Combinatorial and distinct roles of α₅ and α₄ integrins in stress erythropoiesis in mice

During normal hematopoiesis, cells committed to specific lineages complete all their differentiation steps within hematopoietic tissues, and only the most mature cells are released into circulation. Thus, cues provided by the microenvironment (ME) in bone marrow (BM) or spleen, either through ligand/receptor interactions or through secreted cytokines and/or engagement of signaling molecules, are critical for lineage-committed cells and their differentiated descendants to complete their differentiation program. This interaction is uniquely exemplified in studies of erythroid lineage maturation. Close association of erythroid cells with cellular components of the ME seems to influence their differentiation; and among them, the macrophage (Mϕ) has received the most attention. Through physical contact with erythroid cells, Mϕs create a functionally interactive unit, the erythroblastic island (EI), which appears to exert major influence on later stages of erythroid differentiation/maturation. The EI was first described by Bessis et al,¹  but its functional influence was only recently analyzed by in vitro and in vivo approaches. Several proteins on the surface on Mϕs in partnership with proteins on erythroid cells appear to mediate these effects.^2-6  Furthermore, secreted Mϕ proteins, like Gas6, seem to interact with their cognate receptors present in erythroid cells enhancing integrin-dependent binding.⁷  Whether these proteins have a redundant function or work in complexes cooperating with each other, or become functional at different stages of differentiation and under different conditions in vivo, is presently unclear.

In addition to erythroid cell/Mϕ interactions within the EI, direct interactions of erythroid cells with extracellular matrix components, like fibronectin, have been implicated in the regulation of erythropoiesis. The presence of fibronectin counter-receptors on erythroid cells,⁸  like members of the beta₁ integrins (α₄β₁ and α₅β₁), is thought to mediate these interactions. Both α₄ and α₅ integrins are widely expressed in hematopoietic cells and have been implicated in several functional aspects, such as proliferation, survival, maturation of erythroid cells, or in homing and proliferation of hematopoietic progenitor cells.^9-15  However, in vitro and in vivo data have not been consistent, so that the exact role of these 2 integrins has remained inconclusive. For example, adhesion to fibronectin is dependent on both α₄β₁ and α₅β₁, and it was found to influence stem cell homing of hematopoietic progenitors to BM or spleen, but the results have been controversial for α₅ integrin.^8,11,14  Other in vitro experiments comparing effects of α₄ and α₅ have shown that only α₄β₁, not α₅β_1, influences proliferation and protection from apoptosis of erythroid cells,¹⁶  whereas opposite conclusions were reached by others,¹⁷  or effects were selective for BM and not spleen.¹⁸  Yet, genetic deletion of all β₁ integrins showed no effects on erythroid differentiation, only effects on colonization of hematopoietic tissues during development (fetal liver, spleen, and BM)¹⁹ and no effects on baseline or stress erythropoiesis in β₁-conditionally deleted adult animals.²⁰  Thus, the β₁ genetic data are in contrast with differentiation defects described for fetal erythroblasts in α₄β₁ knockout mutants²¹  and in α₄ chimeras,²²  or with impairment of stress hematopoiesis/erythropoiesis in adult animals with α₄ conditional deletion.²³  These diverse and controversial data have been difficult to reconcile and prompted us to reexamine the phenotype of β₁ conditionally deleted mice and compare them to α₄ or VCAM-1 deleted mice under conditions of stress.

Our data dissecting the influence exerted by α₄ versus α₅ integrins in response to erythroid stress provide novel information on the role of β₁ integrins in erythropoiesis. Their effects are distinctly manifested in different hematopoietic environments (ie, BM vs spleen) and seem to be directed at discreet stages of erythroid differentiation. Further, transplantation experiments have indicated that the great majority of the effects seen in β₁^Δ/Δ animals are, by and large, hematopoietic cell autonomous, but as yet undefined effects from a β₁-deficient ME may provide contributory function.

MxCre+β₁^f/f mice were obtained in our laboratory by breeding β₁^f/f mice²⁴  with MxCre⁺ transgenic mice (both from The Jackson Laboratory). β₁ integrin deletion was induced by 6 intraperitoneal injections of polyriboinosinic acid/polyribocytidylic acid (Sigma-Aldrich), and mice were used at least 4 weeks later. Polyriboinosinic acid/polyribocytidylic acid-treated Cre-negative littermates served as controls. α₄ integrin-deficient mice were generated in our laboratory.^23,25  VCAM-1-deficient mice were described previously.²⁶ 

To induce acute anemia, mice were injected with phenylhydrazine (PHZ; Sigma-Aldrich) at 60 mg/kg intraperitoneally on 2 consecutive days. Peripheral blood (PB) parameters were determined using a HEMAVET950 (Drew Scientific). To study the role of macrophages, mice received a single intravenous injection of 200 μL of clodronate-loaded or empty liposomes. Animals were housed at the University of Washington Comparative Medicine Specific Pathogen-Free Vivarium, with irradiated chow and autoclaved water ad libitum. All procedures were done according to protocols approved by the University of Washington Institutional Animal Care and Use Committee.

B6 × 129.F1 mice (The Jackson Laboratory) were used as recipients. Mice were lethally irradiated (1150 cGy) and injected intravenously with 0.25 × 10⁶ (for colony-forming unit-spleen [CFU-S]) or 10 × 10⁶ β₁^Δ/Δ or α₄^Δ/Δ BM cells for long-term repopulation. In the latter, after complete reconstitution of hematopoiesis, acute hemolytic anemia (PHZ) was induced. Mice reconstituted with wild-type β₁⁺ or α₄^+/+ BM cells served as controls.

To study integrin expression in cells from PB, BM, and spleen at steady state or after PHZ treatment, we used the following antibodies: CD44, CD45, β₁, α₂, α₅, α₄β₇, active caspase (BD Biosciences), α₄ (Southern Biotechnology), α₆, F4/80 (AbD Serotec), and α₉ (R&D Systems) and analyzed by fluorescence-activated cell sorting (FACS).

The committed progenitors of all lineages, colony-forming unit-colony (CFU-C), were assessed in methylcellulose cultures.²⁶  For “stress” burst-forming units-erythroid (BFU-E) in spleen, the cells were cultured in the presence of erythropoietin (Epo) only.^27,28  All colony types were counted at day 7, except for CFU-E, which were counted on day 3.

Nonirradiated wild-type (α₄^+/+, β₁^+/+) mice were injected with 20 × 10⁶ control (α4^f/f, β₁^f/f) or β₁^Δ/Δ BM cells and were killed 24 hours later (supplemental Figure 3, available on the Blood Web site; see the Supplemental Materials link at the top of the online article).

Information on CFU-S assay is contained in Figure 5.

Information on adhesion to spleen sections in vitro is contained in supplemental Figure 4.

Bones and spleens fixed with 4% paraformaldehyde were treated as previously described.^29,30  Frozen sections were labeled with anti-CD31 (MEC, 13.3) and anti-CD29 (9EG7, BD Biosciences, and MB1.2, Chemicon/Millipore), anti–bone morphogenetic protein 4 (BMP4; ab 39973, Abcam), anti-CD68 (FA-11, AbD Serotec), and F4/80 (CI:A3-1, Fitzgerald Industries International) and visualized with 2° Alexa 594–conjugated fluorochromes (Invitrogen).

Total RNA was isolated from the spleens of control and β₁^Δ/Δ mice treated with PHZ (d4) and reverse-transcribed using random primers. Obtained cDNA was used in quantitative polymerase chain reaction with primer sets for BMP4 (F-ctcccaagaatcatggactg, R-aaagcagagctctcactggt) and b-actin (F-atcctcaccctgaagtaccc, R-atttcccgctcggccgtggt).

Generation of these mice has been previously described,²⁴  although a detailed evaluation of their erythropoiesis and expression of different β₁ heterodimers in hematopoietic tissues have not been presented. Partial expression data were also presented in a similar β₁ conditional model.³¹  Ablation of β₁ in BM cells was more than 95% (95.5% ± 0.84%, n = 16). Specific β₁ integrin heterodimer expression is shown in Figure 1A. Drastic reduction in α₅β₁ expression is seen, whereas other dimers, like α₄ and α₆, are well expressed, probably because of their association with alternate β₁ partners (α₄β₇ or α₆β₃). Overall dramatic differences in the 2 most abundant integrins, α₄ and α₅, were seen in BM hematopoietic cells of these mice. Integrin α₅ was virtually absent in all BM cells, but α₄ was expressed at high levels, with a significant overexpression of α₄β₇.³²  Cellularity in hematopoietic tissues (BM, spleen) was normal to increased in β₁ mice (Figure 1B-C left panel). Significant differences in progenitor content were present in both BM and PB; however, no significant differences were recorded in the spleen (Figure 1C right panel). Hematocrit and hemoglobin levels were not statistically different from controls, but platelets were higher in β₁ mice (P < .05).

These mice have been described in detail in several prior publications.^23,25  Differences in the expression of α₄ and α₅ integrins in β₁^Δ/Δ and α₄^Δ/Δ spleens are shown in Figure 1D. Of particular note is the fact that the abundance of α₄ and α₅ in erythroid cells is different at progressing differentiation stages: α₅ expression is higher in earlier cells,⁸  whereas α₄ is widely expressed throughout nucleated erythroid cells (Figure 1E).

The phenotype of these mice has been described by us and others.^26,33,34  VCAM-1 represents the major and preferred ligand for α₄β₁ integrin in endothelial cells and also binds α₉β₁. These mice share many functional phenotypic features with α₄^Δ/Δ mice, but there have also been distinct differences.^26,34 

Overall, the aforementioned 3 models are characterized by normal basal erythropoiesis but display different integrins in their erythroid cells, and their functional responses to erythroid stress have not been studied in detail before.

Both β₁^Δ/Δ and control mice received 2 injections of PHZ and were studied at days 4 and 6 after the first injection. All control mice survived the hemolytic challenge, but half of the β₁^Δ/Δ mice died by day 6 (Figure 2A). Hematocrit levels reached a nadir in control animals 2 days after the last injection (day 4) and began to recover from that point on. β₁^Δ/Δ mice had similar initial and nadir hematocrit levels, but in contrast to controls, hematocrit levels continued to drop (supplemental Figure 1). Dramatic differences between controls and β₁^Δ/Δ mice were seen when splenic weights and cellularities were compared at days 4 and 6 (Figure 2A).

To further explore differences between control and β₁^Δ/Δ mice after PHZ, detailed studies on spleen cell populations, differentiation of erythroid cells, and progenitor cell content were carried out (Figures 2,Figure 3–4). TER119⁺ erythroid cells were analyzed by FACS using coexpression with CD71³⁵  or CD44.³⁶  Immature cells, especially R1 (Figures 2B, 3), outnumbered the more mature types (R1:R2 ratio at day 6 was 1:8 in controls vs 1:0.4 in β₁^Δ/Δ) and all subsets of erythroid cells in spleen were diminished compared with controls (Figure 4 top panels; supplemental Figure 2). Not only were erythroid cells very low in β₁^Δ/Δ spleen, but also progenitor cells of all classes (Figure 2C). Specifically BFU-E, “stress” BFU-E and CFU-E were all drastically diminished (Figure 2C). Importantly, there was an increased proportion of apoptotic cells in total splenocytes and in CD71⁺/TER119⁺ cells by activated caspase-3 evaluation (Figure 5A-B).

In contrast to spleen, BM revealed a different picture in which both cellularity and total CFU-C content were no different between controls and β₁^Δ/Δ mice (Figure 4). However, erythroid cells, albeit sufficient in numbers, exhibited a deficiency toward more mature cell numbers, so that, as in spleen, immature erythroid cells outnumbered their mature counterparts, suggesting a differentiation/maturation deficit (Figure 4 upper and middle right panels; supplemental Figure 2). The proportion of active caspase-3⁺/TER119⁺ cells in spleen was significantly increased over controls but to a lesser degree in BM (Figure 5B). Of interest, there was also an increase in circulating progenitor cells after PHZ in both sets of animals, but β₁-deficient mice had much higher numbers, including BFU-E and CFU-E, suggesting no impairment in generating BM progenitors in β₁^Δ/Δ mice (Table 1).

To test whether the lack of splenic response to PHZ challenge was dependent on environmental cells in the spleen or only on the absence of β₁ integrin in hematopoietic cells, we compared responses of nontransplanted β₁^Δ/Δ mice to irradiated wild-type recipient mice reconstituted with β₁^Δ/Δ donor cells. Transplanted controls were wild-type mice reconstituted with normal cells. Approximately 8 weeks after transplantation, after verification that the phenotype of PB cells was as expected (ie, in either β₁⁺ or β₁^Δ/Δ; data not shown), all transplanted mice were subjected to PHZ challenge. As seen in Figure 4, the spleens of mice reconstituted with β₁^Δ/Δ cells not only failed to respond as the transplanted controls, but the same pattern of differentiation failure as in nontransplanted β₁^Δ/Δ mice was seen (Figure 4 right top panel).

Some notable differences were present: the splenic weight (122 ± 12 g in nontransplanted β₁^Δ/Δ mice, n = 6 vs 168 ± 15 g in wild-type recipient mice reconstituted with β₁^Δ/Δ donor cells, n = 5) and total number of R1 cells in spleen were higher in β₁^Δ/Δ reconstituted versus nontransplanted β₁^Δ/Δ mice (Figure 4 upper panel). However, the erythroid cells in the recipients of β₁^Δ/Δ cells, in both BM and spleen, showed the same differentiative defect as the nontransplanted β₁^Δ/Δ mice (Figure 4 upper and middle panels). These data suggest that recipients of β₁^Δ/Δ cells do generate a few more TER119⁺ cells than the nontransplanted mice but display the same overall phenotypic response.

The modest quantitative differences between mice reconstituted with β₁^Δ/Δ cells and nontransplanted β₁^Δ/Δ mice would suggest an ameliorating effect of the normal splenic environment in the former group. Ideally, the reverse transplantation (ie, normal [β₁⁺] cells into β₁^Δ/Δ recipients) would test this issue. However, this type of experiment is not practically feasible because of the poor health (nonhematopoietically related) of β₁^Δ/Δ animals exacerbated by the toxicity of irradiation. Instead, we used CFU-C in nonirradiated normal spleens (Figure 5C; supplemental Figure 3), which tests the ability of β₁^Δ/Δ cells to settle in the nonirradiated environment of a normal spleen. In the second approach, the read-out of β₁^Δ/Δ-derived CFU-S (size and colony number) in the irradiated normal spleen tests the ability of normally lodged β₁^Δ/Δ progenitors to form CFU-S. The data from these experiments are presented in Figure 5D (bottom panels). It is evident from these experiments that the transient homing of β₁^Δ/Δ progenitor cells to spleen is not impaired, but their subsequent development into CFU-S is impaired, suggesting a postlodgment defect in the expansion of β₁^Δ/Δ progenitors in a normal splenic environment. Assuming a similar stress environment after irradiation and after PHZ, the data would suggest that β₁^Δ/Δ cells cannot efficiently expand and differentiate in the erythroid stress environment of the normal spleen, consistent with the picture in β₁^Δ/Δ transplanted mice. To provide an insight on the erythroid maturation defect, we tested the ability of both control and β₁^Δ/Δ TER119⁺ cells taken from spleens on day 6 after PHZ to adhere to control or β₁^Δ/Δ spleen frozen sections. Normal erythroblasts adhere less to β₁^Δ/Δ spleen than to normal spleen (294 ± 23 adhered cells and 641 ± 122 adhered cells, respectively, P < .05) and β₁^Δ/Δ erythroblasts adhered, as expected, less to normal spleen (supplemental Figure 4B). If adhesion is a requisite step for erythroid maturation, our data would suggest that the splenic environment or its altered architecture after PHZ in β₁^Δ/Δ mice may fail to provide an additional support mechanism for proper red cell maturation. However, to what extent these in vitro adhesion-dependent data together with in vitro EI formation using β₁^Δ/Δ or control Mϕs for attachment of erythroid cells (supplemental Figure 4C-D) reflects the in vivo picture is unclear.

Both α₄β₁ and α₅β₁ integrins have been implicated in proliferation of erythroid cells, and α₄ integrin particularly has been considered a participant in EI interactions.³⁷  To characterize the in vivo effects of these integrins in adult hematopoiesis and avoid the use of antibody treatment, we compared the response to stress erythropoiesis among the 3 genetic models: β₁^Δ/Δ mice with a deficiency primarily of α₅β₁ in erythroid cells, α₄^Δ/Δ mice lacking both α₄β₁ and α₄β₇ (the latter being overexpressed in β₁^Δ/Δ mice), and VCAM-1^Δ/Δ mice lacking the major ligand of α₄ integrin in ME cells. Survival and detailed spleen and BM responses were assessed in these models in a similar way as described earlier for β₁^Δ/Δ mice.

Whereas occasional death was seen in α₄^Δ/Δ mice after PHZ treatment, 90% of animals tested survived and responded to PHZ stress (supplemental Figure 1). A transient deficit in progenitor content was seen at day 4 in both BM and spleens of α₄^Δ/Δ compared with controls, but there was a quick recovery at day 6 and beyond (Figure 6). By day 6 after PHZ treatment, their response in spleen was normal in cellularity, progenitor content, and total erythroid cells compared with their controls, unlike the response of β₁^Δ/Δ mice (Figure 6). However, a defect in maturation profile was evident in both spleen and BM of these mice (Figure 6A-B right panels). This was probably responsible for their delayed hematocrit response to PHZ (supplemental Figure 1) and alluded to in our earlier publication.²³  No differentiation defect was seen in the spleen or BM of VCAM-1^Δ/Δ mice, in contrast to the 2 other models (Figure 6C).

Overall, α₄^Δ/Δ mice share the terminal maturation defects present in BM and spleen with β₁^Δ/Δ mice. The data are supportive of the concept that α₅β₁ exerts a critical influence on the accumulation and survival of hematopoietic progenitors in spleen (ie, on the Epo-dependent stage of erythroid cell differentiation), whereas α₄ alone or together with α₅ in β₁^Δ/Δ mice influences the Epo-independent maturation stages of erythropoiesis.

Dramatic differences in erythroid stress responses are seen in the spleen versus BM of both β₁^Δ/Δ mice and mice transplanted with β₁^Δ/Δ cells. To explain some of these differences, we tested whether differences in nonhematopoietic cell β₁ ablation in these 2 tissues may be responsible for this outcome. Three candidate populations (fibroblasts, endothelial cells, and resident Mϕs) were tested for β₁ ablation after polyriboinosinic acid/polyribocytidylic acid treatment. Efficient ablation in Mϕs, but only partial ablation in fibroblastic cells, was found (supplemental Figure 5). The data for cells in spleen were similar.

To examine the expression of β₁ integrin in endothelial cells in situ both in BM and spleen, we labeled them with anti-β₁ and with 9EG7, a monoclonal antibody reacting only with activated α₄β₁.³⁸  As seen in Figure 7, endothelial cells in BM and spleen were similarly labeled in control and β₁^Δ/Δ tissues. Megakaryocytes were labeled only in controls, not in β₁^Δ/Δ, but were labeled in all mice with anti-CD31 (data not shown). Thus, among the 3 populations tested, we found no evidence for better ablation of β₁ integrin in spleen versus BM to explain differences in response. Despite phenotypic similarities, however, it is possible that cells may behave differently in the 2 environments.

To further understand the mechanisms responsible for observed impairment in stress erythropoiesis in spleen, we assessed the expression of BMP4, a signaling molecule implicated in stress erythropoiesis^27,28  and produced by stromal and hematopoietic cells.³⁹  By quantitative reverse-transcribed polymerase chain reaction (data not shown) and as supported by immunofluorescent staining, no gross differences in BMP4 expression were found in the spleens of β₁^Δ/Δ versus control mice (Figure 7I-J). Furthermore, we assessed the role of Mϕs by treating mice with clodronate (supplemental Figure 6). Although the response of clodronate-treated animals to PHZ challenge was severely compromised (supplemental Figure 6 top panels), the interpretation of these experiments is confounded by the effects of clodronate treatment in spleen beyond Mϕ purging (supplemental Figure 6 bottom panels).

Studies using several genetic mutant mice have shown impaired responses to erythroid stress, although erythropoiesis is fairly normal at homeostasis (ie, ICAM,³  Gas-6,⁷  and Stat5a,b³⁵  Rac1/2).¹⁸  Furthermore, in several models of stress erythropoiesis, impairment of response is observed in both BM and spleen or only in one of these tissues. For example, in the absence of the glucocorticoid receptor, erythropoiesis is compromised both in BM and spleen.⁴⁰  Yet, in other models (ie, in Rb null mice), prominent changes are seen only in the spleen.^41,42  In contrast to Rb, the absence of Rac1/Rac2 signals leads to defective early erythropoiesis in BM, but in the spleen erythropoiesis is completed successfully, as the splenic environment apparently circumvents the need of Rac1/Rac2 in early erythropoiesis.

Our results with β₁^Δ/Δ mice have some features in common to previously described mutant mice, but others appear to be unique to our model. Similar to Rb null mice⁴²  or the TRalpha null mice⁴³  and FAK^−/− mice⁴⁴  and in sharp contrast to Rac1/Rac2-deficient mice,¹⁸  defects in β₁^Δ/Δ mice are prominent only in the spleen, leading to failure to survive stress (Figure 2). Precise mechanistic differences between BM and spleen responses have not been explored in most of the cases, although several insights have been obtained. For example, it was previously shown that convergence of Sonic hedgehog (Shh), BMP4, and stem cell factor-dependent signaling in spleen is necessary for the development and expansion of BMP4-resposive stress erythroid progenitors.^27,28  Thus, Smo^Δ/Δ mice,²⁸  with absence of Shh signaling, do not recover from a second PHZ treatment,²⁸  resulting from failed splenic response.

In our β₁^Δ/Δ mice, the generation of progenitor cells by BM and their egress to PB, which is necessary for subsequent colonization of the spleen as it occurs in normal animals, was not impaired (Figure 4; Table 1). But why did these progenitors not accumulate in β₁^Δ/Δ spleens? To explain these data, we entertained the following mechanistic scenarios

An attractive possibility is that progenitors generated in BM and delivered to circulation, as shown by their increase in PB after PHZ (Table 1), cannot home and/or cannot be firmly retained in the spleen if they lack β₁ integrin. Such homing/retention properties may be dependent primarily on the presence of α₅β₁ because in the other 2 related mutants (ie, the α₄^Δ/Δ or VCAM-1^Δ/Δ mice in which fibronectin/α₅ interactions are present), homing of progenitors to spleen is not impaired.²³  Consistent with this concept, previous data with fetal liver β₁ knockout cells have shown failure of BM and splenic colonization by fetal liver β₁ null cells⁴⁵ ; and in another study, there was a small decrease in homing to irradiated spleen with anti-α₅ integrin antibody-treated donor cells.¹⁵  However, our present data, showing no significant impairment of splenic homing in nonirradiated recipients given adult ablated β₁ cells (Figure 5; supplemental Figure 1), appear inconsistent with these data. In this context, it is important to emphasize that homing in some of the previous data was evaluated by successful engraftment (colonization), whereas in our studies we have dissociated homing/lodgment events from subsequent expansion/engraftment events with β₁^Δ/Δ cells. Despite the normal homing, we found that the development of CFU-S was significantly impaired in β₁^Δ/Δ mice as shown by the reduced number and size of CFU-S (Figure 5), suggesting impairment in survival and/or expansion of homed progenitors. Thus, the absence of β₁ signaling in the context of fibronectin-dependent interactions in the spleen has a profound impact on progenitor cell survival and proliferation. Activation of signaling pathways intersecting β₁ integrin signaling⁴⁶  may promote expression of unique sets of genes responsible for cell proliferation/survival, and our data in β₁-decifient PHZ-treated spleens are consistent with this thesis.

The connection of β₁ or α₅β₁ signaling with survival and/or proliferation is supported by a host of previous data:

1. Fibronectin, and not VCAM-1, is the preferential ligand for cycling cells, and α₅ integrin is up-regulated in cycling cells, in contrast to α₄, which is not different in cycling versus noncycling cells.⁴⁷ 

2. In vivo treatment of mice with C274, the RGD-dependent recombinant fibronectin, reduced splenic progenitor growth (HPP-CFC).¹²  By contrast, treatment with CH296 fibronectin containing all binding sites, including the CS-1 VLA4 binding site, did not decrease splenic progenitors.

3. Donor cells treated with a rabbit anti-β₁ antibody resulted in reduced numbers of CFU-S in irradiated normal spleen.⁴⁸ 

4. Wv mutant mice, which have an impaired splenic response to stress after PHZ or after transplantation, are not able to activate α₅.¹² 

5. Impairment of β₁-dependent FAK in conditional mutant mice led to a picture of PHZ response similar to our β₁^Δ/Δ mice with drastically reduced splenic response and decreased proliferation and survival of progenitor cells.⁴⁴ 

6. α₅β₁ integrin influenced Shh signaling in intestinal epithelia proliferation, and the phenotypic changes of β₁ deletion in intestinal epithelial cells were strikingly similar to mice with defective Shh expression or signaling. In aggregate, these findings, together with our own, support the concept of impaired survival and proliferation of α₅β₁-deficient progenitors in the stress splenic environment.

But why would an intrinsic defect like β₁-deficiency lead to impaired interactions selectively within the splenic environment? It is possible that in the spleen, in contrast to BM, retention and subsequent development of erythroid cells are mainly an Arg-Gly-Asp (RGD)-dependent process (RGD being the best known recognition sequence in fibronectin for α₅β₁ integrin). This process will fail in β₁^Δ/Δ mice in which erythroid cells lack α₅β₁. However, in BM other redundant signals on erythroid cells (ie, CD44, α₄β₇, and β₇ integrin) can interact with environmental ligands unique to BM (ie, tenascin-C, laminin species, osteopontin, thrombospondin, syndecan-4, or heparan sulfate proteoglycans),^46,49  and usurp or substitute for the RGD effects on developing erythroid cells. Furthermore, beyond RGD-dependent effects, it is possible that other ME cells or matrix in BM display distinct functional behavior compared with spleen. Such a functional difference could be attributed to differences in the efficiency of the ablation of β₁ integrin between BM versus splenic stromal cells. We have obtained no evidence that cultured fibroblasts are better ablated in BM versus the spleen, whereas endothelial cells in both BM and spleen are, like normal cells, positive for β₁ and9EG7, an antibody that reacts with the activated form of β₁³⁸  (Figure 7), suggesting no effective β₁ ablation. The latter data are consistent with difficulties previously encountered with ablation of β₁ integrin in adult endothelial cells.⁵⁰  By contrast, Mϕs or F4/80⁺ cells in the BM and the great majority of those in the spleen are ablated for β₁ integrin. However, despite phenotypic appearances and depending on the signals they receive from their MEs, it is possible that the Mϕs are functionally different between these 2 tissues as previously shown for other tissues.⁵¹  Nevertheless, it is doubtful that only the functional impairment of β₁^Δ/Δ macrophages can explain the severe phenotype in β₁^Δ/Δ spleens, as all classes of progenitors were affected. Furthermore, the functional role of Mϕs in proliferation/maturation/enucleation of erythroid cells has not been fully elucidated, and controversial information is currently available regarding the molecular interactions between erythroid cells and Mϕs (ie, ICAM-4/α_vβ₃).⁵²  Although many of the maturational defects have been cell-intrinsic (glucocorticoid receptor, thyroid receptor, β₁ integrin), the specific interactions of these structures with ME cells or their effects on apoptosis pathways specifically activated in spleen in erythroid stress have been unclear.

Although progenitor defects in spleen quickly recovered in PHZ-treated α₄^Δ/Δ mice, defects on terminally differentiated cells were present in both α₄- and β₁-deficient animals. Because the differentiation defects were similar in α₄^Δ/Δ and β₁^Δ/Δ mice, it is possible that interactions of both α₄ and α₅ through fibronectin or the interaction of α₄ with both fibronectin and Mϕs are needed for efficient terminal erythroid differentiation. It is tempting to speculate that the effect in β₁^Δ/Δ mice is a combination of absent α₅ together with hypofunctional α₄, compromising adhesive interactions with both fibronectin and Mϕs in spleen. The presence of α₄β₇ did not seem to provide any protective effect in β₁^Δ/Δ mice. It is of interest that, despite prior implications of VCAM-1,⁵  there was absence of differentiation defects in VCAM-1^Δ/Δ mice in our experiments (Figure 6) that were consistent with experiments in fetal liver, in which the effects of VCAM-1 were dissociated from those of α₄β₁.⁵³  Moreover, using K562 cells as a model, it was shown that prolonged engagement to fibronectin throughVLA₅ (α₅β₁) triggered the induction of α₄ in K562 cells (normally lacking α4) and the up-regulated α₄ influenced erythroid differentiation by activating p38, an effect blocked by the RGD peptide.¹¹  Thus, the weight of evidence from previous and present data seems to support the notion that, at least in adults, α₄β₁ interactions (with fibronectin or other partners) may be important for influencing terminal erythroid differentiation, probably within the context of EI, and may act in parallel or in cooperation with α_v integrins (interacting with ICAM4 and VLA4 on erythroid cells).³  However, further studies are needed to clarify some of these issues.

Transplanted mice reconstituted with β₁^Δ/Δ cells had a somewhat better erythroid response (Figure 4), suggesting that the stroma in the spleens of transplanted mice may have contributed to this. A cellular candidate of host origin that might have made a difference in transplanted animals is the F4/80⁺ Mϕs in the spleen. Although the total number of F4/80⁺ cells was not significantly different in nontransplanted versus transplanted animals, the proportion of β₁⁺ F4/80⁺ cells was higher (23% vs < 10% in β₁^Δ/Δ) in transplanted animals, indicating incomplete reconstitution of Mϕ population by donor cells. Thus, the presence of β₁⁺ Mϕs (ie, α_vβ₁⁺) could engage β₁^Δ/Δ erythroid cells through ICAM-4³  or through other undefined ways to promote erythroid expansion and terminal maturation. Intrinsic functional defects in β₁^Δ/Δ Mϕs are also possible. In this context, it was recently noted that, using Lysozyme M-mediated ablation of β₁ in Mϕs, β₁ integrin is important for Mϕ maturation, as F-actin assembly in these cells was impaired through diminished Rac expression.⁵⁴  A similar defect in EMP null Mϕs led to failure of enucleation in erythroid cells.⁶ 

In conclusion, our data evaluating stress erythropoiesis in 3 genetically deficient animals (β₁conditional, α₄, and VCAM-1 with Tie-2Cre-mediated ablation) redefine the functional role of α₅ and α₄ integrins in this process and their distinctive interactions in BM versus spleen.

The authors thank Peirong Wang for his assistance with the photographs. Clodronate was supplied by Roche Diagnostics (Germany), and encapsulated liposomes were from Dr Nico van Rooijen, Department of Cell Biology and Immunology, Faculty of Medicine, Vrije Universiteit, Amsterdam, The Netherlands.

This work was supported by the National Institutes of Health (grants HL058734 and HL46557).

National Institutes of Health

Contribution: T.U. performed experiments and wrote relevant parts of the manuscript; Y.J., S.P., and B.N. performed experiments; and T.P. designed the experiments, evaluated data, and wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Thalia Papayannopoulou, Department of Medicine, Division of Hematology,1705 NE Pacific Street, Box 357720, Seattle, WA 98195; email: thalp@uw.edu.