Transmembrane Protein Aptamer Induces Cooperative Signaling by the EPO Receptor and the Cytokine Receptor β-Common Subunit.

The erythropoietin receptor (EPOR) plays an essential role in erythropoiesis and other cellular processes by forming distinct signaling complexes composed of EPOR homodimers or hetero-oligomers between the EPOR and another receptor, but the mechanism of heteroreceptor assembly and signaling is poorly understood. We report here a 46-residue, artificial transmembrane protein aptamer, designated ELI-3, that binds and activates the EPOR and induces growth factor independence in murine BaF3 cells expressing the EPOR. ELI-3 requires the transmembrane domain and JAK2-binding sites of the EPOR for activity, but not the cytoplasmic tyrosines that mediate canonical EPOR signaling. Instead, ELI-3-induced proliferation and activation of JAK/STAT signaling requires the transmembrane and cytoplasmic domains of the cytokine receptor β-common subunit (βcR) in addition to the EPOR. Moreover, ELI-3 fails to induce erythroid differentiation of primary human hematopoietic progenitor cells but inhibits nonhematopoietic cell death induced by serum withdrawal.

Introduction

Many aspects of cell behavior are controlled by cell surface receptors that receive extracellular signals and orchestrate the cellular response. The formation and activation of alternative receptor complexes with different subunits and signaling properties can dictate receptor output (e.g., Kovacs et al., 2015). The cytokine erythropoietin (EPO) can activate alternative complexes of the EPO receptor (EPOR), a transmembrane (TM) cell surface protein lacking intrinsic kinase activity. Binding of EPO to the EPOR can trigger the homodimerization of EPORs in a productive orientation, leading to the transphosphorylation of Janus kinase 2 (JAK2), which is constitutively associated with the EPOR (Constantinescu et al., 1999a, Constantinescu et al., 2001, Watowich et al., 1999). Activated JAK2 phosphorylates multiple tyrosines in the intracellular domain of EPOR, allowing the recruitment and phosphorylation of downstream signaling proteins, including signal transducer and activator of transcription 5 (STAT5) (Barber et al., 1997, Kuhrt and Wojchowski, 2015, Sawyer and Penta, 1996, Lodish et al., 2009). This signaling pathway is essential for the survival, proliferation, and differentiation of erythroid progenitors.

In addition to erythropoiesis, the EPOR can mediate non-erythroid outcomes in response to EPO treatment, including a tissue-protective response that prevents apoptosis and promotes proliferation in non-hematopoietic cells subjected to injury or metabolic stress (Acharya et al., 2010, Brines, 2010, Jubinsky et al., 1997, Siren and Ehrenreich, 2001, Siren et al., 2001a, Siren et al., 2001b, Jelkmann et al., 2009), reviewed in Jelkmann et al. (2009). The protective effect of EPO appears to require the activation of a heteroreceptor composed of EPOR and the cytokine receptor β-common subunit (β-common receptor [βcR] also known as CD131). In addition to constitutively binding EPOR, βcR also binds the α-chain of interleukin (IL)-3 receptor, granulocyte-macrophage colony-stimulating factor (GM-CSF) receptor (GM-CSFR), and IL-5 receptor (Blake et al., 2002, Jubinsky et al., 1997, Hercus et al., 2013, Lopez et al., 1992). βcR plays an essential role in signaling by these receptors, which lack JAK2 binding or a significant cytoplasmic domain, by providing bound JAK2 and cytoplasmic tyrosines for phosphorylation (Hansen et al., 2008, Hercus et al., 2013). βcR−/− mice lack EPO-induced tissue protection but retain normal hematopoiesis, showing that βcR is required for tissue protection but not for erythroid differentiation in at least some settings (Weber et al., 2005, Brines et al., 2004). In addition, certain modified versions of EPO, such as lysine-carbamylated EPO, specifically induce the tissue-protective, but not the erythroid, effects of EPO (Erbayraktar et al., 2009, Leist et al., 2004, Murphy and Young, 2006, Yamanaka et al., 2018). These results suggest that the tissue-protective effect of EPO is mediated by an EPOR/βcR heteroreceptor, and not by EPOR homodimerization (Bohr et al., 2015). However, the role of the heteroreceptor in tissue protection remains controversial (see citations in Cheung Tung Shing et al. 2018), and in some situations a classical EPOR homodimer can provide a protective signal (Um et al., 2007). Notably, the elements on the EPOR and βcR required for heteroreceptor formation and the molecular mechanism of signaling by the βcR/EPOR heteroreceptor are unknown.

Various receptors, including the EPOR, can be activated through interactions involving their TM domain (TMD). The murine spleen focus-forming virus envelope protein gp55-P specifically binds to the TMD of the mouse EPOR (mEPOR), triggering EPOR activation, erythroid cell proliferation, and polycythemia (Li et al., 1990, Constantinescu et al., 1999b). The platelet-derived growth factor β receptor (PDGFβR) can be activated by the bovine papillomavirus E5 oncoprotein, a 44-residue TM protein that binds specifically to the TMD of the PDGFβR (DiMaio and Petti, 2013, Petti and DiMaio, 1992, Petti et al., 1991). We developed a genetic approach to isolate small biologically active TM proteins in which we construct libraries expressing up to millions of different, small, artificial TM proteins (termed traptamers, for TM aptamers) with randomized, hydrophobic segments. Traptamer libraries are expressed in mammalian cells, and active traptamers are recovered from cells selected for particular biological activities, with the rationale that, by chance, rare traptamers interact with cellular TM proteins and modulate their activity or expression (Cammett et al., 2010, Freeman-Cook et al., 2004, Freeman-Cook and DiMaio, 2005, Scheideman et al., 2012). We have isolated traptamers that specifically activate human EPOR (hEPOR) or mEPOR and cause EPOR-dependent proliferation of murine BaF3 cells (Cammett et al., 2010, Cohen et al., 2014, He et al., 2017). These traptamers bind the TMD of the EPOR and induce hEPOR homodimerization and tyrosine phosphorylation of EPOR and JAK2.

Here, we isolate and characterize a new traptamer, ELI-3, that induces proliferation of BaF3 cells that express the EPOR. ELI-3 interacts with the hEPOR and, unlike EPO or previously isolated traptamers that activate the EPOR, does not require intracellular hEPOR tyrosines, but instead requires the endogenously expressed βcR in addition to the EPOR. ELI-3 does not support differentiation in erythroid cells and inhibits serum withdrawal-induced apoptosis in non-hematopoietic cells. These results show that small TM proteins can specifically activate either the EPOR homodimer or the EPOR/βcR heteroreceptor, with distinct biological outcomes. Our results also demonstrate that the EPOR in the EPOR/βcR heteroreceptor uses a non-canonical mechanism to generate a proliferative signal.

Results

Isolation of a Traptamer that Confers Growth Factor Independence in Cells Expressing hEPOR

The strategy used to isolate new traptamers that cooperate with the EPOR is shown in Figure 1A. We used the YX4 traptamer expression library, in which the TMD of the bovine papillomavirus E5 protein is replaced with a 24-residue stretch of randomized, primarily hydrophobic amino acids (Figure 1B) (Scheideman et al., 2012). The traptamers also contain an N-terminal hemagglutinin (HA) epitope tag. The YX4 library was packaged into retrovirus and used to infect BaF3 cells expressing the hEPOR (BaF3/hEPOR cells) at a low MOI so that most cells received a single infectious retrovirus particle. As a control, cells were infected with empty retrovirus vector, MSCVpuro. BaF3/hEPOR cells normally require IL-3 for proliferation, but EPO or proteins that activate the hEPOR can replace EPO. After puromycin selection, transduced cells were incubated in medium lacking growth factors. As expected, cells expressing MSCVpuro died, but cultures infected with the YX4 library proliferated. After 8 days in medium lacking growth factors, genomic DNA was extracted from proliferating cells and the retroviral inserts were amplified and subjected to next-generation sequencing, which produced over 4 million proper read pairs consistent with the design of the library. We focused on the 278 most abundant sequences. Sequences that lacked frameshift mutations were sorted into 105 groups based on sequence similarity. The frequency of sequences in each group ranged from 0.01% to 6.99%. Sequencing of the starting library plasmid DNA showed no abundant sequences.

Figure 1

Isolation of a New Traptamer that Cooperates with the EPOR to Confer Growth Factor Independence

(A) Scheme to isolate traptamers that cooperate with the hEPOR. BaF3/hEPOR cells were infected with retroviruses expressing the YX4 traptamer library, selected with puromycin, and then incubated in medium lacking growth factors. Genes encoding traptamers were recovered from genomic DNA isolated from live cells after selection and then subjected to deep sequencing. Abundant sequences were synthesized and tested for activity. Black bars represent exogenous hEPOR; black and gray “X”s represent traptamers.

(B) The design of the YX4 library is shown in the single-letter amino acid representation. Randomized residues are colored red. The X's represent randomized positions, each with an 80% probability of encoding a hydrophobic amino acid. The Z's represent randomized amino acids with an ∼30% chance of being a stop codon. The N-terminal HA tag is underlined.

(C) Abundant sequences recovered from growth factor-independent BaF3/hEPOR cells. The randomized regions are colored red. The invariant YW are colored black. The frequency of the sequence (and closely related sequences) and its proportion among all sequences obtained are listed.

(D) BaF3 and BaF3/hEPOR cells stably expressing empty vector (MSCVp) or a traptamer listed in (C) were incubated in medium lacking IL-3. The number of live cells 4 days after IL-3 removal is shown for a representative experiment.

(E) BaF3 cells expressing hEPOR, mEPOR, PDGFβR, stem cell factor receptor (SCFR), or human thrombopoietin receptor (hTPOR) were infected with MSCVp or MSCVp expressing ELI-3. After puromycin selection, cells were incubated in medium lacking IL-3. Where indicated, cells expressing MSCVp were also incubated in medium containing the cognate ligand: EPO for hEPOR and mEPOR, PDGF-BB for PDGFβR, stem cell factor for SCFR, and TPO for hTPOR. The number of live cells 4 days after IL-3 removal is shown. The averaged results and standard deviation of three independent experiments are shown. Statistical significance was evaluated by two-tailed Student's t test with unequal variance.

We expressed seven of the most abundant selected sequences (Figure 1C) in parental BaF3 cells and in BaF3/hEPOR cells. Most of these constructs conferred growth factor independence in both cell lines (Figure 1D), suggesting that they acted through a protein expressed in parental cells. In contrast, the 46-residue ELI-3 traptamer conferred growth factor independence in BaF3/hEPOR cells, but not in parental BaF3 cells, demonstrating that ELI-3 required the hEPOR for activity.

To examine the specificity of ELI-3, we introduced it into BaF3 cells expressing hEPOR, mEPOR, the human thrombopoietin receptor, PDGFβR, or stem cell factor receptor. After IL-3 removal, cells expressing each receptor proliferated in response to its ligand, but not in the absence of ligand (Figure 1E). Notably, ELI-3 induced IL-3 independence only in cells expressing hEPOR or mEPOR (Figure 1E), indicating that ELI-3 activity was specific to EPOR and that it cooperated with either hEPOR or mEPOR, whose TMDs differ at only three residues and adopt a similar α-helical structure (Li et al., 2015).

ELI-3 Forms a Stable Complex with the hEPOR and Requires Specific Residues in the EPOR Transmembrane Domain

We used co-immunoprecipitation to determine if ELI-3 and hEPOR were present in a stable complex. First, the HA epitope tag at the N terminus of ELI-3 was replaced with a FLAG epitope tag to generate F-ELI-3. The FLAG tag did not affect the ability of ELI-3 to induce IL-3 independence in BaF3/hEPOR cells (Figure S1). Detergent extracts were prepared from BaF3/hEPOR cells expressing either MSCVpuro or F-ELI-3 and immunoprecipitated with anti-FLAG antibody. After gel electrophoresis and transfer, membranes were immunoblotted with an anti-HA antibody, which recognizes HA-tagged hEPOR. As shown in Figure 2A, anti-FLAG co-immunoprecipitated the hEPOR from cells expressing F-ELI-3, but not from cells expressing MSCVpuro, demonstrating that F-ELI-3 and the hEPOR co-existed in a physical complex.

Figure 2

ELI-3 Interacts with the Transmembrane Domain of hEPOR

(A) Extracts prepared from BaF3/hEPOR and BaF3/hEPOR(mPR) cells growing in IL-3 expressing MSCVp (V), ELI-3, or FLAG-tagged ELI-3 (F-ELI-3) were subjected to gel electrophoresis either directly (middle and bottom panels) or after immunoprecipitation with anti-FLAG agarose beads (top panel). After transfer, membranes were immunoblotted with anti-HA antibody to detect EPOR or anti-pan-actin antibody as a loading control.

(B) BaF3/hEPOR and BaF3/hEPOR(mPR) cells stably expressing MSCVp or ELI-3 were incubated in medium lacking IL-3. Where indicated, cells expressing MSCVp were incubated in medium containing EPO. The number of live cells 4 days after IL-3 removal is shown. The averaged results and standard deviation of three independent experiments are shown. Statistical significance was evaluated by two-tailed Student's t test with unequal variance.

(C) The sequences of EBC5-16 and ELI-3. Randomized hydrophobic segments are shown in red.

(D) BaF3 cells expressing the wild-type hEPOR, hEPOR mutant L234A, or mutant L241A were infected with retroviruses to express MSCVp, EBC5-16, or ELI-3. After puromycin selection, cells were incubated in medium lacking IL-3. The number of live cells 4 days after IL-3 removal is shown. The averaged results and standard deviation of three independent experiments are shown. Statistical significance was evaluated by two-tailed Student's t test with unequal variance, comparing cell number in cells expressing wild-type receptor to cell number in cells with mutant receptor expressing the same traptamer. *p < 0.05, **p < 0.01.

See also Figure S1.

To determine whether the TMD of hEPOR is required for ELI-3 function, a chimeric hEPOR was used in which the TMD of hEPOR was replaced with the TMD of mouse PDGFβR (designated hEPOR(mPR)). BaF3/hEPOR(mPR) cells were able to grow when IL-3 in the medium was replaced with EPO, demonstrating that hEPOR(mPR) was functional (Figure 2B). However, ELI-3 did not confer growth factor independence on BaF3/hEPOR(mPR) cells (Figure 2B), and F-ELI-3 did not co-immunoprecipitate with hEPOR(mPR) (Figure 2A), showing that the TMD of hEPOR was critical for ELI-3 activity and complex formation between ELI-3 and EPOR.

We next identified hEPOR TMD residues required for the activity of ELI-3 and the previously described traptamer EBC5-16, which activates hEPOR but not mEPOR (Cohen et al., 2014). The TMD sequences of EBC5-16 and ELI-3 are entirely different (Figure 2C). We tested two hEPOR TMD mutants, L234A and L241A. As expected, these receptor mutants did not confer IL-3 independence in BaF3 cells lacking traptamer expression but allowed the cells to proliferate in response to EPO (data not shown). EBC5-16 cooperated well with both EPOR mutants to confer growth-factor independence, whereas ELI-3 failed to cooperate with L234A and displayed markedly reduced activity with L241A (Figure 2D). These results showed that the traptamers required different amino acids in hEPOR TMD, suggesting that the two traptamers interact with the TMD of the hEPOR in distinct manners.

The experiments described above imply that ELI-3 recognizes the TMD of the hEPOR. We conducted biophysical experiments to explore this possibility in more detail. We expressed recombinant ELI-3 in bacteria, purified it, and subjected it to circular dichroism analysis in detergent micelles. As shown in Figure S2, ELI-3 displayed minima at 208 and 222 nm, characteristic of α-helical structure, as expected. We then mixed ELI-3 with purified 15N-labeled TMD plus flanking sequences of the hEPOR (residues 217–252) and conducted solution NMR spectroscopy in the presence of different concentrations of the detergent, 1,2-dihexanoyl-sn-glycero-3-phosphocholine (DHPC) (Figure S3, red peaks). For comparison, the same analysis was also performed for 15N-labeled hEPOR217-252 in the absence of ELI-3 (Figure S3, black peaks). The addition of ELI-3 caused a detergent-sensitive perturbation of the majority of the hEPOR-TMD chemical shifts toward the dimeric state (Figures S3A and S3B). This suggests that the presence of ELI-3 stabilizes a dimeric state of the hEPOR-TMD, as we showed previously with other hEPOR-specific traptamers (He et al., 2017). This effect likely occurs through direct interactions between these two proteins, because detergent alone could not populate the dimer to the same extent as did ELI-3. Because of the inherent challenges in using NMR spectroscopy to study the interaction of these hydrophobic peptides, we focused our further efforts on analyzing ELI-3 activity in cells.

ELI-3 Does Not Induce Erythroid Differentiation

We previously showed that EBC5-16 supported erythroid differentiation of CD34+ human hematopoietic progenitor cells (hHPCs) in vitro in the absence of EPO (Cohen et al., 2014). Here, we used a more quantitative assay to assess whether EBC5-16 or ELI-3 promoted erythroid differentiation in primary human megakaryocyte-erythroid progenitor (MEP) cells, which give rise to colonies containing erythroid or megakaryocytic cells (or both) when cultured in vitro with a cocktail of cytokines including EPO. MEP cells isolated on the basis of expression of cell surface markers (see Methods, Sanada et al., 2016) were infected with MSCVpuro or with retroviruses expressing EBC5-16 or ELI-3. Transduced cells were plated as single cells in medium containing puromycin supplemented with stem cell factor, IL-3, IL-6, and thrombopoietin with or without EPO. After 12–14 days, colonies were stained with antibodies recognizing glycophorin A and CD41a (markers of erythroid and megakaryotic differentiation, respectively). Colonies were classified as megakaryocytic-only (CFU-Mk), erythroid-only burst forming unit (BFU-E), or megakaryocytic/erythroid (CFU-Mk/E) (Xavier-Ferrucio et al., 2018). As shown in Figure 3A, in the presence of EPO, all cultures differentiated into erythroid lineage, megakaryotic lineage, and mixed colonies. In the absence of EPO, >50% of colonies induced by EBC5-16 were BFU-E or CFU-Mk/E, consistent with its ability to induce erythroid differentiation of hHPCs. In contrast, fewer than 5% of the colonies induced by ELI-3 in the absence of EPO were BFU-E or CFU-Mk/E, comparable with control cells lacking traptamer expression. These results demonstrated that ELI-3, unlike EBC5-16, does not promote erythroid commitment and differentiation in human MEP cells. We also note that ELI-3 does not interfere with the ability of EPO to induce erythroid differentiation.

Figure 3

Biological Consequences of ELI-3-Induced EPOR Signaling

(A) Human MEP cells were infected with retrovirus expressing empty vector MSCVp (v), EBC5-16 (5–16), or ELI-3. After puromycin selection, cells were plated in medium supplemented with a cytokine cocktail with or without EPO, as indicated. After 12–14 days, the colonies were stained with anti-GpA and anti-CD41a antibodies and scored by fluorescence microscopy as megakaryocyte-only (CFU-Mk, blue), erythroid-only burst forming unit (BFU-E, red), or megakaryocyte/erythroid (CFU-Mk/E, purple). Top panel, numbers of each type of colony are shown. The averaged results and standard deviation of three independent experiments are shown. Bottom panel, the same data from top panel are shown as the relative percentage of each type of colony.

(B) Top left panel, P19 cells were infected with MSCVp empty retrovirus vector (Vec) or MSCVp expressing ELI-3. After puromycin selection, cells were plated in the presence or absence of serum for 24 h. Statistical significance was evaluated by two-tailed Student's t test with unequal variance. Where indicated, cells were treated with 2 U/mL rhEPO as described in Methods. Cells were then stained with DAPI and examined by fluorescence microscopy. Each symbol represents the fraction of cells displaying fragmented nuclei in an independent experiment. The mean ± standard deviation for each condition is shown. Top right panel, P19 cells were treated as above. Twenty-two hours later, cells were detached from the plate with trypsin, stained with fluorescein isothiocyanate-annexin V, and PI, and analyzed by flow cytometry. Each symbol represents the fraction of PI-negative cells that displayed annexin V staining in an independent experiment. The mean ± standard deviation for each condition is shown. Bottom panel, P19 cells were treated as above, except JAK2 inhibitor IV was added where indicated at time of starvation. Cells were analyzed by flow cytometry as mentioned above.

See also Figure S7.

The Cytokine Receptor β-Common Subunit Is Required for ELI-3-Induced Growth Factor Independence

Because ELI-3 did not induce erythroid differentiation, we considered the possibility that ELI-3 utilized a non-canonical EPOR signaling pathway to induce growth factor independence in BaF3 cells. EPOR and βcR can constitutively associate in the absence of EPO (Brines et al., 2004). We hypothesized that ELI-3 might activate the EPOR/βcR complex to induce proliferation of BaF3/hEPOR cells. We first confirmed that βcR was endogenously expressed in BaF3 cells, consistent with published results (Sakamaki et al., 1992) (Figure S4A, bottom panel, lanes 1 and 2). We next used co-immunoprecipitation to determine if EPOR was in complex with βcR. As shown in Figure S4A (top panel, lanes 7 and 8), the anti-βcR antibody co-immunoprecipitated hEPOR from BaF3/hEPOR cells in the presence or absence of ELI-3, showing that EPOR and βcR were in a physical complex even in the absence of ELI-3.

To assess the role of βcR in ELI-3 activity, we used CRISPR-Cas9 to knockout the endogenous Csf2rb gene, which encodes βcR, in BaF3 cells expressing the hEPOR. BaF3/hEPOR cells were infected by lentiviruses expressing Cas9 and one of four different single guide RNAs (sgRNAs) targeting Csf2rb. Csf2rb knockout by each sgRNA in clonal cell lines was confirmed by immunoblotting with an antibody recognizing the C terminus of βcR (Figure S4B, top panel) and by deep DNA sequencing (data not shown).

The activity of ELI-3 was determined in four βcR knockout cell lines (termed BaF3/h-βcKO cells), each generated by a different sgRNA. As shown in Figures 4A and S4C, EPO and EBC5-16 induced IL-3 independence in BaF3/h-βcKO cells, demonstrating that βcR was not required for proliferation in response to these agents. In sharp contrast, ELI-3 did not induce growth factor independence in βcR knockout cells, but re-expression of wild-type βcR in the knockout cells rescued the activity of ELI-3 (Figure 4A). These results demonstrate that βcR is necessary for ELI-3-induced cell proliferation.

Figure 4

ELI-3-Induced Growth Factor Independence Requires βcR, but Not Cytoplasmic Domain of hEPOR

(A) BaF3/hEPOR cells, BaF3/h-βcKO cells (expressing hEPOR but knocked-out for βcR [βcR knockout]), and BaF3/h-βcKO cells reconstituted with the wild-type βcR gene (βcR add-back) were infected with empty MSCVhyg vector (no traptamer) or MSCVhyg expressing EBC5-16 or ELI-3. After hygromycin selection, cells were incubated in medium lacking IL-3, and the number of live cells was counted 6 days after IL-3 removal. Where indicated, EPO was added. The average results and standard deviation of three independent experiments are shown. Statistical significance for all panels in this figure evaluated by two-tailed Student's t test with unequal variance.

(B) Parental BaF3 cells and cells expressing the wild-type hEPOR or hEPOR mutants lacking eight (F8) or nine (F9) cytoplasmic tyrosines were infected with retroviruses to express MSCVp, EBC5-16, or ELI-3. After puromycin selection, cells were incubated in medium lacking IL-3. Where indicated, cells expressing MSCVp were incubated in medium containing EPO. The number of live cells 4 days after IL-3 removal was counted. The averaged results and standard deviation of three independent experiments is shown.

(C) BaF3 cells expressing the wild-type hEPOR or an hEPOR truncation mutant were infected with retroviruses to express MSCVp or ELI-3. After puromycin selection, cells were incubated in medium lacking IL-3. Where indicated, cells expressing MSCVp were incubated in medium containing EPO. The number of live cells is shown 4 days after IL-3 removal. The averaged results and standard deviation of three independent experiments are shown.

(D) BaF3/h-βcKO cells were infected with MSCVzeo empty vector or MSCVzeo expressing wild-type βcR or a βcR truncation mutant. After zeocine selection, cells were infected with MSCVhyg (no traptamer) or MSCVhyg expressing EBC5-16 or ELI-3. After hygromycin selection, IL-3 independence assays were performed as in (A). Where indicated, EPO was added.

See also Figures S4 and S5.

The Cytoplasmic Tyrosines of hEPOR Are Not Required for ELI-3-Induced Cell Proliferation but the Cytoplasmic and Transmembrane Domains of the βcR Are Required

We next identified elements in hEPOR and βcR required for ELI-3-induced growth factor independence. The cytoplasmic domain of the hEPOR contains eight conserved tyrosines that are phosphorylated by JAK2 in response to EPO and serve as docking sites for signaling proteins. To determine whether ELI-3 required these tyrosines, we constructed an F8 hEPOR mutant in which all of them were mutated to phenylalanines. Parental BaF3 cells, BaF3/hEPOR cells, and BaF3/F8 cells expressing MSCVpuro, EBC5-16, or ELI-3 were cultured in the absence of IL-3 (Figure 4B). As expected, in all cases parental BaF3 cells died and BaF3/hEPOR cells incubated with EPO or expressing either traptamer grew robustly. BaF3/F8 cells grew poorly in response to EPO or EBC5-16, also as expected. Surprisingly, ELI-3 induced robust factor-independent growth of BaF3/F8 cells, indicating that ELI-3-induced mitogenic signaling did not require the conserved cytoplasmic tyrosines in the EPOR. In addition to the eight conserved tyrosines, the hEPOR cytoplasmic domain contains a non-conserved tyrosine at position 285 (Arcasoy and Karayal, 2005). We also tested whether ELI-3 conferred growth factor independence in cells expressing the F9 mutant, in which tyr285 in F8 was replaced with phenylalanine. As shown in Figure 4B, BaF3/F9 cells grew robustly in response to ELI-3, showing that tyr285 was also not essential for ELI-3 activity.

We also tested C-terminal truncation mutants of hEPOR lacking various portions of the cytoplasmic domain. As shown schematically in Figure S4D, three truncation mutants (Δ259, Δ289, and Δ310) were constructed deleting all sequences downstream of trp258, gly288, and leu309, respectively. Δ259 and Δ289 mutants are defective for JAK2 binding, whereas Δ310 retains JAK2 binding. We assessed the effect of EPO or ELI-3 in cells expressing these truncation mutants. As expected, EPO did not induce IL-3 independence in cells expressing any of the hEPOR truncation mutants (Figure 4C). Similarly, ELI-3 did not confer growth factor independence in BaF3/Δ259 or BaF3/Δ289 cells, which do not bind JAK2. Strikingly, however, ELI-3 (but not EBC5-16, data not shown) induced growth factor independence in BaF3/Δ310 cells, confirming that ELI-3, unlike EPO or EBC5-16, does not require the conserved cytoplasmic tyrosines or any other sequences downstream of position 309. These results also suggest that ELI-3 requires JAK2 binding to the EPOR. Similarly, ELI-3 cooperated with an mEPOR mutant lacking most of its cytoplasmic domain (data not shown).

We also tested whether the cytoplasmic domain or TMD of the βcR was required for ELI-3 activity. We constructed two C-terminal βcR truncation mutants, βcRΔ452 and βcRΔ514, which lacked most of the cytoplasmic domain and intracellular tyrosines of βcR (Figure S5A). Δ452 removed the JAK2-binding site, whereas Δ514 left the JAK2-binding site intact (Quelle et al., 1994, Sakamaki et al., 1992). Expression of the truncated βcR in βcR knockout cells expressing wild-type hEPOR was confirmed by blotting for the myc tag (Figure S5B, middle panel). As shown in Figure 4D, unlike wild-type βcR, the truncation mutant did not rescue the activity of ELI-3, whereas EBC5-16 activity was not affected by wild-type or mutant βcR. These results show that elements in the cytoplasmic domain of βcR are required for ELI-3 to cooperate with hEPOR to induce growth factor independence in BaF3 cells. Finally, replacing the TMD of βcR with the TMD of PDGFβR eliminated its ability to cooperate with ELI-3, even though this chimeric receptor still cooperated with EPO (Figure S6A).

ELI-3 Activates JAK2, STAT5, and Mitogen-Actiavted Protein Kinase Signaling

We next determined whether ELI-3 expression induced tyrosine phosphorylation of JAK2 and its downstream signaling proteins. BaF3/Δ259, BaF3/Δ310, and BaF3/hEPOR cells stably expressing MSCVpuro or ELI-3 were starved of IL-3 overnight and then either left untreated or acutely stimulated with 5 units/mL EPO for 5 min at 37°C. Cell lysates were subjected to SDS-polyacrylamide gel electrophoresis and immunoblotting with antibodies recognizing the phosphorylated forms of JAK2, STAT5, MEK, and ERK1/2. Membranes were then stripped and re-probed to determine the total amounts of these proteins.

As expected, JAK2 and its downstream signaling proteins STAT5, MEK and ERK1/2 were phosphorylated in response to EPO in BaF3/hEPOR cells (Figure 5A, lanes 7 and 8). In BaF3/Δ259 cells, none of these proteins were phosphorylated upon EPO treatment, because of the lack of the JAK2-binding sites on the mutant EPOR (lanes 1–3). Similarly, ELI-3 induced robust phosphorylation of JAK2 and its downstream signaling proteins in BaF3/hEPOR cells, but not in BaF3/Δ259 cells (lanes 3 and 9). Importantly, in BaF3/Δ310 cells, STAT5, MEK, and ERK1/2 were robustly phosphorylated in response to ELI-3 expression but displayed minimal phosphorylation upon EPO stimulation (Figure 5A, lanes 5 and 6). There were only minor differences in the total amounts of any of these proteins. Thus, phosphorylation of STAT5, MEK, and ERK1/2 in response to ELI-3 does not require EPOR sequences downstream of the JAK2-binding sites. JAK2 itself was phosphorylated in Δ310 cells treated with EPO or (to a lesser extent) expressing ELI-3 (Figure 5A, lanes 5 and 6).

Figure 5

Cytoplasmic Tyrosines of hEPOR Are Not Required for ELI-3 Signaling

(A) Extracts were prepared from starved BaF3/Δ259, BaF3/Δ310, and BaF3/hEPOR cells expressing MSCVp (V) or ELI-3. Where indicated, cells expressing MSCVp were acutely treated with EPO. Extracts were electrophoresed and immunoblotted with anti-phospho-JAK2 (P-JAK2), anti-phospho-STAT5 (P-STAT5), anti-phospho-MEK (P-MEK), and anti-phospho-ERK1/2 (P-ERK) antibodies (top panel in each pair). Membranes were stripped and re-probed for total JAK2, STAT5, MEK, and ERK1/2 (bottom panel in each pair).

(B) BaF3 cells expressing the hEPOR and tTA tetracycline transactivator and ELI-3 expressed from a tetracycline-responsive promoter were incubated for 48 h at the indicated concentration of doxycycline and starved of IL-3 for 3 h. Extracts were electrophoresed and subjected to western blot with anti-phospho-STAT5 antibody or anti-HA antibody (to detect ELI-3). Membranes were stripped and re-probed with antibody recognizing total STAT5.

See also Figures S4 and S5.

We also expressed ELI-3 under the control of a doxycycline-regulated promoter in BaF3/hEPOR cells and tested its ability to induce STAT5 tyrosine phosphorylation. As shown in Figure 5B, ELI-3 caused a dose-dependent increase in STAT5 phosphorylation without affecting the level of total STAT5. Thus, STAT5 phosphorylation is a relatively rapid and dose-dependent response to ELI-3 expression, suggesting that it is directly induced by ELI-3.

JAK-STAT Inhibitors Inhibit ELI-3-Induced Growth Factor Independence

We used chemical inhibitors to test the importance of JAK/STAT signaling for ELI-3 activity. BaF3/hEPOR and BaF3/Δ310 cells expressing either MSCVpuro or ELI-3 were transferred to IL-3-free medium and cultured in the presence or absence of EPO. To test the requirement for JAK2, 7.5 μg/mL JAK2 inhibitor IV was added to the IL-3-free medium, and cells were counted on day 4. As expected, growth of BaF3/hEPOR cells in the presence of EPO was reduced ∼90% by JAK2 inhibition (Figure 6A). JAK2 inhibition also greatly reduced the ability of ELI-3 to support growth factor-independent growth in BaF3/Δ310 cells and, to a lesser extent, in BaF3/hEPOR cells. ELI-3 activity was also inhibited by JAK inhibitor I (data not shown).

Figure 6

JAK2 and STAT Inhibitors Block ELI-3-Induced Growth Factor Independence

(A) BaF3/Δ310 and BaF3/hEPOR cells expressing MSCVp or ELI-3 were incubated in medium lacking IL-3. On day 0, cells were treated with DMSO (−) or 7.5 μM JAK2 inhibitor IV (+). Where indicated, BaF3/hEPOR cells expressing MSCVp were incubated in medium containing EPO. The number of live cells 4 days after IL-3 removal is shown. The averaged results and standard deviation of three independent experiments are shown. Statistical significance was evaluated by two-tailed Student's t test with unequal variance.

(B) BaF3/Δ310 cells expressing ELI-3 were incubated in medium lacking IL-3 for 2 h and then treated for 30 min with DMSO (D) or the indicated concentrations of JAK2 inhibitor IV. Cell extracts were electrophoresed and immunoblotted with anti-phospho-JAK2 (P-JAK2), anti-phospho-STAT5 (P-STAT5), anti-phospho-MEK (P-MEK), and anti-phospho-ERK1/2 (P-ERK) antibodies. Membranes were then stripped and re-probed for total JAK2, STAT5, MEK, and ERK1/2.

(C) Cells were treated and analyzed as in (A), except STAT5 inhibitor SH-4-54 was used.

(D) As in (B), except SH-4-54 was used.

We next determined the effect of the JAK2 inhibitor on phosphorylation of downstream signaling proteins. BaF3/Δ310 cells expressing ELI-3 were starved overnight in IL-3-free medium, and then treated with either DMSO or JAK2 inhibitor IV for 3 hours. Cell lysates were analyzed by immunoblotting with antibodies recognizing phosphorylated JAK2, STAT5, MEK, and ERK1/2. As shown in Figure 6B, the JAK2 inhibitor did not inhibit JAK2 phosphorylation but caused dose-dependent reduction in phosphorylation of STAT5, MEK, and ERK. The total amounts of these proteins were largely unaffected by the inhibitor. These data indicated that the JAK2 activity is important for ELI-3 signaling.

Similarly, cells were treated with 3 μM SH-4-54, a STAT3/5 inhibitor. As expected, growth of the BaF3/hEPOR cells cultured in IL-3-free medium containing EPO was abolished by SH-4-54 (Figure 6C). ELI-3-induced growth in the absence of IL-3 was reduced by ∼80% in BaF3/hEPOR or BaF3/Δ310 cells, suggesting an important, but not absolute, requirement for STAT5 in ELI-3-induced cell proliferation. Figure 6D showed that STAT5 inhibition greatly reduced the phosphorylation of STAT5, MEK, and ERK in BaF3/hEPOR cells without affecting the overall abundance of these proteins, suggesting that STAT5 is upstream of the mitogen-activated protein kinase pathway in ELI-3-induced signaling.

Requirements for Complex Formation between ELI-3, hEPOR, and βcR

To explore the requirement for assembly of a signaling complex, we first showed that the EPOR truncation mutants constitutively associated with βcR (Figure S4A, top panel). Thus, hEPOR sequences downstream of trp258, including the JAK2-binding sites, were not required for this interaction. Similarly, the Δ452 cytoplasmic truncation mutant of βcR retained the ability to form a complex with hEPOR (Figure S5C). In contrast, heteroreceptor formation was inhibited by replacing the TMD of either hEPOR or βcR with a foreign TMD (Figures S6B and S6C).

To determine if βcR was required for complex formation between ELI-3 and EPOR, we expressed MSCVpuro or F-ELI-3 in BaF3, BaF3/h-βcKO, and BaF3/hEPOR cells. Protein extracts were immunoprecipitated with anti-FLAG antibodies and immunoblotted with antibodies recognizing the HA epitope on hEPOR. As expected, anti-FLAG antibodies co-immunoprecipitated little hEPOR in cells lacking ELI-3 expression, presumably due to non-specific sticking of EPOR to the anti-FLAG beads (Figure 7A, lanes 1 and 3). In contrast, anti-FLAG antibodies co-immunoprecipitated abundant hEPOR from cells expressing F-ELI-3, whether or not the βcR was present (Figure 7A, lanes 4 and 6), showing that complex formation between ELI-3 and the hEPOR did not require the βcR.

Figure 7

EPOR and βcR Requirements for Complex Formation and Signaling

(A and B) Extracts prepared from BaF3, BaF3/h-βcKO (clone 12-1), and BaF3/hEPOR cells expressing MSCVhyg or F-ELI-3 were immunoprecipitated with anti-FLAG magnetic beads (FLAG IP) or directly subjected to gel electrophoresis (input). The membranes were immunoblotted with anti-HA antibody to probe for hEPOR (A) or with anti-βcR antibody (B). Arrow in (B) indicates the band of co-immunoprecipitated βcR. Non-specific bands at ∼120 kDa are marked with an asterisk.

(C) Extracts were prepared from IL-3-starved BaF3 cells (lane 6), BaF3/hEPOR cells (lanes 1 to 4), and BaF3/h-βcKO cells (clone 12-1) (lane 5) expressing MSCVhyg (V) or ELI-3. Where indicated, BaF3/hEPOR cells expressing MSCVhyg were also acutely treated with EPO or IL-3. Extracts were immunoprecipitated with anti-βcR antibody and immunoblotted with anti-phosphotyrosine antibody PY100 (top panel). The same membrane was stripped and re-probed with anti-βcR antibody (bottom panel).

(D) Extracts prepared from starved BaF3/Δ259 cells expressing MSCVhyg vector (lanes 1–3) or ELI-3 (lane 4) were immunoprecipitated with anti-βcR antibody and immunoblotted with anti-phosphotyrosine antibody (top panels). Cells were treated with EPO or IL-3 as indicated. Membranes were stripped and re-probed with anti-βcR antibody (bottom panel). An irrelevant lane was excised as indicated (between lanes 3 and 4).

(E) Extracts were prepared from IL-3-starved BaF3/hEPOR (lanes 1–4), BaF3/h-βcKO (lane 5), and parental BaF3 (lane 6) cells expressing ELI-3, treated with EPO or IL-3, or left untreated, as indicated. Extracts were immunoprecipitated with anti-HA to precipitate hEPOR and then blotted with anti-phosphotyrosine antibody (top panel) or with anti-HA antibody to visualize tyrosine-phosphorylated or total hEPOR (bottom panel).

(F) Extracts prepared from starved BaF3/hEPOR (intact βcR) and the indicated clonal BaF3/h-βcKO cells expressing MSCVhyg (V) or ELI-3 were immunoblotted with anti-phospho-JAK2 (P-JAK2), anti-phospho-STAT5 (P-STAT5), anti-phospho-MEK (P-MEK), and anti-phospho-ERK1/2 (P-ERK) antibodies (top panels in each pair). In lane 2, BaF3/hEPOR cells expressing MSCVhyg were also acutely treated with EPO. The membranes were then stripped and re-probed for the total JAK2, STAT5, MEK, and ERK1/2 (bottom panels).

We also determined whether ELI-3 and βcR were in a complex. Anti-FLAG immunoprecipitates were immunoblotted with an antibody recognizing the βcR. As shown in Figure 7B, anti-FLAG immunoprecipitated a small amount of βcR from cells expressing hEPOR (lane 6), but not from cells that did not express hEPOR (lane 5). The βcR antibody also reacted with a major non-specific band migrating at ∼120 kDa in the immunoprecipitated samples, even in the βcR knockout cells. These results indicated that ELI-3 and the βcR are present in a stable complex and that complex formation required co-expression of the hEPOR.

βcR and hEPOR Are Mutually Required for Signaling in Response to ELI-3

To examine βcR phosphorylation, lysates prepared from parental BaF3 cells, BaF3/hEPOR cells, and BaF3/h-βcKO cells were immunoprecipitated with anti-βcR antibody and immunoblotted with a broadly reactive anti-phosphotyrosine antibody. As expected, βcR was phosphorylated at only a low level in cells expressing MSCVhyg in the presence or absence of EPO treatment (Figure 7C, lanes 1 and 2), but was phosphorylated in cells stimulated with IL-3 (Figure 7C, lane 3). Importantly, phosphorylation of βcR was also observed in cells co-expressing ELI-3, hEPOR, and βcR (Figure 7C, lane 4), but not in cells expressing ELI-3 in the absence of hEPOR (Figure 7C, lane 6), indicating that the hEPOR was required for ELI-3-induced βcR activation. Furthermore, ELI-3 did not induce βcR tyrosine phosphorylation in cells expressing hEPOR/Δ259 (Figure 7D, lane 4), suggesting that the JAK2-binding boxes on hEPOR were required for βcR phosphorylation in response to ELI-3 (but not in response to IL-3 [Figure 7D, lane 3]). Even though the cytoplasmic tyrosines of EPOR are not required for ELI-3 activity, phosphotyrosine blotting showed that ELI-3 induced phosphorylation of hEPOR, but only when βcR was co-expressed (Figure 7E, lanes 4 and 5).

We also examined phosphorylation of downstream signaling proteins. Extracts were prepared from IL-3-starved cells (treated, where indicated, with EPO). ELI-3 induced phosphorylation of JAK2, STAT5, MEK, and ERK1/2 in cells with intact βcR (Figure 7F, lane 4), but phosphorylation was eliminated in all four βcR knockout cell lines (Figure 7F, lanes 6 and 8, and data not shown). βcR knockout did not affect phosphorylation induced by EBC5-16 (Figure 7F, lanes 5 and 7). Thus, downstream signaling by ELI-3 required βcR.

ELI-3 Confers Tissue Protection

Finally, we assessed the tissue protection activity of ELI-3 in mouse P19 teratocarcinoma cells, which undergo apoptosis when cultured in medium lacking serum (Galli and Fratelli, 1993, Siren et al., 2001a). Apoptosis is reduced by treating the cells with high concentrations of EPO or with EPO derivatives that activate the EPOR/βcR heteroreceptor (Erbayraktar et al., 2003). Here, we removed serum from P19 cells expressing ELI-3 or MSCVpuro. Twenty-four hours later, nuclei were stained with DAPI and cells were examined by fluorescence microscopy. As shown in Figure 3B, top left panel, ∼20% of control cells lacking ELI-3 expression underwent apoptosis in the absence of EPO, as assessed by nuclear fragmentation. Consistent with published reports, 2 U/mL EPO caused an ∼50% reduction in apoptosis. P19 cells expressing ELI-3 in the absence of EPO were protected against apoptosis to a similar extent. We also used flow cytometry for annexin V staining to test the ability of ELI-3 to protect P19 cells from apoptosis. Cells treated as above were stained with annexin V and propidium iodide (PI) and analyzed by flow cytometry. Apoptotic cells were scored as cells showing annexin V staining to the outer leaflet of the plasma membrane in the absence of PI uptake (Figure S7). As shown in Figure 3B, top right panel, EPO and ELI-3 caused about a 2-fold reduction in apoptotic cells in this assay as well. JAK2 inhibitor IV abrogated the ability of EPO or ELI-3 to protect cells (Figure 3B, bottom panel). These results showed that ELI-3 can protect non-hematopoietic cells from stress-induced apoptosis and suggest that protection requires JAK2 activity.

Discussion

Although EPO is best known for its role in erythropoiesis, it can activate distinct EPOR complexes with different biological outcomes. A homodimeric form of the EPOR binds EPO with high affinity and drives production of red blood cells, and a heteromeric complex containing both the EPOR and the βcR binds EPO with lower affinity, is inactive in erythropoiesis, and appears to mediate tissue protection. The molecular basis underlying cooperation between the EPOR and βcR deserves attention because induction of the tissue protective response may provide opportunities to limit tissue damage following injury, but little is known about how the heteroreceptor complex forms or initiates signaling. We report here the characterization of a small TM protein that induces EPOR/βcR signaling by interacting with the TMD of the EPOR. We demonstrated that most of the cytoplasmic domain of the EPOR is not required for heteroreceptor signaling in response to ELI-3 and identified elements in EPOR and βcR that are required for formation of the heteroreceptor.

Both ELI-3 and EBC5-16 target the EPOR TMD and induce growth factor independence in BaF3 cells, but they have completely different hydrophobic sequences and use fundamentally different mechanisms to trigger EPOR signaling. EBC5-16, like EPO, required cytoplasmic tyrosines on the EPOR, was independent of βcR, and supported erythropoiesis in vitro, whereas ELI-3 was active in the absence of these tyrosines but required βcR as well as hEPOR, failed to induce erythroid differentiation, and conferred tissue protection. ELI-3 induced phosphorylation of EPOR and downstream substrates only if βcR is present, indicating that ELI-3 does not cause productive EPOR homodimerization, consistent with its lack of erythropoietic activity. Although hEPOR can productively couple to murine βcR, we have not tested whether hEPOR can cooperate with human βcR to mediate ELI-3 activity. In addition to βcR, mouse cells express a closely related IL-3-specific β-chain receptor, but the inability of βcR knockout cells to respond to ELI-3 indicates that the IL-3-specific isoform cannot cooperate with hEPOR to support ELI-3-induced proliferation or signaling in BaF3 cells.

This work provides important new insight into cooperative signaling by the EPOR and βcR. First, little if any of the cytoplasmic domain of either receptor is required for heteroreceptor formation in the absence of ELI-3, but the TMDs of both proteins are required. In addition, the extracellular domains of the EPOR and βcR do not physically interact in vitro (Cheung Tung Shing et al., 2018). These findings suggest that the TMD and/or the TMD-proximal segments of the EPOR and βcR mediate heteroreceptor formation. Second, EPOR/βcR signaling can be activated by proteins that interact non-covalently with the TMD of EPOR, in contrast to previously known activators of this complex, which bind to the ligand-binding domain (Brines et al., 2004). Third, productive signaling is dependent on the intracellular domain of βcR and intact JAK2-binding sites on the EPOR. Importantly, EPOR/βcR-mediated proliferation induced by ELI-3 does not require phosphorylated tyrosines or most of the cytoplasmic domain of the EPOR. Previous reports studying phosphotyrosine-null mutants of EPOR suggested that the receptor can initiate phosphotyrosine-independent signaling. Yoon and Watowich showed that EPOR can provide a phosphotyrosine-independent survival signal in 32D cells (Yoon and Watowich, 2003), and EPOR-HM, a mutant removing all cytoplasmic tyrosines, induced attenuated signaling in primary hematopoietic progenitor cells (Li et al., 2003). Mice expressing EPOR-HM maintained steady-state erythropoiesis but were impaired for stress erythropoiesis (Menon et al., 2006, Zang et al., 2001). It has been proposed that STAT5 can bind directly to phosphorylated JAK2 bound to EPOR to mediate some of these responses (Fujitani et al., 1997), but the role of βcR in these situations was not assessed.

Complex formation between ELI-3 and EPOR does not require βcR, and ELI-3 does not associate with βcR in the absence of the EPOR. In addition, ELI-3 does not interact with EPOR containing a foreign TMD and point mutations in the EPOR TMD can inhibit ELI-3 activity. Finally, our NMR experiments conducted in the absence of other protein components suggest that peptides composed of ELI-3 and hEPOR TMD interact in vitro. Taken together, these findings suggest that ELI-3 and the EPOR TMD contact one another directly, as has been shown for other small TM proteins that activate the hEPOR or PDGFβR (Edwards et al., 2013, He et al., 2017).

After GM-CSF binds to the GM-CSF receptor α-chain in complex with βcR, JAK2 associated with βcR phosphorylates multiple cytoplasmic tyrosines on βcR that then serve as docking sites for signaling factors, including STAT5 (Brizzi et al., 1994, Quelle et al., 1994, Sakamaki et al., 1992). Based on structural and biochemical studies of GM-CSFR, Hansen et al. proposed that the active GM-CSFR/βcR heteroreceptor is a dodecameric structure containing four βcR molecules, four GM-CSFR molecules, and four molecules of GM-CSF ligand (Hansen et al., 2008). This complicated architecture appears to be required to bring two JAK2 molecules into juxtaposition to autophosphorylate because GM-CSFR itself does not contain JAK2-binding sites, and the JAK2 sites in a βcR dimer are otherwise too far apart to allow autophosphorylation (Carr et al., 2001). EPOR/βcR heteroreceptor signaling does not necessarily require this complex arrangement, because hEPOR itself contains JAK2-binding sites, which are required for ELI-3 signaling. Furthermore, the EPOR/βcR heteroreceptor does not utilize the cytoplasmic tyrosines in the EPOR to signal, suggesting that the overall architecture of the EPOR/βcR signaling complex is profoundly different from the active EPOR homodimer.

We propose that the TMD of ELI-3 binds directly to the TMD of EPOR in the EPOR/βcR heteroreceptor and that ELI-3 binding recruits another EPOR molecule into the complex or causes a conformational change in the EPOR/βcR heteroreceptor. This allows JAK2 to autophosphorylate and phosphorylate tyrosines in the cytoplasmic domain of βcR, thereby generating the docking platform that assembles the signaling complex. Consistent with this model, ELI-3-induced phosphorylation of the receptors and downstream substrates requires both EPOR and βcR, and the membrane-distal cytoplasmic segment of βcR, but not EPOR, is required for ELI-3 activity. Thus, signaling in response to ELI-3 requires true cooperation between EPOR and βcR: EPOR provides the binding site for ELI-3 as well as required JAK2-binding sites, and βcR provides tyrosines to serve as docking sites. Further analysis of the ability of additional EPOR and βcR mutants to support ELI-3 action is required to refine and test this model.

EPOR/βcR heteroreceptor signaling has previously been implicated in the tissue-protective effects of EPO (Bohr et al., 2015, Brines et al., 2004, Kahn et al., 2013; reviewed in Brines, 2010). Our results confirm that the activated EPOR/βcR complex lacks erythropoietic activity but confers tissue protection. In this regard, ELI-3 is similar to carbamylated EPO and short fragments of EPO that activate EPOR/βcR signaling and confer tissue protection but do not induce erythropoiesis. It is not clear whether EPOR/βcR signaling induced by ELI-3 is the same as that induced by soluble molecules that bind to the extracellular domain of the EPOR, or whether ELI-3 has revealed the existence of a previously unknown EPOR output.

Our inhibitor studies indicate that JAK2/STAT5 signaling is important for growth factor independence and tissue protection in response to ELI-3. EPO also activates additional signaling pathways, including GATA1 and nuclear factor-κB signaling, some of which have been implicated in the tissue-protective effects of EPO (e.g., Digicaylioglu and Lipton, 2001). Further analysis of signaling induced by ELI-3 and EBC5-16 will determine whether these pathways are triggered by traptamers and provide new insights into the signal transduction pathways that mediate these important cellular responses.

This work also highlights the power of specific TM interactions to regulate cell behavior. We show here that specific TM interactions involving different traptamers with the same target TMD can activate different receptor complexes. We previously showed that the ability of traptamers to distinguish between the hEPOR and the mEPOR can be determined by differences as minimal as the position of a single side-chain methyl group in a traptamer (He et al., 2017). The chemical basis for the ability of traptamers to distinguish between closely related targets and to activate different receptor complexes remains to be determined.

Limitations of the Study

Further experiments are required to establish whether ELI-3 binds directly to the EPOR TM domain, refine the model of EPOR/βcR heteromeric receptor activation, and determine the signaling pathways required for ELI-3-induced tissue protection and proliferation.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.