Augmentation of IFN-γ+ CD8+ T cell responses correlates with survival of HCC patients on sorafenib therapy (original) (raw)
Patient characteristics. We enrolled 39 patients for the study between November 2013 and May 2017 at Roswell Park Comprehensive Cancer Center, and 30 patients were evaluable for all biomarker endpoints (Figure 1). Clinical characteristics of patients are summarized in Table 1.
Description of cohorts for biomarker study. We recruited 39 patients for the study. Radiographically, biochemically, or biopsy-proven, measurable HCC that was not previously treated with systemic therapy and was Child-Pugh Class A or B were the eligibility criteria. The observation period for each patient was the time from the start of therapy with sorafenib to withdrawal of consent, death, or the last visit. Of the 39, 30 patients met eligibility criteria. One pretreatment and 2 posttreatment blood samples were collected for the biomarker study. Patients were nonevaluable for the primary biomarker endpoint if they provided fewer than 2 biomarker samples or came off the study before response assessment. Of 30 eligible patients, 22 provided 3 blood samples for biomarker study and 8 provided 2 blood samples. Response, dosing, and toxicity information were collected in a database, and deidentified biomarker data were analyzed in a blind fashion. PFS, progression-free survival.
Baseline characteristics of patients treated with sorafenib (n = 30)
Clinical outcomes. Median 6-month PFS (primary endpoint for biomarkers) for Child-Pugh Class A and B was 0.42 (0.22–0.60) and 0.29 (0.09–0.52); median OS for Child-Pugh A and B was 9.2 (2.5–21.6) months and 4.6 (1.3–11.3) months, respectively. No clinical predictors were significant at P value less than 0.05 in this small data set in predicting outcomes. These clinical outcomes in all 30 patients with paired biomarker samples were used for correlation with immune biomarkers as shown below. Study status and survival in relation to Child-Pugh score are summarized in Supplemental Tables 1 and 2, respectively; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.130116DS1). No meaningful changes were seen in viral titers of patients with elevated titers at baseline, consistent with what others have reported, and no patients were on antivirals during sorafenib treatment (16).
Reduction in the frequency and number of Tregs and MDSCs following sorafenib treatment is associated with downregulation of ERK phosphorylation. The levels of circulating CD4+Foxp3+ T cells and CD14−HLADR−CD11b+CD33+ MDSCs in patients with HCC were measured by flow cytometry before sorafenib therapy as well as at early (28–35 days) and late treatment (160 days) stages (Figure 2, A and D). Statistically significant reduction in the frequency (7.3% vs. 5.5%, P = 0.001, Figure 2B) and absolute number (9537 vs. 7603, P = 0.05, Figure 2C) of CD4+Foxp3+ Tregs and in the frequency (6% vs. 4.3% P = 0.001, Figure 2E) and absolute number (5836 vs. 4350, P = 0.01, Figure 2F) of CD14−HLADR−CD11b+CD33+ MDSCs was observed in blood samples collected at 28–35 days of sorafenib treatment as compared with pretreatment levels. Decreases in Tregs (7.3% vs. 5.7% P = 0.01, Figure 2B; 9537 vs. 6591, P = 0.05, Figure 2C) and MDSCs (6% vs. 3.8%, P = 0.001, Figure 2E; 5836 vs. 3727, P = 0.01, Figure 2F) were sustained up to after 160 days of treatment and were also statistically significant compared with baseline values. Further decreases in frequencies and absolute numbers did not occur beyond that measured at 28–35 days of treatment, except for the absolute number of MDSCs, which further decreased significantly (Figure 2, B, C, E, and F).
Reduction in immunosuppressive cell subsets after sorafenib therapy. Frequencies of various immune T cell subsets were calculated on live CD3+CD4+ or CD3+CD8+ T cell populations and MDSCs are CD14−HLA–DR−CD11b+CD33+. Absolute number is cells per milliliter. (A) Representative histogram offset showing the frequency of CD4+Foxp3+ Tregs measured at pretreatment and after 28–35 days and 160 days of sorafenib treatment. (B) Frequency and (C) absolute numbers of CD4+Foxp3+ Tregs pretreatment and after 28–35 days and 160 days (n = 22). (D) Representative histogram offset showing frequency of CD11b+CD33+ MDSCs measured at pretreatment and after 28–35 days and 160 days of sorafenib treatment. (E) Frequency and (F) absolute numbers of MDSCs pretreatment and after 28–35 days and 160 days (n = 22). (G) Ratio of CD4+CD127+ T cells to CD4+Foxp3+ T cells (CD4+CD127+ T cells/CD4+Foxp3+ T cells) pretreatment and after 28–35 days and 160 days (n = 22). (H) Ratio of CD8+CD127+ T cells to CD4+Foxp3+ T cells pretreatment and after 28–35 days and 160 days (n = 22). (I) Representative histogram offset, showing frequency of Foxp3+flt-3+p-ERK+ Tregs measured at pretreatment and after 28–35 days and 160 days of sorafenib treatment. (J) Frequency and (K) absolute numbers of Foxp3+flt-3+p-ERK+ Tregs measured at pretreatment and after 28–35 days and 160 days of sorafenib treatment (n = 22). (L) Representative histogram offset, showing frequency of flt-3+p-ERK+ MDSCs measured at pretreatment and after 28–35 days and 160 days of sorafenib treatment. (M) Frequency and (N) absolute numbers of flt-3+p-ERK+ MDSCs pretreatment, 28–35 days, and 160 days (n = 22). Each symbol represents an individual HCC patient. Error bars represent mean ± standard deviation (SD). ****P < 0.0001; ***P < 0.001; **P < 0.01; *P < 0.05; paired t test.
In addition to the decline in Treg frequencies and absolute numbers, importantly, there was a significant increase in the ratios of both CD4+CD127+ (5.1 vs. 7.4, P = 0.001, Figure 2G) and CD8+CD127+ (6.1 vs. 8.6, P = 0.001, Figure 2H) effector T cells (Teffs) to Tregs after sorafenib treatment. Enhancement in the ratios of CD4+ Teffs and CD8+ Teffs to Tregs was sustained for a long period after sorafenib treatment (5.1 vs. 7.9, P = 0.01, Figure 2G; 6.1 vs. 8.6, P = 0.05, Figure 2H). Mitigation of Tregs and MDSCs and subsequent enhancement in the Teff/Treg ratio following sorafenib therapy that were long-lasting likely contribute to antitumor immune responses in these patients.
Coupled with diminution of Tregs and the MDSC population, a significant reduction in the levels of ERK phosphorylation was observed on flt-3+ Tregs and MDSCs after sorafenib treatment. We found that the frequency (65.5% vs. 50.5%, P = 0.001, Figure 2, I and J) and absolute number (383.5 vs. 233, P = 0.06, Figure 2K) of flt-3+ p-ERK+ Tregs and frequency (87.1% vs. 73.9%, P = 0.001, Figure 2, L and M) and absolute number (396 vs. 244.7, P = 0.01, Figure 2N) of flt-3+p-ERK+ MDSCs were significantly decreased in samples collected 28–35 days after initiation of sorafenib treatment. Decreased ERK phosphorylation levels on Tregs (65.5% vs. 37.62%, P = 0.001, Figure 2J; 383.5 vs. 169.4, P = 0.01, Figure 2K) and MDSCs (87.1% vs. 70.5%, P = 0.001, Figure 2M; 396 vs. 176, P = 0.01, Figure 2N) were also monitored at the second follow-up measurement after 160 days of sorafenib treatment. p-ERK is a key downstream target of sorafenib and could therefore serve as a potential biomarker of its efficacy. Our studies show that signaling via flt-3 receptor on Tregs and MDSCs is inhibited by sorafenib, resulting in the downregulation of ERK phosphorylation, implicating this as a potential mechanism by which sorafenib inhibits suppressive Tregs and MDSCs in patients with HCC. Eight of thirty patients had died before the second posttreatment blood draw (160 days), and immune profiles of these patients at 28–35 days combined with surviving patients (n = 22 + 8) have been summarized in Supplemental Figure 1. A comparison of the peripheral blood immune cell population with intrahepatic immune cell subsets could not be performed in this small cohort of 30 patients with advanced disease who presented with poor liver function because tumor biopsies for translational studies could have affected their current standard of care or quality of life. We acknowledge this as a limitation of our study.
Reduction in Foxp3+ Tregs coexpressing CTLA-4 and neuritin after sorafenib treatment. Our characterization of the pre- and posttreatment HCC patient samples demonstrated that the frequency of Foxp3+CTLA-4+ Tregs was significantly reduced after sorafenib therapy (4.3% vs. 3.5%, P = 0.01, Figure 3, A and B). However, absolute number of Foxp3+CTLA-4+ Tregs did not show any significant decrease (Figure 3C). In addition to reducing the frequency of Tregs expressing CTLA-4, in a subset of patients, sorafenib treatment also decreased the frequency of Tregs coexpressing the neurotrophic factor known as neuritin on their surface (40.5% vs. 24.5%, P = 0.01, Supplemental Figure 2, A and B), a molecule implicated as a Treg-associated factor and relevant to an activated Treg phenotype. The frequency of neuritin+ Tregs in HCC patients’ pretreatment samples was also significantly higher than age-matched healthy subjects (40.5% vs. 14.3%, P = 0.001, Supplemental Figure 2B). Thus, neuritin may be an additional marker of Tregs responsible for dampening antitumor immunity in HCC patients who are also susceptible to therapeutic disruption of VEGF/VEGFR signaling.
Decrease in T cell exhaustion markers after sorafenib treatment. Immunophenotypic analysis of T cells was performed after stimulation of PBMCs in vitro using anti–CD3/CD28 at pretreatment and after 28–35 days and 160 days of sorafenib treatment by multicolor flow cytometry. Frequencies of various immune cell subsets were calculated on live CD3+CD4+ or CD3+CD8+ T cell populations, and absolute number is cells per milliliter. (A) Representative histogram offset showing frequency of Foxp3+CTLA-4+ Tregs measured at pretreatment and after 28–35 days and 160 days of sorafenib treatment. (B) Frequency and (C) absolute numbers of Foxp3+CTLA-4+ Tregs pretreatment, 28–35 days, and 160 days (n = 22). (D) Representative histogram offset showing frequency of CD4+Ki67+PD-1+ T cells measured at pretreatment and after 28–35 days and 160 days of sorafenib treatment. (E) Frequency and (F) absolute number of CD4+Ki67+PD-1+ T cells pretreatment, 28–35 days, and 160 days (n = 22). (G) Representative histogram offset showing frequency of CD8+Ki67+PD-1+ T cells measured at pretreatment and after 28–35 days and 160 days of sorafenib treatment. (H) Frequency and (I) absolute number of CD8+Ki67+PD-1+ T cells pretreatment, 28–35 days, and 160 days (n = 22). Each symbol represents an individual HCC patient. Error bars represent mean ± standard deviation (SD). ****P < 0.0001; ***P < 0.001; **P < 0.01; paired t test.
Reduction in T cell exhaustion following treatment with sorafenib. We further examined the expression levels of immune checkpoint receptor PD-1, a molecular signature of T cell exhaustion on activated T cells, during sorafenib treatment in patients with HCC. A significant reduction in the frequency (31.5% vs. 26%, P = 0.001, Figure 3, D and E) as well as absolute number (13,508 vs. 10,460, P = 0.01, Figure 3F) of CD4+Ki67+PD-1+ T cells and in the frequency (29.8% vs. 21%, P = 0.001, Figure 3, G and H) and absolute number (5025 vs. 3643, P = 0.01, Figure 3I) of CD8+Ki67+PD-1+ T cells was observed 28–35 days after sorafenib treatment as compared with pretreatment levels. Of note, the frequency of CD4+PD1+ T cells (31.5% vs. 23.6%, P = 0.001, Figure 3E) and CD8+PD1+ T cells (29.8% vs. 20.5%, P = 0.001, Figure 3H) remained significantly low in the follow-up blood samples collected after 160 days of treatment, demonstrating that sorafenib-mediated downregulation of this inhibitory checkpoint on T cells was sustained for a prolonged period. Profiles of immune checkpoint receptors at 28–35 days of 8 patients who had died before the second posttreatment blood draw (160 days) combined with surviving patients (n = 22 + 8) have been summarized in Supplemental Figure 3. Our results underscore that immune checkpoint inhibitor therapy combined with sorafenib could generate a synergistic effect and may favor better clinical outcome in patients with HCC.
In addition to Foxp3+ Tregs, expression of neuritin has also been linked to anergic CD8+ and CD4+ T cell populations. We found that in patients with HCC, the frequency of CD4+PD-1+ T cells coexpressing neuritin was significantly higher than in normal healthy subjects, suggesting a role for these cells in liver cancer–associated immune dysfunction (33% vs. 21.3%, P = 0.05, Supplemental Figure 2, C and D). Notably, these cells were significantly reduced after sorafenib treatment (33% vs. 19.6%, P = 0.05, Supplemental Figure 2D). Neuritin expression on CD8+PD-1+ T cells was similarly reduced after sorafenib therapy (35.3% vs. 17.8%, P = 0.01, Supplemental Figure 2, E and F). This decline in PD-1+neuritin+ T cells, potentially exhausted or anergic populations capable of attenuating antitumor immune responses in HCC, may be partly attributed to a hitherto-unanticipated susceptibility of these populations to VEGFR signaling antagonism.
Enhanced IFN- γ production and proliferation of effector CD4+ and CD8+ T cells in patients treated with sorafenib. Teff proliferation and IFN-γ production were evaluated at 3 time points during sorafenib treatment: before treatment initiation and at 28–35 days and 160 days after commencement of treatment. Stimulation of PBMCs in vitro using anti–CD3/CD28 resulted in the generation of significantly increased frequency (7.9% vs. 15.3%, P = 0.001, Figure 4, A and B) as well as absolute number of CD4+Ki67+IFN-γ+ T cells (3101 vs. 8314, P = 0.001, Figure 4C) and significantly high frequency (9.9% vs. 17.5%, P = 0.001, Figure 4, D and E) and absolute number of CD8+Ki67+IFN-γ+ T cells (1094 vs. 3731, P = 0.001, Figure 4F) after 28–35 days of sorafenib treatment as compared with the pretreatment sample. We observed significantly higher frequencies as well as absolute numbers of proliferating CD4+Ki67+ (7.9% vs. 16.7%, P = 0.001, Figure 4B; 3101 vs. 10,865, P = 0.001, Figure 4C) and CD8+Ki67+ T cells (9.9% vs. 18.2%, P = 0.001, Figure 4E; 1094 vs. 4353, P = 0.001, Figure 4F) producing IFN-γ even at a later stage of treatment in surviving patients, highlighting the durability of the sorafenib-induced immune modulatory effect in HCC. Profiles of T cell IFN-γ production by 8 patients who had passed away before the second posttreatment blood draw combined with surviving patients (n = 22 + 8) have been summarized in Supplemental Figure 4.
Effect of sorafenib treatment on Teffs. PBMCs from patients with HCC were stimulated with anti–CD3/CD28 in vitro for 48 hours and immunophenotypic analysis was performed. Frequencies of various immune cell subsets were calculated on live CD3+CD4+ or CD3+CD8+ T cell populations, and absolute number is cells per milliliter. (A) Representative histogram offset showing frequency of CD4+Ki67+IFN-γ+ T cells measured at different time points of sorafenib treatment. (B) Frequencies and (C) absolute number of CD4+Ki67+IFN-γ+ T cells pretreatment, 28–35 days, and 160 days (n = 22). (D) Representative histogram offset showing frequency of CD8+IFN-γ+ T cells measured at different time points of sorafenib treatment. (E) Frequencies and (F) absolute number of CD8+Ki67+IFN-γ+ T cells pretreatment, 28–35 days, and 160 days (n = 22). (G) Representative histogram offset showing frequency of CD8+Ki67+GRB+ T cells measured at different time points of sorafenib treatment. (H) Frequencies and (I) absolute number of CD8+Ki67+GRB+ T cells pretreatment, 28–35 days, and 160 days (n = 22). Each symbol represents an individual HCC patient. Error bars represent mean ± standard deviation (SD). ****P < 0.0001; ***P < 0.001; **P < 0.01; paired t test.
Increased expression of granzyme B by proliferating cytotoxic T cells in patients treated with sorafenib. Further, we investigated the expression of granzyme B, a mediator of target cell lysis by CTLs in cell-mediated immune responses, at different time intervals during the course of treatment. Activated CD8+ T cells showed a significant increase in the frequency as well as absolute numbers of Ki67+granzyme B+ CTLs in blood samples collected 28–35 days (27.2% vs. 46.6%, P = 0.001, Figure 4, G and H; 2870 vs. 6631, P = 0.01, Figure 4I) and 160 days (27.2% vs. 48.4%, P = 0.001, Figure 4H; 2870 vs. 10,910, P = 0.001, Figure 4I) after initiation of sorafenib treatment as compared with pretreatment samples. Profiles of granzyme B+CD8+ T cells at 28–35 days from 8 patients who had died before the second posttreatment blood draw combined with surviving patients (n = 22 + 8) are summarized in Supplemental Figure 4.
Predictive and prognostic immune correlates of survival in patients with HCC. OS of patients was significantly affected by the absolute numbers of IFN-γ–producing CD8+Ki67+ T cells before treatment initiation (log-rank P = 0.01, Figure 5A). Both the frequencies and absolute numbers of this phenotype showed significant association with PFS (log-rank P = 0.005, Figure 5B; log-rank P = 0.003, Figure 5C). Furthermore, patients with an increase in the absolute numbers of this phenotype quantified after sorafenib therapy had improved PFS (log-rank P = 0.03, Figure 5D). Importantly, high frequencies of CD8+Ki67+IFN-γ+ T cells were also associated with significantly reduced risk of death over time (hazard ratio = 0.33; P = 0.005). Collectively, our results highlight that CD8+Ki67+ T cells producing IFN-γ may represent a key immune subset defining response to sorafenib treatment.
Kaplan-Meier plots showing both the predictive immune correlates of response to and efficacy of sorafenib treatment in patients with HCC. Plots A–C show the correlation between CD8+Ki67+IFN-γ+ T cells before sorafenib treatment (baseline values) and patient survival. The association between CD8+Ki67+IFN-γ+ T cells and OS or PFS was calculated as described in Methods. (A) Absolute numbers of baseline values of CD8+Ki67+IFN-γ+ T cells and OS. (B) Frequency of CD8+Ki67+IFN-γ+ T cells at baseline and PFS. (C) Absolute numbers of CD8+Ki67+IFN-γ+ T cells and PFS. (D) Log change in absolute numbers of CD8+Ki67+IFN-γ+ T cells (log of the ratio [post/pre] and PFS).
Patients with high pretreatment Teff/Treg ratio achieved significant improvement in OS (log-rank P = 0.03, Figure 6A). In corroboration of this finding, we also show that patients with low numbers of baseline Foxp3+CTLA-4+ Tregs had improved PFS (log-rank P = 0.007, Figure 6B).
Kaplan-Meier plots showing the baseline immune correlates predictive of response to sorafenib treatment in patients with HCC. Plots A–C show the correlation between immune parameters before sorafenib treatment and patient outcome. The association between immune markers and OS or PFS was calculated as described in Methods. (A) Ratio of CD4+CD127+ T cells/CD4+Foxp3+ Tregs and OS. (B) Absolute numbers of CD4+Foxp3+CTLA-4+ T cells and PFS. (C) Absolute numbers of flt-3+p-ERK+ MDSCs and PFS.
Pretreatment levels of the absolute number of flt-3+p-ERK+ MDSCs showed significant correlation with PFS of patients (log-rank P = 0.02, Figure 6C). In concordance with earlier studies reporting increased responsiveness of HCC patients with high baseline p-ERK+ tumors to sorafenib treatment, our HCC patients with increased numbers of baseline p-ERK+ MDSCs may be more susceptible to VEGFR signaling antagonism. Although there are multiple strategies that tumor cells use, MDSCs are a key driver of tumor-mediated immune evasion, and targeting MDSCs in immunotherapy appears to be a promising strategy with clinical relevance for patients with HCC.