IFN-γ directly inhibits the activity of erythropoietin in human erythroid progenitors

To the Editor,

Anemia of inflammation (AI) is prevalent in chronically ill patients. An understanding of the pathogenesis of AI primarily coheres around a mechanism of impaired iron availability in maturing red blood cells due to increased production of hepcidin, but inflammatory suppression of the biological activity of erythropoietin (EPO), the primary regulator of red blood cell production, also contributes to the pathogenesis of AI [1]. To explain how inflammation mechanistically alters response of erythroid progenitors to EPO, a pathway was proposed in which iron deprivation reduces EPO receptor (EPOR) surface delivery and EPO activation of the JAK-STAT5 signaling pathways by downregulating Scribble, a master regulator of receptor trafficking and signaling [2]. However, there is likely an additional component to this mechanism since JAK-STAT5 activation is only partially diminished by iron restriction, and a recent study showed reduction in EPO-mediated signaling in erythroid cells obtained from patients with AI without concomitant alteration in the number of EPOR [3]. Laboratory and clinical observations have suggested the possibility of a direct iron-independent effect of inflammatory cytokines, such as interferon-γ (IFN-γ), on EPO responsiveness [3,4], but this possibility has not been confirmed [1]. We recently reported that IFN-γ suppressed hematopoietic stem and progenitor cell (HSPC) activity in acquired aplastic anemia by specifically forming heteromers with thrombopoietin (TPO), the primary regulator of HSPCs [5]. TPO complexed to IFN-γ showed impaired binding to its receptor, contributing to the observed perturbation of TPO-induced signaling and decreased HSPC survival in the presence of IFN-γ. Because EPO shares significant sequence and structural homology with TPO (Fig. S1), we hypothesized that IFN-γ may also limit response of erythroid progenitors to EPO using a similar mechanism of EPO:IFN-γ heteromer formation, receptor binding occlusion and blunting of EPO signaling pathways.

To investigate whether IFN-γ can directly impair responsiveness of erythroid cells to EPO, we first quantified erythroid progenitors in CFU assay after ex vivo culture of human CD34+ HSPCs for 7 days in erythroid differentiation medium (containing 1 IU/mL EPO), in the presence or absence of IFN-γ (Fig. 1A). As previously shown, formation of red cell progenitors (BFU-E and CFU-E) was significantly inhibited by IFN-γ relative to control cultures without IFN-γ (Fig. 1B and C). Consistent with these data, we found a significant increase in apoptosis of erythroid progenitors in the presence of IFN-γ compared to control cultures (Fig. S2). Notably, higher concentrations of EPO (16 and 64 IU/mL) could overcome this inhibition (Figs. 1B and C, and S2). This effect does not reflect a compensatory increase in colony formation by additional erythroid progenitors stimulated by increased EPO levels since higher concentrations of EPO did not alter the maximum number of erythroid colonies obtained at EPO 1 IU/mL in the absence of IFN-γ (Fig. 1C).

Fig. 1.

High concentrations of EPO overcome IFN-γ-mediated inhibition of erythroid progenitor cell formation and EPO-mediated JAK-STAT signaling. (A) Overview of the experimental procedure. Human CD34+ cells were collected by apheresis of normal volunteers after G-CSF mobilization. Cells were expanded in erythroid differentiation medium (containing SCF, IL-3, and EPO 1, 16 or 64 IU/mL), with or without IFN-γ for 7 days. Following differentiation, cells were collected and the impact of IFN-γ on erythroid progenitors (BFU-E and CFU-E) was assessed by CFU assay. (B) Representative images of CFU assay plates demonstrating a marked reduction in erythroid colonies in the presence of IFN-γ and restoration of erythroid progenitor cell formation with higher concentrations of EPO in culture. (C) Number of erythroid colonies obtained after 7-day culture of human CD34+ HSPCs in medium supplemented with various concentrations of EPO, in the presence or absence of IFN-γ (n = 8). (D) Phosphorylation state of STAT5 measured using Phospho-Flow 15 min after stimulation of UT-7/EPO cells with various concentrations of EPO in the presence or absence of IFN-γ. Left panel presents a summary of pSTAT5 mean fluorescence intensity (MFI, n = 5). Dotted line represents MFI of isotype control. Right panels show representative flow cytometry histograms. Results are displayed as mean ± standard error of the mean (SEM); **P < 0.01, *P < 0.05, ns, not significant, by unpaired t-test (panels C and D).

To understand the potential mechanism by which IFN-γ impaired erythroid progenitor formation and survival, we examined the impact of IFN-γ on the primary intracellular signaling pathway (JAK-STAT5) induced upon binding of EPO to its receptor. In EPOR+ UT-7 cells (UT-7/EPO), IFN-γ markedly impaired signaling induced by low concentrations of EPO (1–10 IU/mL), as measured by decreased phosphorylation of STAT5 (Fig. 1D). This effect was not due to down-regulation of EPOR in the presence of IFN-γ (Fig. S3). Higher concentrations of EPO (50–100 IU/mL) had no significant impact on the intensity of JAK-STAT5 signaling in control cells not exposed to IFN-γ, likely due to saturation of EPOR, but ameliorated IFN-γ-mediated blockade of EPO signaling. Hence, inhibition of a critical pathway of growth factor cell signaling by IFN-γ might explain the impaired erythropoiesis observed under inflammatory conditions in ex vivo cultures.

To clarify the mechanism underlying the observed inhibition of EPO activity by IFN-γ in erythroid cells, we first quantified the EPO:EPOR binding affinity in the presence or absence of IFN-γ using microscale thermophoresis (MST). We previously confirmed that dissociation constants (K_D) measured by MST corroborate with K_D determined by traditional radiolabeled (¹²⁵I) ligand assays in primary cells [5]. As formerly reported using alternative in vitro analytical methods [6], crystal structures of EPO/EPOR [7], and ¹²⁵I ligand assays in Ba/F3 cells expressing EPOR [8], EPO and EPOR displayed a 2-site interaction with K_D of 0.1 μM and 7.8 μM, characterizing a high- and low-affinity binding site, respectively (Fig. 2A, Table S1). The high-affinity interaction involves the initial contact between EPO and one EPOR subunit; the low-affinity site engages the second EPOR subunit into the complex to form a signaling-competent receptor conformation. Notably, addition of IFN-γ did not perturb the EPO:EPOR high-affinity interaction but weakened, in a dose-dependent fashion, the low-affinity site K_D from 7.8 μM to 39.9 μM in the presence of 1000-fold molar excess of IFN-γ (Fig. 2A, Table S1). This weakening could result from a partial obstruction of the low-affinity EPO:EPOR binding site. This effect was specific to IFN-γ; other hematopoietic cytokines had no detectable impact on the EPO:EPOR interaction (Fig. 2B, Table S1). In addition, disruption of the EPO:EPOR low-affinity binding site was not due to competitive binding of IFN-γ to EPOR (Fig. 2C). Because IFN-γ induced a dose-dependent perturbation of EPO binding to its receptor without directly interacting with EPOR, we reasoned that IFN-γ might instead bind to EPO, forming EPO:IFN-γ heteromeric complexes that sterically occlude the low-affinity EPO:EPOR binding site, akin the TPO:IFN-γ heteromerization previously reported in HSPCs [5]. Remarkably, we determined a specific interaction between EPO and fluorescently labeled IFN-γ (Fig. 2D) with an affinity (K_D = 0.2 μM, Table S2) superior to the EPO:EPOR low-affinity interaction (K_D = 7.8 μM, Table S1). By contrast, no interaction was observed between IFN-γ and other cytokines (Fig. 2D, Table S2). When EPO was used as the labeled partner, we detected a similar high-affinity EPO:IFN-γ interaction (K_D = 0.2 μM) and no evidence of complex formation between EPO and other cytokines (Fig. 2E, Table S3).

Fig. 2.

IFN-γ forms heteromeric complexes with EPO and disrupts the low-affinity binding interaction between EPO and EPOR. (A) Evaluation of high and low affinity EPO:EPOR binding sites by MST in the presence or absence of IFN-γ (n = 3). Addition of IFN-γ resulted in specific partial disruption of the low-affinity site in a dose-dependent manner. Dissociation constants (K_D) are listed in Table S1. (B) Evaluation of high and low affinity EPO:EPOR binding sites by MST in the presence or absence of 1000-fold molar excess concentration of hematopoietic cytokines (n = 3). The EPO:EPOR interaction was unaffected by SCF or Flt3L. K_D are listed in Table S1. (C) Evaluation of the binding affinity between IFN-γ and EPOR by MST. IFN-γ did not interact with EPOR. (D) Evaluation of the binding affinity between NT647-labeled IFN-γ and EPO, SCF and Flt3L. Specific IFN-γ:EPO heteromer formation was detected. K_D are listed in Table S2. (E) Evaluation of the binding affinity between NT647-labeled EPO and IFN-γ, SCF and Flt3L. Specific EPO:IFN-γ heteromer formation was detected. K_D are listed in Table S3. Results are displayed as mean ± SEM; curve fit by polynomial nonlinear regression (panels A–C) or Langmuir isotherm nonlinear regression (panels D, E).

Our study proposes an additional mechanism for the inflammatory block to EPO responsiveness in AI, implicating EPO:IFN-γ heteromer formation, weakened EPO binding to EPOR, and interference with intracellular signaling cascades, resulting in increased apoptosis of erythroid progenitors. While both low- and high-affinity binding sites are competent for signaling [8], complete receptor engagement is required to allow the most favorable receptor orientation for maximal signaling of intracellular kinase pathways [7]. Hence, although inhibition of the EPO:EPOR low-affinity interaction by IFN-γ was only partial, it was sufficient to impair intracellular signaling likely caused by an aberrant receptor dimer conformation. This understanding may facilitate the development of rational assays for high-throughput screening of novel small molecules capable of overcoming the inflammatory block in AI. Availability of a new therapeutic would be significant in view of the widespread nature of AI, the associated morbidity and mortality, and the inefficacy of standard treatments.