU-E colonies from spleen 36 hours after anemia induction exhibited altered morphology. The colonies were larger and had more small satellite colonies associated with the BFU-Es. They also grew faster, such that spleen colonies grown for 5 days resembled bone marrow colonies that had grown for 7 days, and spleen BFU-Es were routinely grown for 5 days out of convenience. In many ways these erythroid progenitors resembled fetal liver erythroid progenitors, which are known to exhibit a faster cell cycle than bone marrow BFU-Es. Fetal liver BFU-Es can also develop in media containing only Epo, without an added BPA.25 Given that f/f mice also have a defect in fetal liver, we repeated the analysis of bone marrow, peripheral blood, and spleen BFU-Es following induction of acute hemolytic anemia; however, this time cells were cultured in media containing only Epo. The bone marrow contained very few cells that could form BFU-E colonies in Epo-only media (Figure 1C). In the control and f/+ spleens, however, the expansion of BFU-Es at 36 hours was completely recapitulated when the cells were plated in Epo-only media (Figure 1D). If fact, more splenic BFU-E colonies developed in this media. Similar to the initial observations, the f/f mice exhibited a delay in the expansion of BFU-Es with the maximum number of colonies observed at 4 days after anemia induction. The number of spleen BFU-Es observed under the Epo-only culture conditions responded in a linear manner when increasing numbers of cells were plated, which suggests that the spleen cells are not producing BPA (data not shown). Once again we did not identify any BFU-Es in the peripheral blood at any of the time points in the f/f, f/+, or control mice indicating that these BFU-Es are resident in the spleen.

In addition to shorter cell cycling times, fetal liver erythroid progenitors also exhibit an increased sensitivity to Epo.25 We tested the Epo sensitivity of splenic erythroid progenitors. We observed that these progenitor cells were actually less sensitive to Epo than bone marrow cells (Figure 1E). Decreased Epo sensitivity is an ideal property of a stress progenitor because differentiation of these progenitors would be dependent on the high serum Epo concentrations only present during the response to acute anemia. Taken together, these results suggest that the spleen contains a distinct population of erythroid progenitors that are poised to respond to acute erythroid stress. These progenitors, which we will refer to as “stress BFU-Es,” form large burst colonies in 5 days, require only Epo at relatively high levels to develop, and are resident in the spleen.

f/f mice have a mutation in the Madh5 gene

The flexed-tail locus is located on mouse chromosome 13.26 We generated a panel of 408 F2 progeny using a F1(C57BL/6-f/f X BALB/c) intercross. F2 progeny were scored at birth for anemia by hematocrit and for the presence of siderocytes by staining blood smears for iron deposits. We constructed a high-resolution genetic linkage map of the f locus and initially localized the gene 0.6 cM distal to the microsatellite marker D13MIT13. Further analysis of markers showed that the f locus cosegregated with the marker D13Mit208 (Figure 2A). Our linkage mapping results differ from the recent work from Fleming et al.15 They mapped the f locus to a more proximal position on chromosome 13 and identified a mutation in the sideroflexin 1 (sfxn1) gene, which they proposed caused the f/f mutant phenotype. Because there is only a single allele of the f mutation, all f/f mice must carry the same mutation.27 Like Fleming et al,15 our colony of C57BL/6J-f mice was derived from C57BL/6J-f mice obtained from the Jackson Laboratory. However, when we scored our f/f progeny for the presence of the mutation in exon 2 of Sfxn1, we identified f/f progeny that exhibited severe anemia and siderocytes at birth, but were heterozygous for the insertion mutation (Supplementary Figure 1; see the Supplemental Figures link on the Blood website at the top of the online article). These results suggest that in our colony, the f mutation has been separated from the mutation in Sfxn1 by recombination and, thus, mutation of Sfxn1 cannot be the cause of the f/f mutant phenotype.

Figure 2.

Genetic linkage map of the locus and molecular analysis of Madh5 transcripts in f/f, f/+, and control mice. (A) Linkage map of the f locus on mouse chromosome 13. The position of the markers on chromosome 13 according to NCBI m33 mouse genome assembly is indicated. The number of recombinants in each interval among the scored intercross F2 progeny is indicted at the right. The positions of Madh5 and Sfxn1 are shown. (B) The coding region of the Madh5 was cloned by RT-PCR of spleen RNA from the indicated mice. The arrows indicate the position of the wild-type and mutant mRNAs. The asterisk indicates a nonspecific background band. The exon structure of the wild-type and f/fmRNAs is indicated at the right. The f/fmouse shown here is an example of a mutant mouse that expresses very little wild-type Madh5 mRNA.

To identify candidate genes for the f locus, we initially took advantage of the fact that one of our flanking markers, D13Mit13, is located in the IL-9 gene.28 This region of mouse chromosome 13 is homologous to human chromosome 5q31. Comparison of the human and mouse gene maps in the region immediately surrounding IL9 revealed MADH5 was located in this region. The possibility that Madh5 was encoded by the f locus was supported by the recent mouse genome sequence release that showed that D13MIT208 is located in the Madh5 gene. Previous work in Xenopus and mice has demonstrated that BMP4/Madh5-dependent signals play a key role in the development of erythroid cells.18,29 Madh5 is highly expressed in the fetal liver during development,30 and we have observed Madh5 expression in the spleen of mice recovering from PHZ-induced acute anemia (data not shown).

To determine whether Madh5 is mutated in f/f mice, we cloned the entire coding region of Madh5 by reverse transcription-PCR (RT-PCR) from spleen RNA isolated from C57BL/6J-+/+, C57BL/6J-f/+, and C57BL/6J-f/f mice. Only the expected product was observed in wild-type mice; however, in both f/+ and f/f mice an additional band was observed (Figure 2B). The majority of the mRNA in f/+ mice is the wild-type fragment, whereas in f/f mice the majority of the mRNA is a truncated mRNA. The level of wild-type message in f/f mutant mice varies, but most express less than 30% wild-type mRNA. Analysis of the sequence of the truncated mRNA showed that it is a mixture of 2 mis-spliced mRNAs. The consistent feature of these mutant mRNAs was the deletion of exon 2, which contains the AUG initiator codon. In one of the mis-spliced mRNAs we also observed deletion of exon 4 and insertion of a 12-nucleotide sequence at the splice junction between exons 6 and 7. In the other mis-spliced mRNA, we observed an aberrant splice into the middle of exon 3 (Figure 3). There were no other alterations in the coding sequence of Madh5 observed in f/f mice. Southern blot analysis of the Madh5 genomic region did not identify any deletions or rearrangements in the Madh5 locus suggesting that the alterations in the Madh5 mRNA in f/f mice are due to defects in mRNA splicing (data not shown). We have sequenced the entire Madh5 transcription unit from f/f and control mice and have not identified a consistent mutation, which suggests that the defect may lie in the promoter or 3′ to the Madh5gene. We are currently investigating this possibility.

Figure 3.

Analysis of BMP4 expression during the recovery from acute anemia. (A) RT-PCR analysis of BMP4 expression in C57BL/6-+/+ control (left) and C57BL/6-f/f (right) mice. Arrows indicate the positions of the BMP4-specific band and the hypoxanthine guanine phosphoribosyl transferase 1 (HPRT) control band and the number of PCR cycles is indicated at the right. (B) Spleen sections from C57BL/6-+/+ (top row) and C57BL/6-f/f (bottom row) mice stained with anti-BMP4 antibodies at the indicated times following PHZ-induced acute anemia. BMP4 staining was visualized using an Olympus FV300 confocal laser scanning microscope equipped with a UPLANFL 100×/0.3 objective lens (Olympus, Melville, NY). Images were acquired using Fluoview version 4.3 software (Olympus). (C) RT-PCR analysis of BMP4 expression in MSS31 spleen stromal cells grown at normoxic (20%) and hypoxic (6%) conditions (top). The position of the putative hypoxia-inducible element in the BMP4 gene is indicated and an alignment of this sequence from mouse, human, and rat is presented (bottom).

BMP4 expression is induced in the spleen just prior to the expansion of “stress BFU-Es”

The identification of aberrantly spliced Madh5 mRNA in f/f mice suggests a role for the BMP2, 4, and 7 signaling pathways in the response to acute anemia.16,31 We investigated the expression of BMP2, 4 and 7 in the spleen during the response to acute anemia by RT-PCR. BMP2 is not expressed in the spleen at any time point during the recovery from acute anemia, whereas BMP7 is expressed at low levels at all times tested (data not shown). BMP4 is not expressed in the spleen of untreated mice; however, expression is initiated at 12 hours and peaks at 24 hours with lower levels at 36 and 48 hours after anemia induction. We also observed low levels at 6 and 8 days after anemia induction. The highest expression observed at 24 hours after anemia induction is just prior to the expansion of “stress BFU-Es” in the spleen (Figure 3A). Staining of spleen sections with anti-BMP4 antibodies showed that BMP4 protein was not present in untreated mice; however, high levels of BMP4 were observed throughout the red pulp of the spleen at 24 and 36 hours after anemia induction, but expression is severely decreased by 48 hours and essentially off by 96 hours after anemia induction (Figure 3B). BMP4 was excluded from the white pulp.

The expression of BMP4 is tightly regulated during the response to acute anemia. During the analysis of the BMP4 expression by RT-PCR, we also tested the expression in f/f mice. Surprisingly, BMP4 expression is expanded in the mutant mice. Untreated mice and all the time points during the response to acute anemia exhibit BMP4 expression (Figure 3A). The expression of BMP4 in untreated f/f correlates with the observation of stress BFU-Es in these mice (Figure 1D). Despite the constitutive mRNA expression, BMP4 protein expression was not observed at all time points, suggesting that BMP4 expression is regulated post-transcriptionally. These observed differences in BMP4 expression in f/f mice suggest that the regulation of BMP4 may require a Madh5-dependent signal to inhibit the expression of BMP4 in the spleens of untreated mice.

The increase in serum Epo concentration that occurs during the response to acute anemia is regulated at the transcriptional level by the hypoxia-inducible transcription factor complex, HIF-1.32 Given that Epo expression is regulated by hypoxia we tested whether BMP4 expression is also induced by hypoxia. MSS31 spleen stromal cells,33 which support erythroid progenitor cell expansion in vitro, were grown at normoxic (20% O2) and hypoxic (6% O2) conditions. At low oxygen concentration, the expression of BMP4 was significantly increased (Figure 3C). Analysis of the BMP4 gene revealed the presence of a putative HIF element in the 3′-untranslated region (UTR), which is conserved in the human, rat, and mouse BMP4 genes.

BMP4 causes the differentiation of an immature progenitor cell into an Epo-responsive stress BFU-E

Given that BMP4 was induced in the spleen just prior to the expansion of stress BFU-Es, we next tested whether culturing spleen cells from untreated mice in Epo and BMP4 could induce the expansion of stress BFU-Es. Spleen cells from untreated f/f and control mice were plated in methylcellulose media containing 3 U/mL Epo and various concentrations of BMP4. Control spleen cells responded in a dose-dependent manner to BMP4 with a 6.1-fold increase in the number of stress BFU-Es seen at a 15-ng/mL BMP4 dose (Figure 4A). The f/f spleen cells failed to respond to BMP4 except at the highest concentration, which is consistent with their defect in Madh5. BMP4 had very little effect on the number of BFU-Es in the bone marrow, suggesting that there are very few BMP4-responsive erythroid progenitor cells in the bone marrow (Figure 4A). We did not detect any BMP4-responsive cells in the peripheral blood in untreated or in mice treated with PHZ to induce anemia (data not shown).

Figure 4.

Analysis of the ability of BMP4 to induce the formation of stress BFU-Es in spleen cells from untreated mice. (A) Bone marrow (top) and spleen (bottom) cells from untreated f/f (▪) and control mice (▦) were plated in methylcellulose media containing Epo (3 U/mL) and the indicated concentration of BMP4. (B) Spleen cells from C57BL/6-+/+ mice were preincubated without (▪) or with BMP4 (15 ng/mL; ▦) for 24 hours, washed, and then plated in the methylcellulose media containing either Epo (3 U/mL) alone or Epo plus BMP4 (15 ng/mL). Error bars indicate standard deviation.

BMP4 can induce the expansion of stress BFU-Es, but the mechanism by which BMP4 affects these cells is not clear. One might imagine 2 possible roles for BMP4. First, BMP4 could synergize with Epo to promote the differentiation of stress BFU-Es, much like SCF synergizes with Epo to increase the number and size of bone marrow BFU-Es.34 Alternatively, BMP4 may act on an earlier cell inducing it to differentiate into an Epo-responsive stress BFU-Es. This possibility is similar to the situation in Xenopus embryos where BMP4 can induce mesodermal cells to become erythroid progenitors.18 To test these possibilities, we preincubated spleen cells from untreated mice with and without BMP4 for 24 hours, washed the cells, and then plated them in methylcellulose media containing either Epo alone or Epo and BMP4. The cells preincubated with BMP4 and plated in Epo alone gave rise to as many stress BFU-E colonies as the cells plated in Epo and BMP4 (Figure 4B). These results suggest that a short exposure to BMP4 promotes the differentiation of an immature BMP4-responsive (BMP4R) cell into an Epo-responsive stress BFU-E.

f is a gain-of-function allele of Madh5

Our analysis shows that the f mutation maps to the Madh5 locus, f/f mice express mis-spliced mRNAs, and spleen cells from f/f fail to respond to BMP4. Although all of these data are consistent with f being a mutation in Madh5, we crossed f mice with Madh5+/ mice20 to generate f/Madh5 mice to test whether f was allelic to Madh5. Figure 5A shows the expansion of stress BFU-Es in f/Madh5 and +/Madh5 during the recovery from acute anemia. Both genotypes exhibit an altered recovery when compared to control (compare Figure 5A with Figure 1D). The peak expansion of stress BFU-Es in f/Madh5 mice is delayed until 48 hours, whereas +/Madh5 mice exhibit an increase in stress BFU-Es at 36 hours, and they continue to expand at 48 hours. Analysis of BMP4 mRNA expression in these mice showed that both f/Madh5 and +/Madh5 expressed BMP4 at all time points during the recovery (Figure 5B). These data are similar to what is observed in f/f and f/+ mice, suggesting that a Madh5-dependent signal is required for the regulation of BMP4 in the spleen. These results show that the phenotype of f/Madh5 is more severe than f/+ and +/Madh5, which demonstrates that f is allelic to a targeted mutation in Madh5. Classical genetic analysis, however, would predict that if f was a hypomorphic or loss of function mutation in Madh5 then f/Madh5 should have a more severe phenotype than f/f. We observed the opposite that the f/f phenotype was the most severe. These data suggest that the f mutation in Madh5 is a neomorphic or gain-of-function mutation in Madh5. To test this possibility, we expressed each of the 2 mis-spliced Madh5 mRNAs identified in f/f mice in a cell line that responds to BMP4 and analyzed whether these mRNAs could affect BMP4-dependent responses. W-20-17 is a mouse osteoblast cell line that differentiates in response to BMP4.24 The mutant transcripts as well as controls were cloned into the MSCV-neo retroviral vector and W-20-17 cells were infected with recombinant retroviruses. Although the mis-spliced messages lack the initiator ATG of wild-type Madh5, both messages contain in-frame ATGs that could be used to initiate the translation of truncated forms of the Madh5 protein (Figure 6A). Treatment of W-20-17 cells with BMP4 induces the osteoblast differentiation program as measured by an increase in alkaline phosphatase (AP) activity. W-20-17 cells that express either f mutant message have a profound defect in BMP4-dependent induction of AP activity with mutant message no. 1 exhibiting the most severe defect (Figure 6B). These results show that the mis-spliced Madh5 mRNAs present in f/f mice dominantly suppress BMP4-dependent signals. Furthermore, W-20-17 cells do not express endogenous Madh5 (data not shown), but rather rely on Madh1 and Madh8 to transmit BMP4 signals, which suggests that the mis-spliced mRNAs also inhibit signaling through Madh1 and Madh8. These results explain why f/Madh5 mice have a less severe phenotype than f/f because the f/Madh5 mice express lower levels of the mis-spliced mRNAs.

Figure 5.

Analysis of the recovery from acute anemia in f/Madh5 and+/Madh5 mice. (A) Analysis of the expansion of stress BFU-Es in the spleen of C57BL/6-f/Madh5 (□) and C57BL/6-+/Madh5 mice (▪). Spleen cells were plated in methylcellulose media containing 3 U/mL Epo. (B) Expression of BMP4 in the spleen of C57BL/6-f/Madh5 and C57BL/6-+/Madh5 mice. The BMP4 and HPRT control bands are indicated by the arrows. These results are from 25 cycles of PCR. Error bars represent SEM.

Figure 6.

Analysis of the effect of overexpression of the f mis-spliced Madh5 mRNAs on BMP4 signaling in W-20-17 osteoblast cells. (A) Schematic representation of the f mis-spliced Madh5 mRNAs and control Madh5 mRNA. The position of the endogenous ATG is indicated in the wild-type mRNA. The asterisks indicate the positions of putative in-frame ATGs in the mis-spliced mRNAs. The arrowhead indicates the 12-nucleotide insertion between exons 6 and 7. (B) Induction of alkaline phosphatase activity by BMP4 in control W-20-17 cells and W-20-17 cells expressing wild-type or mis-splice Madh5 mRNAs. Alkaline phosphatase activity was normalized to protein concentration and expressed in relative units. Error bars represent SEM.

Spleen MEPs are the BMP4R cells

To further characterize the BMP4R cell, we fractionated spleen cells and assayed for response to BMP4. Initially we sorted cells based on their expression of lineage-restricted markers. BMP4R cells were observed in the Lin population (Figure 7A). Previous work had demonstrated that mice with a mutation in the Kit receptor, Dominant white spotting (W) mice, failed to respond rapidly to acute anemia, which suggested that the BMP4R cell may also be Kit+.35 Analysis of LinKit+ spleen cells confirmed this hypothesis because BMP4R cells were not detected in LinKit spleen cells (Figure 7B). In the bone marrow, erythroid progenitors are derived from the common myeloid progenitors (CMPs) and the MEPs.19 CMPs and MEPs are Kit+, but differ in their expression of CD34. We analyzed the expression of CD34 in LinKit+ spleen cells and observed that they were CD34, suggesting that the spleen does not contain CMPs (data not shown). Isolation of MEPs from spleen showed that they responded to BMP4 (Figure 7C). Interestingly, MEPs isolated from bone marrow failed to respond to BMP4. These results suggest that the unique microenvironment of the spleen alters the properties of MEPs, rendering them responsive to BMP4, and further underscores our assertion that distinct erythroid progenitors are present in the spleen poised to respond to acute erythroid stress.

Figure 7.

Identification of the subpopulation of progenitor cells from untreated spleen that responds to BMP4. (A) Unfractionated, Lin+ and Lin cells from untreated wild-type spleen were plated in 3 U/mL Epo without (▪) or with 15 ng/mL BMP4 (▦) and the induction of stress BFU-Es was analyzed. (B) Kit+Lin and KitLin cells from untreated wild-type spleen were plated in 3 U/mL Epo without (▪) or with 15 ng/mL BMP4 (▦) and the induction of stress BFU-Es was analyzed. (C) MEPs (LinSca1-IL-7RαKit+CD34-FcγRlow) isolated from bone marrow and spleen of wild-type mice were plated in 3 U/mL Epo without (▪) or with 15 ng/mL BMP4 (▦) and the induction of stress BFU-Es was analyzed. N.D. indicates none detected. Error bars represent SEM.


Acute anemia induces a robust erythroid response that occurs in the spleen. Our data demonstrate that in response to PHZ-induced anemia, dedicated stress progenitors resident in the spleen respond to a BMP4-dependent signal, which leads to the rapid expansion of erythroid progenitors. These data differ from the previous model that suggested that high serum Epo levels induced by tissue hypoxia mobilized erythroid progenitors from the bone marrow to spleen where they expanded and differentiated.1,2 This early model was based on the observation that BFU-Es were present in the peripheral blood during the recovery from acute anemia.1 However, we did not observe migration of erythroid progenitors in our studies and all of our data suggest that stress BFU-Es are resident in the spleen. One reason for the differences in our data could be that earlier experiments cultured BFU-Es for 10 days rather than the 5 to 7 days used in our studies. The longer culture conditions may have allowed more immature cells to develop. In addition, we used a modified protocol to induce anemia, which used a single injection of a high dose of PHZ. This protocol results in a synchronous and reproducible expansion of erythroid progenitors. The earlier experiments used multiple injections of a low dose of PHZ. We have found that multiple low doses of PHZ did not allow us to look at early events during the recovery and did not produce reliable results (L.E.L. and R.F.P., unpublished observations, 1999).

The earlier model suggested that bone marrow progenitors develop in the spleen during the recovery. Our data, however, demonstrate that specialized BMP4R cells are resident in the spleen. BMP4 induces these cells to differentiate, leading to the expansion of stress BFU-Es in the spleen but not in the bone marrow. Stress BFU-Es exhibit properties that are distinct from bone marrow BFU-Es in that they form large burst colonies in 5 days rather than 7 days and they require only Epo without any added BPA for BFU-E formation. Fractionation of spleen cells revealed that BMP4R cells are contained within the MEP population. However, only spleen MEPs are able to respond to BMP4. This observation suggests that the spleen microenvironment may provide a specialized signal that enables spleen MEPs to respond to BMP4.3 This situation would be similar to chondrogenesis, where presomitic mesoderm cells require BMP signals to differentiate into chondrocytes but are unable to respond to BMP unless they first encounter a sonic hedgehog signal.36

The notion that f/f mice have a defect in a spleen progenitor was first suggested almost 30 years ago. Gregory et al showed that f/f mice had normal numbers of BFU-Es and erythroid colony-forming units (CFU-Es), but were defective in transient endogenous colony-forming units (TE-CFUs).12 This population of cells was defined by an in vivo assay that identifies progenitor cells that form endogenous spleen colonies following sublethal irradiation and stimulation of erythropoiesis by Epo injection or bleeding.37 BMP4R cells exhibit many of the properties expected in a putative TE-CFU. They are resident in the spleen, rapidly expand at times of great erythropoietic need, and require high levels of Epo for differentiation. In addition to TE-CFUs, previous work has identified other stress erythroid progenitors in the spleen. Mice with mutations in the glucocorticoid receptor are slow to respond to acute anemia and it has been suggested that a CD34+Kit+TER119+ population of cells fails to expand in these mutant mice.38 Another group identified a 4A5+TER119+ bipotential megakaryocyte-erythroid progenitor (MEP) that expands in the spleen following PHZ-induced anemia.39 However, both of these progenitors express the lineage-restricted marker, TER119, which is not present on BMP4R cells or BFU-Es in general.40 The relationship between these progenitors and the BMP4R cells and the stress BFU-Es is not understood, but the most likely possibility is that they are derived from BMP4R cells.

Our analysis of f/f mice has identified a mutation in Madh5, which causes aberrant splicing. The mis-spliced Madh5 mRNAs exert a dominant-negative effect on BMP4 signaling, which suggests that f represents a gain-of-function allele of Madh5. These mis-spliced mRNAs inhibit BMP4-dependent signals in W-20-17 osteoblast cells, which do not express endogenous Madh5, suggesting that the f mutant mRNAs inhibit BMP4 signaling mediated by Madh1 and or Madh8. This possibility is consistent with our observation that f/f exhibits a more severe phenotype than f/Madh5 mice. In this case, f/f mice express higher levels of mis-spliced mRNAs and thus would have a more significant impairment of BMP4-dependent signaling than f/Madh5 mice.

Previous work from Fleming et al suggested that f was a mutation in the putative mitochondrial transporter Sfxn1.15 We show that the mutation in Sfxn1has been separated from the f locus by recombination in f/f mice in our colony. All of our observations characterizing the BMP4-dependent expansion of stress BFU-Es in the spleen during the recovery from acute anemia and the defective response in f/f mice support the idea that the f locus encodes Madh5. In addition, other phenotypes associated with f/f mice, tail flexures, and white belly spots, can easily be explained by defects in the BMP4 signaling pathway. The tail flexures are caused by defects in chondrogenesis that result in vertebrae fusion.41 BMP4 plays a key role in regulating chondrocyte development.36 Furthermore, white belly spots are caused by defects in the migration of neural crest-derived melanocytes.42 Inhibition of BMP4 signaling in chick embryos impairs the ability of neural crest cells to migrate.43 Thus, the combination of the observed defects in BMP4-dependent signals in f/f mice with the genetic interactions of f and Madh5 alleles demonstrates that the f locus encodes Madh5.


We thank M. Yon, A. Lariviere, and P. Porayette for technical assistance and for discussion of the manuscript; C. Deng for the Madh5–/+ mice; R. J. Lechleider for help with genotyping the Madh5 mice; and A. Bernstein and C. Pawson for their help and support during the early phases of this work.


  • Reprints:
    Robert F. Paulson, 115 Henning Bldg, Pennsylvania State University, University Park, PA 16802; e-mail: rfp5{at}
  • Prepublished online as Blood First Edition Paper, December 9, 2004; DOI 10.1182/blood-2004-02-0703.

  • Supported by a Seed grant from College of Agricultural Sciences at the Pennsylvania State University (R.F.P.) and National Institutes of Health grant RO1 HL70720 (R.F.P.). L.E.L. and J.M.P. contributed equally to this work.

  • The online version of this article contains a data supplement.

  • The publication costs of this article were defrayed in part by page charge payment. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 U.S.C. section 1734.

  • Submitted February 24, 2004.
  • Accepted December 2, 2004.


View Abstract