Lnk inhibits erythropoiesis and Epo-dependent JAK2 activation and downstream signaling pathways

Erythropoietin (Epo) is essential for red cell production by preventing committed erythroid progenitors from undergoing apoptosis and allowing them to proliferate and differentiate. Epo^-/- and Epo receptor (EpoR)^-/- mice die at embryonic day 13.5 (e13.5) due to severe fetal anemia.¹  The EpoR belongs to the type I cytokine receptor family, characterized by a single transmembrane domain and a cytoplasmic tail lacking a kinase domain. A ligand-induced receptor homodimer conformational change leads to trans-phosphorylation and activation of Janus kinase 2 (JAK2).^2,3  Erythroid development in JAK2-deficient embryos is arrested earlier and fetal anemia is more severe than that of EpoR^-/- embryos, indicating JAK2 is essential for signaling downstream of multiple cytokine receptors, in addition to EpoR, that are important for primitive erythroid development.^4,5 

Activated JAK2 phosphorylates key tyrosine residues in the EpoR cytoplasmic domain, thereby providing docking sites for SH2 domain-containing downstream signaling molecules. EpoR activates signal transducers and activators of transcription 5 (Stat5), Ras/mitogen-activated protein kinase (MAPK), and phosphoinositide-3 kinase (PI-3K)/Akt pathways.^2,3  These signaling modules have been implicated in hematopoiesis. Stat5 plays an important role in maintaining a high erythropoietic rate during fetal development and during stress responses in adult mice, as shown by studies using stat5a^-/- stat5b^-/- mice. Stat5ab-deficient erythroid progenitors exhibit high levels of apoptosis and are less sensitive to Epo.⁶  Moreover, the antiapoptotic role of Stat5 in EpoR signaling is mediated through its direct induction of Bcl-X_L in erythroid cells.⁷ 

Intricate cytokine receptor signaling networks rely heavily on adaptor proteins. Lnk is a member of a newly discovered adaptor protein family. Lnk and other family members, APS and SH2-B, share a common protein organization. They do not possess a kinase domain but contain several protein-protein interaction domains: a proline-rich amino-terminus, a pleckstrin homology (PH) domain, a Src homology 2 (SH2) domain, and a conserved tyrosine near the carboxyl-terminus.⁸ 

Among mice nullizygous for Lnk family members, Lnk-deficient mice show the most profound perturbation in hematopoiesis. Mice nullizygous for Lnk revealed an expansion of pro/pre and immature B cells.⁹  Moreover, overexpression of Lnk in aorta-gonad-mesonephros (AGM) primary cultures suppresses the emergence of CD45⁺ hematopoietic cells, via inhibition of the stem cell factor (SCF)/c-Kit signaling pathways.¹⁰  We recently demonstrated that in both bone marrow and spleen, Lnk-deficient mice exhibit increased numbers of megakaryocytes with increased ploidy. Lnk-deficient megakaryocytes derived from bone marrow (BM) and spleen display enhanced sensitivity to thrombopoietin (Tpo) during in vitro culture. Lnk overexpression in BM progenitor cells attenuated Tpo-dependent megakaryocyte proliferation and endomitosis. Importantly, Lnk-deficient megakaryocytes show enhanced and prolonged activation of Tpo-induced signaling pathways, indicating Lnk is a physiologic negative regulator of Tpo-mediated signaling and megakaryocytopoiesis.¹¹  However, a role for Lnk in erythroid development or Epo/EpoR signaling has not been established.

Lnk-deficient mice have near-normal steady-state hematocrits, even though there are increased numbers of erythroid colony-forming unit (CFU-e) progenitors in the spleen.¹²  We hypothesized that any disturbance in erythropoiesis in Lnk-nullizygous mice might be disguised due to a large erythropoietic reserve capacity in adult animals. To test this, we challenged adult Lnk^-/- mice with erythropoietic stress, and found that they showed superior recovery with an enhanced erythropoietic rate. Lnk-deficient splenic erythroid progenitor cells exhibited enhanced Epo activation of the Akt and MAPK signaling pathways. Importantly, Lnk-deficient CFU-e progenitors displayed enhanced sensitivity to Epo in forming erythroid colonies. To dissect the mechanism by which Lnk modulates EpoR signaling, we found that overexpression of Lnk, but not the R364E mutant with a disrupted SH2 domain, inhibited Epo-dependent growth and signaling in 32D/EpoR cells. Extending this result using a novel primary culture of erythroid progenitors, we demonstrated that the Lnk SH2 domain was essential for blocking Epo-dependent erythroid differentiation and inducing erythroblast apoptosis. Furthermore, our results indicate that Lnk directly attenuates EpoR phosphorylation and JAK2 activation, thereby inhibiting major signaling pathways initiated by Epo/EpoR.

Lnk-deficient mice were generously provided by Dr Tony Pawson (Samuel Lunenfeld Research Institute, Toronto, Canada), and have been backcrossed to the C57/Bl6 background for 7 generations. Epo was generously provided by Amgen (Thousand Oaks, CA). Interleukin 3 (IL-3)-dependent 32D cells were maintained as described in Tong and Lodish.¹¹ 

pcDNA-Lnk was generously provided by Dr Satoshi Takaki (University of Tokyo, Japan). Lnk cDNA was digested with BamH1 and EcoR1 sites and ligated to the MSCV-IRES-hCD4 (MICD4) vector linearized by BglII and EcoR1. The Lnk constructs in the MSCV-IRES-GFP (MIG) vector were made as described.¹¹ 

Fetal liver cells were isolated from e13.5 or e14.5 Balb/c mice (Jackson Laboratories, Bar Harbor, ME), and Ter119^- progenitor-rich erythroid cells were purified as described earlier.¹³  Cells were then plated at 1 × 10⁵/mL to 2 × 10⁵/mL in a 12-well plate and spin-infected for one hour with the desired retroviral supernatant. Transduced cells were then cultured in Iscoves modified Dulbecco medium (IMDM) containing 15% fetal bovine serum (FBS), 30 ng/mL SCF, 20 ng/mL Flt3L, and 10 ng/mL IL-6 for 6 hours, to allow expression of transduced genes and maintenance of the progenitor state. The cells were subsequently washed and resuspended in 1 mL media containing 2 U/mL Epo on fibronectin-coated plates.¹³  Some experiments included controls where Epo was omitted. On the second day, the media were replaced by 2 mL erythroid differentiation media (EDM) as described elsewhere.¹³ 

Cultured fetal liver cells were washed and resuspended in phosphate-buffered saline (PBS) with 2% FBS at a concentration of 3 × 10⁵/mL. Cells (100 uL per slide) were centrifuged onto slides at 1000 rpm for 5 minutes (Cytospin 3; Thermo Shandon, Pittsburgh, PA). The slides were air dried and stained with 3,3′ diaminobenzidine (Sigma, St Louis, MO) and Wright-Giemsa (Sigma) according to the manufacturer's recommendation. Cytology images were taken using an Olympus BH-2 microscope (Olympus, Melville, NY) with Splan FL2 100× objective lenses and a C-35 AD-2 camera. Numerical aperture was 1.25 Oil. Images were then scanned using Nikon Super Coolscan 4000ED (Nikon, Melville, NY) and acquired with Adobe Photoshop software (Adobe Systems, San Jose, CA).

One day or 2 days after the culture of retroviral-transduced erythroid progenitor cells, the resulting erythroblasts were removed from the culture plates. Cells were then stained for hCD4, Ter119, and CD71 as described earlier.¹³  For apoptosis analysis, the erythroblasts were stained first for hCD4, then incubated with phycoerythrin (PE)-conjugated annexin V and 7-AAD (7-actinomycin D; BD Pharmingen, San Diego, CA) for 15 minutes at room temperature according to the manufacturer's protocol. Fluorescence activated cell sorting (FACS) analysis was carried out using a Becton Dickinson FACSCalibur.

To detect CFU-e colonies, 1 × 10⁵ BM and 1 × 10⁶ spleen cells from wild-type mice, and 1 × 10⁵ BM and 5 × 10⁵ spleen cells from Lnk^-/- mice were plated in duplicate in semisolid methylcellulose (M3334; StemCell Technologies, Vancouver, BC, Canada) according to the manufacturer's protocol. Benzidine-positive (Sigma) colonies were counted after 2 days in culture. To detect erythroid burst-forming unit (BFU-e) colonies, 1.5 × 10⁴ BM and 3 × 10⁵ spleen cells from wild-type mice and 7.5 × 10³ BM and 1 × 10⁵ spleen cells from Lnk^-/- mice were plated in duplicate in methylcellulose (M3434; StemCell Technologies) according to the manufacturer's protocol. The BFU-e colonies were counted after 8 to 10 days in culture. To test Epo sensitivity of CFU-e progenitors, cells from wild-type and Lnk-deficient spleens were isolated and plated in duplicate in methylcellulose (M3234; StemCell Technologies) containing varying concentrations of Epo. Same numbers of cells were plated as described above.

Wild-type and Lnk-nullizygous mice (7 to 9 weeks old; 22 g to 28 g per mouse) were injected intraperitoneally with phenylhydrazine (PHZ) at 60 mg/g body weight on days 0 and 1. Mice of both genotypes were divided into 2 subgroups, and retro-orbital vein bleeds were obtained from each subgroup at alternative intervals, in order to prevent phlebotomy-induced anemia. Hematocrits were measured on a hematocrit centrifuge (Micro-MB; International Equipment, Needham, MA). Freshly drawn blood was stained with new methylene blue “N” (J.T. Baker, Phillipsburg, NJ) for detecting RNA in reticulocytes. Reticulocyte count was subsequently measured by counting the portion of reticulocytes in total red cells on blood smears. The corrected reticulocyte count was calculated as described in the legend to Figure 1.

After sorting 32D/EpoR cells for the green fluorescent protein-positive (GFP⁺) population expressing either vector alone or a vector encoding Lnk, we starved the cells in RPMI containing 1% bovine serum albumin (BSA) and subsequently stimulated them with 10 U/mL Epo for 0, 10, 30, and 120 minutes. The cell lysates were Western blotted with the indicated antibodies: anti-Stat5 (1:500, C-17; Santa Cruz Biotechnology, Santa Cruz, CA), anti-p-Stat5 (pTyr 694), p-MAPK (pThr202/Tyr204), and p42/44 MAPK antibodies (1:1000; Cell Signaling Technology, Beverly, MA).

Ter119^- erythroid progenitor cells were transduced and cultured as described in “Fetal liver cell preparation and culture,” or directly cultured in the presence of Epo. At 14 to 16 hours after culturing in Epo-containing media, erythroblasts were starved in IMDM containing 1% BSA for 2 hours and stimulated with 10 U/mL Epo for 0, 10, 30, and 120 minutes. The protein lysates were Western blotted as described above. For immunoprecipitation assays, the cells were stimulated with 10 U/mL Epo for 7 minutes, and lysed in 10 mM Tris-Cl (pH 7.4) and 150 mM NaCl, and 1% NP-40, containing phosphatase and protease inhibitors. The protein supernatant was precipitated with anti-EpoR or JAK2 serum (1:1000 for EpoR antibody and 1:250 for JAK2 antibodies; Upstate Cell Signaling Solutions), or anti-Lnk antibodies (M20, Santa Cruz Technology). The precipitates were blotted with anti 4G10 (1 μg/mL; Upstate Cell Signaling Solutions, Charlottesville, VA), anti-EpoR (0.5 μg/mL), JAK2 (0.5 μg/mL), or anti-Lnk (1:500) antibodies.

At day 4 or 5 after PHZ treatment, spleens from wild-type and Lnk-deficient mice were isolated. Mature red cells were lysed using ammonium chloride solution, and CD71⁺ erythroid progenitor cells were purified using fluorescein isothiocyanate (FITC)-CD71 antibodies followed by EasySep FITC magnetic beads purification (StemCell Technology). CD71⁺ erythroid progenitor cells were subsequently starved for one hour in IMDM containing 1% BSA and stimulated with 0, 1, 10, and 100 U/mL Epo for 10 minutes. Western blot analysis was performed as described above.

We found that Lnk-deficient mice displayed increased erythroid progenitor numbers. As shown in Table 1, while the numbers of CFU-e progenitors showed only slight differences in Lnk-deficient BM, BFU-e progenitor numbers (per 2 × 10⁴ BM cells) increased 1.7-fold compared with wild-type controls (Table 1, column 4). Taking into account that there are 30% more BM cells in Lnk-deficient mice relative to normal mice (Table 1, column 4), the total number of BFU-e progenitors in Lnk-deficient BM is 2.2-fold greater than in wild-type mice (Table 1, column 5). Spleens from Lnk-deficient animals displayed more profound differences than did BM. Splenic BFU-e and CFU-e progenitors increased 1.7- and 2.4-fold in frequency respectively, compared with wild-type mice, when equal numbers of cells were analyzed (Table 1, column 4). Since there are twice (1.9 ×) as many spleen cells in Lnk-deficient mice relative to wild-type mice, the total numbers of BFU-e and CFU-e progenitors are increased 3.2- and 4.6-fold, respectively, in Lnk-deficient spleen (Table 1, column 5). Thus, Lnk deficiency resulted in an abnormal increase in erythroid progenitor numbers in both BM and spleen.

Even though Lnk-deficient mice have increased numbers of erythroid progenitors, they exhibit near-normal steady-state hematocrits (Figure 1A and Velazquez et al¹² ). Since adult mice have a large erythropoietic reserve capacity, quantitative differences in erythropoietic rate might be masked at the steady state. We therefore tested Lnk nullizygous mice for their ability to generate high erythropoietic rates under stress. We challenged the mice with a chemical-induced hemolytic anemia by phenylhydrazine (PHZ) injection at days 0 and 1 (Figure 1A). The hematocrits of mice from both groups dropped sharply to 26% at day 2 (Figure 1A). Wild-type mice continued decreasing their hematocrits, reaching a nadir (22%) at day 4 (Figure 1A). In contrast, Lnk-deficient mice quickly recovered their hematocrits to 32% by day 4. At day 6, Lnk-deficient mice still exhibited significantly higher hematocrits than those of wild-type mice (Figure 1A). By days 8 to 11, both wild-type and Lnk-deficient mice recovered their normal hematocrits (Figure 1A).

Reticulocytes are red cells newly generated from bone marrow and are prematurely released into the circulation under stress. Therefore, the proportion of reticulocytes in the peripheral blood red cell population is indicative of the erythropoietic rate. The basal reticulocyte count is less than 2% (Figure 1B). Under erythropoietic stress, the reticulocyte count in wild-type mice reached a maximal level of 19% at day 4 (Figure 1B). This rapid increase in reticulocyte count was more pronounced in Lnk-deficient mice, reaching a maximal level of 31% at day 4 (Figure 1B). In summary, Lnk nullizygous mice exhibited a less severe PHZ-induced anemia and a faster recovery from erythropoietic stress.

At days 4 to 7 after PHZ treatment, spleens are enriched for Epo-responsive erythroid progenitors, and these progenitor cells are mostly CD71⁺. We purified CD71⁺ splenic erythroid progenitor cells and stimulated them with different concentrations of Epo (Figure 1C). With increasing concentrations of Epo (0 U/mL to 100 U/mL), wild-type CD71⁺ erythroblasts showed increasing activation of Akt and p42/44MAPK (Figure 1C), with maximal activation of both signaling pathways at 10 U/mL to 100 U/mL Epo. Strikingly, Lnk-deficient CD71⁺ erythroblasts exhibited higher activation of Akt and MAPK at all Epo concentrations (Figure 1C). Thus, the absence of Lnk results in enhanced signaling pathways induced by Epo.

To determine whether Lnk deficiency results in increased Epo sensitivity at the cellular level, erythroid progenitors from either wild-type or Lnk^-/- mice were exposed to a range of Epo concentrations, and erythroid colony formation was monitored. Figure 2 shows that Lnk-deficient CFU-e progenitors were indeed more responsive to lower concentrations of Epo than wild-type controls. Interestingly, BM CFU-e progenitors from these mice showed similar sensitivity to Epo (data not shown).

To study the mechanism by which Lnk inhibits Epo/EpoR signaling pathways, we overexpressed Lnk in IL-3-dependent 32D hematopoietic cells. To this end, we first established a stable cell line expressing the EpoR, designated as 32D/EpoR cells. The effect of Lnk in Epo-dependent 32D cell growth was then examined by retroviral-transduction of wild-type Lnk using a bicistronic MSCV-IRES-GFP (MIG) vector. As GFP expression is tightly correlated with the expression of the gene cloned upstream of the internal ribosome entry site (IRES),¹⁴  we were able to identify cells expressing Lnk by analyzing GFP fluorescence. We introduced either vector alone or Lnk into 32D/EpoR cells and determined the fraction of GFP⁺-infected cells 2 days later. As shown in Figure 3A, while Lnk expressing 32D/EpoR cells declined in number during the 4-day experiment period, control cells exhibited exponential growth.

In order to investigate the role of different domains of Lnk in growth inhibition, we generated point mutations in the Lnk cDNA to ablate individual domains and overexpressed these mutant forms of Lnk in 32D/EpoR cells (Figure 3B). We determined the initial fraction of GFP⁺-infected cells 2 days after infection. We then cultured the cells in Epo and measured the GFP⁺ fraction every 3 days, as the cells divide, relative to that 2 days after infection (Figure 3B). The fraction expressing wild-type Lnk was dramatically reduced to 20% after 3 days and to 3% after 6 days, whereas the fraction of control vector-expressing GFP⁺ cells was unchanged (100%) over the 9-day culture period (Figure 3B). In contrast, a mutation that disrupts the Lnk SH2 domain (R364E) completely abolished its ability to inhibit cell growth (Figure 3B), whereas mutation of the conserved tyrosine at the C-terminus (Y536F) or a mutation that disrupts the functional PH domain (W191A) only moderately compromised Lnk inhibitory function (Figure 3B).

We next studied downstream signaling molecules affected by Lnk in 32D cells. We purified 32D/EpoR cells transduced with the vector alone, wild-type Lnk, or Lnk mutants by flow cytometric sorting, and subjected them to Western blotting analysis following Epo stimulation. In vector-transduced 32D/EpoR cells, Epo induced maximal Stat5 phosphorylation at 10 to 30 minutes, and the level of Stat5 phosphorylation dropped at 120 minutes (Figure 3C, lanes 1-4). Overexpression of Lnk abolished Stat5 activation induced by Epo (Figure 3C, lanes 5-8), but total Stat5 protein levels were unchanged (Figure 3C, lower panel). Similarly, in control cells, Epo induced maximal p44/42 MAPK phosphorylation at 10 minutes, and the level of MAPK phosphorylation dropped thereafter (Figure 3D, lanes 1-4). Lnk overexpression diminished this induction (Figure 3D, lanes 5-8). Total MAPK protein levels were unaffected (Figure 3D, lower panel).

Figure 3C-D also shows that the Lnk SH2 domain is essential for the inhibition of Stat5 and MAPK activation by Lnk. Lnk (R364E)-transduced cells showed similar Stat5 and MAPK activation to control cells (Figure 3C-D, lanes 9-12). In contrast, Lnk (W191A)- and Lnk (Y536F)-expressing cells attenuated the activation of Stat5 and MAPK to an extent similar to that by wild-type Lnk (Figure 3C-D, lanes 13-20). Thus, the Lnk SH2 domain is essential for inhibiting EpoR signaling and Epodependent cell growth in 32D cells, whereas neither the conserved tyrosine at the C-terminus nor the PH domain is required for these inhibitory functions.

In order to investigate whether Lnk affects Epo function in red cell development, we examined primary fetal liver erythroid progenitors. Ter119^- erythroid progenitor cells were purified from e13.5 or e14.5 fetal livers, and retrovirally transduced with either vector alone or Lnk. Ter119^- progenitor cells, when cultured in the presence of Epo, divide 4 to 5 times, and differentiate into enucleated reticulocytes, a process remarkably similar to that seen in vivo.¹³  Since infection rates routinely reach 90%, no further purification of infected cells is necessary.

We first quantitated the effect of Lnk on Epo-dependent primary erythroblast expansion (Figure 4A). By the end of a 2-day culture in the presence of Epo, control cells increased about 15-fold in number, whereas control cells cultured in the absence of Epo increased only 4-fold (Figure 4A). Lnk expressing erythroblasts cultured in the presence of Epo also increased only 4-fold, similar to control cells cultured in the absence of Epo (Figure 4A).

We next analyzed the mechanism responsible for Lnk-mediated inhibition of Epo-dependent cell expansion. At the end of a 2-day culture, those infected with the control vector contained 6% apoptotic (annexin V⁺ and 7-AAD^-) and 6% dead cells (annexin V⁺ and 7-AAD⁺; Figure 4B). Overexpression of Lnk resulted in 33% apoptotic and 25% dead cells, numbers similar to those of control cells cultured in the absence of Epo (30% apoptotic and 21% dead cells; Figure 4B). However, Lnk did not block cell cycle progression of proerythroblasts after 16 to 18 hours of culture in the presence of Epo. Control cells exhibited exponential growth with 31%/57%/12% G₁/S/G₂ populations, which is similar to the cell-cycle progression of Lnk-expressing cells, 33%/59%/8% G₁/S/G₂ populations. Interestingly, the lack of Epo also did not block cell-cycle progression; control cells grown in the absence of Epo showed 34%/56%/10% G₁/S/G₂ populations. Thus, the cell number reduction in Lnk-expressing cells is due to increased apoptosis.

Consistent with the results obtained from 32D cells, the Lnk SH2 mutant R364E did not affect erythroblast survival when overexpressed in Ter119^- erythroid progenitor cells (Figure 4). In contrast, the PH domain mutation (W191A) or Y536F mutation moderately disrupted Lnk inhibitory function (Figure 4). Thus, the Lnk SH2 domain is crucial for, and the PH domain and the conserved tyrosine contribute to, its inhibitory function in Epo-mediated erythroid cell survival.

We next examined whether Lnk affects erythroid differentiation in the primary erythroid culture system as described earlier.^7,13  Cultured erythroblasts were stained for erythroid-specific marker Ter119 and nonerythroid-specific transferrin receptor CD71. Five distinct populations can be defined by their characteristic staining patterns: Ter119^lowCD71^med (primitive progenitor cells and proerythroid cells, Figure 5A, R1); Ter119^lowCD71^high (proerythroblasts and early basophilic erythroblasts, Figure 5A, R2); Ter119^highCD71^high (basophilic erythroblasts, Figure 5A, R3); Ter119^highCD71^med (chromatophilic and orthochromatophilic erythroblasts, Figure 5A, R4), and Ter119^highCD71^low (late orthochromatophilic erythroblasts and reticulocytes, Figure 5A, R5). Therefore, these cell populations correspond to progressive developmental stages, with R1 being the least and R5 being the most differentiated erythroid cells.

In freshly isolated e14.5 fetal liver cells, about 15% are Ter119^- (Figure 5A). After magnetic bead purification, Ter119^- populations are routinely enriched to about 95% with most of them being early erythroid progenitor cells (R1 and R2, Figure 5B). After 16 to 18 hours of culture of Ter119^- progenitors in the presence of Epo, control erythroblasts (R1 and R2 cells) differentiate into R2 and R3 cells, shown as CD71^high and Ter119^low to high (Figure 5D, left panels).¹³  Following a second day of culture, the erythroblasts further differentiate into benzidine-positive R3-R5 cells (Figure 5D, right panels).

In contrast, control erythroblasts cultured in the absence of Epo did not show hyper-upregulation of CD71/TfR (Figure 5E, left panels); they showed only 50% of the expression levels of CD71/TfR relative to control cells cultured in the presence of Epo for 16 to 18 hours (Figure 5C). Instead, they prematurely expressed intermediate levels of Ter119 and this abnormal Ter119 expression did not associate with hemoglobin expression as determined by benzidine-Giemsa staining (Figure 5E, left panels). Interestingly, Lnk-expressing erythroblasts cultured in the presence of Epo differentiated in an indistinguishable manner compared with control cells cultured in the absence of Epo (Figure 5F, left panels). They expressed comparable levels of CD71/TfR as control cells lacking Epo (Figure 5C). Importantly, erythroblasts overexpressing the Lnk SH2 mutant (R364E) showed normal levels of CD71/TfR expression (Figure 5C). Therefore, Lnk expression blocks Epo-dependent hyper-up-regulation of CD71/TfR expression as early as 16 to 18 hours after culture, and the Lnk SH2 domain is essential for this inhibition.

Due to rapid expansion of control erythroblasts in the presence of Epo (15-fold increase shown in Figure 4A), nonerythroid cells (R0) that reside in the Ter119^- fraction at the start of the culture (15%-25%) only make up 2% to 8% of the population at the end of a 2-day culture (Figure 5D, right panel, dashed line region). In contrast, when cultured in the absence of Epo, the total erythroblast number increased only 4-fold (Figure 4A); among them, the nonerythroid cell population (R0) became more visible (20%-40%) due to proportionally fewer cells of erythroid lineage (Figure 5E, right panel, dashed line region). Cytologic studies with benzidine-Giemsa staining recapitulated this finding (Figure 5D-E, right panels). Specifically, at the end of a 2-day culture, vector-transduced erythroid cells cultured in the presence of Epo showed large quantities of orthochromatophilic erythroblasts and enucleated reticulocytes (Figure 5D, right panel, arrowheads). However, there were markedly fewer cells of the erythroid lineage in vector controls cultured in the absence of Epo (Figure 5E, right panel). In addition, as determined by morphology, the nonerythroid cell population was significantly higher (Figure 5E, right panel, thin long arrows), compared with controls cultured in the presence of Epo. The nonerythroid populations are mostly myeloid (Mac-1⁺/Gr-1⁺) cells, and some B cells (B220⁺) (data not shown).

Interestingly, the differentiation profile of Lnk-expressing erythroblasts is very similar to that of control cells cultured in the absence of Epo, determined by both FACS analysis and cytology (Figure 5F; compare to Figure 5E). The reticulocytes in Lnk-expressing cells cultured in Epo and control cells cultured in the absence of Epo appear to be abnormal: they exhibited less benzidine staining and were smaller in size than normal.

In summary, the Lnk SH2 domain is essential to block Epo-dependent erythroid differentiation by preventing CD71/TfR upregulation.

To investigate the mechanism by which Lnk inhibits Epo-dependent erythroblast survival, we examined Epo-induced signaling pathways in primary fetal liver erythroid cells transduced with either the control vector or the vector encoding Lnk. Epo stimulated maximal phosphorylation of p44/42MAPK, Akt, and Stat5 in 10 minutes in control erythroblasts (Figure 6A-C, top lanes). Lnk overexpression attenuated the activation of p44/42MAPK and Akt, and abolished the activation of Stat5 (Figure 6, top lanes). In addition, the Lnk SH2 mutant (R364E) that did not affect erythroblast survival, did not affect Epo-induced MAPK, Akt, and Stat5 activation in primary erythroblasts (Figure 6, top lanes). In contrast, erythroblasts expressing Lnk (W191A) or Lnk (Y536F) showed similar blunted activations of Epo-induced signaling pathways to that of erythroblasts expressing wild-type Lnk (Figure 6, top lanes). The total protein levels remained unchanged among all the samples (Figure 6A, middle panel; 6B-C, bottom panels) and Lnk expression levels also were similar in cells expressing the different Lnk mutants (Figure 6A, bottom panel). The SH2 domain of Lnk is therefore crucial for Lnk's inhibitory function in Epo-mediated signaling pathways in primary erythroblasts; whereas the PH domain and the conserved C-terminal tyrosine (Y536) are not.

Since Lnk abrogated all 3 major signaling pathways originating from the EpoR, we analyzed whether Lnk also affects the phosphorylation of the EpoR and its associated JAK2 kinase. In control cells, Epo induced EpoR (Figure 7A) and JAK2 (Figure 7B) phosphorylation in 7 to 10 minutes; however, in wild-type Lnk-, but not Lnk (R364E)-transduced erythroid cells, both EpoR and JAK2 phosphorylation were markedly reduced (Figure 7A-B). In fact, Lnk itself becomes tyrosine-phosphorylated shortly following Epo stimulation, and this phosphorylation was abolished when the Lnk SH2 domain was disrupted (Figure 7C). Therefore, the Lnk SH2 domain is required for its own phosphorylation, and is essential for attenuating EpoR and JAK phosphorylation induced by Epo.

Among mice nullizygous for Lnk family members, Lnk-deficient mice show the most profound perturbation in hematopoiesis. In addition to an abnormal expansion of immature B cells⁹  and enhanced BM repopulating activity,¹⁵  Lnk-deficient animals exhibit elevated numbers of megakaryocytes with increased ploidy.¹¹ 

In this study, we demonstrated that Lnk-deficient mice have increased numbers of erythroid progenitors in the BM and spleen. In addition, CFU-e progenitors from Lnk-deficient spleens are hypersensitive to Epo compared with wild-type controls. Lnk-deficient mice are less severely affected by and recover faster from erythropoietic stress, and this is due to an enhanced erythropoietic rate in Lnk-deficient mice after PHZ treatment. A principal if not the sole action of Lnk in erythropoiesis is the direct inhibition of Epo/EpoR signaling, since CD71⁺ erythroid progenitors from PHZ-treated Lnk-deficient spleens exhibited increased activation of Epo-induced signaling pathways. Thus, one of our most important results is that the adaptor protein Lnk is a physiologic negative regulator for Epo-mediated erythropoiesis.

Although Lnk^-/- mice show superior recovery from erythropoietic stress and have increased erythroid progenitor numbers, they exhibit normal steady-state hematocrits. One possibility is that Lnk^-/- mice have decreased Epo levels, which is technically difficult to quantify at normal or subnormal serum levels. The fact that Lnk deficiency does not result in enhanced EpoR signaling at the steady state (data not shown) may partially explain the lack of elevated red blood cell counts in these mice. Nonetheless, how the differentiation of BFU-e to CFU-e progenitors and subsequently to mature red blood cells is compensated in Lnk^-/- mice remains to be elucidated. We have similar findings to Dr Tony Pawson's group that in Lnk^-/- mice Ter119⁺ cells were decreased in the BM but increased in the spleen.¹²  The balance or dynamics between BM and extramedullary hematopoiesis in Lnk^-/- mice is unclear. Whether the BM environment is perturbed in Lnk-deficient mice, or abnormal lymphoid or myeloid homeostasis in these mice have an effect on BM erythropoiesis, remains to be determined.

Lnk overexpression in primary erythroid progenitors blocks Epo/EpoR-dependent erythroid development, and the first manifestation (16 hours) is the abrogation of Epo-dependent hyper-up-regulation of the transferrin receptor CD71/TfR (R1 to R2). This occurs before Lnk-induced apoptosis, and in the absence of cell cycle arrest. By the end of a 2-day culture in the presence of Epo, Lnk overexpression resulted in fewer terminal differentiated erythrocytes, which appear to have less hemoglobin than control cultures. The fact that Lnk-expressing erythroblasts display an almost identical “differentiation” profile and cytologic morphology compared with control cells cultured in the absence of Epo, suggests that Lnk specifically inhibits Epo/EpoR signaling.

CD71/TfR, which acts in the cellular uptake of the iron-transferrin complex, is up-regulated in many cell types during proliferation. Due to an exceptionally large demand for iron to support hemoglobin production, erythroid cells express 100-fold higher levels of CD71/TfR than nonerythroid cells. The expression levels of CD71/TfR in erythroid cells couple with cell differentiation as well as proliferation. This hyper-up-regulation of CD71/TfR in early precursors (R1 to R2) is completely Epo-dependent,¹⁶  and mediated via TfR gene transcriptional and posttranscriptional mechanisms.¹⁷  In agreement with Epo-dependent CD71/TfR up-regulation, overexpression of constitutive active Stat5 in primary erythroblasts, one of the major signaling molecules induced by Epo, led to Epo-independent CD71/TfR up-regulation (data not shown).

Epo is believed to serve as a survival factor for committed CFU-e progenitors and allows them to undergo proliferation and terminal differentiation.¹⁸  Largely, apoptosis in erythroid progenitors deprived of Epo occurs during the G₁ and S phases of the cell cycle, without growth arrest.¹⁹  Consistent with this, mice deficient for the antiapoptotic protein Bcl-X_L, which is a Stat5 target and induced by Epo in late erythropoiesis, lack mature erythrocytes.²⁰  Conversely, exogenous expression of Bcl-X_L in primary erythroblasts allows terminal differentiation without Epo.²¹  Our findings in primary fetal liver cultures also agree with this antiapoptotic role of Epo. Lnk overexpression, which blocks Epo signaling, leads to a dramatic cell death but not cell-cycle arrest.

Erythroid progenitor cells possess a vast heterogeneity in their sensitivity to Epo. This intrinsic difference of erythroblasts in their sensitivity to Epo might be tightly regulated through negative signaling molecules associated with the EpoR. One example is the suppressor of cytokine signaling (SOCS) proteins.^22,23  SOCS1-deficient fetal liver CFU-e progenitors exhibit an enhanced sensitivity to Epo in forming erythroid colonies.²⁴  Similarly, we found that Lnk-deficient spleens are also hypersensitive to Epo in generating CFU-e colonies. In primary cultures of fetal liver cells, overexpression of SOCS1 in erythroid progenitors caused a dramatic cell death (data not shown), which is very similar to what we observed after Lnk overexpression. Taken together, our data suggest that Lnk is a negative regulator of Epo/EpoR signaling that controls the threshold of Epo-dependent survival.

C-Kit was previously reported to be an Lnk target.¹⁵  We recently identified a second Lnk target, the Tpo receptor mpl. We showed that Lnk-deficient megakaryocytes exhibited enhanced and prolonged activation of Tpo/mpl-induced signaling proteins.¹¹  In this study, we discovered a third Lnk target, the EpoR. We demonstrated that Lnk abrogates Epo-induced Stat5, Akt, and MAPK activation in both hematopoietic cell lines and primary erythroblasts. Consistently, PHZ-treated Lnk-deficient splenic erythroid progenitor cells showed enhanced signaling transduction induced by Epo.

One common feature identified in Lnk regulation of these 3 receptors is that the Lnk SH2 domain is crucial for its inhibitory function. Therefore, Lnk may negatively regulate signaling pathways originating from multiple cytokine receptors through a similar mechanism. Interestingly, the conserved tyrosine near the C-terminus of Lnk, Y536, is dispensable for the negative regulatory roles of Lnk in lymphocyte development²⁵  and Tpo-mediated megakaryocytopoiesis.¹¹  In contrast, we found that Y536 is important, but not required, for modulating Epo-dependent 32D cell proliferation or erythroblast survival. The Lnk PH domain contributes but is not essential for inhibiting both Epo- and Tpo-mediated hematopoiesis, whereas it is dispensable for inhibiting SCF-dependent hematopoietic differentiation in AGM culture.¹⁰ 

Taking advantage of the accessibility and abundance of fetal liver erythroid cells, we further demonstrated that Lnk itself becomes tyrosine phosphorylated following Epo administration and inhibits Epo-induced EpoR phosphorylation and JAK2 activation. Interestingly, Lnk does not inhibit EpoR cell-surface expression (Lily Huang and W.T., unpublished data, October 2003); neither does it induce EpoR or JAK2 degradation (Figure 7). Our data support several possible mechanisms by which Lnk downregulates JAK2 activity. One is that Lnk may disrupt the binding of positive regulators of JAK2, such as SH2-Bβ.²⁶  Second, Lnk may recruit other JAK2 inhibitors, such as SOCS1²³  or PTP-1B.²⁷  Third, binding of Lnk to JAK2 may cause a conformational change that keeps JAK2 in a kinase-inactive state. Furthermore, the Lnk SH2 domain is essential for inhibiting EpoR and JAK2 phosphorylation, and the Lnk SH2 mutant lost its ability to be phosphorylated in response to Epo. These findings suggest that one or more tyrosine-phosphorylated proteins interact with the Lnk SH2 domain, and are crucial for phosphorylating Lnk and conferring Lnk inhibitory functions.

Taken together, we demonstrate that Lnk through its SH2 domain negatively regulates Epo-mediated erythropoiesis and Epo/EpoR signaling pathways by down-modulating JAK2 activation. Our findings likely provide a generalized mechanism for Lnk inhibitory function in the signaling of many type 1 cytokine receptors. This represents a new mechanism for rapidly down-regulating cytokine receptor signaling. Our work adds Lnk to the existing list of negative regulators of cytokine signaling including the SOCS proteins and SHP1 phosphatase.

We are grateful to Drs Tony Pawson and Laura Velazquez for Lnk-deficient mice and Dr Satoshi Takaki for Lnk cDNA constructs. We thank Ferenc Reinhardt for animal technical supports.