Treatment with ipilimumab improves overall survival (OS) in patients with metastatic melanoma. Because ipilimumab targets T lymphocytes and not the tumor itself, efficacy may be uniquely sensitive to immunomodulatory factors present at the time of treatment. We analyzed serum from patients with metastatic melanoma (247 of 273, 90.4%) randomly assigned to receive ipilimumab or gp100 peptide vaccine. We quantified candidate biomarkers at baseline and assessed the association of each using multivariate analyses. Results were confirmed in an independent cohort of similar patients (48 of 52, 92.3%) treated with ipilimumab. After controlling for baseline covariates, elevated chemokine (C-X-C motif) ligand 11 (CXCL11) and soluble MHC class I polypeptide–related chain A (sMICA) were associated with poor OS in ipilimumab-treated patients [log10 CXCL11: HR, 1.88; 95% confidence interval (CI), 1.14–3.12; P = 0.014; and log10 sMICA quadratic effect P = 0.066; sMICA (≥ 247 vs. 247): HR, 1.75; 95% CI, 1.02–3.01]. Multivariate analysis of an independent ipilimumab-treated cohort confirmed the association between log10 CXCL11 and OS (HR, 3.18; 95% CI, 1.13–8.95; P = 0.029), whereas sMICA was less strongly associated with OS [log10 sMICA quadratic effect P = 0.16; sMICA (≥247 vs. 247): HR, 1.48; 95% CI, 0.67–3.27]. High baseline CXCL11 and sMICA were associated with poor OS in patients with metastatic melanoma after ipilimumab treatment but not vaccine treatment. Thus, pretreatment CXCL11 and sMICA may represent predictors of survival benefit after ipilimumab treatment as well as therapeutic targets. Cancer Res; 75(23); 5084–92. ©2015 AACR.

Ipilimumab is a human monoclonal antibody targeting CTLA-4 (cytotoxic T lymphocyte antigen-4). CTLA-4 is expressed on activated T cells, has structural similarities to the costimulatory molecule CD28, and binds to the same ligands as CD28 albeit with higher affinity. Binding of CTLA-4 to CD80/CD86 inhibits T-cell activation by limiting IL2 production and expression of the IL2 receptor (CD25; ref. 1). Ipilimumab prevents CTLA-4 from binding its ligands, thus promoting activation of effector T cells via prolonged CD28 signaling (2). In addition, anti-CTLA-4 antibodies can deplete intratumoral regulatory T cells, subverting yet another mechanism of immunosuppression (3).

Ipilimumab improved overall survival (OS) in patients with metastatic melanoma in a randomized, double-blinded phase III clinical trial (4, 5). Patients with metastatic melanoma who received ipilimumab monotherapy achieved a 28.4% four-year OS rate (6). Despite the success of ipilimumab, the majority of the patients on this study died as a consequence of melanoma. With an increasing number of treatment options available and increased use of targeted therapies, predictive biomarkers that identify those patients most likely to benefit from a specific treatment are needed (7, 8).

Unlike traditional cancer therapies, immunotherapeutics act primarily upon cells of the immune system. The requirement for the immune system as a third-party mediator of the drug's activity suggests the balance of positive and negative regulators of the immune response at the time of therapy may be a critical determinant of efficacy for any immunotherapy. Cytokines, chemokines, and soluble receptors regulate the survival, activity, and location of immune effector cells and thus represent potential players in determining drug efficacy. Of particular interest are soluble factors involved in the recruitment and regulation of effector T cells representing the most readily measurable clinical biomarkers.

To identify candidate soluble factor(s) predictive of improved survival following ipilimumab treatment, we analyzed pretreatment sera from treatment (ipilimumab) and “active control” (gp100 vaccine) patients from the pivotal phase III clinical trial of ipilimumab (4) for a variety of factors and correlated their levels with OS. A hypothesis-guided panel of candidate biomarkers was selected, including biomarkers previously reported to associate with response to ipilimumab. Each analyte was assessed in univariate and multivariate models for its correlation with OS. Correlative biomarkers identified in the initial screen were further validated by testing sera from an independent cohort of ipilimumab-treated patients at our institution.

Clinical trials

Detailed information regarding the phase III clinical trial of ipilimumab (NCT00094653) was reported elsewhere (4). Briefly, patients with metastatic melanoma having failed at least one prior therapy that may have included IL2, dacarbazine, and/or temozolomide were enrolled excluding those with ocular melanoma. All patients were HLA-A*0201+ as the restricting element for the gp100 peptides used. All ipilimumab-treated patients received ipilimumab alone at 3 mg/kg every 3 weeks for 4 treatments. In the gp100 group, patients received two peptides (1 mg each), injected subcutaneously as an emulsion with incomplete Freund's adjuvant (Montanide ISA-51). Peptide injections were given immediately after 90-minute intravenous infusion of placebo. Tumor burden was assessed by the treating physician as previously described (4).

Serum samples were also obtained from patients treated on an expanded access program at the Earle A. Chiles Research Institute (EACRI cohort). Detailed information regarding this Compassionate Use Trial for Unresectable Melanoma with Ipilimumab is available elsewhere (NCT00495066). All patients received ipilimumab alone (3 or 10 mg/kg every 3 weeks for 4 treatments) with no exclusions for ocular primary melanomas or HLA type.

All patients provided written informed consent and all studies were carried out in accordance with the Declaration of Helsinki under good clinical practice and Institutional Review Board approval.

Serum cytokine analysis

Serum was collected and stored at −80°C. Chemokine (C-C motif) ligand 2 (CCL2), CCL3, CCL4, CCL8, CCL18, CCL26, chemokine (C-X-C motif) ligand (CXCL9), CXCL10, CXCL11, CXCL13, and VEGF were measured using a bead-based multiplexed immunoassay (R&D Systems). Soluble MHC class I polypeptide–related sequence A (sMICA), sMICB, soluble UL16-binding protein (sULBP)-1, sULBP-2, sULBP-3, and sULBP-4 were measured using a custom multiplex bead array (R&D Systems). Bead-based assays were analyzed using the Luminex-based Bio-Plex system (BIO-RAD). Soluble CD25 (sCD25) and soluble lymphocyte-activation gene 3 (sLAG-3) were measured by ELISA (R&D Systems). Serum sHLA-G was measured by ELISA (Exbio Vestec). Only serum cytokines having statistical significance in univariate analyses of OS were reported.

Statistical considerations

Differences in patient baseline characteristics between treatment groups (ipilimumab vs. gp100) or trials (phase III vs. compassionate use) were evaluated using a t test for age and χ2 tests for gender, Eastern Cooperative Oncology Group (ECOG) performance status, lactate dehydrogenase (LDH), prior IL2 therapy, and prior immunotherapy. Differences in baseline serum biomarkers between study and treatment groups were tested with Wilcoxon rank-sum tests because of skewed distributions.

Analysis of OS was conducted in the phase III trial and separately in the confirmatory EACRI cohort due to the differences in the patient populations, study design, and study protocol. In the phase III trial, differences within treatment group were of primary interest and thus tested in separate models. Survival was defined as time from beginning of ipilimumab treatment to date of death, censoring at date of last follow-up. To calculate median follow-up time, deaths were censored.

Univariate survival analysis was performed for each treatment subgroup using Cox proportional hazards regression. Effects of CXCL11, sMICA, sMICB, sCD25, VEGF, absolute lymphocyte counts (ALC), tumor burden, and effect of LDH [≤ vs. > upper limit of normal (ULN)] were shown. Models of quadratic effects to examine possible nonlinear effects and models of linear effects of continuous variables were tested. When the quadratic effect was not significant or the linear effect was more strongly significant, main effects model results were reported. For sMICA and VEGF, quadratic effect was significant in some models. To present an HR, results are also reported for a categorized variable, with the cutoff point determined as the quintile where a threshold effect was observed in the phase III trial. Kaplan–Meier plots used for determining cutoff points are shown in Supplementary Fig. S1. In multivariate analyses, model results of the remaining variables (other than sMICA), however, are from the model containing the continuous form of the variable with the quadratic effect. Continuous measures were approximately log normal and analyzed as log10-transformed.

In multivariate analysis of OS, Cox proportional hazards regression was used to test effects of biomarker candidates on survival after controlling for other biomarkers and baseline patient characteristics. Only CXCL11 and sMICA were included, as they were significant in univariate survival models of the ipilimumab group but not the gp100 group. Covariates in models for both studies were age, gender, ECOG status, prior immunotherapy, LDH, and ALC. Tumor burden was also included in multivariate model for the phase III trial cohort, but not the EACRI cohort as these data were not captured.

Analyses were performed using SAS 9.3 (SAS Institute Inc.). Forest plots were prepared using Forest Plot Viewer (9) and edited using Adobe Illustrator. GraphPad Prism was used for depicting some Kaplan–Meier plots.

Patient characteristics of phase III study

Demographics for the phase III study were previously reported (4). Briefly, 676 patients were enrolled with 137 selected to receive ipilimumab monotherapy (treatment group), 136 to receive gp100 monotherapy (control group), and 403 treated with the combination of these agents. Biomarker analysis was restricted to the monotherapy groups. Baseline characteristics were similar between monotherapy groups (Table 1), except that a higher proportion of patients received prior immunotherapy in the gp100 alone group (P = 0.036; Table 1, column D). Patients were followed for a median of 31 months (range, 27–43 months). OS of the ipilimumab group was 45.6% at 12 months, 33.2% at 18 months, and 23.5% at 24 months, with a median OS of 10.1 months [95% confidence interval (CI), 8.0–13.8]. OS of the gp100 group was 25.3% at 12 months, 16.3% at 18 months, and 13.7% at 24 months with a median OS of 6.4 months (95% CI, 5.5–8.7). Analysis of soluble immunomodulatory proteins was performed on serum collected prior to treatment. Baseline CXCL11 concentrations were comparable between ipilimumab (median, 38; range, 2–1,027 pg/mL) and gp100 (median, 39; range, 2–911 pg/mL) groups. Similarly, baseline levels of sMICA were consistent between ipilimumab (median, 115; range, 13–1,573 pg/mL) and gp100 (median, 121; range, 13–2,074 pg/mL) groups.

Table 1.

Baseline patient characteristics

ABCDE
Phase III trial cohortEACRI cohortStatistics
Demographic or clinical characteristicIpilimumab monotherapy (n = 124)gp100 monotherapy (n = 123)Ipilimumab monotherapy (n = 48)P: A vs. BaP: A vs. Ca
Age, y    0.99 0.51 
 Median 57 57 60   
 Range 23–90 19–88 36–81   
Male, % 61.3 54.5 60.4 0.28 0.92 
ECOG, %    0.85d 0.22d 
 0 51.6 52.9 41.7   
 1 47.6 43.9 52.1   
 2 0.8 3.2 6.3   
LDH > ULN, % 37.1 36.4 66.0 0.91 0.0007 
ALC, ×109/L    0.27 0.61 
 Median 1.3 1.2 1.3   
 Range 0.4–3.3 0.3–2.8 0.3–4.1   
Prior IL2 therapy, % 23.4 25.2 60.4 0.74 <0.0001 
Prior immunotherapy, %b 39.5 52.9 77.1 0.036 <0.0001 
CXCL11, pg/mL    0.51 0.0003 
 Median 38 39 14   
 Range 2–1027 2–911 3–153   
sMICA, pg/mL    0.99 <0.0001 
 Median 115 121 299   
 Range 13–1573 13–2074 12–2456   
Median survival, moc 10.1 6.9 8.6   
ABCDE
Phase III trial cohortEACRI cohortStatistics
Demographic or clinical characteristicIpilimumab monotherapy (n = 124)gp100 monotherapy (n = 123)Ipilimumab monotherapy (n = 48)P: A vs. BaP: A vs. Ca
Age, y    0.99 0.51 
 Median 57 57 60   
 Range 23–90 19–88 36–81   
Male, % 61.3 54.5 60.4 0.28 0.92 
ECOG, %    0.85d 0.22d 
 0 51.6 52.9 41.7   
 1 47.6 43.9 52.1   
 2 0.8 3.2 6.3   
LDH > ULN, % 37.1 36.4 66.0 0.91 0.0007 
ALC, ×109/L    0.27 0.61 
 Median 1.3 1.2 1.3   
 Range 0.4–3.3 0.3–2.8 0.3–4.1   
Prior IL2 therapy, % 23.4 25.2 60.4 0.74 <0.0001 
Prior immunotherapy, %b 39.5 52.9 77.1 0.036 <0.0001 
CXCL11, pg/mL    0.51 0.0003 
 Median 38 39 14   
 Range 2–1027 2–911 3–153   
sMICA, pg/mL    0.99 <0.0001 
 Median 115 121 299   
 Range 13–1573 13–2074 12–2456   
Median survival, moc 10.1 6.9 8.6   

aP values shown from the t test for age, Wilcoxon log-rank test for CXCL11 and sMICA, and χ2 tests for all other variables.

bIncluding IL2.

cMedian survivals times calculated from Kaplan–Meier estimates.

dECOG 1–2 versus 0.

CXCL11, sMICA, and OS

Univariate analysis of ipilimumab-treated patients showed that a 10-fold increase in CXCL11 was associated with double the risk of death (HR, 2.08; 95% CI, 1.40–3.11; P = 0.0003; Fig. 1), whereas CXCL11 was not associated with OS in the gp100 group (HR, 1.21; 95% CI, 0.87–1.68; P = 0.2597). The effect of CXCL11 on OS was significantly different for the ipilimumab group versus the gp100 group (P = 0.040). In the univariate analysis of log10 sMICA, higher sMICA was associated with decreased survival in the ipilimumab group [log10 sMICA quadratic effect P < 0.0001; sMICA (≥247 vs. <247): HR, 3.46; 95% CI, 2.16–5.56 with P < 0.0001] but not in the gp100 group (log10 sMICA HR, 0.91; 95% CI, 0.61–1.36; P = 0.6373). Elevated sMICB, LDH, tumor burden, and sCD25 were all associated with poorer survival regardless of treatment (Fig. 1). Elevated VEGF was also associated with decreased survival in both groups, although marginally so for the ipilimumab-treated group (Fig. 1). Higher numbers of lymphocytes (ALC) at baseline were associated with better OS in both treatment groups (Fig. 1). These univariate analyses suggest that CXCL11 and sMICA are potential predictors of OS in ipilimumab-treated melanoma patients, whereas sMICB, sCD25, VEGF, LDH, tumor burden, and ALC represent putative prognostic biomarkers.

Figure 1.

Univariate analysis of biomarker effects on OS for patients from the phase III clinical trial. HR and CI for association with OS of patients treated with ipilimumab (ipi) or gp100. Cox proportional hazards regression was used for univariate analysis of biomarker effects on OS. HR is numerator versus denominator. Among 124 patients analyzed, 35 were censored in ipilimumab-treated group. Among 123 patients analyzed, 13 were censored in gp100-treated group. Missing data: LDH, ALC, tumor burden 1–2 missing in gp100 group; CXCL11, sCD25, VEGF 5 missing in gp100 group, 11 missing in ipilimumab group. *, in quadratic effects model of ipilimumab group, (log10 sMICA)2P < 0.0001. §, in gp100 group, (log10 VEGF)2P = 0.0002.

Figure 1.

Univariate analysis of biomarker effects on OS for patients from the phase III clinical trial. HR and CI for association with OS of patients treated with ipilimumab (ipi) or gp100. Cox proportional hazards regression was used for univariate analysis of biomarker effects on OS. HR is numerator versus denominator. Among 124 patients analyzed, 35 were censored in ipilimumab-treated group. Among 123 patients analyzed, 13 were censored in gp100-treated group. Missing data: LDH, ALC, tumor burden 1–2 missing in gp100 group; CXCL11, sCD25, VEGF 5 missing in gp100 group, 11 missing in ipilimumab group. *, in quadratic effects model of ipilimumab group, (log10 sMICA)2P < 0.0001. §, in gp100 group, (log10 VEGF)2P = 0.0002.

Close modal

Multivariate analyses were also conducted focusing on CXCL11 and sMICA, as these two biomarkers were identified in the univariate analysis as correlating with ipilimumab but not gp100 treatment. Models were used to test the independent effects of CXCL11 and sMICA after adjusting for each other and the covariates LDH, ALC, tumor burden, age, sex, and ECOG status. Within the ipilimumab-treated group CXCL11 and LDH, but not tumor burden, ALC, age, sex, or ECOG score, were associated with OS (log10 CXCL11 HR, 1.88; 95% CI; 1.14–3.12; P = 0.014: LDH HR, 2.99; 95% CI, 1.78–5.02; P < 0.0001; Fig. 2A). sMICA was also associated with OS [log10 sMICA quadratic effect P = 0.0659; sMICA (≥247 vs. <247): HR, 1.75; 95% CI, 1.02–3.01 with P = 0.0420] but less strongly than CXCL11 (Fig. 2A). In the gp100-treated group, only LDH was independently associated with OS (LDH: HR, 2.24; 95% CI, 1.36–3.69; P = 0.0016; Fig. 2B). These multivariate results again suggest that CXCL11 and sMICA are potential predictive biomarkers of OS in ipilimumab-treated melanoma patients, whereas LDH represents a prognostic biomarker for patients with melanoma irrespective of treatment.

Figure 2.

Multivariate analysis of biomarker effects on OS for patients from the phase III clinical trial. HR and CI for association of potential biomarker with OS of patients treated with ipilimumab (A) or gp100 (B). Cox proportional hazards regression was used for multivariate analysis of biomarker effects on OS. Among the 113 total patients analyzed, 34 were censored in ipilimumab-treated group. Among total 115 patients analyzed, 13 were censored in gp100-treated group. *, in quadratic effects model of ipilimumab group, (log10 sMICA)2P = 0.0659.

Figure 2.

Multivariate analysis of biomarker effects on OS for patients from the phase III clinical trial. HR and CI for association of potential biomarker with OS of patients treated with ipilimumab (A) or gp100 (B). Cox proportional hazards regression was used for multivariate analysis of biomarker effects on OS. Among the 113 total patients analyzed, 34 were censored in ipilimumab-treated group. Among total 115 patients analyzed, 13 were censored in gp100-treated group. *, in quadratic effects model of ipilimumab group, (log10 sMICA)2P = 0.0659.

Close modal

CXCL11 and sMICA in an independent ipilimumab-treated cohort

We analyzed sera from patients with melanoma (48 of 52, 92.3%) collected prior to treatment with ipilimumab in an expanded access program at our institution (EACRI cohort). When comparing patient characteristics between the ipilimumab-treated phase III trial cohort and the EACRI cohort, we found more patients in the EACRI cohort with elevated LDH (P = 0.0007), prior IL2 therapy (P < 0.0001), and prior immunotherapy (P < 0.0001; Table 1, column A, C, and E). This comparison suggests that the EACRI cohort included more patients with advanced disease and poorer prognosis. This discrepancy may account for shorter median survival of the EACRI cohort (8.6 months) relative to that of ipilimumab-treated phase III trial cohort (10.1 months). In the EACRI cohort, median follow-up was 39 months (range, 0.8–40 months).

Univariate analyses showed that elevated pretreatment concentrations of CXCL11, sCD25, and LDH were associated with an increased risk of death (Fig. 3). Similar to phase III study findings, a 10-fold increase in CXCL11 was associated with a 3.7-fold increase in the risk of death (HR, 3.74; 95% CI, 1.71–8.22; P = 0.0010). sMICA and VEGF effects were nonlinear, depicted by the threshold effect as seen in the Kaplan–Meier plots of survival (Supplementary Fig. S2). Elevated sMICA was also associated with increased risk of death [log10 sMICA quadratic effect: P = 0.0244; sMICA (≥247 vs. <247): HR, 2.06; 95% CI, 1.06–4.00 with P = 0.0324; Fig. 3]. Elevated VEGF was associated with decreased survival and sMICB and ALC were not associated with survival.

Figure 3.

Univariate analysis of the ipilimumab-treated EACRI cohort. HR and CI for association of potential biomarker with OS of patients treated with ipilimumab. Cox proportional hazards regression was used for univariate analysis of biomarker effects on OS. HR is numerator versus denominator. Among 48 total patients analyzed, 8 were censored. *, in quadratic effects model, (log10 sMICA)2P = 0.0244. §, (log10 VEGF)2P = 0.0200.

Figure 3.

Univariate analysis of the ipilimumab-treated EACRI cohort. HR and CI for association of potential biomarker with OS of patients treated with ipilimumab. Cox proportional hazards regression was used for univariate analysis of biomarker effects on OS. HR is numerator versus denominator. Among 48 total patients analyzed, 8 were censored. *, in quadratic effects model, (log10 sMICA)2P = 0.0244. §, (log10 VEGF)2P = 0.0200.

Close modal

Multivariate analysis showed that CXCL11 and LDH were associated with OS (log10 CXCL11: HR, 3.18; 95% CI, 1.13–8.95; P = 0.0288 and LDH: HR, 2.24; 95% CI, 1.02–4.95; P = 0.0457) after controlling for each other, gender, age, ECOG status, and prior immunotherapy (Fig. 4). sMICA may be associated with OS in this cohort [log10 sMICA quadratic effect: P = 0.1589; sMICA (≥247 vs. <247): HR, 1.48; 95% CI, 0.67–3.27 with P = 0.3284; Fig. 4], a result due in part to adjusting for CXCL11 and LDH and somewhat small cohort size. Thus, we confirmed the predictive association between CXCL11 and OS in ipilimumab-treated melanoma patients but found a weaker association between sMICA and OS in the EACRI cohort. We also confirmed the association between LDH and OS in the EACRI cohort, compatible with the notion that LDH is a prognostic marker for patients with metastatic melanoma.

Figure 4.

Multivariate analysis of the ipilimumab-treated EACRI cohort. HR and CI for association of potential biomarker with OS of patients treated with ipilimumab. Cox proportional hazards regression was used for multivariate analysis of biomarker effects on OS. HR is numerator versus denominator. Among 47 total patients analyzed, 8 were censored. *, in quadratic effects model, (log10 sMICA)2P = 0.1589.

Figure 4.

Multivariate analysis of the ipilimumab-treated EACRI cohort. HR and CI for association of potential biomarker with OS of patients treated with ipilimumab. Cox proportional hazards regression was used for multivariate analysis of biomarker effects on OS. HR is numerator versus denominator. Among 47 total patients analyzed, 8 were censored. *, in quadratic effects model, (log10 sMICA)2P = 0.1589.

Close modal

Kaplan–Meier survival curves

To illustrate the effects of CXC11 and sMICA on OS in the phase III trial, Kaplan–Meier survival plots are shown (Fig. 5) for the biomarker high or low groups on the basis of selected cutoff points (the median of CXC11; 35 pg/mL, the 80th percentile of sMICA; 247 pg/mL, Fig. 5A and B). Because of the quadratic association of sMICA with survival, the 80th percentile was the cutoff point chosen on the basis of the approximate threshold value seen in the sMICA quintile plot (Supplementary Fig. S1). Kaplan–Meier survival plots for the EACRI cohort are also shown using the same cutoff points as used for the phase III study data plot (Fig. 5C and D). The distribution of baseline CXCL11 was lower and the distribution of sMICA levels was higher in the confirmatory cohort (Table 1, column E). Nonetheless, both cutoff points successfully dichotomize patients treated with ipilimumab into patients with poor or better OS.

Figure 5.

Kaplan–Meier curves for OS according to pretreatment CXCL11 or sMICA status. Curves for OS obtained by applying selected cut points for CXCL11 (A and C) and sMICA (B and D) to the phase III trial cohort (A and B) or the EACRI cohort (C and D). Numbers of subjects at risk at each 10-month interval are listed below each graph. The difference in treatment effect [ipilimumab (ipi) or gp100] according to CXCL11 or sMICA concentration is represented in the top right of each graph (log-rank).

Figure 5.

Kaplan–Meier curves for OS according to pretreatment CXCL11 or sMICA status. Curves for OS obtained by applying selected cut points for CXCL11 (A and C) and sMICA (B and D) to the phase III trial cohort (A and B) or the EACRI cohort (C and D). Numbers of subjects at risk at each 10-month interval are listed below each graph. The difference in treatment effect [ipilimumab (ipi) or gp100] according to CXCL11 or sMICA concentration is represented in the top right of each graph (log-rank).

Close modal

We found that high baseline serum CXCL11 and sMICA were associated with poor OS in patients with metastatic melanoma treated with ipilimumab but not in patients treated with a “control” gp100 vaccine. This association was validated in an independent cohort of ipilimumab-treated melanoma patients, strongly suggesting that measurement of pretreatment serum CXCL11 and sMICA levels may identify patients most likely to benefit from ipilimumab. Because a minority of patients with melanoma benefit from ipilimumab, avoiding treatment of these refractory patients would reduce exposure to inefficient therapy, eliminate their risks for adverse effects and lower overall costs of therapy.

Potential predictors of responsiveness to ipilimumab treatment have been identified in previous reports. For instance, elevated levels of several candidate biomarkers [e.g., C-reactive protein (CRP; refs. 10, 11), erythrocyte sedimentation rate (ESR; ref. 12), LDH (10–14), S100 protein (12), sCD25 (15), and VEGF (16)] were thought to associate with reduced benefit following ipilimumab treatment. In contrast, increased baseline ALC were associated with improved OS upon ipilimumab treatment (10–14, 17). Interestingly, each of these candidate biomarkers were at some time reported as prognostic biomarkers for melanoma (18, 19). We were fortunate to have serum samples from the phase III study comparing an ipilimumab-treated cohort with a control cohort allowing us to differentiate predictive versus prognostic biomarkers (20). We also used samples from a second independent cohort of ipilimumab-treated patients to validate our basic findings, which together with multivariate analyses, diminished the possibility of coincidental influence from other covariates. These analyses have demonstrated the predictive nature of CXCL11 and sMICA for patients treated with ipilimumab, and the prognostic nature of sMICB, VEGF, sCD25, LDH, and ALC for all patients with melanoma independent of treatment with ipilimumab.

Our data revealed a strong association between elevated serum CXCL11 protein and reduced OS upon ipilimumab treatment in patients with metastatic melanoma. CXCL11, along with CXCL9 and CXCL10, binds to CXCR3, a critical chemokine receptor for directional migration of TH1 and cytotoxic T cells (21). In contrast to our findings with serum CXCL11, tissue expression of CXCL11 mRNA correlates with T-cell infiltration into tumors and improved prognosis (22). Microarray analysis of mRNA expression in melanoma tissues from ipilimumab-treated patients also associated baseline and posttreatment tissue expression of CXCL11 with presence of T cells in tumor and favorable clinical responses (23). The discrepancy between these results and ours may be attributed to sample source (tissue vs. serum) and/or assay targets (mRNA vs. protein). Paired analysis of mRNAs and proteins in tissue and serum might provide evidence to resolve this discrepancy. As CXCL11 has distinct immunoregulatory functions, in contrast to immunostimulatory functions mediated by CXCL9 and CXCL10 (24, 25), we prefer the following nonmutually exclusive explanations of how CXCL11, even in the presence of CXCL9 and CXCL10, may limit T-cell effector function as a part of negative feedback loop: (i) disrupting chemokine gradients for directional migration of T cells (26), (ii) preventing CXCR3–CXCL9/10 interaction by promoting receptor internalization (24), (iii) suppressing T-cell responses through induction and/or recruitment of regulatory T cells (25, 27), and (iv) promoting growth and metastasis of tumors expressing CXCR3 (28–31).

Elevated levels of sMICA and sMICB have been reported in several types of malignant diseases and implicated in cancer immunoevasion (32). While the cellular stress response in transformed cells induces expression of membrane-bound MICA/MICB, the cleavage of these molecules produces soluble forms of MICA and MICB capable of inhibiting interactions between membrane-bound MICA/MICB and NKG2D, thus desensitizing the activation signal through NKG2D in effector T cells and natural killer (NK) cells (33). The cleavage of membrane-bound MICA/MICB is promoted by the metalloprotease ADAM10 and is enhanced within hypoxic tumor environments (34). As hypoxia also promotes expression of immune inhibitory molecules (e.g., PD-L1 and LAG-3) and favors accumulation of regulatory immune cells (35), sMICA may be a biomarker that reflects this immunosuppressive tumor environment.

The greatest significance of these findings may ultimately be in the identification of CXCL11 and sMICA as immunotherapeutic targets. Agents that inhibit the immunosuppressive activity of CXCL11 and sMICA without preventing interaction of the receptors (CXCR3 and NKG2D, respectively) with immunopotentiating ligands (CXCL9/CXCL10 and membrane-bound MICA, respectively) may have therapeutic activity in patients with cancer, either alone or with other agents, including chemotherapeutics, radiation, or other immunotherapeutics. To explore this idea, future studies should address whether CXCL11 and sMICA directly interfere with ipilimumab-enabled effector T cells or are merely elevated as a result of the immunosuppressive environment found in patients refractory to ipilimumab therapy.

Y. Koguchi, F.R. Bahjat, H.M. Hoen, and A.J. Korman have ownership interest in a pending patent. C. Milburn is an employee of Bristol-Myers Squibb. F.R. Bahjat is the co-founder of NeurAlexo, VP Pharmalocalogy at Oncovir and reports receiving other commercial research support from Oncovir, Bristol Myers Squibb and is also a consultant/advisory board member of Bristol Myers Squibb. A.J. Korman has ownership interest in Bristol-Myers Squibb. K.S. Bahjat reports receiving commercial research grant from Bristol-Myers Squibb. No potential conflicts of interest were disclosed by the other authors.

Conception and design: Y. Koguchi, C. Milburn, A.J. Korman, K.S. Bahjat

Development of methodology: C. Milburn, K.S. Bahjat

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Koguchi, S.A. Bambina, R.K. Fuerstenberg, B.D. Curti, A.J. Korman, K.S. Bahjat

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Koguchi, H.M. Hoen, B.D. Curti, F.R. Bahjat, K.S. Bahjat

Writing, review, and/or revision of the manuscript: Y. Koguchi, H.M. Hoen, M.D. Rynning, R.K. Fuerstenberg, B.D. Curti, W.J. Urba, C. Milburn, F.R. Bahjat, A.J. Korman, K.S. Bahjat

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H.M. Hoen, K.S. Bahjat

Study supervision: B.D. Curti, A.J. Korman, K.S. Bahjat

The authors thank Gwen Kramer for assistance with sample preparation, Michael Gough, Marka Crittenden, and Will Redmond for helpful discussions, and the dedication and skill of the clinical research team at the Providence Cancer Center.

These studies were supported by institutional funding from the Providence Portland Medical Foundation and a sponsored-research grant from Bristol-Myers Squibb (BMS).

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.

1.
Wolchok
JD
,
Hodi
FS
,
Weber
JS
,
Allison
JP
,
Urba
WJ
,
Robert
C
, et al
Development of ipilimumab: a novel immunotherapeutic approach for the treatment of advanced melanoma
.
Ann N Y Acad Sci
2013
;
1291
:
1
13
.
2.
Rudd
CE
,
Taylor
A
,
Schneider
H
. 
CD28 and CTLA-4 coreceptor expression and signal transduction
.
Immunol Rev
2009
;
229
:
12
26
.
3.
Simpson
TR
,
Li
F
,
Montalvo-Ortiz
W
,
Sepulveda
MA
,
Bergerhoff
K
,
Arce
F
, et al
Fc-dependent depletion of tumor-infiltrating regulatory T cells co-defines the efficacy of anti-CTLA-4 therapy against melanoma
.
J Exp Med
2013
;
210
:
1695
710
.
4.
Hodi
FS
,
O'Day
SJ
,
McDermott
DF
,
Weber
RW
,
Sosman
JA
,
Haanen
JB
, et al
Improved survival with ipilimumab in patients with metastatic melanoma
.
N Engl J Med
2010
;
363
:
711
23
.
5.
Page
DB
,
Postow
MA
,
Callahan
MK
,
Allison
JP
,
Wolchok
JD
. 
Immune modulation in cancer with antibodies
.
Annu Rev Med
2014
;
65
:
185
202
.
6.
Wolchok
JD
,
Weber
JS
,
Maio
M
,
Neyns
B
,
Harmankaya
K
,
Chin
K
, et al
Four-year survival rates for patients with metastatic melanoma who received ipilimumab in phase II clinical trials
.
Ann Oncol
2013
;
24
:
2174
80
.
7.
Sharma
P
,
Allison
JP
. 
The future of immune checkpoint therapy
.
Science
2015
;
348
:
56
61
.
8.
Curti
BD
,
Urba
WJ
. 
Clinical deployment of antibodies for treatment of melanoma
.
Mol Immunol
2015
;
67
:
18
27
.
9.
Boyles
AL
,
Harris
SF
,
Rooney
AA
,
Thayer
KA
. 
Forest plot viewer: a new graphing tool
.
Epidemiology
2011
;
22
:
746
7
.
10.
Wilgenhof
S
,
Four Du
S
,
Vandenbroucke
F
,
Everaert
H
,
Salmon
I
,
Liénard
D
, et al
Single-center experience with ipilimumab in an expanded access program for patients with pretreated advanced melanoma
.
J Immunother
2013
;
36
:
215
22
.
11.
Simeone
E
,
Gentilcore
G
,
Giannarelli
D
,
Grimaldi
AM
,
Caracò
C
,
Curvietto
M
, et al
Immunological and biological changes during ipilimumab treatment and their potential correlation with clinical response and survival in patients with advanced melanoma
.
Cancer Immunol Immunother
2014
;
63
:
675
83
.
12.
Kelderman
S
,
Heemskerk
B
,
van Tinteren
H
,
van den Brom
RRH
,
Hospers
GAP
,
van den Eertwegh
AJM
, et al
Lactate dehydrogenase as a selection criterion for ipilimumab treatment in metastatic melanoma
.
Cancer Immunol Immunother
2014
;
63
:
449
58
.
13.
Di Giacomo
AM
,
Danielli
R
,
Calabrò
L
,
Bertocci
E
,
Nannicini
C
,
Giannarelli
D
, et al
Ipilimumab experience in heavily pretreated patients with melanoma in an expanded access program at the University Hospital of Siena (Italy)
.
Cancer Immunol Immunother
2010
;
60
:
467
77
.
14.
Delyon
J
,
Mateus
C
,
Lefeuvre
D
,
Lanoy
E
,
Zitvogel
L
,
Chaput
N
, et al
Experience in daily practice with ipilimumab for the treatment of patients with metastatic melanoma: an early increase in lymphocyte and eosinophil counts is associated with improved survival
.
Ann Oncol
2013
;
24
:
1697
703
.
15.
Hannani
D
,
Vétizou
M
,
Enot
D
,
Rusakiewicz
S
,
Chaput
N
,
Klatzmann
D
, et al
Anticancer immunotherapy by CTLA-4 blockade: obligatory contribution of IL-2 receptors and negative prognostic impact of soluble CD25
.
Cell Res
2015
;
25
:
208
24
.
16.
Yuan
J
,
Zhou
J
,
Dong
Z
,
Tandon
S
,
Kuk
D
,
Panageas
KS
, et al
Pretreatment serum VEGF is associated with clinical response and overall survival in advanced melanoma patients treated with ipilimumab
.
Cancer Immunol Res
2014
;
2
:
127
32
.
17.
Ku
GY
,
Yuan
J
,
Page
DB
,
Schroeder
SEA
,
Panageas
KS
,
Carvajal
RD
, et al
Single-institution experience with ipilimumab in advanced melanoma patients in the compassionate use setting
.
Cancer
2010
;
116
:
1767
75
.
18.
Mouawad
R
,
Spano
J-P
,
Khayat
D
. 
Old and new serological biomarkers in melanoma: where we are in 2009
.
Melanoma Res
2010
;
20
:
67
76
.
19.
Rochet
NM
,
Kottschade
LA
,
Grotz
TE
,
Porrata
LF
,
Markovic
SN
. 
The prognostic role of the preoperative absolute lymphocyte count and absolute monocyte count in patients with resected advanced melanoma
.
Am J Clin Oncol
2015
;
38
:
252
8
.
20.
Simon
R
. 
Clinical trials for predictive medicine: new challenges and paradigms
.
Clin Trials
2010
;
7
:
516
24
.
21.
Van Raemdonck
K
,
Van den Steen
PE
,
Liekens
S
,
Van Damme
J
,
Struyf
S
. 
CXCR3 ligands in disease and therapy
.
Cytokine Growth Factor Rev
2014
;
26
:
311
27
.
22.
Galon
J
,
Angell
HK
,
Bedognetti
D
,
Marincola
FM
. 
The continuum of cancer immunosurveillance: prognostic, predictive, and mechanistic signatures
.
Immunity
2013
;
39
:
11
26
.
23.
Ji
R-R
,
Chasalow
SD
,
Wang
L
,
Hamid
O
,
Schmidt
H
,
Cogswell
J
, et al
An immune-active tumor microenvironment favors clinical response to ipilimumab
.
Cancer Immunol Immunother
2011
;
61
:
1019
31
.
24.
Sauty
A
,
Colvin
RA
,
Wagner
L
,
Rochat
S
,
Spertini
F
,
Luster
AD
. 
CXCR3 internalization following T cell-endothelial cell contact: preferential role of IFN-inducible T cell chemoattractant (CXCL11)
.
J Immunol
2001
;
167
:
7084
93
.
25.
Zohar
Y
,
Wildbaum
G
,
Novak
R
,
Salzman
AL
,
Thelen
M
,
Alon
R
, et al
CXCL11-dependent induction of FOXP3-negative regulatory T cells suppresses autoimmune encephalomyelitis
.
J Clin Invest
2014
;
124
:
2009
22
.
26.
Griffith
JW
,
Sokol
CL
,
Luster
AD
. 
Chemokines and chemokine receptors: positioning cells for host defense and immunity
.
Annu Rev Immunol
2014
;
32
:
659
702
.
27.
Koch
MA
,
Tucker-Heard
G
,
Perdue
NR
,
Killebrew
JR
,
Urdahl
KB
,
Campbell
DJ
. 
The transcription factor T-bet controls regulatory T cell homeostasis and function during type 1 inflammation
.
Nat Immunol
2009
;
10
:
595
602
.
28.
Billottet
C
,
Quemener
C
,
Bikfalvi
A
. 
CXCR3, a double-edged sword in tumor progression and angiogenesis
.
Biochim Biophys Acta
2013
;
1836
:
287
95
.
29.
Robledo
MM
,
Bartolome
RA
,
Longo
N
,
Rodríguez-Frade
JM
,
Mellado
M
,
Longo
I
, et al
Expression of functional chemokine receptors CXCR3 and CXCR4 on human melanoma cells
.
J Biol Chem
2001
;
276
:
45098
105
.
30.
Monteagudo
C
,
Martin
JM
,
Jorda
E
,
Llombart-Bosch
A
. 
CXCR3 chemokine receptor immunoreactivity in primary cutaneous malignant melanoma: correlation with clinicopathological prognostic factors
.
J Clin Pathol
2007
;
60
:
596
9
.
31.
Longo-Imedio
MI
,
Longo
N
,
Treviño
I
,
Lázaro
P
,
Sánchez-Mateos
P
. 
Clinical significance of CXCR3 and CXCR4 expression in primary melanoma
.
Int J Cancer
2005
;
117
:
861
5
.
32.
Le Bert
N
,
Gasser
S
. 
Advances in NKG2D ligand recognition and responses by NK cells
.
Immunol Cell Biol
2014
;
92
:
230
6
.
33.
Raulet
DH
,
Gasser
S
,
Gowen
BG
,
Deng
W
,
Jung
H
. 
Regulation of ligands for the NKG2D activating receptor
.
Annu Rev Immunol
2013
;
31
:
413
41
.
34.
Baginska
J
,
Viry
E
,
Paggetti
J
,
Medves
S
,
Berchem
G
,
Moussay
E
, et al
The critical role of the tumor microenvironment in shaping natural killer cell-mediated anti-tumor immunity
.
Front Immun
2013
;
4
:
490
.
35.
Labiano
S
,
Palazon
A
,
Melero
I
. 
Immune response regulation in the tumor microenvironment by hypoxia
.
Semin Oncol
2015
;
42
:
378
86
.