The thrombopoietin receptor S264F mutant drives MPN with marrow fibrosis through a dual mechanism

Abstract

The thrombopoietin (TPO)/TPO receptor (TPOR/MPL) axis is fundamental for megakaryocyte (MK) differentiation, platelet production, and maintenance of hematopoietic stem cells. TPO is produced in the kidney, liver, and bone marrow and binds the extracellular domain (ECD) of TPOR on the cell surface, leading to receptor dimerization and activation of the JAK-STAT pathway.¹ Upon binding its receptor on MKs and platelets, TPO is internalized and degraded, regulating its level via a negative feedback mechanism. Hence, in normal human hematopoiesis, TPO levels correlate inversely with platelet count.²

TPOR is encoded by the MPL gene on chromosome 1. Activating germline mutations of TPOR have been implicated in hereditary thrombocytosis.¹ Somatic TPOR mutations are phenotypic driver mutations in myeloproliferative neoplasms (MPN), occurring in 5%–7% essential thrombocythemia (ET) cases, classically at positions 515 and 505 in exon 10, which covers the transmembrane domain (TMD),³ leading to constitutive JAK-STAT activation^{4, 5} and an MPN phenotype in mouse models with an increase in MK and platelet numbers.^{6, 7}

Conversely, homozygous knockout of MPL in mice causes significant thrombocytopenia and high TPO levels due to a lack of receptor-mediated clearance.¹ In humans, inherited inactivating TPOR mutations are implicated in congenital amegakaryocytic thrombocytopenia (CAMT), an autosomal recessive disease characterized by thrombocytopenia and progression to trilineage aplasia; these patients also exhibit high TPO levels. Most mutations in CAMT are missense mutations, predominantly affecting TPO-binding residues 102 and 104, causing absent cell surface expression or inability to bind TPO, respectively.⁸ Intriguingly, reduced surface TPOR due to germline mutations can paradoxically result in thrombocytosis.¹ For example, the CAMT mutation MPL R102P exerts the opposite phenotype of thrombocytosis when heterozygous. It is thought that the partial trafficking defect leads to insufficient clearance of TPO, which drives thrombocytosis via the remaining wild-type (WT) receptor.⁹

We present a somatic MPL mutation not previously reported in MPN, identified in a 51-year-old white British male, who presented clinically well but with an incidental thrombocytosis of 840 × 10⁹/L. Other blood parameters were normal (Supporting Information S1: Figure S1). Blood film examination confirmed thrombocytosis but was otherwise unremarkable. Bone marrow trephine demonstrated MPN features with grade MF1–MF2 fibrosis but no excess of blasts. Using Continuous Indexing of Fibrosis (CIF) analysis,¹⁰ this trephine lies in the prefibrotic primary myelofibrosis (pre-PMF) region of PCA space (Figure 1A,B). He had a normal karyotype. A 67-gene next-generation sequencing panel did not identify canonical driver mutations but detected the MPL c.791C>T p.(S264F) missense mutation at a variant allele frequency (VAF) of 44%. This variant has been reported before in hematological malignancy, but only in pediatric acute myeloid leukemia with dysplasia.¹¹ Analysis of T cells demonstrated a much lower VAF (2.4%), supporting somatic origin. He commenced pegylated interferon, achieving a sustained complete hematological response after 13 months of 45 μg every other week. There have been serial reductions in VAF to 32% after 13 months and 24% after 23 months.

Plasma TPO level was 81.3 pg/mL (Human Thrombopoietin Quantikine ELISA Kit; R&D Systems). Plasma TPO levels in healthy individuals are usually undetectable (79%), with a mean detectable level of 91.8 pg/mL for this assay (range: not detected-196 pg/mL; R&D Systems). With the thrombocytosis exhibited in this case, 81.3 pg/mL is inappropriately high.

The 3.4 Å resolution cryo-electron microscopy structure of the extracellular human TPO/TPOR complex was recently reported.¹² The high- and low-affinity interfaces, referred to as regions binding TPO sites 1 and 2, identified serine 264 as a TPO-binding residue shared in both regions. S264 is a residue of the second fibronectin type 3-like domain (D2) of TPOR and is part of a loop facing TPO when binding its receptor.

Functional experiments using dual-luciferase STAT5 reporter gene assays in transiently transfected HEK293T cells demonstrated that the TPOR S264F mutant, unlike S264A, induces constitutive activation of the JAK-STAT pathway. This was evidenced by increased STAT5 transcriptional activity under unstimulated conditions compared to WT TPOR, at levels comparable to the canonical activating mutant S505N (Figure 1C). Upon TPO stimulation, however, S264F showed only a very modest additional increase in STAT5 activity (Figure 1C). A similar activation pattern was observed in a STAT3 reporter assay (Supporting Information S1: Figure S2). Protein expression levels of the WT and different TPOR mutants were confirmed by western blot (Supporting Information S1: Figure S3). A 7-day proliferation assay using stably transduced Ba/F3 cells showed that S264F supports IL-3-independent growth, albeit to a lesser extent than S505N (Figure 1D).

To assess ligand binding, we performed a flow cytometry-based TPO-HaloTag (TPO-HT)/TPOR binding assay in stably transduced Ba/F3 cells using Alexa Fluor 660 (AF660)-labeled TPO-HT. Expression of TPO-HT produced in HEK293T cell supernatant was confirmed and semi-quantitatively assessed by western blot (Supporting Information S1: Figure S4). Its biological activity was validated using a STAT5 luciferase assay (Supporting Information S1: Figure S5). Specific binding was further confirmed by competition with an excess of untagged TPO, which reduced the AF660 signal only in WT TPOR-expressing cells (Supporting Information S1: Figure S6). WT TPOR-expressing cells displayed saturable specific binding with an apparent Kd of ~3 nM. In contrast, cells expressing TPOR S264F and untransduced control cells showed no detectable specific binding or saturation across the tested concentration range (Figure 1E). However, we cannot exclude very low-affinity binding that is not detected by our assay.

Western blot and flow cytometry analyses confirmed comparable total protein expression levels and cell surface localization of WT and S264F TPOR, indicating that the mutation does not impair protein expression or trafficking (Figure 1F and Supporting Information S1: Figure S7). In WT TPOR-expressing cells, increasing concentrations of TPO-HT were associated with a decrease in surface TPOR levels, consistent with ligand-induced receptor internalization. In contrast, TPOR S264F surface levels remained stable across all conditions, consistent with impaired ligand binding (Supporting Information S1: Figure S8).

Notably, consistent with the STAT5 luciferase assay, S264F retained a detectable response in a TPO dose-response assay compared to WT (Supporting Information S1: Figure S9). This suggests residual low-level ligand binding below the detection limit of the binding assay, yet sufficient to trigger downstream signaling at high concentrations.

Eltrombopag, a small-molecule TPOR agonist, interacts with juxtamembrane extracellular residues to induce an alpha-helical conformation in a region preceding H499, and thereby promotes receptor dimerization and activation similar to S505N.⁵ We conducted a dose-response assay with eltrombopag and observed increased sensitivity to the drug, reflected by a lower EC50 (51.4 ng/mL−91.0 nM) compared to WT TPOR (327 ng/mL−579 nM) (Figure 2A). This finding is compatible with a primed predimerized state induced by S264F. In line with this, TPOR S264F induced an increased response to eltrombopag in the STAT5 reporter assay (Figure 1C).

To determine whether S264F promotes TPOR dimerization, we performed a NanoBRET assay using TPOR constructs fused to a C-terminal NanoLuc donor and HaloTag acceptor. Our data show that the S264F mutation resulted in enhanced dimerization of TPOR intracellular domains, as evidenced by a higher BRET ratio than WT TPOR (Figure 2B). TPOR constructs showed comparable total protein and surface expression levels (Supporting Information S1: Figure S10 and S11). The JAK2 V617F mutant was included as a positive control for TPOR dimerization (Figure 2B).

Computational modeling using AlphaFold3¹³ suggested a local aromatic interaction network within the TPOR ECD involving residues F104, F164, and the mutant F264 (Figure 2C). We measured centroid-to-centroid distances between aromatic rings across five independent AlphaFold3 models. Distances compatible with π–π interactions were consistently observed between intra-protomer F264 and F164 (mean 5.62 Å), whereas distances involving F104 or inter-protomer F164/F264 were more variable and often exceeded the expected range for stable interactions. To complement this analysis, we used RING¹⁴ to identify noncovalent contacts. Intraprotomer π–π interactions between F264 and F164 were detected across all models, while interactions involving F104 or inter-protomer F164/F264 were less frequently observed. Accordingly, introduction of the F164A mutation in the S264F background abolished the constitutive activation of the JAK-STAT pathway induced by the S264F single mutant as detected by the STAT5 luciferase reporter assay (Supporting Information S1: Figure S12), without reduced cell surface localization between receptor variants (Supporting Information S1: Figure S13). In contrast, the F104A mutation did not suppress S264F-mediated TPOR activation, further supporting our model. Collectively, these data support a model in which the introduction of an aromatic residue at position 264 favors π–π interactions, contributing to local structural stabilization within the ECD. These constraints may influence the transmembrane alpha-helical dimer interface toward a conformation permissive for TPO-independent activation, enabling spontaneous dimerization and downstream signaling.

In conclusion, our findings demonstrate the functional relevance of the TPOR S264F mutation in the patient's disease, acting via a dual mechanism. First, S264F impairs TPO binding and receptor-mediated clearance, resulting in relatively high plasma TPO levels despite thrombocytosis, which may contribute to paradoxical ligand-driven thrombopoiesis through residual WT receptor. Second, the mutation appears to induce constitutive TPOR signaling, probably through an extracellular-domain conformational change that propagates to the transmembrane dimer interface.

Many patients with MPN lack clear pathogenic variants and are therefore classified as triple-negative. Here, we demonstrate that TPOR S264F is likely the driver of pre-PMF in this patient. Investigating non-canonical TPOR mutations, particularly those affecting the ECD, is essential as they reveal critical extracellular regions that prevent autonomous receptor activation and could identify novel therapeutic targets.

Charlotte Lees: Writing—original draft; investigation; writing—review and editing; visualization; formal analysis; methodology; project administration. Emile Brandt: Writing—original draft; writing—review and editing; investigation; methodology; formal analysis; visualization; validation. Emmanuel Dutranoix: Investigation; writing—review and editing; visualization; methodology. Gabriel Levy: Methodology; investigation; writing—review and editing. Bogdan I. Iorga: Methodology; writing—review and editing; visualization; investigation. Hosuk Ryou: Investigation; methodology; writing—review and editing; formal analysis; visualization; software. Jens Rittscher: Methodology; investigation. Daniel Royston: investigation; methodology; writing—review and editing; formal analysis; visualization. Stefan N. Constantinescu: Conceptualization; supervision; writing—review and editing; resources. Adam J. Mead: Conceptualization; supervision; writing—review and editing; resources.

Stefan N. Constantinescu is co-founder of MyeloPro Diagnostics and Research GmbH, Vienna, Austria. Jens Rittscher is a co-founder of Ground Truth Labs. Other authors declare no conflict of interests.

The patient gave written consent to the INForMeD Study (REC reference 16/LO/1376).

Funding to SNC is acknowledged from the Ludwig Institute for Cancer Research, Fondation contre le Cancer Belgium, Fondation Salus Sanguinis, Fondation “Les Avions de Sébastien,” Actions de Recherche Concertées (ARC) 16/21-073, Fonds de la Recherche Scientifique—FNRS (grants PDR T.0043.21 and U.G008.22), and Wel Research Institute WELBIO grant F 44/8/5—MCF/UIG—10955. EB holds a PhD fellowship from the Fondation “Les Avions de Sébastien.” CL holds a clinical research training fellowship from the Medical Research Council (MRC), UK. This work was supported by the National Institute for Health Research (NIHR) Oxford Biomedical Research Centre. Computation used the Oxford Biomedical Research Computing (BMRC) facility, a joint development between the Wellcome Centre for Human Genetics and the Big Data Institute, supported by Health Data Research UK and the NIHR Oxford Biomedical Research Centre. J.R. is an adjunct professor of the Ludwig Oxford Branch. This work was supported by a CRUK Senior Cancer Research Fellowship to AJM. (Grant number C42639/A26988) and a CRUK Discovery Programme Award to AJM. (Grant number DRCNPG-Nov24/100003). The authors would like to acknowledge the CRUK Oxford Centre (CTRQQR-2021\100002) and the National Institute for Health Research (NIHR) Oxford Biomedical Research Centre (BRC). The views expressed are those of the authors and not necessarily those of the National Health Service (NHS), the NIH, the NIHR, or the Department of Health and Social Care.

The data that support the findings of this study are available from the corresponding author upon reasonable request.

The author has provided the required Data Availability Statement, and if applicable, included functional and accurate links to said data therein.