The tumor immune microenvironment (TIME) has an important impact on response to cancer immunotherapy using immune checkpoint inhibitors. Specifically, an “infiltrated-excluded”/“cold” TIME is predictive of poor response. The antidiabetic agent metformin may influence anticancer immunity in esophageal squamous cell carcinoma (ESCC).
We analyzed matched pre- and posttreatment ESCC specimens in a phase II clinical trial of low-dose metformin treatment (250 mg/day) to evaluate direct anti-ESCC activity and TIME reprogramming. Follow-up correlative studies using a carcinogen-induced ESCC mouse model were performed with short-term (1 week) or long-term (12 weeks) low-dose metformin (50 mg/kg/day) treatment.
In the clinical trial, low-dose metformin did not affect proliferation or apoptosis in ESCC tumors as assayed by Ki67 and cleaved caspase-3 immunostaining. However, metformin reprogrammed the TIME toward “infiltrated-inflamed” and increased the numbers of infiltrated CD8+ cytotoxic T lymphocyte and CD20+ B lymphocyte. Further, an increase in tumor-suppressive (CD11c+) and a decrease in tumor-promoting (CD163+) macrophages were observed. Metformin augmented macrophage-mediated phagocytosis of ESCC cells in vitro. In the ESCC mouse model, short-term metformin treatment reprogrammed the TIME in a similar fashion to humans, whereas long-term treatment further shifted the TIME toward an active state (e.g., reduction in CD4+ FoxP3+ regulatory T cells) and inhibited ESCC growth. In both humans and mice, metformin triggered AMPK activation and STAT3 inactivation, and altered the production of effector cytokines (i.e., TNFα, IFNγ, and IL10) in the immune cells.
Low-dose metformin reprograms the TIME to an activated status and may be a suitable immune response modifier for further investigation in patients with ESCC.
The tumor immune microenvironment (TIME) has an important impact on response to immune checkpoint inhibitor therapy for cancer. Therefore, therapeutic strategies that can reshape TIME toward a more active state may improve the treatment outcome. Here, we described the first phase II clinical trial of low-dose metformin (250 mg/day) in human esophageal squamous cell carcinoma (ESCC) that demonstrated the impact of metformin on B cells, T cells, and macrophages in TIME. Our mouse model data corroborated and extended the findings that the impact on the TIME is sustained with long-term low-dose metformin treatment. The current discovery highlights that low-dose metformin reprograms the TIME to an activated state and may be a safe and efficacious pretreatment/combination option to boost the effectiveness of immunotherapy (e.g., CD47 blocking agents) in future clinical trials. Low-dose metformin may be a suitable immune response modifier in patients with ESCC that can be easily integrated in routine care.
Complex microenvironments have evolved during tumorigenesis to facilitate cancer growth (1). The tumor microenvironment contains many types of immune cells: dendritic cells (DC), natural killer (NK) cells, macrophages, granulocytes, and mast cells from the innate immune system, and T and B lymphocytes from the adaptive immune system (2–4). This tumor immune infiltrate contributes to define a pathologically active and specialized cancer niche, the tumor immune microenvironment (TIME).
The TIME in solid tumors can be characterized as three states: active (A-TIME), equilibrated (E-TIME), and suppressive (S-TIME; refs. 5–8). A hallmark of A-TIME is strong infiltration with CD8+ effector T cells and tumor-suppressive macrophages. The S-TIME is characterized by increased CD4+ helper and regulatory FoxP3+ T cells and tumor-promoting macrophages (1, 9). A major hallmark of the S-TIME is T-cell exhaustion induced by activation of inhibitory immune checkpoints, such as PD-1/PD-L1 (6). The E-TIME is characterized by equal infiltration of immune effector cells and immunosuppressive cells (10–12). Patients with A-TIME were found to have more favorable clinical outcomes compared with patients with S-TIME (10). Patients with E-TIME also had longer survival than those with S-TIME (10–12). Therefore, therapeutic strategies that can reshape the TIME toward a more reactive phenotype would be expected to improve the outcome of cancer immunotherapy.
Metformin, a widely used drug for type II diabetes, is being investigated in many clinical trials for human cancers, with doses ranging from 250 to 2,000 mg/day (13–17). Previously, we reported that metformin also inhibited the growth and progression of esophageal squamous cell carcinoma (ESCC) in preclinical models (18, 19). A recent retrospective study revealed that metformin benefited patients with ESCC with type II diabetes (20). In recent studies, there is evidence that metformin affects the TIME, an immunomodulatory effect that is yet to be completely defined (21–23). Recent evidence also suggests that the antitumor activity of metformin can, at least partly, be attributed to immune activation (24). Thus, metformin appears to have immunomodulatory effects that may be harnessed for cancer immunotherapy.
Here, we performed the first randomized phase II clinical trial in patients with ESCC with a low-dose (250 mg/day) and short-term treatment with metformin to determine: (i) the direct impact of metformin on ESCC cells and (ii) the impact of metformin on the TIME status. The human data were further corroborated and extended using an orthotopic ESCC mouse model and in vitro immune functional experiments.
Materials and Methods
We recruited 128 patients in a phase II trial (registration number: ChiCTR-ICR-15005940) from the Cancer Hospital of Shantou University Medical College (CHSUMC) between September 2014 and September 2016. Written informed consents were obtained from all participants in accordance with the principles established by the Helsinki Declaration. The clinical protocol was approved by the Institution Ethics Committee and Institutional Review Board of CHSUMC (2014060938). The inclusion criteria were: (i) 18–75 years of age, (ii) clinical diagnosis of ESCC, and (iii) scheduled to undergo surgical resection without neoadjuvant chemotherapy. The exclusion criteria were: (i) renal insufficiency, (ii) pregnant or lactating females, (iii) history of other malignancies, (iv) unstable angina, uncontrolled ischemic cardiomyopathy, or congestive heart failure with symptoms (e.g., New York Heart Association functional class III or IV), (v) diabetics receiving insulin, metformin, or sulfonylurea medications, (vi) history of lactic acidosis, (vii) chronic liver disease or cirrhosis, (viii) allergic to or intolerant of metformin, and (ix) inability to provide written informed consent.
Clinical study design
In this double-blind study, enrolled patients underwent endoscopic biopsy of esophageal tumors, and participants with a pathologic diagnosis of ESCC were randomized to receive metformin 250 mg orally per day or placebo tablets of identical appearance in the same regimen for a convenient duration of 7 to 14 days before surgery. Surgical samples of cancer from study subjects were freshly collected for evaluation. Fasting blood glucose was checked before and after drug treatment. The primary objective of this trial was to investigate the impact of metformin on cell proliferation and apoptosis markers, Ki67 and cleaved caspase-3, in ESCC tissues. The secondary objective was to evaluate the impact of metformin on immune components in ESCC tissues. The study subjects were staged according to the Union for International Cancer Control Tumor–Node–Metastasis staging system (7th edition; ref. 25).
TIME status was analyzed and categorized with some modifications according to previous reports (1, 9, 10). In brief, the median of the percentage of marker-positive cells was used as a cutoff value to dichotomously classify each immune cell type into a rich or poor status. Markers used were CD8, CD4, and FoxP3 for T-cell populations, CD20 for B cells (and CD19 for the mouse studies), and CD68, CD11c, and CD163 for macrophages (with F4/80 and CD206 for mouse studies; refs. 1, 9, 10). The strong infiltration of the lesion with CD8+ effector T cells (1, 9, 21, 26), tumor-suppressive macrophages (CD68+ and CD11c+; refs. 1, 9, 27), and B cells (CD20+; refs. 28–30) was reported to be the hallmark of inflamed status, whereas the immunosuppressed status was characterized by rich CD4+ helper T cells (1, 9, 31), regulatory FoxP3+ T cells (1, 9, 32), and tumor-promoting macrophages (CD163+; refs. 1, 9). Here, the TIME was considered as activated (A-TIME) if a majority (≥4) of 7 markers contributing to an inflamed status (CD8-rich, CD4-poor, FoxP3-poor, CD20-rich, CD68-rich, CD11c-rich, and CD163-poor) were detected simultaneously. The TIME was considered as suppressive (S-TIME) if a majority (≥4) of 7 markers contributed to an immunosuppressed status (CD8-poor, CD4-rich, FoxP3-rich, CD20-poor, CD68-poor, CD11c-poor, and CD163-rich). In all other cases, the TIME was considered as in equilibrium (E-TIME). The TIME status before and after treatment was analyzed as aforementioned, and treatment-induced changes were classified as: (i) no change, (ii) positive (anticancer) change, and (iii) negative (procancer) change.
All protocols and procedures for animal experiments were reviewed and approved by Ethics Committee of Shantou University Medical College (SUMC) and the Chancellor's Animal Research Committee at SUMC (SUMC2014–148). Six-week-old female C57BL/6 mice (Vital River Lab Animal Technology Co Ltd.) were given the carcinogen 4-Nitroquinoline N-oxide (4-NQO, Cat. N8141; Sigma) in drinking water (100 μg/mL) for 16 weeks to induce ESCC. The mice were randomized into four groups (10/group). To investigate the short-term effects of low-dose metformin on ESCC, metformin (50 mg/kg) or PBS (same volume) was injected intraperitoneally for 1 week and then the mice were euthanized. To investigate the long-term effects, an additional two groups were treated with metformin or PBS the same way except for 12 weeks.
SPSS 20.0 statistical software package (IBM Corp.) and R Version 3.5.3 (The R Project for Statistical Computing, http://www.r-project.org) were used. Values before and after treatment for the same study subjects were compared by the paired t test. Comparisons between independent groups were performed with the independent t test or Wilcoxon test, where appropriate, for numeric values. Ratios and proportions were analyzed with the χ2 test or Fisher exact test where appropriate. Bonferroni correction was applied for multiple testing where appropriate. Correlation between immune makers was examined by Pearson's correlation test. A multivariate regression analysis was conducted to determine the relationship between immune markers expression and clinicopathologic variables.
Details for biochemical assays are included in Supplementary Materials and Methods.
Low-dose metformin treatment does not affect blood glucose balance nor directly affects human ESCC cells
In the clinical trial, subjects with pathologically confirmed ESCC were randomized into two arms with 46 receiving 250 mg metformin per day and 42 receiving placebos (Supplementary Fig. S1A). Subjects in both arms took the study medication for a convenient duration of 7 to 14 days until tumor resection (Supplementary Fig. S1B). Eventually, 38 evaluable patients from each group were analyzed. The mean treatment duration for the metformin and placebo groups was 10.05 and 10.34 days, respectively. The demographic and clinicopathologic characteristics were summarized (Supplementary Table S1). Neither treatment with metformin nor placebo changed fasting blood glucose level (Supplementary Fig. S2A). Further, metformin or placebo treatment did not change cell proliferation in ESCC specimens as determined by Ki67 staining of cancer in surgical specimens (Supplementary Fig. S2B). Similarly, no impact on the amount of cleaved caspase-3 (a marker of apoptosis) was detected after treatment with metformin or placebo (Supplementary Fig. S2C). Thus, low-dose metformin neither had an impact on blood glucose homeostasis nor a direct effect on proliferation and apoptosis of ESCC cells.
Low-dose metformin increased CD8 T-cell and CD20 B-cell infiltration in human ESCC
Next, we determined whether low-dose metformin had any impact on the TIME by determining the composition of immune cell infiltrates before and after treatment. To profile the impact of metformin on adaptive immunity, CD8, CD4, FoxP3, and CD20 were immunostained in pre- and posttreatment tumor samples, with no significant difference in any of these markers between placebo and metformin-treated groups in the pretreatment samples (Fig. 1A and B). However, the infiltration of CD8+ T cells was significantly increased in the metformin group, but not placebo group, in posttreatment samples (P < 0.001; Fig. 1A). Correspondingly, in posttreatment conditions, a significantly higher percentage of CD8+ T cells was detected in the metformin group compared with the placebo group (P < 0.001; Fig. 1A). For the infiltration of CD4+ and FoxP3+ T cells, no significant change was detected in pre- versus posttreatment specimens in either study arm, although a significant decrease in CD4+ T cells was detected after treatment in the metformin group compared with placebo group (Fig. 1A). Finally, metformin treatment, but not placebo, significantly increased CD20+ B-cell infiltration (P = 0.002; Fig. 1B), although no significant change was detected after treatment between metformin and placebo groups (Fig. 1B). These IHC results were confirmed using multicolor fluorescence microscopy (Fig. 1C). CD8+ T cells and CD20+ B cells were increased in metformin-treated patients (P = 0.035 and P < 0.001, respectively; Fig. 1C), but not CD4+ T cells. Although PD-L1 is a crucial checkpoint protein associated with T-cell exhaustion and S-TIME (33, 34), no change in PD-L1 expression was detected either with IHC or immunofluorescence in either study arm (Fig. 1D and Supplementary Fig. S3). Collectively, these data showed that short-term low-dose treatment with metformin altered the composition of adaptive immune cells infiltrated in ESCC. Importantly, in multivariate regression analyses, metformin treatment remained an independent factor for change in both CD8 and CD20 staining (P < 0.001 and P = 0.001, respectively; Supplementary Fig. S4).
Low-dose metformin altered macrophage composition in human ESCC
In addition to adaptive immune cells, innate cells such as macrophages are critical players in the TIME. Tumor-promoting macrophages are associated with S-TIME and tumor-suppressive macrophages with A-TIME. In line with the results on adaptive immunity, none of the myeloid markers tested (CD68, CD11c, and CD163) were significantly differently expressed between the placebo and metformin groups in the pretreatment samples (Fig. 1E). Further, the total infiltration of macrophages (defined by CD68+ cells) was not changed in either the metformin or placebo group in pre- versus posttreatment tumor samples (Fig. 1E). However, in the metformin group, but not placebo group, the posttreatment samples had a significant increase in CD11c+ tumor-suppressive macrophages (P < 0.001; Fig. 1E). Correspondingly, a significantly higher percentage of CD11c+ tumor-suppressive macrophages was detected in posttreatment samples of metformin-treated patients compared with placebo-treated patients (P < 0.001; Fig. 1E). Reversely, metformin treatment decreased the percentage of CD163+ tumor-promoting macrophages in posttreatment samples (P < 0.001; Fig. 1E), whereas again no significant difference was detected in pre- versus posttreatment samples upon placebo treatment (Fig. 1E). Correspondingly, a significantly lower percentage of CD163+ tumor-promoting macrophages was detected in posttreatment samples of metformin-treated patients compared with placebo-treated patients (P < 0.001; Fig. 1E). Metformin remained an independent factor for change in CD11c and CD163 in multivariate regression analysis (P = 0.001 and P < 0.001, respectively; Supplementary Fig. S4).
Low-dose metformin shifts the TIME toward an anticancer state
In order to define the cumulative impact of the changes in individual markers, the TIME status was calculated for pre- and posttreatment samples and defined as S-TIME, E-TIME, or A-TIME. Treatment with metformin clearly altered the TIME composition, with 17 patients (45%) undergoing a favorable change in TIME after treatment compared with only 3 patients (8%) in placebo control (Fig. 2A; Supplementary Table S2). Indeed, metformin-treated tumors were statistically significantly more likely to undergo a positive TIME change than placebo with an OR of 9.44 (95% confidence interval, 2.47–36.12, P = 0.001) of positive TIME change versus no or negative TIME change in metformin versus placebo.
When assessing matched pre- and posttreatment TIME status, only minimal positive changes and many negative changes in TIME status were detected in placebo-treated patients (Fig. 2B). In contrast, in metformin-treated patients, a high number of positive TIME changes were detected (Fig. 2B). Specifically, in placebo-treated patients, a positive change from S-TIME or E-TIME was detected in only a few patients (Fig. 2C; 2 samples, 20% and 1 sample, 6%, respectively), whereas in metformin-treated patients, a large proportion of patients with pretreatment S-TIME or E-TIME underwent a positive TIME change (Fig. 2C; 10 samples, 77% and 7 samples, 47%, respectively). Of note, in pretreatment samples with an E-TIME composition, a negative change from E-TIME to S-TIME was detected with a similar frequency for placebo and metformin treatment (Fig. 2B; 5 samples each; 29% vs. 33%). However, a negative change from A-TIME to S-TIME was only detected in the placebo group (Fig. 2B; 3 samples, 27%). Taken together, low-dose metformin had a strong positive effect on TIME status, with a predominant shift toward a more antitumoral composition.
Specific correlations among immune cells were investigated in the two study arms. Particularly in the metformin arm, increased CD8 expression was accompanied by decreased CD163 in majority of the posttreatment samples of metformin-treated patients (22 out of 38 patients, 57.89%) compared with pretreatment with a strong negative correlation (Fig. 2D; P = 0.001). Further, increased CD8 expression was accompanied by increased CD20 in the majority of the posttreatment samples of metformin-treated patients (20 out of 38 patients, 52.63%) compared with pretreatment with a positive correlation (Fig. 2D; P = 0.038).
Treatment of metformin in carcinogen-induced ESCC mouse model mirrors metformin-induced TIME changes in patients with ESCC
Next, we confirmed these clinical findings in an ESCC mouse model in which the carcinogen 4-NQO was provided in the drinking water for 16 weeks (Fig. 3A). Treatment with a low dose of metformin (50 mg/kg/day) for 1 week (week 16 to week 17) did not significantly affect ESCC proliferation or the number of tumors per mouse compared with SHAM (PBS)–treated mice (Supplementary Fig. S5). However, tumors from mice in the metformin-treated group had significantly more A-TIME status than the SHAM-treated group (60% vs. 10%, P = 0.041; Fig. 3B). Specifically, CD8+ T cells and CD19+ B cells were increased in the metformin group compared with the control group (P = 0.005 and P = 0.011, respectively; Fig. 3C–E), with no significant change in CD4+ T cells, FoxP3+ regulatory T cells (Treg), or expression of PD-L1. Further, metformin treatment did not significantly increase the total number of F4/80+ macrophage populations (Fig. 3F), but CD11c+ tumor-suppressive macrophages increased (P = 0.005; Fig. 3F) along with a concomitant decreased level of CD206+ tumor-promoted macrophages (P = 0.007; Fig. 3F). Collectively, these data indicate that low-dose and short-term metformin treatment in this ESCC mouse model closely recapitulated the effect of metformin in patients with human ESCC.
When low-dose treatment was extended to 12 weeks, ESCC cell proliferation as well as the number of tumors per metformin-treated mouse was reduced compared with SHAM-treated mice (P = 0.005, Fig. 4A; P = 0.030, Fig. 4B). Moreover, the number of mice with A-TIME status was increased in metformin-treated mice compared with SHAM-treated mice (70% vs. 10%, P = 0.013; Fig. 4C).
Similar to the results from the short-term metformin experiment, long-term treatment also significantly increased the percentage of CD8+ T cells and CD19+ B cells (P = 0.002, Fig. 4D; P = 0.002, Fig. 4E). In addition, this long-term metformin treatment did reduce the percentage of CD4+ T cells and particularly of FoxP3+ T cells (P = 0.033 and P = 0.035, respectively; Fig. 4D). Moreover, expression of PD-L1 was significantly decreased after long-term metformin treatment compared with SHAM treatment (P < 0.001; Fig. 4F). Regarding macrophage content, the long-term metformin treatment also increased CD11c+ tumor-suppressive macrophages and severely suppressed CD206+ tumor-promoting macrophages, yielding a net decrease in total number of F4/80+ macrophages (P = 0.005, P < 0.001, and P = 0.034, respectively; Fig. 4G). Thus, long-term metformin treatment changed TIME toward E- and A-TIME and suppressed tumor growth in the mouse model.
Metformin augments macrophage phagocytic activity
One of the more prominent TIME changes in all analyses was the increase in tumor-suppressive macrophages and simultaneous decrease in tumor-promoted macrophages, suggesting metformin may augment antitumor innate immunity. To substantiate this finding in in vitro phagocytosis assays, ESCC cell line KYSE140 was pretreated with different concentrations of metformin (1, 3, and 5 mmol/L) for 48 hours, with no induction of apoptosis at 1 mmol/L and a slight trend toward apoptosis at 3 mmol/L in KYSE140 (Supplementary Fig. S6A). Interestingly, in mixed culture phagocytosis assay, a significant increase in phagocytic uptake of tumor cells was observed when macrophages pretreated with metformin for 48 hours with or without KYSE140 cells pretreated with metformin for 48 hours (Supplementary Fig. S6B and S6C). Similar results were observed in another ESCC cell line (Supplementary Fig. S7). These findings indicate that low-dose short-term metformin exposure specifically activated macrophage phagocytic activity.
Low-dose metformin triggered metabolic signaling and altered the immune cell cytokine expression profile
To identify metformin-mediated activity and modulation of the TIME, we evaluated phosphorylation of AMPK as a well-established sensor of metformin-mediated activity (32, 35, 36) and phosphorylation of STAT3, which has recently been identified as regulated by metformin (19, 37, 38). In patients with ESCC, metformin treatment significantly increased the percentage of CD8+ T cells positive for phospho-AMPK (p-AMPK), with an increase from approximately 16% to 49% in pretreatment versus metformin treatment (P < 0.001; Supplementary Fig. S8A). Similarly, the percentage of CD11c+ tumor-suppressive macrophages positive for p-AMPK increased from approximately 14% to 55% (P < 0.001; Supplementary Fig. S8B). In contrast, no change in p-AMPK was detected in placebo-treated groups for either CD8- or CD11c-positive cells (Supplementary Fig. S8). Conversely and in line with recent publications, treatment with metformin significantly decreased the percentage of CD8+ T cells and CD11c+ macrophages positive for phospho-STAT3 (p-STAT3), from 63% to 10% and 61% to 9%, respectively (P < 0.001; Supplementary Fig. S8). Again, in placebo-treated patients, no change was detected in p-STAT3 in pre- or posttreatment samples (Supplementary Fig. S8).
In close agreement with these findings, the low-dose and short-term metformin treatment in the ESCC mouse model significantly increased the percentage of p-AMPK–positive CD8+ T cells and CD11c+ macrophages compared with SHAM-treated mice (P < 0.001; Supplementary Fig. S9) while decreasing the percentage of p-STAT3–positive cells (P < 0.001; Supplementary Fig. S9). These effects on phosphorylation of AMPK and dephosphorylation of STAT3 became even more pronounced in the low-dose and long-term metformin-treated ESCC mice (P < 0.001; Supplementary Fig. S10). Thus, low-dose metformin metabolically activated AMPK and inactivated STAT3 in both cytotoxic T cells and tumor-suppressive macrophages.
As modulation of AMPK and STAT3 is known to alter the cytokine secretion profile of immune cells (26, 31, 39), we determined effector cytokine expression in CD8+ T cells and CD11c+ macrophages. In placebo-treated patients, no significant changes were detected in either CD8- or CD11c-positive cells for any of the cytokines analyzed (Fig. 5). However, metformin treatment significantly increased the proportion of CD8+ T cells expressing TNFα from approximately 12% to 46% (P < 0.001; Fig. 5A), whereas no effect was detected on IFNγ (Fig. 5A). A similar increase in terms of TNFα-positive CD11c+ macrophages was detected, from approximately 14% to 47% (P < 0.001; Fig. 5B). Further, metformin treatment significantly decreased the percentage of IL10-positive CD11c+ tumor-suppressive macrophages form approximately 46% to 7% (P < 0.001; Fig. 5B). Again, these results were faithfully replicated in mice treated with low-dose and short-term metformin, with a significant increase in CD8+ and CD11c+ cells expressing TNFα (P < 0.001; Supplementary Fig. S11), a significant decrease in IL10-positive CD11c+ tumor-suppressive macrophages (P < 0.001; Supplementary Fig. S11B), and no change in IFNγ expression (Supplementary Fig. S11A). In mice treated with low-dose metformin for long term, the changes in cytokine were analogous yet more pronounced (P < 0.001; Supplementary Fig. S12). Furthermore, these more pronounced changes in TNFα and IL10 were accompanied by a significant increase in CD8+ T cells expressing IFNγ, from approximately 17% to 62% (P < 0.001; Supplementary Fig. S12A).
Taken together, these data indicate that metformin treatment increased AMPK signaling, leading to an increase in proinflammatory cytokine expression in key immune effector cell types in the tumor. These results were closely mirrored in the mouse model clearly indicating that short-term metformin treatment triggered a proinflammatory and antitumoral reshaping of the TIME. Long-term treatment with metformin further augmented this antitumoral TIME shift.
In the current study, short-term administration of low-dose metformin reprogrammed the TIME in patients with ESCC from unfavorable S-TIME (immune-suppressive) to E-TIME (equilibrated) or A-TIME (activated). The changes in patients with ESCC were almost identical to those observed in the 4-NQO–induced ESCC mouse model. Long-term metformin treatment in this mouse model further resulted in a more robust TIME alteration and inhibition of tumor progression. To the best of our knowledge, this is the first reported phase II clinical trial investigating the usage of low-dose metformin in human ESCC and the first report detailing the impact of low-dose metformin on the TIME in ESCC.
The pro- or antitumoral state of the immune infiltrate in cancer is held to be of crucial importance for the efficacy of cancer (immune) therapy (1–3). For instance, S-TIME not only provides a driving force for tumor progression, but also facilitates drug resistance (40), whereas A-TIME associates with better prognosis. Our results indicate that metformin is able to trigger such TIME shifts at a low dose and in a short treatment schedule, which would make this metformin protocol fully compatible with integration into standard-of-care treatment in patients with ESCC.
Although initial focus was on direct anticancer activity of metformin, metformin-mediated immunomodulatory effects have started to be recognized as well. In this respect, high-dose metformin treatment (1,000 mg twice daily) for a mean duration of 13.6 days was recently reported to increase CD8+ T cells and FoxP3+ T cells in patients with head and neck squamous cell carcinoma (13). Such a high dose of metformin is in line with the majority of clinical studies in patients with cancer, with dosing typically ranging from 500–2,000 mg per day (14, 15, 41). Further, metformin improves antitumor T-cell immunity in patients with ovarian cancer (21). Correspondingly, immune-mediated anticancer effects were uncovered using immunocompetent mouse models (TRAMP mice with prostatic cancer and BALB/c mice bearing RLmale1 leukemia cells; refs. 23, 26). Further, metformin improves antitumor T-cell immunity in preclinical models of renal cell carcinoma, melanoma, fibrosarcoma, leukemia, or hepatocellular carcinoma (26, 32, 42), either by boosting CD8+ T-cell immunity or attenuating the ability of naive CD4+ cells to differentiate into FoxP3+ Tregs. Notably, although isolated reports on immune modulation in humans and various mouse models exist, the overall impact of metformin on TIME is unclear nor is there a clear correlation established between preclinical and clinical effects of metformin. Our immunocompetent mouse model clearly demonstrates a large spectrum of immune-mediated antitumor effects of metformin even at a low dosage of metformin. These effects on immunity precede an impact on ESCC cells, suggesting TIME remodeling is perhaps the primary mode of action of this low dose of metformin. More importantly, there is a remarkable uniformity in TIME compositional changes observed in patients with ESCC and the mouse model we employed. Thus, this mouse model may well serve as a relevant tool to predict response for potential metformin-based cancer immunotherapy combinations.
One of the prominent effects of metformin was an increase in antitumoral tumor-suppressive macrophages and a decrease in protumoral macrophages. Interestingly, in vitro metformin-treated macrophages proved to have higher phagocytic activity toward ESCC cells, together indicating that metformin may (re)activate macrophage-mediated immunity. These data are in line with previous studies in which metformin activated macrophage-mediated immune responses in hepatocellular carcinoma and glioma (42, 43). These findings suggest that metformin treatment reprograms macrophages into an antitumor mode and sets the stage for potential combination innate immune-targeting strategies such as combination with CD47-blocking or CD24-blocking antibodies that remove CD47/SIRP-alpha- or CD24/Siglec-10–mediated “don't eat me” signaling (44). The mechanism for this metformin-mediated shift in innate immune balance remains to be determined, but at least in part appears to be due to direct impact on macrophage biology.
Although there is no clear molecular mechanism through which metformin modulates tumor-associated immune cells, it has been widely appreciated that AMPK activation correlates with metformin treatment (35). Metformin-mediated activation of AMPK may induce the downregulation of NF-kB and STAT signaling (26, 45), although AMPK-independent regulation of such inflammatory signaling by metformin has been reported in some settings (19, 37, 38). Our results derived from both humans and mice clarified that low-dose metformin treatment activated AMPK and inactivated STAT3 in tumor-suppressive macrophages and cytotoxic T cells. Compared with short-term treatment, long-term treatment in the mouse model induced a more robust and sustained response of these signaling pathways. Given that both AMPK and STAT3 are central regulators of cellular metabolism and inflammation, our data favor the notion that metformin mediates the metabolic-inflammatory signaling pathways. Short-term metformin treatment also triggered an increase in TNFα and a decrease in IL10 expression in tumor-suppressive macrophages, and increased expression of TNFα in cytotoxic T cells. Long-term treatment with metformin in addition increased IFNγ expression in cytotoxic T cells. These cytokines are clearly associated with antitumoral activity of immune cells. These data provide insights with regard to metformin-dependent functional differentiation of immune cells.
Of note, we also identified a higher proportion of B cells in patients with ESCC and the mouse model after metformin treatment. Tumor-infiltrating B cells have been reported in mouse cancer models (46) and human solid tumors (30), but the role of tumor-infiltrating B cells remains unclear. Future studies are being set up to examine the specific composition and function of tumor-infiltrating B cells upregulated by metformin, e.g., by assessing the impact of CD20-mediated B-cell depletion in the ESCC mouse model.
Our studies demonstrate that low-dose metformin treatment clearly shifts the TIME in patients with ESCC from S-TIME toward a more activated E-TIME and A-TIME. To gain insight into the clinical relevance of this finding, follow-up prospective studies that correlate TIME alterations with relapse rate, progression-free survival, and overall survival are warranted. Moreover, development and potential expansion of tumor-reactive T-cell clones due to metformin treatment should be evaluated, e.g., by prospective monitoring of tumor-reactive T-cell clones. In this respect, blood of patients with ESCC is known to contain NY-ESO-1– and PRAME-reactive T cells.
Several limitations exist for our study. First, the clinical trial was designed as a randomized controlled trial with a relatively small number of participants to examine pharmacodynamic response biomarkers. Although our mouse model data showed that a longer treatment duration resulted in a more robust and lasting response, the pharmacodynamic effects of long-term low-dose metformin treatment in patients with ESCC needs to be recapitulated. Second, the TIME was highly heterogeneous among patients because TIME can be influenced by many factors. In addition to the main composition of TIME (T cells, B cells, and macrophages) investigated in our study, other immune cells such as DCs, NK cells, granulocytes, and mast cells were not included in our analyses. Whether metformin affected polarization of macrophages and which B-cell subsets were altered by metformin remain significant questions for further investigation.
In conclusion, with this phase II clinical trial and correlative mouse model studies, we demonstrate that low-dose metformin treatment shifts the TIME in ESCC from a protumoral state toward a more antitumoral state (Fig. 6). These results provide the basis for designing future clinical trials for evaluating the potential impact of metformin on long-term outcome as well as set the stage for clinical trials with cancer immunotherapeutics such as CD47 or CD24 blocking agents, in which metformin treatment provides a safe immunomodulatory strategy.
Disclosure of Potential Conflicts of Interest
S.-C.J. Yeung reports grants from Depomed, Inc., grants from Bristol-Myer Squibb, and personal fees from Celgene, Inc. outside the submitted work. No potential conflicts of interest were disclosed by the other authors.
S. Wang: Data curation, software, formal analysis, validation, investigation, visualization, methodology, writing-original draft, project administration, writing-review and editing. Y. Lin: Data curation, software, formal analysis, validation, investigation, visualization, methodology, writing-original draft, writing-review and editing. X. Xiong: Data curation, software, formal analysis, validation, methodology. L. Wang: Data curation, software, formal analysis, validation, visualization, methodology. Y. Guo: Resources, data curation, investigation. Y. Chen: Resources, data curation, investigation. S. Chen: Resources, data curation, investigation. G. Wang: Resources, data curation, investigation. P. Lin: Data curation, software, formal analysis, validation, investigation, visualization, methodology. H. Chen: Data curation, investigation, methodology. S.-C.J. Yeung: Conceptualization, data curation, software, formal analysis, supervision, validation, methodology, project administration, writing-review and editing. E. Bremer: Software, formal analysis, validation, visualization, methodology, writing-review and editing. H. Zhang: Conceptualization, resources, data curation, software, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing-original draft, project administration, writing-review and editing.
We thank the surgeons, nurses, pathologists and physicians, and patients who participated in these studies. This work was supported by the following funding agencies: National Natural Science Foundation of China (81773087 and 81572876 to H. Zhang) and Clinical Research Enhancement Initiative of Shantou University Medical College (201421).
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.