Blockade of JAK2-mediated extrinsic survival signals restores sensitivity of CML cells to ABL inhibitors

Letter to the Editor

The introduction of imatinib just over ten years ago revolutionized CML therapy, making long-term survival a reality for the majority of newly diagnosed patients. Unfortunately, residual disease remains detectable in most patients, and even patients with undetectable BCR-ABL (complete molecular response) have a substantial risk of disease recurrence upon discontinuation of imatinib(1). Studies to understand the mechanism(s) underlying CML persistence in spite of ongoing, successful TKI-based disease management have mainly focused on cell-intrinsic mechanisms(2-4). However, there is compelling evidence that the microenvironment plays an important role in resistance to chemotherapy by providing critical survival cues to cells(5). Therefore, we hypothesized that extrinsic factors may also contribute to CML disease persistence despite effective inhibition of BCR-ABL kinase activity by imatinib and second generation ABL inhibitors.

To investigate these extrinsic survival factors in a more physiologic context, we used immortalized human bone marrow stromal cells as a model of the microenvironment(6). We found that conditioned media (CM) from HS-5, but not HS-23 or HS-27a, enhanced survival of CML cell lines in the presence of imatinib, dasatinib and nilotinib (Supplemental Figure 1), consistent with previous reports(7, 8). Furthermore, we found that co-culture of CML CD34⁺ cells with HS-5 cells, or culture in transwells over HS-5 cells, also attenuated the effects of imatinib on colony formation (Supplemental Figure 1). These data suggested that soluble factors may be responsible for HS-5-mediated protection from TKIs. We quantified the relative abundance of several cytokines in CM from the three human stromal cell lines to identify putative protective factors and found high concentrations of interleukin-6 (IL-6), interleukin-8 (IL-8), growth-related oncogene 1 (GRO1), monocyte chemotactic protein 1 (MCP-1), monocyte chemotactic protein 3 (MCP-3), granulocyte colony-stimulating factor (G-CSF) and granulocyte macrophage colony-stimulating factor (GM-CSF) in HS-5 CM compared to HS-23 and HS-27a CM (data not shown). Since several of these cytokines (G-CSF, GM-CSF, and IL-6) signal through JAK2(9), we used CYT387, a JAK1/JAK2 inhibitor, and TG101209, a relatively specific JAK2 inhibitor(10, 11), to interrogate the role of JAK2 in protection of CML cells by HS-5 CM.

We treated CML cell lines with 1 μM imatinib in regular media (RM) or CM, combined with increasing concentrations of each JAK2 inhibitor and found that inclusion of 5 μM CYT387 or 1 μM TG101209 abolished the anti-apoptotic effect of CM (Figure 1A,B and Supplemental Figure 2), while JAK2 inhibitors alone induced little apoptosis. To determine if JAK2 inhibition also abolished the protective effect of CM on primary cells, CML CD34⁺ cells were co-cultured with HS-5 cells or in serum-free media for four days with the indicated concentrations of imatinib, CYT387 or both inhibitors, followed by clonogenic assays. Consistent with our results in K562 cells, HS-5 co-culture significantly increased CFU-GM formation in the presence of imatinib (23±3.9% vs. 9±3.8%, p=0.018, Figure 1C). Inclusion of CYT387 in addition to imatinib attenuated this difference in a dose-dependent manner (Figure 1C). However, in contrast to the CML cell line data, CYT387 alone also reduced the number of CFU-GM colonies formed by CML CD34⁺ cells, and this effect was more pronounced in HS-5 co-cultures than in cytokine-free media (17±10.8% vs. 41±31.2%, p<0.01, Figure 2C), suggesting that primary cells respond more dramatically to JAK2 inhibition than CML cell lines, especially in the presence of stroma.

Figure 1. Protection by HS-5 CM is abrogated by JAK2 inhibition.

(A) K562 cells were treated with the indicated combinations of imatinib and TG101209, or (B) CYT387, for 60 hours in RM or CM and apoptosis was assessed by Annexin-V staining. (C) CML CD34⁺ cells were cultured in 24-well plates in media alone or in direct contact with irradiated HS-5 cells for four days, then removed and re-plated in Methocult for 15 days, and CFU-GM colonies counted. Results are normalized to the number of colonies from each untreated condition and combined results from six CML patient samples are shown (N=6). Error bars represent standard error of the mean. Statistically significant differences are represented by *, which signifies p<0.05 by paired t-test.

Figure 2. Combination of ABL and JAK2 inhibitors in an in vivo mouse model of CML reduces BCR-ABL⁺ cells but also normal hematopoietic cells.

(A) Kaplan-Meier curves show survival of mice treated with the indicated inhibitors. The table indicates inhibitor dosage and the number of mice per treatment group. One mouse in the high-dose combination cohort died from gavage and was removed from the survival analysis, *. (B, C) Box-and-whisker plots demonstrate spleen and liver weights at the time of autopsy. Mice treated with the high-dose combination were sacrificed for autopsy and analysis on day 27 due to concerns of toxicity. Mice treated with nilotinib monotherapy and low-dose combination were sacrificed for autopsy and analysis on day 42 (D) Representative histopathology of spleen, liver, and bone marrow at the time of autopsy. Hematoxylin and eosin (H&E) staining was performed on formalin fixed tissues. (E, F) Box-and-whisker plots show the percentage of GFP-positive cells by FACS in the spleen and bone marrow at the time of autopsy.

Study of downstream JAK2 signaling molecules identified phospho-STAT3 (pSTAT3) as a likely target of JAK2 (Supplemental Figures 3,4). Although we found that HS-5 CM increased phospho-STAT5 in LAMA-84, KBM-5 and CML progenitors, we found increased pSTAT3 in all CML cell lines and primary CD34⁺ progenitors tested (Supplemental Figure 4), suggesting a more broad role for STAT3, consistent with the data of others(7). Of note, imatinib treatment increased pSTAT3 in all CML cell lines tested and the amount of pSTAT3 was further enhanced in the presence of HS-5 CM, suggesting that inhibition of BCR-ABL induces a shift to an adaptive survival pathway that is substantially reinforced within the context of the microenvironment (Supplemental Figure 4). This adaptive JAK2-STAT3 survival pathway can be attenuated by addition of the JAK2 inhibitors CYT387 and TG101209, which then restores sensitivity to ABL inhibitors (Figure 1).

To test the effects of JAK2 and ABL inhibitors in vivo, we used a retroviral transduction/transplantation model of CML(12, 13). Given that imatinib has limited efficacy in this model, these studies were performed with 75 mg/kg/d nilotinib, a more potent ABL inhibitor(14). We used TG101209 because it is more selective for JAK2 compared to CYT387, at a maximum dose of 200 mg/kg/d based on previously published studies(10). Mice were divided into five cohorts: vehicle control, TG101209 monotherapy (200 mg/kg/d), nilotinib monotherapy (75 mg/kg/d), and nilotinib (75 mg/kg/day) combined with either low-dose (50 mg/kg/day) or high-dose (200 mg/kg/day) TG101209.

Vehicle-treated mice died within two days of initiating treatment, demonstrating the aggressive nature of this CML model(12, 13). Mice treated with TG101209 monotherapy demonstrated slightly prolonged survival (median survival of 20.5 days vs. 15.5 days for the control, p=0.065), but eventually all mice succumbed to leukemia (Figure 2A-D). The mice in the high-dose combination cohort initially fared well, however, two mice died unexpectedly on day 26 without any evidence of leukemia. Since this data suggested toxicity, the remainder of the high-dose combination cohort was terminated on day 27 for detailed autopsy. The nilotinib monotherapy and low-dose combination cohorts were continued on treatment until day 42. At this time, there were no significant differences in their peripheral blood counts or circulating GFP⁺ cells (data not shown), and the animals were sacrificed for detailed analysis.

Autopsy of untreated and TG101209 monotherapy cohorts revealed diffuse leukemic infiltration of the spleen and liver (Figure 2D). The nilotinib monotherapy and low-dose combination cohorts had smaller spleens (Figure 2B), with restoration of normal splenic architecture (Figure 2D). In contrast, the mice treated with the high-dose combination exhibited spleens that were nearly devoid of leukocytes and normal follicles (Figure 2D). Similarly, mice treated with the high-dose combination had reduced bone marrow cellularity compared to the other cohorts (Figure 2D). Mononuclear cells were isolated from all mice at time of death or sacrifice, and GFP⁺ cells from the spleen and bone marrow were measured by FACS to quantify residual disease. As shown in Figure 2E, there was a trend towards reduced splenic disease burden in the high-dose combination cohort when compared to nilotinib alone (4.1±4.1% GFP⁺ cells compared to 11.7±10.4% for nilotinib monotherapy, p=0.047). A less pronounced decrease of GFP⁺ cells was observed in the bone marrow (8.7±4.9% compared to 13.8±12% for nilotinib monotherapy, p=0.22).

JAK2 has long been implicated as an important signaling molecule in CML, although its exact role remains unclear(15-18). In our murine CML model, JAK2 inhibition alone only moderately prolonged survival, and all mice eventually succumbed to BCR-ABL⁺ myeloproliferative disease, suggesting that monotherapy merely delays disease progression. Combination treatment with nilotinib and high-dose TG101209 was more effective against BCR-ABL⁺ cells than nilotinib alone (Figure 2E), but also suppressed normal hematopoietic cells, suggesting that any synergistic effect of JAK2 and ABL inhibition on BCR-ABL⁺ cells may be nullified by toxic effects to non-leukemic cells at higher doses of JAK2 inhibitors (Figure 2 and similar results in non-leukemic control mice shown in Supplemental Figure 5). In contrast, combination treatments with a low-dose TG101209 component did not demonstrate any advantage over nilotinib alone (Figure 2), suggesting a very narrow therapeutic window for combination JAK2 and ABL inhibitor therapy. More extensive drug combinations of imatinib and CYT387 and TG101209 in vitro with normal and CML CD34⁺ cell colony formation also did not identify a combination that was able to preferentially suppress CML CD34⁺ cell colony formation over normal CD34⁺ cell colony formation (Supplemental Figure 6). Considered together, these in vitro and in vivo results suggest that combinations of JAK2 and ABL inhibitors may insufficiently discriminate between normal and CML cells, limiting their therapeutic use. Since primary CD34⁺ cells mainly reflect a progenitor population rather than true stem cells, it remains possible that a clinically relevant differential effect occurs in more primitive cells, although identifying the optimal pharmacokinetics and dosing will be challenging.

A number of other potential targets are currently being explored to eliminate CML disease persistence, including Wnt/β-catenin, Hedgehog and FOXO3a. However, like JAK2, these molecules are also utilized by normal HSCs, raising the possibility that similar problems may be encountered when combination therapy is attempted in vivo. However, despite their many commonalities, CML cells expressing pharmacologically inactivated BCR-ABL are not identical to normal cells and it is conceivable that BCR-ABL inhibition may render a previously redundant survival pathway essential, thereby generating a new CML-specific vulnerability that spares normal cells. Identification of such a pathway would provide a rational target to eliminate CML stem cells and eradicate disease.