The identification of JAK2/MPL mutations in patients with myeloproliferative neoplasms (MPN) has led to the clinical development of JAK kinase inhibitors, including ruxolitinib. Ruxolitinib reduces splenomegaly and systemic symptoms in myelofibrosis and improves overall survival; however, the mechanism by which JAK inhibitors achieve efficacy has not been delineated. Patients with MPN present with increased levels of circulating proinflammatory cytokines, which are mitigated by JAK inhibitor therapy. We sought to elucidate mechanisms by which JAK inhibitors attenuate cytokine-mediated pathophysiology. Single-cell profiling demonstrated that hematopoietic cells from myelofibrosis models and patient samples aberrantly secrete inflammatory cytokines. Pan-hematopoietic Stat3 deletion reduced disease severity and attenuated cytokine secretion, with similar efficacy as observed with ruxolitinib therapy. In contrast, Stat3 deletion restricted to MPN cells did not reduce disease severity or cytokine production. Consistent with these observations, we found that malignant and nonmalignant cells aberrantly secrete cytokines and JAK inhibition reduces cytokine production from both populations.
Significance: Our results demonstrate that JAK–STAT3-mediated cytokine production from malignant and nonmalignant cells contributes to MPN pathogenesis and that JAK inhibition in both populations is required for therapeutic efficacy. These findings provide novel insight into the mechanisms by which JAK kinase inhibition achieves therapeutic efficacy in MPNs. Cancer Discov; 5(3); 316–31. ©2015 AACR.
See related commentary by Belver and Ferrando, p. 234
This article is highlighted in the In This Issue feature, p. 213
Myeloproliferative neoplasms (MPN) are clonal hematopoietic disorders characterized by the proliferation of one or more myeloid lineages (1). Patients with MPN develop progressive splenomegaly, thrombosis, bleeding, and/or infection. Patients with MPN are also at cumulative risk to develop progressive bone marrow failure and/or transform to acute myeloid leukemia (AML), conditions which are associated with dismal clinical outcome (2). The discovery of cytokine-independent colony formation of MPN progenitors suggested that constitutive cytokine signaling contributes to MPN pathogenesis (3). The observation of somatic activating JAK2V617F mutations in patients with polycythemia vera (PV), essential thrombocythemia (ET), and primary myelofibrosis (PMF) provided the first insight into the molecular basis of MPN (4–7). Subsequent studies identified somatic JAK–STAT pathway mutations in JAK2V617F-negative MPN, most commonly in the CALR gene (8, 9) and the thrombopoietin receptor (MPLW515L; ref. 10). These data underscore the central importance of genetic alterations in the JAK–STAT signaling pathway in MPN pathogenesis.
The discovery of JAK2/MPL mutations in the majority of patients with MPN provided the rationale for the clinical development of JAK kinase inhibitors for patients with MPN and subsequently for other malignancies. Clinical studies with JAK kinase inhibitors have shown that these agents improve splenomegaly, systemic symptoms, and overall survival (11). On the basis of these data, ruxolitinib, a JAK1/JAK2 kinase inhibitor, was approved for patients with myelofibrosis, and several other JAK inhibitors are in late-phase clinical trials. Although JAK inhibitors offer substantial clinical benefit to patients with MPN, the mechanisms by which these agents achieve clinical efficacy have not been fully delineated. Patients with MPN have significantly elevated circulating levels of proinflammatory cytokines, and increased circulating cytokine levels are associated with adverse survival in myelofibrosis (12). It has been hypothesized that the cytokine-driven inflammatory state in MPN contributes to the constitutional symptoms, bone marrow fibrosis, extramedullary hematopoiesis, and disease progression characteristic of these myeloid malignancies. JAK inhibitor therapy with ruxolitinib is associated with a reduction in the level of proinflammatory cytokines (13); however, the role of aberrant cytokine production in myelofibrosis pathogenesis and in the response to JAK inhibitors remains to be delineated. We therefore sought to elucidate the role of aberrant cytokine production in MPN pathogenesis and in the response to JAK kinase inhibitors. We used a novel microfluidic single-cell profiling technique to examine the cytokine secretion profiles of myelofibrosis cells on a single-cell level and show that a significantly greater degree of multifunctionality and heterogeneity in cytokine production is a characteristic feature of myelofibrosis cells. Moreover, we show that JAK–STAT signaling in nonmutant hematopoietic cells contributes to MPN pathogenesis and that inhibition of JAK–STAT signaling in both mutant and nonmutant cells is required to reduce inflammatory signaling and to achieve clinical benefit in MPNs.
Proinflammatory Cytokines Are Elevated in Myelofibrosis Mice and Reversed with JAK1/2 Inhibitor Treatment
To identify cytokines that are altered in myelofibrosis, we measured the serum levels of 32 cytokines in the MPLW515L bone marrow transplant myelofibrosis model (14) using multiplex bead-based Luminex technology. We identified a set of inflammatory cytokines, including IL6, CXCL9, and CCL2, which are elevated in the serum of MPLW515L-mutant diseased mice (Fig. 1A), similar to the alterations in circulating cytokines observed in patients with myelofibrosis (12, 13). Ruxolitinib treatment (90 mg/kg, twice a day) normalized cytokine levels, consistent with the effects seen with chronic JAK inhibition in patients with MPN (Fig. 1A and Supplementary Fig. S1; ref. 13). Circulating cytokine levels were also elevated in myelofibrotic (6-month-old) Jak2V617F;Vav-Cre knockin mice (Fig. 1B; ref. 15), and ruxolitinib treatment (60 mg/kg, twice a day) normalized cytokine levels in mice transplanted with Jak2V617F-mutant cells (Fig. 1C). Short-term ruxolitinib treatment (3 doses, 90 mg/kg, twice a day) reduced cytokine production to a similar extent to that observed with 14 days of ruxolitinib treatment (90 mg/kg, twice a day; Fig. 1D), consistent with the rapid improvements in symptoms and splenomegaly seen with JAK inhibitor therapy (13) and with a direct effect of JAK kinase inhibition on cytokine secretion. The majority of cytokines (7 of 8) were also increased in the bone marrow supernatant (Fig. 1E) of MPLW515L-mutant diseased mice, suggesting that aberrant cytokine production in myelofibrosis is, at least in part, derived from bone marrow cells.
Bone Marrow Myelofibrosis Cells Feature Aberrant Cytokine Secretion Profiles
To evaluate whether bone marrow cells are the source of aberrant cytokine production in myelofibrosis, we optimized a microchip system that allowed us to perform multiplexed measurements of up to 15 secretory cytokines from primary murine and human bone marrow cells at single-cell resolution (Fig. 2A; refs. 16–18). Hierarchical cluster analysis of the single-cell secretomic profiles delineated multiple distinct populations that displayed heterogeneous secretion signatures and revealed marked differences between myelofibrosis and control bone marrow cells (Fig. 2A). We observed a significant increase in the fraction of cytokine-secreting cells and in the extent of cytokine secretion per single myelofibrosis cell (Fig. 2B). These data suggest the increased cytokine production in myelofibrosis results from increased per-cell cytokine secretion and from an increase in the fraction of cytokine-secreting cells (Fig. 2C). The proportion of cells secreting at least one cytokine was significantly higher in myelofibrosis mice compared with control mice (67.6% vs. 46.6%; P < 0.001; Fisher exact test). The majority of cytokine-secreting control bone marrow cells (73%) secreted less than two cytokines, consistent with physiologic secretion of one cytokine per cell. In contrast, we found that bone marrow cells from myelofibrosis mice were composed of a significantly elevated frequency of multifunctional cells that cosecrete multiple cytokines: 43.9% of cells from myelofibrosis mice secreted two or more cytokines and 15.6% of cells secreted four or more cytokines compared with 1.2% in control mice (Fig. 2D).
To determine which cytokines were most frequently cosecreted by myelofibrosis cells, we calculated a mutual exclusivity P value for each pair of secreted cytokines (Fig. 2E). CCL3, CCL4, and TNFα were commonly cosecreted in myelofibrosis cells, and secretion of these cytokines was inversely correlated with the secretion of other cytokines (Fig. 2F). We observed cosecretion of all other cytokine combinations (IL6, IL12, IL10, CCL2, CCL5, CXCL9, and CXCL1) in myelofibrosis cells, consistent with the presence of at least two distinct populations of cytokine-secreting cells in myelofibrosis. In line with these observations, principal component analysis (PCA) demonstrated that a large proportion of normal bone marrow cells did not secrete any protein and other smaller subsets with distinct secretion patterns produced only one or two cytokines, most notably, CCL3 and CCL4 (Fig. 2G and Supplementary Fig. S2). In contrast, PCA maps of myelofibrosis bone marrow cells showed multiple, large populations with heterogeneous secretion signatures. We then used the Simpson diversity index to quantify the extent of heterogeneity in cytokine secretion, and found that this index increased significantly from 0.68 in normal cells to 0.85 in myelofibrosis cells (P < 0.001), consistent with a marked increase in the heterogeneity of single-cell cytokine secretion in myelofibrosis.
Mature Myeloid and Progenitor Cells Contribute to Aberrant Cytokine Levels In Vivo
We next sought to define which hematopoietic compartments contribute to aberrant cytokine secretion in myelofibrosis. We used the single-cell platform to profile mature myeloid cells (CD11b/Gr1 double-positive) and megakaryocyte/erythroid progenitor (MEP) cells from myelofibrosis and control mice. Similar to unfractionated bone marrow cells, GFP-positive myelofibrosis myeloid cells and MEP cells had an increase in the fraction of cytokine-secreting cells and increased cytokine secretion per cell (Supplementary Fig. S3). PCA analysis showed that each population had distinct cytokine secretion profiles (Fig. 3A). IL6 and IL10 were mainly secreted by MEPs, and mature myeloid cells were the main source of TNFα and CCL3 (Fig. 3B and Supplementary Fig. S3). These data indicate that different cell types have distinct cytokine secretion profiles and potentially distinct roles in myelofibrosis pathogenesis.
Aberrant Single-Cell Cytokine Secretion in PMF Cells
We next performed single-cell profiling of circulating granulocytes from patients with myelofibrosis and healthy individuals (Supplementary Table S1). The average secretion level of different cytokines was significantly elevated in myelofibrosis cells, resulting from both an increase in the fraction of cells secreting specific cytokines and the secretion level of individual cytokines per cell (Fig. 3C and D). Six out of eight cytokines (IL6, IL10, IL12, TNFα, CCL2, and CCL5), which were aberrantly secreted in our murine myelofibrosis model, were also secreted at a much higher level by PMF patient cells. We observed that IL8 was most highly secreted by myelofibrosis granulocytes; previous studies have shown that increased serum IL8 levels are associated with adverse outcome in myelofibrosis (12). Immunohistochemical analysis confirmed increased IL8 expression in myelofibrosis bone marrow sections, whereas control bone marrow cells showed only weak expression of IL8 (Supplementary Fig. S4). The proportion of cells secreting at least one cytokine was significantly increased in human myelofibrosis cells compared with control cells (41.4% vs. 14.6%; P < 0.001; Fisher exact test; Fig. 3E), demonstrating similar patterns of cytokine secretion in patients with myelofibrosis and murine models. Altogether, our single-cell profiling studies highlight the striking heterogeneity and multifunctionality in cytokine secretion profile in myelofibrosis.
Deletion of Il6 from the Mouse System Shows Only Minor Effects on MplW515L-Driven Disease
Previous studies of cytokine signaling in cancer have largely focused on IL6 as a mediator of inflammation in leukemia and other malignancies (19). Given that we found a significant increase in IL6 secretion in both MPN models and myelofibrosis patient cells, we first investigated the effects of Il6 deletion on MPLW515L-induced myelofibrosis in vivo. We assessed the specific role of IL6 in myelofibrosis in vivo by transplanting IL6-deficient myelofibrosis cells into IL6-deficient recipients. Il6 deletion led to a modest reduction in white blood cell (WBC) count, but did not reduce spleen weight or the extent of myeloproliferation in vivo (Supplementary Fig. S5A and S5B). No significant differences were observed in the proportion of CD11b/Gr1 double-positive cells in the target organs (Supplementary Fig. S5C and S5D). Consistent with these data, we found that the serum levels of other proinflammatory cytokines remained elevated in IL6-deficient myelofibrosis mice, and we observed compensatory increases in IL4 and IL5 serum levels in the absence of IL6 production (Supplementary Fig. S5E and S5F). Although the specific role of other cytokines in myelofibrosis requires further investigation, these data suggest that inflammatory signaling in MPN is driven by multiple cytokines, and that inhibiting secretion or signaling of an individual cytokine cannot attenuate the cytokine-signaling loop contributing to MPN pathogenesis.
Deletion of Stat3 Reduces Cytokine Production and Ameliorates Myelofibrosis In Vivo
We next aimed to investigate whether there are specific signaling pathways driving cytokine production in vivo in myelofibrosis. First, to better characterize oncogenic signaling pathways activated by JAK2/MPL alleles in vitro and identify potential signaling effectors, we generated isogenic cell lines expressing the most common JAK2 and MPL mutations observed in patients with MPN. Notably, Ba/F3 cells expressing MPLW515L or MPLW515L–JAK2V617F showed evidence of constitutive STAT3 activation (Supplementary Fig. S6). We also observed increased STAT3 phosphorylation in splenocytes from MPLW515L-diseased mice and in the bone marrow of patients with PMF (Supplementary Fig. S7). JAK inhibitor therapy reduced STAT3 phosphorylation in vivo (Fig. 4A).
Previous studies have shown that STAT3 signaling contributes to cytokine production in different malignant and inflammatory contexts (20); however, the role of STAT3 signaling in MPN inflammatory signaling and pathogenesis has not yet been elucidated. We thus performed MPLW515L bone marrow transplantations using hematopoietic-specific conditional Stat3-knockout mice or respective littermate controls as donors (Supplementary Fig. S8). Stat3 deletion improved survival, reduced disease severity, and reduced cytokine-mediated inflammation, similar to the effects observed with ruxolitinib therapy. Stat3 deletion resulted in lower white blood counts, lower spleen weights, and a reduced degree of reticulin fibrosis (Fig. 4B–F and Supplementary Fig. S6), and decreased the proportion of MEP cells in spleen and bone marrow (Fig. 4G). Similar effects of Stat3 deletion were seen with Mx-Cre and Vav-Cre, and with somatic deletion of Stat3 after bone marrow transplantation (Supplementary Fig. S9). Most importantly, hematopoietic specific Stat3 deletion normalized circulating cytokine levels (Fig. 4H), with similar reductions in cytokine levels as with ruxolitinib therapy. These data demonstrate that STAT3 activation is required for cytokine production in myelofibrosis, and Stat3 deletion phenocopies the effects of ruxolitinib on cytokine production and on disease sequelae in vivo.
Mutant-Specific Stat3 Deletion Does Not Reduce Inflammatory Signaling or Myelofibrosis In Vivo
It is currently not known if JAK inhibitors achieve clinical benefit solely from target inhibition in MPN cells, or if JAK–STAT inhibition in nonmalignant cells contributes to therapeutic efficacy. We therefore examined whether restricting Stat3 deletion to mutant hematopoietic cells would result in similar effects as pan-hematopoietic Stat3 deletion. We transplanted lethally irradiated CD45.1-positive recipient mice with MPLW515L-positive, Stat3-deleted bone marrow (CD45.2-positive) along with Stat3 wild-type support marrow (CD45.1-positive; Fig. 5A and B). In contrast to the significant effects seen with complete hematopoietic Stat3 deletion, MPN-specific Stat3 deletion did not reduce disease severity (Fig. 5C and D and Supplementary Fig. S10). Consistent with these data, MPN-specific Stat3 deletion did not significantly attenuate cytokine production, in contrast to the effects observed with pan-hematopoietic Stat3 deletion (Fig. 5E). These data are consistent with a requirement for STAT3 signaling in both malignant and nonmalignant hematopoietic cells in myelofibrosis.
Cytokine Production in Myelofibrosis Originates in Malignant and Nonmalignant Hematopoietic Cells
The genetic studies described above suggested that aberrant cytokine production in myelofibrosis emanates from malignant and nonmalignant hematopoietic cells. We therefore analyzed cytokine mRNA expression in sorted MPLW515L-mutant (GFP-positive) and wild-type (GFP-negative) bone marrow cells using NanoString technology. We found that MPLW515L-mutant myelofibrosis cells express high mRNA levels of a subset of inflammatory cytokines, including Il6, consistent with tumor-derived cytokine production (Fig. 6A). In contrast, we found that some cytokines, including Ccl2 and Tnfa, were derived from both GFP-positive and GFP-negative populations, and other cytokines, including Il12 and Cxcl9, were largely derived from nonmutant cells (Fig. 6A).
We then performed single-cell analysis of sorted mutant and wild-type subpopulations to further delineate the contribution of malignant and nonmalignant cells to cytokine production. Single-cell analysis of lineage-positive (mature) and lineage-negative [hematopoeitic stem/progenitor (HSPC)] cells revealed that GFP-positive and GFP-negative cells from mature and HSPC compartments aberrantly secrete cytokines in myelofibrosis (Fig. 6B and C). We observed an increase in cytokine secretion and in the degree of single-cell cytokine secretion heterogeneity in both GFP-positive and GFP-negative cells from stem/progenitor and differentiated cells compared with control bone marrow cells (Supplementary Fig. S11). MPLW515L-mutant HSPCs were the largest source of IL6, whereas nonmutant mature cells were the primary source of CXCL9 (Fig. 6B). Mutant and wild-type cells both secreted TNFα, IL10, and CCL2 consistent with production by multiple populations (Fig. 6C). We observed an increase in total cytokine production and an increase in the proportion of cells which secreted multiple cytokines in GFP-positive/negative mature and GFP-positive HSPC from myelofibrosis mice (Fig. 6D and Supplementary Figs. S11 and S12). Importantly, ruxolitinib treatment normalized cytokine expression from GFP-positive and GFP-negative cell populations (Fig. 6E), demonstrating that JAK inhibition reduces cytokine production from both tumor and nontumor populations in vivo.
We next sought to extend our findings to JAK2V617F-mutant MPN. We transplanted whole bone marrow of Jak2V617F;Vav-Cre knockin mice (CD45.2) with CD45.1 wild-type support marrow into CD45.1-recipient mice (MPN mice). After all mice were engrafted with JAK2V617F-positive disease, we performed single-cell cytokine analysis on sorted CD45.2 and CD45.1 cells to elucidate cytokine secretion in mutant and nonmutant cells (Fig. 7A). As a control, we transplanted wild-type CD45.2 bone marrow with CD45.1 support cells into CD45.1-recipient mice (control mice). Single-cell cytokine profiling revealed marked differences in cytokine secretion in bone marrow cells derived from MPN and control mice (Supplementary Fig. S13). Most importantly, this included aberrant cytokine production in CD45.2-mutant and CD45.1-nonmutant cells from MPN mice. We observed an increase in the fraction of cytokine-secreting cells and in the extent of cytokine secretion per single cell in mutant and nonmutant hematopoietic populations from Jak2-mutant MPN mice (Fig. 7B and C and Supplementary Fig. S14). The proportion of cells secreting at least one cytokine was significantly higher within the wild-type population (CD45.1) of myelofibrosis mice compared with the CD45.1 population of control mice (47.1% vs. 27%; P < 0.001; Fisher exact test; Fig. 7D). Similarly, 65% of CD45.2-positive JAK2V617F-expressing cells from myelofibrosis mice secreted at least one cytokine compared with only 28.5% of control CD45.2 cells (P < 0.001; Fisher exact test). Furthermore, Jak2V617F-negative CD45.1-positive and Jak2V617F-positive CD45.2-positive bone marrow cells from myelofibrosis exhibited a marked increase in the proportion of multifunctional cells that cosecrete multiple cytokines (Fig. 7D). Quantification of the extent of heterogeneity using the Simpson diversity index showed that nonmutant CD45.1-positive and Jak2V617F-positive, CD45.2-positive cells of diseased mice featured a significant increase in the heterogeneity of single-cell secretion (CD45.1: 0.64 vs. 0.43, CD45.2: 0.74 vs. 0.44). Consistent with these data and with our data in the MPLW515L model, ruxolitinib treatment normalized cytokine expression from CD45.1-negative and CD45.2-positive cell populations in the Jak2V617F knockin transplantation model (Supplementary Fig. S15). We next sought to investigate whether mutant and nonmutant cells in PMF could contribute to aberrant cytokine secretion. We identified a patient with PMF with a JAK2V617F-mutant allele burden of 38%, consistent with partial involvement of the hematopoietic compartment of JAK2V617F-mutant cells (Supplementary Table S2). Notably, in this patient we found that 59% of the hematopoietic cells secreted one or more cytokines from our measurement (Fig. 7E). Moreover, the frequency of multifunctional cells secreting two or more cytokines was significantly elevated in the patient (Fig. 7F). The proportion of cytokine-secreting cells was greater than expected on the basis of JAK2V617F-mutant allele burden, suggesting that in this patient, a subset of JAK2V617F wild-type has to contribute to cytokine production. Taken together, these data illustrate the heterogeneity of aberrant cytokine secretion in MPN models and patients, and that aberrant inflammatory signaling from more than one population contributes to MPN pathogenesis and therapeutic response.
Although MPNs present as chronic myeloid malignancies, the quality of life and overall survival of patients with PMF is more similar to advanced malignancies, including metastatic epithelial tumors. Patients with PMF have markedly elevated levels of proinflammatory cytokines, which are thought to contribute to PMF-associated symptoms and sequelae (12, 13). Clinical data show that JAK inhibition reduces constitutional symptoms and splenomegaly in patients with PMF and post–PV/ET myelofibrosis, concurrent with a reduction in circulating cytokine levels (13). Although the prognostic value of circulating cytokine levels in myelofibrosis has been previously demonstrated (12), the mechanisms that govern aberrant cytokine production in myelofibrosis and the source of aberrant cytokines in myelofibrosis have not previously been demonstrated. Here, we show that JAK–STAT activation in malignant and nonmalignant cells contributes to myelofibrosis pathogenesis, and that cytokine production by both populations is an important feature of myelofibrosis.
In this study, we demonstrate that JAK1/2 inhibition leads to a rapid, potent reduction in serum cytokine levels, consistent with the rapid clinical benefits seen with JAK inhibitor therapy, and demonstrate that this is a direct effect of JAK kinase inhibition on cytokine production. Single-cell cytokine analysis showed that hematopoietic cells from myelofibrosis models and patients with PMF aberrantly secrete a spectrum of inflammatory cytokines. We performed multiplex, highly sensitive measurements of cytokine secretion from over a thousand captured, viable single cells. The results demonstrated that hematopoietic cells in myelofibrosis show significant upregulation of a spectrum of proinflammatory cytokines, elevation of cellular heterogeneity in cytokine secretion, and increased multifunctional cytokine production, which are not observed in normal hematopoietic cells. Our data on a specific lineage, myeloid, demonstrated that both mature and progenitor myeloid cells contribute to increased cytokine production, and more interestingly that they show distinct cytokine profiles, suggesting their different roles in myelofibrosis pathogenesis. Future studies will delineate the role of additional cytokines in inflammatory signaling in MPN, and use single-cell profiling to analyze nonhematopoietic cell types and previously unexplored hematopoietic lineages, such as lymphoid cells and purified stem cells, in MPN models/patient samples and in other malignancies to determine whether multifunctional cytokine secretion is observed in other malignant contexts.
Many studies have identified JAK2 mutations in patients with MPN and in other malignancies, and several known JAK2 signaling mediators have been linked to MPN disease manifestation and progression (21–23); however, the role of STAT3 signaling in MPN pathogenesis and in inflammatory signaling in myeloid malignancies has not been previously delineated. STAT3 represents a key link between cancer and inflammation, and as such provides an ideal candidate signaling effector driving cytokine production in vivo. In this study, we demonstrate a critical role for STAT3 in inflammatory cytokine production in myelofibrosis. Pan-hematopoietic Stat3 deletion improved survival, reduced disease severity, and reduced cytokine secretion, with similar efficacy as observed with ruxolitinib therapy. In contrast, restricting loss of Stat3 to the malignant clone did not reduce disease severity or cytokine production, demonstrating that STAT3 signaling must be inhibited in malignant and nonmalignant cells to achieve clinical efficacy. Consistent with these findings, we discovered that malignant and nonmalignant cells aberrantly secrete inflammatory cytokines and that JAK inhibition reduces cytokine production from both populations. Our results demonstrate that JAK–STAT3-mediated cytokine production from malignant and nonmalignant cells contributes to MPN pathogenesis and that JAK inhibition in both populations is required for therapeutic efficacy. These data reveal an important, unexpected mechanism of action of JAK inhibition in MPN, such that the target must be inhibited in tumor and nontumor cells to achieve clinical benefit.
Recent studies in MPN models and in other hematologic malignancies have shown that additional, nonhematopoietic populations can influence malignant transformation, including nestin-positive stromal cells, osteoblasts, and other cell types in the hematopoietic niche (24–26). The role of cytokine secretion from these additional populations, particularly at single-cell resolution, has not been explored to date. We believe that single-cell cytokine profiling can be used to elucidate the specific role of different niche populations in normal and malignant hematopoiesis and to assess the impact of JAK inhibitor therapy on the nonhematopoietic microenvironment in model systems and in primary patient samples, which should be the subject of subsequent investigations into cytokine secretion in different malignant states.
Taken together, our data underscore the critical role of aberrant cytokine signaling mediated by STAT3 activation in MPN pathogenesis. Most importantly, our studies support the notion that JAK kinase inhibition in malignant and nonmalignant cells is required to achieve clinical efficacy in myelofibrosis. Recent data have suggested that ruxolitinib improves overall survival in patients with pancreatic cancer with evidence of systemic inflammation. As such, inhibition of cytokine signaling in malignant and nonmalignant cells might offer clinical benefit in other malignancies characterized by aberrant inflammatory signaling. We hypothesize that JAK inhibition may have a broader role in cancer therapy, and that this therapeutic approach may improve outcomes for patients with malignancies characterized by systemic inflammation.
Patient studies were conducted in accordance with the Declaration of Helsinki. Primary patient samples were collected from patients with myelofibrosis and control healthy donors under Memorial Sloan Kettering Cancer Center (MSKCC; New York, NY) Institutional Review Board Protocol 09-141. Ficoll technique was applied to isolate granulocytes from the peripheral blood of patients with myelofibrosis and healthy control individuals. JAK2 allele burden for patients with myelofibrosis was calculated using targeted sequencing approaches (RainDance, MiSeq).
Reagents and Transgenic Mice
Ruxolitinib was provided by Incyte/Novartis and formulated for administration by oral gavage as previously described (27). Conditional Stat3 knockout, Jak2V617F knockin, and germline Il6 knockout mice have been described previously (15, 28). Floxed mice were crossed to the interferon-responsive Mx-Cre and Vav-Cre deleter lines. Congenic mice (CD45.1) used as recipients in transplant studies were purchased from The Jackson Laboratory. Antibodies used for Western blotting included phosphorylated and total JAK2, STAT3, STAT5, and MAPK (Cell Signaling Technology). Mutant (GFP-positive) and wild-type (GFP-negative) populations were separated by FACS using GFP and/or specific cell-surface marker: phycoerythrin (PE)-conjugated CD11b, allophycocyanin (APC)-conjugated Gr1, TER119 (APC-CY7), CD71 (PE-CY7), PE-conjugated CD117, Pacific Blue (PacBlue)–conjugated CD16/CD32, phycoerythrin-Cy7 (PeCy7)–conjugated Sca-1, and APC-conjugated CD34. All plasmids used in this study have been previously described (29). Lineage-positive (LIN+) and lineage-negative (LIN−) subpopulations were separated using a panel of APC-CY7–conjugated antibodies that recognize all mature hematopoietic lineages (CD4, CD19, CD11b, Gr1, TER119, CD3, B220, and NK1.1).
Bone Marrow Transplantation Model
All animal experiments were performed in accordance with our MSKCC Institutional Animal Care and Use Committee–approved animal protocol. Animal care was in strict compliance with institutional guidelines established by the MSKCC, the Guide for the Care and Use of Laboratory Animals (National Academy of Sciences, 1996), and the Association for Assessment and Accreditation of Laboratory Animal Care International. MPLW515L bone marrow transplantation experiments were performed as described previously (14). Briefly, prestimulated c-KIT enriched bone marrow cells were subjected to two rounds of cosedimentation with viral supernatant containing MSCV-hMplW515L-IRES-GFP or empty MSCV-IRES-GFP control vector. A total of 1 × 106 cells (∼25%–40% GFP-positive, MPL-expressing cells) were injected into the tail veins of lethally irradiated syngeneic mice. For secondary Jak2V617F;Vav-Cre bone marrow transplantations, 5 × 106 bone marrow cells harvested from sick Jak2V617F knockin mice were injected into the tail veins of 8-week-old syngeneic mice (15). Engrafted mice were monitored daily for signs of illness, and nonlethal bleeds were performed biweekly to follow up disease progression. Mice showing signs of being moribund, having more than 10% weight loss, or having palpable splenomegaly extending across the midline were sacrificed. Spleens were removed and weighed, and single-cell suspensions were prepared for subsequent cell staining and fractionation, Western blot analysis, or histopathologic analysis. Peripheral blood chimerism was routinely determined 14 days after bone marrow transplantation. Stat3 excision was confirmed by Western blot analysis and quantitative real-time PCR analysis on red cell lysed peripheral blood samples.
At first signs of disease, mice were randomized to begin 14 or 21 days of treatment with the JAK1/2 inhibitor ruxolitinib [90 mg/kg (Balb/c, MplW515L) or 60 mg/kg (C57/BL6, secondary Jak2V617F recipient mice), p.o., twice a day] or vehicle followed by measurement of plasma cytokine levels. Mice were ranked on the basis of baseline WBC count and assigned to treatment groups to achieve congruent WBC profiles. Serum was collected 4 hours after last drug administration. For kinetic studies, mice assigned for short-term drug treatment received only three doses, which were administered simultaneously with the last doses of the long-term treatment arm.
Luminex assays were carried out using the FlexMAP 3D multiplexing platform (Luminex xMAP system). Plasma levels of 32 cytokines were measured simultaneously using the Millipore Mouse Cytokine 32-plex kit. Serum samples were prepared following manufacturer's instructions. xPONENT (Luminex) and Milliplex Analyst Software (Millipore) was applied to convert mean fluorescent intensity values into molecular concentrations by the use of a standard curve (5-parameter logistic fitting method).
Femur and spleen samples were fixed in 4% paraformaldehyde overnight and then embedded in paraffin as previously described (30). Paraffin sections were cut on a rotary microtome (3-μm; Mikrom International AG), spread in a 45°C water bath, mounted on microscope slides (Thermo Scientific), and air-dried in an oven at 37°C overnight. After drying, tissue section slides were processed either automatically for hematoxylin and eosin (H&E) staining (COT20 stainer, Medite) or manually for reticulum staining with silver impregnation method kit (Bio-Optica). Slides were scanned with Scanscope XT slide scanner (Aperio). Pictures were taken at a 20× (H&E and reticulin) magnification by using Imascope viewer (Aperio). For the expression analysis of IL8 and phospho-STAT3, 5-μm unstained sections from cases of bone marrow samples were stained on the Ventana Discovery XT per the manufacturer's instructions. The following antibodies were used: goat anti-IL8 antibody from R&D Systems (1:40 dilution), monoclonal rabbit anti–phospho-STAT3 antibody from Cell Signaling Technology (1:50 dilution). Photomicrographs were taken using an Olympus BX41 with DP20 software.
Single-Cell Cytokine Secretion Assay
A high-density antibody barcode was created using a microchannel-guided flow patterning technique as previously described (16, 17). A polydimethylsiloxane microchip with 20 microfluidic channels for flow patterning was fabricated using a soft lithography technique. After the microchip was bonded with a poly-l-lysine glass slide, the capture antibody for each cytokine was injected into a microfluidic channel and flowed through the chip via air pressure (1.8 psi, 12 hours). Antibodies were then immobilized on the glass slide to form the antibody barcode. The subnanoliter microchamber array for single-cell capture is a one-layer polydimethylsiloxane slab fabricated using a soft-lithography method and contains 7,700 microchambers (16, 31). A single-cell suspension of bone marrow cells was prepared in RPMI-1640 supplemented with 20% FBS, IL3, SCF, and IL6 at a density of 4 × 105 cells/mL. After blocking with 3% BSA/PBS (2 hours), 200 μL of the single-cell suspension was pipetted onto the chip and allowed to settle (5 minutes). Normally approximately 25% of the microchambers contained single cells based on the Poisson distribution (16). The antibody barcode glass slide was then put on top of the microchambers. Assembled microchamber arrays were then imaged (Nikon Eclipse Ti) to determine the number of cells in each microchamber by using a custom-made image-processing algorithm. After incubation (24 hours, 37°C in 5% CO2), the cytokine signals from antibody barcode chips were developed via an immuno-sandwich ELISA assay. The slide was scanned with GenePix 4200A (Axon) microarray scanner to collect fluorescent signals. All fluorescent scanned images were processed with GenePix software and Excel macro to acquire average fluorescent signals from each microchamber for all bars in each antibody barcode set, and the fluorescent signals were matched with the number of cells. Single-cell data were gated on the basis of the background signals to distinguish between cytokine producers and nonproducers. The data were further transformed and normalized to perform multidimensional data analysis as previously described (17, 32–34).
Data Processing and Analysis of Single-Cell Cytokine Assays
For all single-cell secretomic analysis, our custom-built algorithms in Excel, MATLAB, and R-packages were used to process, analyze, and visualize the single-cell cytokine profiles. Briefly, the predetermined cell numbers of the microchambers were matched with fluorescence intensity values for each measured protein from corresponding microchambers. The background signals from all zero-cell (empty) microchambers were used to determine the gate to detect cytokine-producing cells and nonproducers. The gating threshold was calculated by (average intensity of all empty microchambers for a given cytokine) + 2 × (standard deviation of all empty microchambers for a given cytokine) (34). Any cells with fluorescence intensity of a specific cytokine below the gating threshold were considered nonproducers, and their cytokine intensity values were converted to 0. Any cells with fluorescence intensity above the gating threshold were considered cytokine-producers and given the cytokine intensity value of (the measured fluorescence intensity − the average fluorescence intensity of all zero-cell microchambers for a given cytokine). The gated and background-subtracted cytokine intensity values were log-transformed. To eliminate variability in detection sensitivity and profile the secretion of all assayed proteins, each protein was normalized by the average intensity and the standard deviation. The number of all cells above the gates and the sum of fluorescence intensity of a given cytokine were measured to calculate the frequency of cytokine producers and the average cytokine secretion per single cell. For the single-cell analysis of unsorted, whole bone marrow compartment, the data were processed without log transformation. By not conducting log transformation, which significantly suppresses the data range, we could distinguish smaller subsets with unique cytokine profiles from the more heterogeneous and larger unsorted bone marrow cell population. Our statistical analysis and graphical representations of the single-cell secretomic profiles are based on the methods routinely used to analyze large-scale high-dimensional datasets from numerous cell subsets (17, 32–34).
Principal Component, Mutual Exclusivity, and Statistical Analysis
All analyses were performed in R. The PCA was performed using the prcomp R function and visualized with the ggplot2 package. The first two principal components explained 60.1% of the variance. PC3 and PC4 (cumulative proportion of variance 78%) were, similar to PC1 and PC2, mainly defined by secretion of TNFα, CCL3, CCL4, CXCL1, and CCL5 (Supplementary Fig. S1A). Because PC3 to PC4 covered essentially the same cytokines as PC1 and PC2, we focused on the visualization of the first two PCs. PC5 (83%) was dominated by IL10 secretion. We next tested whether two cytokines were less (mutual exclusivity) or more frequently cosecreted than expected by chance. To this end, we created for each pair of cytokines a 2 × 2 contingency table (secreted only in cytokine A; only in B; in both A and B; in neither A nor B) and obtained a P value with the Fisher exact test. P values were adjusted for multiple testing using the Benjamini–Hochberg correction and a false discovery rate (FDR) < 0.05 was considered significant. FDRs were visualized in a heat map, with pairs of cytokines displaying significant cosecretion colored in red, and mutually exclusive pairs in light yellow. Pairs with cosecretion patterns not significantly different from randomly expected patterns were shown in gray. We further tested whether pairs of cytokines are more or less frequently cosecreted in myelofibrosis mice compared with control mice. We again used a Fisher exact test to obtain P values for the corresponding 2 × 2 contingency table (number of cosecreted cells in myelofibrosis, number of non-cosecreted cells in myelofibrosis, number of cosecreted cells in control mice, number of non-cosecreted cells in control mice). Only cells with at least two secreted cytokines were considered in this mutual exclusivity analysis.
The Student t test was used to compare the mean of two groups. Normality tests were used to test the assumption of a normal distribution. Graphs represent mean values ± SEM. Kaplan–Meier survival analysis and the log-rank test were used to compare survival outcomes between the two groups.
Disclosure of Potential Conflicts of Interest
M. Riester is a full-time employee of Novartis. J. Bromberg reports receiving a commercial research grant from AstraZeneca. M. Murakami is an investigator at Novartis and has ownership interest (including patents) in the same. R. Fan has ownership interest (including patents) in IsoPlexis. R.L. Levine reports receiving a commercial research grant from Incyte. No potential conflicts of interest were disclosed by the other authors
Conception and design: M. Kleppe, R.L. Levine
Development of methodology: M. Kleppe, M. Kwak, S. Marubayashi, V. Romanet, R. Fan
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Kleppe, M. Kwak, P. Koppikar, M. Keller, L. Bastian, N. Bhagwat, E. Papalexi, O. Abdel-Wahab, R. Rampal, S. Marubayashi, J.J. Chen, J. Bromberg, J. Teruya-Feldstein
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Kleppe, M. Kwak, P. Koppikar, M. Riester, L. Bastian, T. Hricik, O. Abdel-Wahab, J.J. Chen, J.S. Fridman, M. Murakami, F. Michor, R. Fan, R.L. Levine
Writing, review, and/or revision of the manuscript: M. Kleppe, M. Kwak, M. Riester, M. Keller, O. Abdel-Wahab, S. Marubayashi, J.S. Fridman, J. Teruya-Feldstein, M. Murakami, T. Radimerski, F. Michor, R. Fan, R.L. Levine
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Kleppe, M. Kwak, M. Keller, T. Hricik, A.S. McKenney, J.J. Chen, J.S. Fridman
Study supervision: F. Michor, R. Fan, R.L. Levine
The authors thank the members of the Levine and Fan laboratories, Neal Rosen, and Scott Armstrong for helpful comments and discussion. The authors also thank Benjamin L. Ebert (Harvard Medical School, Boston, MA) and Ann Mullally (Harvard Medical School) for providing Jak2V617F knockin mice. The authors are grateful to the MSK Kravis Center for Molecular Oncology/Geoffrey Beene Core for their assistance with NanoString and targeted sequencing analysis and Irina Linkov (IHC Pathology Core) and Janine Pichardo for technical assistance with pathologic experiments. The authors also acknowledge support from the Yale Institute for Nanoscience and Quantum Engineering (YINQE) and the Yale Nanofabrication Center for allowing use of their microfabrication facilities.
This work was supported by the Dana-Farber Cancer Institute's Physical Sciences Oncology Center (U54CA143798-01), to R.L. Levine, R. Fan, M. Riester, and F. Michor; NCI 1R01CA151949-01 to R.L. Levine; and NCI 1R01CA138234-01 to F. Michor and R.L. Levine, and partly by NIDDK 1U01DK104331-01 to R. Fan and R.L. Levine. M. Kleppe is a fellow of the Leukemia and Lymphoma Society and was previously supported by an EMBO Long-Term Fellowship. A.S. McKenney was supported by a Medical Scientist Training Program grant from the National Institute of General Medical Sciences of the NIH under award number T32GM007739 to the Weill Cornell/Rockefeller/Sloan Kettering Tri-Institutional MD–PhD Program.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.