STAT3 Activation as a Predictive Biomarker for Ruxolitinib Response in Head and Neck Cancer

STAT3, in its active, phosphorylated form, is a transcription factor regulating genes that mediate cell-cycle progression, cellular proliferation and survival, and epithelial-to-mesenchymal transition (1). Overexpression and/or hyperactivation of this oncogene is common in head and neck squamous cell carcinoma (HNSCC) and other solid tumors and has been associated with progression and poor prognosis (2–5). Transcription factors, including STAT3, remain challenging therapeutic targets. Blockade of upstream receptor and non-receptor kinases, such as JAKs, represents a feasible approach to abrogate STAT3 activation in tumors (3, 6).

To date, three orally bioavailable JAK inhibitors have been FDA-approved: (i) the ATP-competitive JAK1/2 inhibitor ruxolitinib for myeloproliferative diseases and graft versus host disease (ii), the JAK2 inhibitor fedratinib for myeloproliferative diseases, and (7) the JAK3 inhibitor tofacitinib for rheumatoid arthritis and inflammatory bowel disease (8–11). In hematological malignancies such as polycythemia vera and myelofibrosis, an activating JAK2 V617F mutation drives hyperactivation of STAT3. Ruxolitinib and, more recently, fedratinib have been shown to decrease STAT3 activation and are viable treatment options for these conditions (12–17). However, the impact of JAK inhibition on solid tumor growth is incompletely understood.

Promising preclinical studies showed that JAK inhibition with the small molecule, ATP-competitive inhibitor AZD1480 decreased STAT3 activation in a variety of solid tumors (18). We previously reported that JAK inhibition by AZD1480 decreased STAT3 activation in HNSCC cell lines and suppressed tumor growth in patient-derived xenografts (PDX; ref. 19). However, clinical trials of AZD1480 were discontinued because of neurotoxicity (20). Ruxolitinib, a well-tolerated JAK1/2 inhibitor, has also been explored in preclinical solid tumor models. Treatment with ruxolitinib decreased levels of phosphorylated STAT3 and inhibited cellular proliferation and survival of hepatocellular carcinoma, tamoxifen-resistant breast cancer, esophageal squamous cell carcinoma, pancreatic adenocarcinoma, colon cancer, and lung adenocarcinoma cell lines and mouse models (21–26).

Studies of ruxolitinib in patients with advanced lung, pancreatic, breast, or colon cancers suggest that this agent is well tolerated, but there are no reports of any significant improvement in overall survival or progression-free survival (7, 27–31). The impact of ruxolitinib has not been reported to date in patients with HNSCC.

In the present study, we investigated the impact of ruxolitinib in HNSCC preclinical models, including PDXs, generated from HNSCC tumors resected from patients enrolled on a window-of-opportunity trial (NCT03153982) of ruxolitinib to assess whether these models can serve as avatars of clinical response to treatment. Ruxolitinib consistently inhibited STAT3 activation, cell growth, and colony formation in HNSCC cell lines. In vivo, ruxolitinib treatment inhibited tumor growth in a PDX derived from a patient who exhibited a clinical response in conjunction with decreased STAT3 activation, but not in a PDX from a patient whose disease progressed on ruxolitinib therapy and where STAT3 activation levels were not inhibited by treatment. In addition, baseline levels of pSTAT3 and total STAT3 were higher in the PDX developed from the patient whose disease progressed on therapy. Collectively, our findings indicate that ruxolitinib abrogates JAK/STAT signaling in HNSCC preclinical models, suggesting that ruxolitinib treatment may be effective in a subset of patients with HNSCC. Assessment of baseline pSTAT3 and total STAT3 levels may be useful for identifying the subset of patients most likely to respond to ruxolitinib.

CAL-33, CAL 27 and FaDu cells were purchased from the ATCC. Cells were cultured in DMEM, 10% FBS and 1% penicillin–streptomycin (P/S), and maintained at 37°C with 5% CO₂. All cell lines were authenticated via Short Tandem Repeat testing using Genetica DNA Laboratories.

The primary antibodies against GAPDH (#5174S), pSTAT3 (#9145S), and total STAT3 (#4904S) were from Cell Signaling Technology. Goat anti-rabbit IgG (H + L)-horseradish peroxidase–conjugate antibody (#1706515) and Protein Assay Dye Reagent Concentrate (#5000006) were obtained from Bio-Rad. PhosSTOP phosphatase inhibitor cocktail tablet (#04906837001) and complete protease inhibitor cocktail tablets (#11697498007) were purchased from Roche Diagnostics (Risch-Rotkreuz). Luminol reagent (#SC-2048) was purchased from Santa Cruz Biotechnology Inc. SuperSignal ELISA Femto Maximum (#37075) was purchased from Thermo Fisher Scientific Pierce. Thiazolyl blue tetrazolium (MTT; #M5655) was purchased from Sigma-Aldrich. One gram of MTT was diluted in 500 mL of PBS and filtered to make a stock solution, then stored at 4°C. Ruxolitinib was provided by Incyte in powder form under a Material Transfer Agreement. For cell line experiments, ruxolitinib was dissolved in DMSO to generate a 20 mmol/L stock solution, aliquoted and stored at −20°C. For in vivo experiments, ruxolitinib was dissolved in vehicle (0.5% methyl cellulose and 0.1% tween-80) and stored for up to one week at 4°C.

Cells were plated at 3.0×10⁵ cells/well in six-well plates and treated the following day with ruxolitinib for 24 hours. The treated cells were harvested using RIPA buffer (150 mmol/L NaCl, 5 mmol/L EDTA, 50 mmol/L Tris pH 8.0, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS) with phosSTOP phosphatase inhibitor cocktail and cOmplete protease inhibitor cocktail. Proteins (40 μg/lane) were electrophoresed on 10% SDS/polyacrylamide gels and transferred to polyvinylidene difluoride membranes, blocked for one hour at room temperature in TBS containing 0.1% Tween-20 (TBST) and 5% milk powder, incubated in primary antibodies overnight at 4°C, then incubated with goat anti-rabbit horse radish peroxidase secondary antibody for one hour at room temperature. Immunoreactive bands were visualized using chemiluminescence and densitometry was performed using ImageJ software.

Cells were seeded at 2.0×10⁴ cells/well in 24-well plates. The cells were treated in triplicate at less than 30% confluence with varying concentrations of ruxolitinib in 5% FBS growth media. All control groups were treated with 0.01% DMSO. After 96 hours of treatment, MTT assays were performed according to the manufacturer's instructions.

CAL-33 cells were seeded at 1.0×10³ cells/well in 12-well plates. The next day, cells were treated in triplicate with 3 and 10 μmol/L of ruxolitinib in 5% FBS growth media (25, 32). All control groups were treated with vehicle (0.01% DMSO). Drug-containing media were changed every four days. After 11 days, colonies were stained with crystal violet. Colonies were identified and counted using ImageJ software if they were 20 pixels or greater.

All in vivo studies with mouse models were approved by the UCSF Institutional Animal Care and Use Committee (protocol #AN187611). A previously published PDX responsive to AZD1480 was passaged into both flanks of ten 5–6 week-old NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice from The Jackson Laboratory (19). PDX models generated from tumor specimens obtained from ruxolitinib-treated patients with HNSCC were derived under the auspices of an ongoing Phase 0 clinical trial (NCT03153982). For PDX models, tumors were resected in the operating room and, after pathologic assessment, a portion was placed in RPMI-1640 media (4°C) and immediately implanted into both flanks of 5 to 6-week-old NSG mice, as previously described (19). After development, each PDX tumor was passaged into both flanks of 10 NSG mice. When the volumes of PDX tumors reached approximately 150 mm³, mice were randomized to be treated with vehicle or 75 mg/kg of ruxolitinib, five days per week via oral gavage. Treatment was continued through day 22. Tumors were measured with digital calipers two to three times per week and volumes calculated using the formula: (length × width × width)/2. Mice were euthanized at indicated time points and tumors harvested. Tumor portions used for immunoblotting were frozen at −80°C; lysates were prepared by extracting protein with a tissue grinder and lysis buffer. Immunoblotting was performed as described above.

The AZD-sensitive PDX was used in the 7th passage, the ruxolitinib progressive disease (PD) PDX was used in the 5th passage and the ruxolitinib stable disease (SD) PDX was used in the 2nd passage.

For air–liquid interface cultures, primary organoids were initiated from minced PDX tumor tissue fragments and embedded in type I collagen gels in an inner Transwell insert. Culture medium in an outer dish diffused via the permeable Transwell into the inner dish and the top of the collagen layer was exposed to air, allowing cells access to a sufficient oxygen supply. Organoid culture medium consisted of Advanced DMEM/F-12 + 10% FBS, supplemented with growth factors, including WNT3A, EGF, NOGGIN, and RSPO1. PDXOs were treated with ruxolitinib at selected concentrations (3, 10, and 30 μmol/L) or DMSO control for 7 days, followed by evaluation of organoid viability. Briefly, the organoid samples were dissociated with 200 U/mL collagenase IV at 37°C for 45 minutes, followed by 50 μg/mL Liberase TL for 15 minutes. The dissociated organoid cells were washed with 100% FBS and resuspended in organoid culture medium. The organoid cells were stained with trypan blue dye to determine the number of live organoids.

The parent clinical trial is ongoing at the University of California San Francisco and the University of Arizona. The protocol was approved by both local institutional review boards, carried out in accordance with the Declaration of Helsinki and Good Clinical Practice, and registered with ClinicalTrials.gov (NCT03153982). All patients gave written informed consent. Key eligibility criteria included: primary or recurrent HNSCC planned for oncologic surgery; age ≥18; ECOG performance status 0–2; adequate end-organ function. Participants are treated with ruxolitinib 20-mg twice daily for 14–28 days from enrollment to the morning of surgery. Given the brief duration on therapy, we used the quantitative change in tumor size as a more informative continuous variable against which to assess biomarkers in our prior window-of-opportunity trials (33, 34). The RECIST system was used to define index lesions and the sum longest diameter.

Densitometry data were analyzed using generalized linear models, if required to meet the assumptions of the models, on the log-transformed scale. Statistical significance of the ratio of pSTAT3 to total STAT3 between the vehicle- and ruxolitinib-treated groups was determined via contrasts with the multiplicity of comparisons adjusted to control the experiment-wise type I error rate at 5%. When multiple doses of ruxolitinib were considered, the step-down Holm method was used to control the overall type I error rate within each experiment (35). When only one dose was considered and comparing the treatment group comparisons conducted for progressive and SD PDXs was the focus, the Bonferroni method was used. For in vivo xenograft experiments, random effects models were used to address variation in the tumor growth rate across the mice within the xenograft experiments. Data were log-transformed to meet the assumptions of the random effects models. Statistical significance of the comparison of total tumor volume change from baseline (fold change on the log scale) was determined via contrasts. For ex vivo organoids experiments, the unpaired two samples t test was used to compare the mean of organoids derived from progressive and SD PDXs. All statistical analyses were performed using sas software version 9.3 (SAS institute, Cary, NC) at a significance level of 0.05.

The data generated in this study are available within the article and its Supplementary Data files. Requests for the raw data may be directed to the corresponding author.

We previously reported that the JAK inhibitor AZD1480 abrogated STAT3 activation in HNSCC cells (19). To assess the impact of ruxolitinib inhibition of JAK1/2 on STAT3 activation, a panel of representative HNSCC cell lines, CAL-33, CAL 27, and FaDu, were treated with increasing doses of ruxolitinib for 24 hours. Expression levels of total and phosphorylated STAT3 levels were assessed via immunoblotting and the pSTAT3/total STAT3 ratio quantified by densitometry. As shown in Fig. 1, ruxolitinib treatment resulted in a dose-dependent decrease of pSTAT3 levels in all three HNSCC cell lines, and the decrease was significant when the ruxolitinib dose was ≥0.1 μmol/L.

Activated STAT3 induces the transcription of pro-proliferative genes. Aberrant or constitutive STAT3 activation can therefore lead to unregulated cell growth, a hallmark of cancer (36, 37). To determine whether ruxolitinib inhibition of JAK1/2 abrogates HNSCC cell growth, CAL-33, CAL 27, and FaDu cells (selected because they have been extensively characterized, are widely available, and exhibit STAT3 activation upon cytokine stimulation) were treated for 96 hours with vehicle or increasing doses of ruxolitinib in three separate experiments, followed by performance of MTT assays. Inhibition of cellular growth (relative to vehicle treatment) was observed in the HNSCC cell lines at higher doses of ruxolitinib (Fig. 2A).

Colony formation is another approach to test the impact of a drug on tumor cell proliferation and survival. To investigate the impact of ruxolitinib on the colony-forming ability of HNSCC cells, CAL-33 cells were treated with two different concentrations (3 and 10 μmol/L) of ruxolitinib in two independent experiments. At the end of the treatment period, colonies were stained with crystal violet and quantified. Decreased colony number was observed with 10 μmol/L ruxolitinib treatment (P = 0.0001; Fig. 2B).

Given the promising in vitro results in HNSCC cell lines, xenograft models were generated in immunodeficient mice to assess the impact of ruxolitinib treatment on tumor growth in vivo. We previously reported that HNSCC PDX growth was inhibited by treatment with AZD1480, a JAK inhibitor whose clinical development was halted because of neurotoxicity in early-phase trials (18–20). Thus, we investigated whether ruxolitinib, a safe and well-tolerated JAK1/2 inhibitor, demonstrates comparable efficacy in this same PDX model. As shown in Fig. 3A, ruxolitinib significantly slowed tumor growth in the AZD1480-sensitive HNSCC PDX model (P = 0.0082). For all in vivo studies of ruxolitinib, we selected a dose based on previous reports in tumor xenografts (38, 39). To determine the impact of ruxolitinib on STAT3 activation in the tumors, the levels of pSTAT3 in tumor lysates were determined by immunoblotting. Ruxolitinib treatment significantly decreased pSTAT3 levels (P = 0.0008) in the PDXs in conjunction with abrogation of tumor growth.

A window-of-opportunity clinical trial (NCT03153982) is ongoing to evaluate the pharmacodynamic impact of ruxolitinib in HNSCC. Tumor volumes before and after treatment are calculated according to RECIST 1.1 criteria and compared (33). Given the short duration of treatment, we describe the proportion of patients whose tumor volumes decrease or increase as defined previously (34, 40). In these earlier window-of-opportunity trials, we elected not to use RECIST definitions, because the short window of exposure is unlikely to result in a partial response. We defined the quantitative change in tumor size as a more informative continuous variable. PDXs were generated from the post-treatment samples as we have described previously (19).

To determine whether clinical responses observed in the window setting were concordant in the PDX models generated from post-treatment tumors, we selected two post-treatment PDXs from a representative patient with SD versus one with PD on ruxolitinib treatment (exhibited >20% increase in sum longest diameter of index lesion; Table 1). The patient whose HNSCC progressed was a 71-year-old woman with a T2N0M0 (stage 2) human papillomavirus (HPV)–negative cancer of the oral tongue who received 15 days of ruxolitinib treatment. The patient whose tumor did not progress on ruxolitinib was a 57-year-old woman with a T3N2bM0 (stage 4) HPV-negative cancer of the oral tongue who was treated for 18 days with ruxolitinib before surgical resection (exhibited no change in sum longest diameter of index lesion). The PDXs were randomized to be treated with either vehicle or ruxolitinib for 3 weeks.

As shown in Fig. 4A, tumors grew approximately linearly on the log-transformed scale within each mouse. We analyzed longitudinal measurements of total tumor volume on the log-transformed scale using a random effects model. The model accommodates between mouse heterogeneity and permits different lines (different intercepts/slopes) to approximate longitudinal change in total tumor volume within the mice. The impact of ruxolitinib on tumor growth is represented as the difference in the average slope (average rate of tumor growth) by treatment. For the ruxolitinib-PD PDX, it took 35 days for the tumors to achieve 150 mm³; for the ruxolitinib SD PDX, it took 20 days for the tumors to achieve 150 mm³. The mice with tumors were randomized into vehicle and ruxolitinib groups when the average tumor volume reached about 150 mm³.

Ruxolitinib significantly inhibited tumor growth in the PDX from the patient who demonstrated SD on treatment (P = 0.0001). In contrast, ruxolitinib did not inhibit growth of the PDX derived from the patient who experienced PD on therapy (P = 0.7599). However, in the absence of PDX models developed from the pre-ruxolitinib treated tumors, it is not possible to determine the role of ruxolitinib treatment in the apparent resistance of the patient with PD.

We did not detect any significant difference in the tumor histology of the two representative patients enrolled on the ruxolitinib clinical trial (Supplementary Fig. S1). Both patients’ tissue sections showed a keratinizing, well to moderately differentiated squamous cell carcinoma. A pre-treatment tumor sample was only available for one of the patients (SD) and although there was no histologic evidence of tumor regression, the post-treatment excision showed slightly more atypia compared with the pre-treatment biopsy with scattered foci of moderately differentiated tumor. The degree of inflammation in the two post-treatment tumors was similar. When comparing the clinical patient samples with the PDX tissues, the histology was very similar. The only significant histologic differences between the patient tissue and the PDX tissue are that the clinical specimen also contains a mild chronic inflammatory infiltrate.

To determine whether ruxolitinib treatment affected STAT3 activation, pSTAT3 levels in the PD and SD PDX tumor specimens were analyzed by immunoblotting. As shown in Fig. 4B, although downregulation of pSTAT3 by ruxolitinib treatment was observed in both PDXs, inhibition of STAT3 activation was more pronounced in the PDX from the patient whose tumor did not grow in the neoadjuvant setting with ruxolitinib treatment (adjusted P value of 0.0532 for the PD PDX and 0.0174 for the SD PDX). Unexpectedly, pSTAT3 levels were increased by ruxolitinib treatment in the PDX derived from the patient whose disease progressed on therapy. Comparison of baseline pSTAT3 levels in the vehicle group PDX tumors (using the same lysates to generate the results in panel 4b) revealed higher expression in the PDX derived from the patient whose disease progressed compared with the PDX from the patient whose disease was stable on ruxolitinib treatment (P = 0.0001; Fig. 4C). Similarly, baseline levels of total STAT3 were higher in the PD PDX compared with the SD PDX (P = 0.0123; Fig. 4C). These results suggest that baseline levels of either pSTAT3 or total STAT3 may serve as predictive biomarkers to identify patients who are likely to respond to ruxolitinib therapy.

To confirm our findings using an ex vivo approach, we generated organoids from the PDXs and tested the impact of ruxolitinib. As shown in Fig. 5, ruxolitinib demonstrated a dose-dependent decrease in cell growth, as measured by vital dye exclusion, in the organoids derived from a patient who experienced the SD clinical response compared with organoids derived from a patient who demonstrated PD. Similar results were seen using a metabolic assay (Supplementary Fig. S3). As observed with the PDX experiments, pSTAT3 expression levels were downregulated in the PDX-derived organoids from the patient with SD compared with the organoids from the patient with PD (Supplementary Fig. S2). These results suggest that antitumor effects can be tested without requiring the use of additional laboratory animals.

Aberrant hyperactivation of STAT3 is associated with disease progression and poor prognosis in HNSCC (2, 41). Directly targeting STAT3 has been challenging, and the possibility of targeting upstream activators such as JAKs with small-molecule inhibitors represents an attractive alternative (3). JAK inhibition with AZD1480 in HNSCC preclinical models showed promising in vitro and in vivo results; however, neurotoxicity detected in a Phase I trial halted the clinical development of this agent (19, 20). Ruxolitinib is a selective JAK1/2 inhibitor that was approved by the FDA for myeloproliferative disorders—many of which are associated with an activating JAK2 V617F mutation—after it was shown to decrease STAT3 activation and symptom burden in these hematological malignancies (14, 16, 42). However, its efficacy in solid tumors, including HNSCC, is incompletely understood. We evaluated the anticancer effects of ruxolitinib in HNSCC preclinical models, including PDXs generated from patients treated with ruxolitinib in an ongoing window-of-opportunity clinical trial.

Preclinical studies investigating JAK inhibition with ruxolitinib in a variety of solid tumors, including HNSCC, have shown promising anti-cancer effects. Treatment of hepatocellular carcinoma cell lines with ruxolitinib inhibited STAT3 activation and cell proliferation and survival at higher concentrations than used in the present study (25). Similar abrogation of STAT3 activation was observed in pancreatic adenocarcinoma and colon cancer cell lines (23, 24). In esophageal squamous cell carcinoma, ruxolitinib treatment inhibited STAT3 activation and decreased cell proliferation in a dose-dependent manner (26). Others reported that ruxolitinib treatment abrogated STAT3 phosphorylation in HNSCC CAL 27 cells, consistent with our results using this cell line (43). In the HNSCC cell line SCC15, ruxolitinib treatment reduced cell growth (44). Overall, our results provide further evidence that ruxolitinib abrogates STAT3 activation and inhibits tumor growth in solid tumor preclinical models, including HNSCC.

Several in vivo studies in solid tumors have demonstrated that ruxolitinib inhibits tumor growth and decreases pSTAT3 levels in cell line-derived xenografts and PDXs. Mice harboring non–small cell lung cancer xenografts treated with ruxolitinib, alone or in combination with cisplatin, exhibited significant abrogation of STAT3 activation and inhibition of tumor growth, compared with the control group (45, 46). The impact of ruxolitinib has been tested in vivo using HNSCC cell line–derived tumor models as well; mice harboring HNSCC tumors derived from SCC9 cells were treated with ruxolitinib. However, there was no effect on tumor growth and proliferation (44). In addition to the PDXs reported in this study, the impact of ruxolitinib on solid tumors has been reported previously in five hepatocellular carcinoma PDXs models. These hepatocellular carcinoma PDX models were generated from tumors with wild-type JAK1 or mutant JAK1 (S703I, N451S, E483D, and A1066S mutations; ref. 47). Of these models, only JAK1 S703I represents an activating mutation; ruxolitinib treatment of this PDX model resulted in a significant inhibition of tumor growth and STAT3 activation (47). In contrast, ruxolitinib treatment of the three PDXs with the non-activating JAK1 mutations did not evoke antitumor effects (47). Although ruxolitinib has not previously been tested in HNSCC PDXs, we reported that treatment with the JAK1/2 inhibitor AZD1480 resulted in inhibition of STAT3 activation and tumor growth in HNSCC PDXs (19). In the present study, the same AZD1480-responsive HNSCC PDX also exhibited decreased tumor growth and reduced STAT3 activation after three weeks of ruxolitinib treatment.

The antitumor effects of ruxolitinib are being studied in a variety of hematological malignancies and solid tumors, as well as PDXs derived from ruxolitinib-treated patients. In a Phase I/II single-arm clinical trial investigating the impact of ruxolitinib on chronic myelomonocytic leukemia (CMML), 36% of enrolled patients responded to treatment, and 40% of patients with splenomegaly had greater than a 50% reduction in splenic size (NCT01776723; ref. 48). PDX models were generated using pre-treatment bone marrow mononuclear cells from three responders and three non-responders with splenomegaly to determine whether responses to ruxolitinib treatment in these models recapitulated findings in the clinical trial. Ruxolitinib-treated PDXs from clinical responders had a significant reduction in spleen and engraftment size compared with vehicle-treated PDXs; this impact was not observed in PDXs from clinical non-responders. Similarly, in our study, ruxolitinib treatment of an HNSCC PDX model generated from a patient with HNSCC who exhibited SD on neoadjuvant ruxolitinib, resulted in tumor growth inhibition in conjunction with decreased STAT3 activation. In contrast, ruxolitinib failed to abrogate both STAT3 activation and tumor growth in an HNSCC PDX derived from a patient whose disease progressed on ruxolitinib treatment. Increased baseline levels of pSTAT3 and total STAT3 were detected in the PDX developed from the PD patient compared with the SD patient. These findings support the possibility that baseline expression of the pathway targets may serve as predictive biomarkers to identify patients most likely to benefit from ruxolitinib treatment.

Overall, our in vitro and in vivo findings are consistent with other studies in solid tumor models. However, the relatively high concentrations required to achieve maximum growth inhibition may limit clinical translation of our findings in preclinical models. Our studies with PDX models suggest that ruxolitinib may be effective in a subset of patients with HNSCC with this malignancy based on expression of total and/or phosphorylated STAT3. For example, a prior report that ruxolitinib inhibited tumor growth only in hepatocellular carcinoma PDXs that carry an activating JAK mutation (47). Finally, we demonstrate that ruxolitinib treatment of PDXs mirrors the clinical response in patients with HNSCC, which was consistent with a similar study in CMML; to our knowledge a similar finding has not been reported to date in other solid tumors (48).

The immunological impact of JAK inhibition with ruxolitinib in HNSCC is an area for further study as this agent has been shown to have immunosuppressive, with anti-inflammatory effects (49). In patients with myeloproliferative neoplasms, ruxolitinib treatment decreased the number of CD4⁺ T cells and pro-inflammatory cytokines (50). Its anti-inflammatory impact makes ruxolitinib a potentially attractive therapy for solid tumors as development of these cancers is associated with chronic inflammation, such as through IL6-mediated STAT3 activation (3, 51). Furthermore, cytokine secretion from tumors themselves has been shown to inhibit activation of cytotoxic T cells, thereby impeding anticancer immune response (51). Treatment of mice harboring pancreatic adenocarcinoma with ruxolitinib not only decreased tumor volume, but also restored cytotoxic T lymphocyte levels in tumors, as evidenced by elevated CD8⁺ cell levels (52). Further studies on the immunological impact of ruxolitinib in immunocompetent HNSCC models are warranted.

In this study, we demonstrated that ruxolitinib exhibits anticancer effects in HNSCC in vitro and in vivo models. Whereas all cell lines tested demonstrated decreased pSTAT3 expression and reduced proliferation upon ruxolitinib treatment, the PDXs derived from patients treated with this agent on a window-of-opportunity are more informative. Our findings in these relevant preclinical models suggest that baseline levels pSTAT3 and total STAT3, as well as modulation of pSTAT3 following a brief course of ruxolitinib, may identify patients most likely to respond to this JAK/STAT inhibitor.

Z. Qureshy reports nonfinancial support from Incyte Corporation and grants from UCSF School of Medicine during the conduct of the study. W.R. Ryan reports personal fees from Olympus, Rakuten Medical, and Intuitive outside the submitted work. S.R. Long reports personal fees from Lunit outside the submitted work. D.E. Johnson reports grants and nonfinancial support from Incyte Corporation during the conduct of the study; personal fees from Bluedot Bio outside the submitted work; and reports a patent for STAT3 decoy issued and licensed to Bluedot Bio. J.R. Grandis reports grants from Incyte Corporation during the conduct of the study. No disclosures were reported by the other authors.

Z. Qureshy: Formal analysis, investigation, writing–original draft. H. Li: Formal analysis, investigation, writing–original draft. Y. Zeng: Validation. J. Rivera: Validation. N. Cheng: Investigation. C.N. Peterson: Validation. M.-O. Kim: Writing–review and editing. W.R. Ryan: Writing–review and editing. P.K. Ha: Writing–review and editing. J.E. Bauman: Writing–review and editing. S.J. Wang: Writing–review and editing. S.R. Long: Writing–review and editing. D.E. Johnson: Conceptualization, writing–original draft, project administration, writing–review and editing. J.R. Grandis: Conceptualization, resources, supervision, writing–original draft, project administration, writing–review and editing.

This work was supported by NIH R35CA231998 (to J.R. Grandis), R01DE028289 (to D.E. Johnson and J.R. Grandis), and a Yearlong Research Fellowship awarded by the University of California San Francisco School of Medicine (to Z. Qureshy).

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.