Mutant Calreticulin Requires Both Its Mutant C-terminus and the Thrombopoietin Receptor for Oncogenic Transformation

Myeloproliferative neoplasms (MPN) are clonal disorders of hematopoiesis arising in the hematopoietic stem cell (HSC) compartment and characterized by an excess production of mature blood cells of the myeloid lineage (1, 2). BCR–ABL-negative MPN comprise three distinct diseases: polycythemia vera (PV), essential thrombocythemia (ET), and myelofibrosis (MF). PV is characterized by myeloid hyperplasia and increased red cell mass; ET, by megakaryocytic hyperplasia and increased platelet counts; and MF, by megakaryocytic hyperplasia in conjunction with bone marrow fibrosis. These disorders share a common molecular basis unified by aberrant cytokine signaling. A V617F-activating mutation in the non-receptor tyrosine kinase JAK2 (JAK2^V617F) results in hyperactive JAK–STAT signaling downstream of multiple hematopoietic cytokine receptors, including the erythropoietin receptor (EPOR) and thrombopoietin receptor (MPL; refs. 3–6) and is present in ∼95% of patients with PV and 50% to 60% of patients with ET or MF. Activating mutations in MPL have also been detected in 1% to 5% of patients with ET or MF (7, 8), and negative regulators of JAK–STAT signaling such as LNK (9), c-CBL (10, 11), or SOCS (12) are also somatically inactivated at low frequency in patients with MPN.

Recent whole-exome sequencing studies have revealed that the majority of the remaining patients with JAK2-unmutated and MPL-unmutated ET or MF harbor somatic mutations within the gene CALR (13, 14). CALR encodes a Ca²⁺ binding chaperone protein, calreticulin, that localizes primarily to the endoplasmic reticulum (ER) and regulates protein folding quality control pathways (15). The wild-type CALR protein comprises three protein domains: (i) a conserved N domain, which contains residues that regulate CALR chaperone activity; (ii) a central P domain, which contains a lectin-like chaperone site; and (iii) a C domain, which includes a string of negatively charged amino acid residues responsible for Ca²⁺ buffering and a C-terminus ER retention signal (KDEL; ref. 16). More than 30 different CALR mutations have been identified in patients with MPN, all of which are small insertions and/or deletions (indels) within exon 9 that lead to a 1 base-pair (+1 bp) reading frameshift and the generation of a mutant-specific 36 amino acid C-terminal tail. The mutant-specific CALR C-terminal tail lacks the KDEL ER-retention signal and contains an abundance of positively charged amino acids (13). The relevance of these features to the oncogenicity of mutant CALR remains unclear.

In this study, we dissect the functional and biochemical activity of the most common CALR mutation observed in patients with MPN (i.e., p.L367fs*46, which results in a 52-bp deletion) and report three major advances in understanding the pathogenesis of CALR-mutant MPN: (i) we provide an explanation for the megakaryocyte lineage-specific clinical phenotype of CALR-mutant MPN, (ii) we delineate the oncogenic role of the CALR mutant–specific C-terminus in the pathogenesis of CALR-mutant MPN, and (iii) we provide a biologic basis for the observation that JAK2^V617F and CALR mutations are typically mutually exclusive in patients with MPN.

The role of BCR–ABL (17, 18), JAK2^V617F (19–22), and MPL^W515L (7) as bona fide MPN driver mutations was demonstrated based on their ability to induce MPN phenotypes in mice (23). However, equivalent data for CALR mutations are currently lacking. In order to test whether expression of mutant CALR alone could similarly induce an MPN phenotype in mice, we performed bone marrow transplantation (BMT) experiments whereby c-KIT–enriched primary mouse bone marrow cells were transduced with retroviruses expressing either an empty vector (EV), a wild-type human CALR cDNA (CALR^WT), or a mutant human CALR cDNA (CALR^MUT), and transplanted the cells into lethally irradiated recipient mice (Fig. 1A). The specific CALR mutation used throughout this study was the 52-bp deletion.

Donor-derived chimerism for both CALR^WT- and CALR^MUT-expressing cells decreased relative to EV-expressing cells, which may indicate a cytotoxic effect associated with CALR overexpression (Supplementary Fig. S1A). However, by 16 weeks after transplantation, mice transplanted with cells expressing CALR^MUT (but not CALR^WT or EV) had developed an MPN phenotype reminiscent of ET, characterized by isolated thrombocytosis (Fig. 1B) and megakaryocytic hyperplasia (Fig. 1C and D; Supplementary Fig. S1B). The megakaryocytes in CALR^MUT mice were markedly enlarged with hyperlobulated nuclei and emperipolesis, consistent with an MPN phenotype (Fig. 1C). Given the megakaryocyte lineage-specific phenotype of CALR^MUT-expressing mice, we next assessed CALR expression in primary MPN bone marrow samples using immunohistochemical analysis. Consistent with recent reports (24, 25), we found that CALR is highly expressed in megakaryocytic lineage cells and in immature myeloid cells in the bone marrow of patients with CALR-mutant MPN, with little or no CALR expression observed in erythroid lineage cells or in mature myeloid cells (Supplementary Fig. S1C).

We also performed flow cytometric analysis to characterize the stem and progenitor cell compartment of the bone marrow in the transplanted mice. We observed a modest expansion of the LIN−SCA1+c-KIT+ (LSK) compartment, which is enriched for HSC activity, in CALR^MUT mice relative to EV mice (P= 0.02) or CALR^WT mice (P= 0.03) (Supplementary Fig. S1D). In contrast, no significant expansion of the LIN−SCA1−c-KIT+ (LK) myeloid progenitor cell compartment was observed (Supplementary Fig. S1D). Together, these data indicate that overexpression of mutant CALR alone is sufficient to engender an MPN phenotype in mice and the disease phenotype that ensues closely recapitulates the clinical features observed in patients with CALR-mutant MPN.

The megakaryocyte lineage–specific phenotype of CALR-mutant MPN prompted us to investigate the role of MPL in disease pathogenesis. For this, we used a standard Ba/F3 cellular transformation assay. Ba/F3 cells are a murine IL3-dependent hematopoietic cell line that can be rendered IL3 independent upon ectopic expression of certain oncogenes, such as BCR–ABL (26). In the case of JAK2^V617F, however, this transformation requires coexpression of a type I cytokine receptor, such as the erythropoietin receptor (EPOR) or MPL (27). We found that CALR^WT or CALR^MUT were both incapable of transforming parental Ba/F3 cells to IL3 independence (Fig. 2A). However, we found that CALR^MUT (but not CALR^WT) was able to successfully transform Ba/F3 cells to IL3 independence upon coexpression of MPL (Fig. 2B). This dependence on MPL was specific, as CALR^MUT was incapable of transforming cells expressing other type I hematopoietic cytokine receptors, such as the EPOR and the granulocyte-colony stimulating factor receptor (G-CSFR) to IL3 independence (Fig. 2C and D). Similarly, we observed that CALR^MUT was able to transform human megakaryocytic UT-7 cells (28) to GM-CSF independence when MPL (but not EPOR or G-CSFR) was stably expressed, demonstrating that this effect is not restricted to Ba/F3 cells (Fig. 2E–G). Of note, JAK2^V617F indiscriminately transformed all type I cytokine receptor–expressing Ba/F3 cells, pointing to an important distinction between the transforming requirements of JAK2^V617F and mutant CALR (Fig. 2B–D).

To confirm the transforming capacity of mutant CALR when expressed at physiologic levels, we used CRISPR/Cas9 gene editing (29). We first stably expressed Cas9 in parental Ba/F3, Ba/F3–MPL, Ba/F3–EPOR, and Ba/F3–G-CSFR cells, then infected each cell line with a small guide RNA (sgRNA) targeting exon 9 of Calr, in the region where the 52-bp deletion occurs in patients with CALR-mutant MPN (Supplementary Fig. S2). To ensure that the observed effects were due to on-target gene editing, we performed separate infections using two distinct sgRNAs (m1 or m2), each targeting exon 9 of Calr, and compared our findings to cells infected with a scramble sgRNA control (Fig. 3A–D). We observed IL3-independent outgrowth of Calr-targeted Ba/F3–MPL-Cas9 cells (Fig. 3B) but not Calr-targeted parental Ba/F3–Cas9 cells (Fig. 3A) or Calr-targeted Ba/F3–Cas9 cells overexpressing EPOR (Fig. 3C) or G-CSFR (Fig. 3D). To confirm that IL3-independent growth in Calr-targeted Ba/F3–MPL-Cas9 cells was a result of an on-target event, we harvested cells 8 days after IL3 withdrawal and extracted genomic DNA. We then PCR-amplified a 422-bp region spanning the target site, performed subcloning of PCR amplicons and sequenced 30 individual clones. Out of 13 subclones from m1 Calr-targeted Ba/F3–MPL-Cas9 cells, 11 were found to contain indels that led to +1-bp frameshift mutations. For m2 Calr-targeted Ba/F3–MPL-Cas9 cells, 10 subclones out of 17 were found to contain indels that led to +1-bp frameshifts, and 3 of the 10 showed a 52-bp deletion within exon 9 (Fig. 3E). These data confirm that MPL is required for mutant CALR to transform Ba/F3 cells and demonstrate that introduction of a +1-bp frameshift to the endogenous Calr locus is sufficient to confer oncogenic activity to CALR. Furthermore, the sequencing data (Fig. 3E) suggest that the introduction of heterozygous +1-bp frameshift mutations to the endogenous Calr locus is sufficient to transform Ba/F3–MPL cells, consistent with the observation that CALR mutations are typically heterozygous in patients with MPN (13, 14).

We next performed whole-transcriptome RNA sequencing to identify gene expression changes associated with mutant CALR-mediated transformation. We found that CALR^MUT-transformed Ba/F3–MPL cells grown in the absence of IL3 for 24 hours display strong enrichment of STAT5 (30, 31) and STAT3 (31, 32) gene expression signatures when compared to cells overexpressing CALR^WT (Fig. 4A; Supplementary Fig. S3A; Supplementary Table S1). In consonance, immunoblotting revealed activation of MPL, JAK2, STAT5, and STAT3 in CALR^MUT-transformed Ba/F3–MPL cells that were starved of IL3 for 24 hours, but no evidence of STAT activation in IL3-starved parental Ba/F3, Ba/F3–EPOR, or Ba/F3–G-CSFR cells expressing CALR^MUT (Fig. 4B and C). We also observed activation of STAT5 in UT-7-MPL cells transformed to GM-CSF independence by ectopic expression of CALR^MUT (Supplementary Fig. S3B) and activation of STAT5 and STAT3 in Ba/F3–MPL cells transformed to IL3 independence through the introduction of +1-bp frameshift mutations to the endogenous Calr locus by CRISPR/Cas9 gene editing (Supplementary Fig. S3C).

Given the increased JAK–STAT activity in CALR^MUT-transformed Ba/F3–MPL cells, we next evaluated the sensitivity of these cells to JAK2 inhibition. We performed shRNA-mediated knockdown of JAK2 (using two separate shRNAs) and found that JAK2 knockdown significantly decreased the proliferative rate of CALR^MUT-transformed Ba/F3–MPL cells (growing in the absence of IL3) relative to a nontargeting shRNA control (P= 0.0001; Fig. 5A). Moreover, we found that the differential activation of STAT3 in these cells was abrogated by treatment with the JAK1/2 inhibitor INCB018424 (ruxolitinib; ref.33; Fig. 5B). Concordant with this, genes downregulated in JAK2 inhibitor signatures were enriched in Ba/F3–MPL cells expressing CALR^MUT as compared with CALR^WT-expressing Ba/F3 MPL cells (Fig. 5C). Together, these data demonstrate that CALR^MUT transforms cells through activation of the JAK–STAT signaling axis downstream of MPL, and that Ba/F3–MPL cells transformed by CALR^MUT are sensitive to JAK2 inhibition.

We next sought to understand the biochemical basis by which the mutant C-terminus of CALR^MUT contributes to transformation. To define the specific domains within CALR^MUT necessary for oncogenic transformation, we generated a panel of mutated versions of CALR^MUT and systematically assessed their ability to transform Ba/F3–MPL cells.

We began by generating a series of CALR domain mutants. We first generated N domain and P domain versions of CALR. The N domain cDNA encodes the first 197 shared amino acids between WT and mutant CALR, whereas the P domain cDNA encodes the middle 111 shared amino acids between WT and mutant CALR. We then generated CALR^WT C domain and CALR^MUT C domain versions of CALR (C domain^WT and C domain^MUT, respectively) that differ in the terminal 36 amino acids that distinguish mutant CALR from WT CALR (Fig. 6A). We found that ectopic expression of the N, P, or C domain^WT domain-only versions of CALR alone was incapable of transforming Ba/F3–MPL cells to IL3 independence (Fig. 6B). Moreover, we found that ectopic expression of the CALR^MUT C domain alone was also unable to transform Ba/F3–MPL cells to IL3 independence (Fig. 6B). Together, these data indicate that the CALR mutant C-terminus is necessary but alone is not sufficient for transformation (Fig. 6A and B).

We next focused on altering the full-length CALR^MUT protein. One of the most prominent features of CALR mutations is that they all result in loss of the terminal KDEL motif necessary for ER retention. We therefore tested whether loss of this sequence alone renders CALR an oncogene. We found that only removing the KDEL sequence from CALR^WT did not lead to transformation of Ba/F3–MPL cells (Fig. 6C). This finding suggests that the mutant CALR C-terminus does not exert its oncogenic activity solely through CALR mislocalization but rather the novel mutant C-terminus likely imparts a gain-of-function, a conclusion consistent with an absence of patients with MPN harboring +2 frameshifts that causes a premature stop codon (13).

Having demonstrated that the introduction of +1-bp frameshift mutations to the endogenous Calr locus is sufficient to transform Ba/F3–MPL-Cas9 cells (Fig. 3B), we noted that differences exist between human mutant CALR and mouse mutant CALR protein sequences (Supplementary Fig. S4A–S4B). This suggested to us that the oncogenic properties of the CALR mutant C-terminus may not reside within a specific residue or sequence of residues but rather reflect a shared property of the neomorphic C-terminal tail of human mutant CALR and mouse mutant CALR. To test this hypothesis, we first synthesized a series of CALR^MUT C-terminus sequence mutants in which blocks of 8 to 10 amino acids of the mutant C-terminus were individually deleted (CALR^MUTΔ0–8, CALR^MUTΔ9–18, CALR^MUTΔ19–26, CALR^MUTΔ27–36; Fig. 6A). We found that all four CALR^MUT C-terminus sequence mutants were capable of transforming Ba/F3–MPL cells with equal efficacy (Fig. 6D). These data are consistent with our hypothesis that the transforming activity of mutant CALR is not contained within specific residues within the CALR^MUT C-terminus.

MPN-associated CALR mutations replace a group of negatively charged amino acids in the C-terminus of CALR^WT protein with a surplus of positively charged amino acids [most often lysine (K) and arginine(R)] in the C-terminus of CALR^MUT protein (13). We therefore next hypothesized that the shared oncogenic property of the neomorphic C-terminal tail of human mutant CALR and mouse mutant CALR may relate to the positive electrostatic charge of the mutant C-terminus. To test this hypothesis, we designed a CALR^MUT C-terminus in which every K and R residue in the C-terminus of mutant CALR was replaced with a neutral glycine (G) (CALR^MUT-neutral; Fig. 6E). As a control, we replaced every non-K and non-R residue with a glycine (CALR^MUT-positive; Fig. 6E). Remarkably, expression of CALR^MUT-positive retained the ability to transform Ba/F3–MPL cells to IL3 independence (Fig. 6E), despite 18 of 36 amino acids within the C-terminus being altered to glycine and the majority of the C-terminus being comprised of only 3 different amino acids. In contrast, the CALR^MUT-neutral was incapable of transforming BaF3–MPL cells to IL3 independence (Fig. 6E).

In aggregate, these findings reveal an essential requirement for the positive electrostatic charge of the CALR^MUT C-terminus in mediating the transforming activity of CALR^MUT, while revealing a striking degeneracy with regard to the sequence requirements.

Given the absolute requirement for MPL in CALR^MUT-mediated transformation, we hypothesized that MPL and CALR^MUT may physically interact. To test this hypothesis, we performed FLAG-immunoprecipitation assays in 293T cells cotransfected with MPL and FLAG-tagged CALR variants. In support of the Ba/F3 transformation assay results, we found that compared to CALR^WT, CALR^MUT displayed differential binding to MPL (Fig. 7A), but not to EPOR (Fig. 7B) or G-CSFR (Fig. 7C). We also found increased binding between MPL and mutant CALR (as compared to wild-type CALR) in Ba/F3–MPL-Cas9 cells targeted with CRISPR/Cas9 to engender endogenous level mutant CALR expression (Supplementary Fig. S5), demonstrating that this differential interaction is not as a result of overexpressing mutant CALR. Moreover, expression of the mutant C-terminus alone was not sufficient to mediate binding in 293T cells (Fig. 7D), suggesting that the tertiary structure of CALR^MUT is important for physical interaction with MPL.

Given the essential requirement for the positive electrostatic charge of the CALR^MUT C-terminus in CALR^MUT-mediated transformation, we next tested the ability of the CALR^MUT charge variants described in Fig. 6E to bind to MPL. We found that binding of CALR^MUT-neutral to MPL was markedly attenuated as compared to the binding of full-length CALR^MUT to MPL, and strikingly that the CALR^MUT-positive retained the ability to bind MPL (Fig. 7E). These findings are in consonance with the Ba/F3 transformation assay results, where CALR^MUT-neutral does not transform Ba/F3–MPL cells but CALR^MUT-positive does (Fig. 6E).

Taken together, these findings reveal a specific physical interaction between MPL and mutant CALR and demonstrate a correlation between the binding of mutant CALR to MPL and the transforming activity of mutant CALR. In aggregate, our findings suggest that binding of mutant CALR to MPL is required for cellular transformation.

The identification of recurrent mutations in CALR in MPN is among the most unexpected findings from recent whole-exome studies in myeloid malignancies (13, 14). CALR, an ER chaperone protein that normally functions to bind misfolded proteins in the ER and prevent their export to the Golgi (34), had never previously been found to be mutated in cancer or to be associated with hematologic disorders.

The presence of recurrent mutations in CALR suggests a key pathogenic role for mutant CALR in MPN (13, 14). Our murine BMT data demonstrate that mutant CALR alone is sufficient to induce an MPN phenotype in vivo and support the proposition that CALR mutations are driver mutations in MPN. These data accord with a recent study, which also used a retroviral BMT mouse model of the CALR 52bp deletion, and observed a similar MPN phenotype (thrombocytosis and megakaryocytic hyperplasia), in addition to the development of myelofibrosis at 6 months or more after transplantation (35). A central role for mutant CALR in early MPN ontogeny is also consistent with clinical sequencing studies showing that CALR mutations are typically mutually exclusive with the other MPN-driver mutations (such as JAK2^V617F and MPL^W515L). Moreover, CALR mutations are detectable in the long-term HSC compartment (13, 14), and clonal analysis of CALR-mutated patient samples has shown that CALR mutations frequently reside in the earliest detectable phylogenetic node (14).

Our data reveal an essential and specific role for MPL in mutant CALR-mediated cellular transformation in multiple experimental contexts. First, we found that ectopic expression of mutant CALR in Ba/F3 hematopoietic cells requires coexpression of MPL for transformation, a finding also demonstrated in recent reports (35, 36). Importantly, our data further demonstrate a specific requirement for MPL under conditions of endogenous level–mutant CALR expression (following CRISPR/Cas9 gene editing in Ba/F3 cells) and in human cytokine-dependent megakaryocytic cells (following lentiviral delivery in UT-7 cells), which more closely recapitulate CALR-mutant MPN. The latter finding has also recently been reported by another group (37). The specificity of the dependence on MPL is in striking contrast to JAK2^V617F, which is more agnostic with respect to cytokine receptor coexpression requirements (27, 38). These differences in cytokine receptor requirements align well with how CALR and JAK2 mutations are distributed within the MPN subtypes. CALR mutations are restricted to ET and MF (39, 40), which exhibit a megakaryocyte predominant disease phenotype and are not found in PV, where the defining clinical feature is erythrocytosis; JAK2^V617F mutations, on the other hand, are found in all three MPN subtypes and JAK2-mutant MPN typically display leukocytosis, erythrocytosis, and thrombocytosis. The high degree of cytokine receptor specificity for mutant CALR is likely to relate physical interaction between mutant CALR and MPL but not with other cytokine receptors, such as EPOR (discussed below).

Consistent with the mutual exclusivity of CALR and JAK2 mutations in MPN, we observed evidence of increased JAK–STAT signaling in CALR^MUT-transformed Ba/F3–MPL cells, which we found to be susceptible to genetic and pharmacologic inhibition of JAK2 signaling. These findings accord with a recent microarray gene expression analysis that similarly depicts increased JAK–STAT pathway activation in CALR-mutant primary MPN granulocytes (41), and with a recent case report describing two patients with CALR-mutant MPN who demonstrated clinical responses to the JAK2 inhibitor fedratinib (42). The clonal selectivity of JAK2 inhibitors for CALR-mutant cells in MPN remains to be determined, but the on-target toxicity of JAK2 inhibitors (e.g., anemia) is likely to be an issue in the treatment of CALR-mutant MPN just as it is in the treatment of JAK2-mutant MPN (43), suggesting that novel approaches aimed at preferentially targeting CALR-mutant cells in MPN will be needed. A recent study demonstrated the use of nonsignaling diabodies to reorient the EPOR into an inactive dimer topology and thus inhibit the erythropoietin-independent but EPOR-dependent proliferation of primary JAK2^V617F-mutant erythroid precursor cells in vitro (44). Analogous strategies to inhibit MPL signaling may be a viable strategy to selectively target CALR-mutant cells in MPN.

The observation that over 30 disparate mutations all generate the same novel C-terminal peptide tail strongly suggests that CALR mutations are gain-of-function or that they confer a neomorphic function on mutant CALR, with the mutant C-terminus playing a critical oncogenic role. Concordant with this, a recent study reported that overexpression of a CALR mutant in which the entire exon 9 was deleted did not induce MPN in mice (35). Through extensive mutagenesis-based structure–function analysis, our data reveal some novel insights into the role the C-terminus of mutant CALR plays in oncogenesis. Our data propose a model whereby the oncogenic activity of mutant CALR is not encoded within a specific sequence of the mutant CALR C-terminal tail. Rather, our data demonstrate a role for the positive electrostatic charge of the mutant C-terminus in influencing the ability of mutant CALR to physically associate with MPL, thus facilitating transformation. This model poses two unanswered questions: (i) How does the mutant CALR C-terminus favor MPL binding over other type I cytokine receptors? (ii) How does this interaction activate JAK–STAT signaling? One possible explanation is that mutant CALR exhibits a different tertiary structure than wild-type CALR, thus facilitating a specific interaction with MPL. It is also possible that the diminished calcium sequestration capacity of mutant CALR due to loss of the negatively charged amino acids (45) may stabilize MPL association. A recent report showed that a glycosylation site within the extracellular domain of MPL is required for mutant CALR-mediated transformation in vitro, suggesting that MPL glycosylation influences its ability to interact with mutant CALR (36). Future studies aimed at further addressing these questions are warranted.

In conclusion, our data provide insights into the molecular mechanism by which mutant CALR transforms hematopoietic cells. Specifically, we uncover an essential requirement for MPL in mutant CALR-mediated transformation. We further show that the oncogenicity of mutant CALR is dependent on the positive electrostatic charge of the C-terminus of the mutant protein, which promotes physical interaction with MPL. Together, our findings elucidate a novel paradigm of cancer pathogenesis and help explain how CALR mutations drive the development of MPN.

The 293T cells were obtained from ATCC in 2011, and their identity was authenticated by short tandem repeat (STR) profiling. Ba/F3 cells and UT-7 cells were purchased from German Collection of Microorganisms and Cell Cultures (DSMZ) in 2014 and were not further authenticated. All cell lines were intermittently tested for Mycoplasma.

Type I cytokine receptor–expressing Ba/F3 and UT-7 cell lines were generated by retroviral transduction. In brief, retroviral supernatants were generated by cotransfection of pMSCV-hygro-hMPL, pMSCV-neo-hEPOR, or pMSCV-neo-hG-CSFR with packaging plasmid in 293T cells. Viral supernatants were collected 24 and 48 hours after transfection. Ba/F3 or UT-7 cells were subjected to spin infection with viral supernatants, followed by 7 days of antibiotic selection. CALR variant lentiviral supernatants were generated by cotransfection of LeGO-iV2 empty vector, LeGO-iV2-CALR wild-type, LeGO-iV2-52-bp deletion, or pMSCV-IRES-GFP-JAK2^V617F with packaging plasmids in 293T cells. Viral supernatants were collected 24 and 48 hours after transfection. Stable Ba/F3 and UT-7 lines stably expressing type I cytokine receptors were then subjected to spin infection with CALR variant lentiviral supernatants. Forty-eight hours after spin infection, cells were sorted for GFP expression using a BD FACSAria cell sorter (BD Biosciences).

Stable knockdown of endogenous JAK2 in Ba/F3-MPL cells was achieved using a pLKO.1-based lentiviral vector shRNA construct targeting JAK2 (shJak2 #1 5′-CCAACATTACAGAGGCATAAT-3′; shJak2 #2 5′ CGTGGAATTTATGCGAATGAT-3′) or control nontargeting shRNA against Luciferase (shLuc 5′-GCTGAGTACTTCGAAATGTCC-3′). The lentiviral backbone vector and packaging plasmids were transfected into 293T cells and the viral supernatant was harvested 24 and 48 hours later. Ba/F3–MPL cells overexpressing CALR^MUT were spin infected with lentiviral supernatants as indicated, then subjected to puromycin selection for 7 days.

Sorted and exponentially growing Ba/F3 or UT-7 cells were washed four times with PBS and seeded in triplicate at 1 × 10⁵ cells/mL in RPMI-1640 medium supplemented with 10% FBS and 5% penicillin/streptomycin in the presence or absence of cytokine for indicated time points. Living cells were counted at each time point using a Beckman Coulter VI-Cell XR Cell Viability Analyzer (Beckman Coulter).

Indel mutations targeting exon 9 of the endogenous Calr locus (focused on the region of the 52-bp deletion found in patients with MPN) were introduced to Ba/F3 cells using CRISPR/Cas9 gene editing. Stable expression of the codon optimized Streptococcus pyogenes Cas9 in parental Ba/F3 cells and Ba/F3 cells overexpressing hMPL, hEPOR, or hG-CSFR was achieved by lentiviral transduction of pLX_TRC311-Cas9 (29) and selected with 5 μg/mL blasticidin. In brief, lentiviral supernatants were generated by co-transfection of pLX_TRC311-Cas9 with packaging plasmid in 293T cells. Viral supernatants were collected 24 hours after transfection. Ba/F3 cells were subjected to spin infection with viral supernatants, followed by 9 days of antibiotic selection. Cas9 activity was confirmed using a reporter assay (29). sgRNAs targeting exon 9 of Calr were designed using the Broad Institute cleavage efficiency predictor, and off-target scores (Supplementary Table S2) were obtained from the MIT Optimized CRISPR Design website (29). For sgRNA cloning, the lentiGuide vector (Addgene plasmid 52963), driven by a U6 promoter, was digested with BsmBI (NEB) and ligated with BsmBI-compatible annealed oligos (Supplementary Table S2). An extra G was added to the 5′ end of sgRNAs that lacked it to allow for U6 transcriptional initiation. Stable mutagenesis of endogenous Calr in Ba/F3 cells overexpressing Cas9 was achieved by lentiviral transduction of lentiGuide harboring sgRNA m1: 5′-GAGGACAAGAAGCGTAAAG-3′; sgRNA m2: 5′-AGGCTTAAGGAAGAAGAA-3′ or a control scrambled vector. The lentiviral backbone vector and packaging plasmids were transfected into 293T cells and the viral supernatant was harvested 24 hours later. Parental Ba/F3 Cas9 cells and Ba/F3 Cas9 cells overexpressing hMPL, hEPOR, or hG-CSFR were subjected to spin infection with viral supernatants, followed by 7 days of puromycin selection.

CRISPR/Cas9 on-target gene editing of IL3-independent Ba/F3–MPL-Cas9 cells was confirmed using Sanger sequencing following pGEM T-easy cloning of the PCR-amplified Calr exon 9–targeted region. First, genomic DNA was extracted (DNeasy Blood and Tissue Kit; Qiagen) from cells 8 days after IL3 withdrawal. PCR amplification of the Calr exon 9 locus was performed using 2x PCR Promega Master Mix (Thermo Fisher Scientific) and the following primers: Calr_Fwd (ACCACCTGTCTTTCCGTTCT) and Calr_Rev (GGCCTCTACAGCTCATCCTT) (IDT). The Calr PCR amplicons were ligated into a pGEM-T Easy vector using T4 DNA ligase (Promega) and incubated for 1 hour at room temperature. Transformation was carried out in JM109-competent cells, and insert-containing white recombinant colonies were selected on LB agar plates containing X-gal and IPTG by incubating at 37°C overnight. The colonies were Sanger sequenced using the universal primers T7 (TAATACGACTCACTATAGGG) and SP6 (ATTTAGGTGACACTATAG) to verify genome editing of the Calr exon 9 region.

Gene expression data are available in the GEO database with the accession number GSE74890.

Ba/F3–MPL cells ectopically expressing CALR variants or JAK2^V617F were cultured in RPMI with 10% FBS and 5% penicillin/streptomycin for 24 hours in the absence of mIL-3, followed by treatment with 1 μmol/L INCB018424 or DMSO for 3.5 hours. Harvested cells were fixed in 3% to 4% paraformaldehyde for 10 minutes, and then permeabilized with ice-cold methanol for 10 minutes on ice or overnight at −20°C. Cells were washed twice in PBS with 1% BSA and staining was carried out in PBS 1% BSA for 30 minutes at 4°C with Alexa Fluor 647 Rabbit Anti pSTAT3 (Y705; Cell Signaling Technology). Cells were washed after staining with PBS with 1% BSA and samples were analyzed using a FACS Canto cytometer (BD Biosciences). Data analysis was performed with FlowJo software V10.0.8.

Six-to-eight-week-old C57BL/6 female mice were purchased from Taconic. Retroviral supernatants were generated by transient cotransfection of 293T cells with MSCV-IRES-GFP empty vector, CALR WT or 52 bp deletion constructs, and EcoPak packaging construct using TransIT LT-1 Reagent (Mirus Bio). Viral supernatant was collected 24 and 48 hours after transfection. Two days before the BMT, bone marrow cells were collected from the femurs, tibias, and spines of donor mice. Cells were then incubated with CD117 (c-KIT) MicroBeads (Miltenyi Biotec) and subjected to positive selection using an autoMACS Pro Separator (Miltenyi Biotec). c-KIT-enriched cells were then cultured overnight in SFEM medium supplemented with 50 ng/mL recombinant murine TPO, 50 ng/mL recombinant murine SCF, 10 ng/mL recombinant murine IL3, and 10 ng/mL recombinant murine IL6. Eighteen hours later, cells were infected with retroviral supernatant by spin infection on RetroNectin coated plates and cultured overnight in SFEM media containing 50 ng/mL recombinant murine TPO, 50 ng/mL recombinant murine SCF, 10 ng/mL recombinant murine IL3, and 10 ng/mL recombinant murine IL6. The following day, cells (1 × 10⁶ per mouse) were resuspended in Hank's Balanced Salt Solution and injected retro-orbitally into lethally irradiated (900 cGy) C57 BL/6 recipient mice. Peripheral blood was collected at indicated time points to monitor relative frequency of GFP-positive cells. Following red blood cell lysis (BD Pharm Lyse; BD Biosciences) and homogenization through a 40-μm filter, samples were analyzed by flow cytometry using the fluorescence-activated cell sorter BD FACSCanto (BD Biosciences).

Mice were bled 0, 4, and 16 weeks after BMT. Blood (75 μL) was collected in EDTA capillary tubes and diluted with 300 μL PBS to produce a 1:5 dilution. Complete blood counts were run on the ADVIA 2120i blood analyzer (Siemens Healthcare Global).

Bone marrow was collected and prepared for staining by red blood cell lysis (BD Pharmlyse; BD Biosciences) and homogenization through a 70-μm filter. All samples were analyzed by flow cytometry using an LSR II (BD Biosciences). All staining steps were performed in ice-cold PBS containing 2% FBS. Post-acquisition analysis of data was performed with FlowJo software V9.2.3 (Treestar). The following antibodies were used: lineage cocktail containing CD3e (145-2C11), CD5 (53-7.3), Ter-119 (TER-119), Gr-1 (RB6-8C5), Mac-1 (M1/70), and B220 (30-F11); Kit (2B8), Sca-1 (D7), CD34 (Ram34), CD16/32 (93). For dead cell discrimination, DAPI was used.

All mouse tissues were fixed in 10% neutral buffered formalin, embedded in paraffin, and stained with hematoxylin and eosin (H&E). Images of histologic slides were obtained on a Nikon Eclipse E400 microscope (Nikon) equipped with a SPOT RT color digital camera model 2.1.1 (Diagnostic Instruments). Bone marrow megakaryocytes were quantified by a pathologist, who was blinded to mouse genotype, and represent an average megakaryocyte count from 10 high-power fields assessed.

All human bone marrow biopsies were fixed overnight in Bouin solution (StatLab Medical Products) and decalcified for 15 minutes using RapidCal-Immuno (BBC Biochemical). Immunohistochemical studies were performed on paraffin sections using a recombinant rabbit monoclonal antibody to wild-type calreticulin (clone EPR3924; Abcam). A synthetic peptide corresponding to residues in the N-terminal domain of calreticulin was used as an immunogen. Staining was performed on the Leica Bond III staining platform (dilution 1:2,000) using the Bond Polymer Refine Detection Kit (Leica Biosystems Inc.). Antigen retrieval was performed using the Bond Epitope Retrieval 1 solution for 30 minutes.

Immunohistochemical studies were performed on deidentified primary MPN bone marrow samples, under an Institutional Review Board-approved protocol (#2013P001110) at Brigham and Women's Hospital.

All comparisons represent two-tailed unpaired t test analysis unless otherwise specified.

See supplementary information for reagents, cell culture, RNA sequencing, and statistical analyses for RNA sequencing.

No potential conflicts of interest were disclosed.

Conception and design: S. Elf, N.S. Abdelfattah, E. Chen, J.-A. Losman, A. Mullally

Development of methodology: S. Elf, N.S. Abdelfattah, A. Mullally

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Elf, N.S. Abdelfattah, E. Chen, E.A. Rosen, A. Ko, F. Peisker, N. Florescu, S. Giannini, O. Wolach, E.A. Morgan, R.K. Schneider, A. Mullally

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Elf, N.S. Abdelfattah, J. Perales-Patón, E.A. Rosen, A. Ko, F. Peisker, N. Florescu, Z. Tothova, J.-A. Losman, R.K. Schneider, F. Al-Shahrour, A. Mullally

Writing, review, and/or revision of the manuscript: S. Elf, N.S. Abdelfattah, E. Chen, J. Perales-Patón, E.A. Rosen, F. Peisker, N. Florescu, S. Giannini, O. Wolach, E.A. Morgan, Z. Tothova, J.-A. Losman, R.K. Schneider, F. Al-Shahrour, A. Mullally

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Elf, N.S. Abdelfattah, A. Ko, Z. Tothova, A. Mullally

Study supervision: S. Elf, N.S. Abdelfattah, F. Al-Shahrour, A. Mullally

The authors thank Drs. Benjamin Ebert and Sagar Koduri and all members of the Mullally and Ebert laboratories for helpful discussions and input, and Professor Harvey Lodish (MIT) for his valuable insight and interest in their work.

This work was supported by the NIH (K08 HL109734 to A. Mullally), a Damon Runyon clinical investigator award (A. Mullally), MPNRF (A. Mullally), and the Jeanne D. Housman Fund for Research on Myeloproliferative Disorders (A. Mullally). S. Elf is a recipient of a T32 molecular hematology training award (NHLBI), E. Chen is a recipient of a Lady Tata Memorial Trust Award, and E.A. Rosen is a recipient of an ASH Physician-Scientist Career Development Award and an ASH HONORS award.