Immune-mediated antitumor effect by type 2 diabetes drug, metformin

As metformin has been reported to decrease the rate of cancer incidence in type 2 diabetic patients, we at first examined whether the drug could protect mice from methylchoranthrene-induced skin carcinogenesis. BALB/c mice were injected with 200 μg of methylchoranthrene on the right back and given 5 mg/mL metformin dissolved in the drinking water throughout the experiment. Significant inhibition of tumor development was observed in metformin-treated nondiabetic mice (Fig. S1A). We next attempted to determine whether metformin would be effective against an established solid tumor. Mice were intradermally injected with X-ray-induced RLmale1 leukemia cells and were provided oral metformin beginning on day 7. The tumors were gradually and completely rejected with no reappearance after metformin withdrawal. A rechallenge with more than twice the original number of the same tumor cells did not yield mass formation (Fig. 1A, Left), suggesting the generation of an immunologic memory response. Moreover, the antitumor effect was completely abrogated in SCID mice (Fig. 1A, Right), clearly demonstrating the necessity of T and/or B cells. Cytotoxic T lymphocytes (CTLs) specific for the tumor antigen peptide pRL1a (30) were generated in mice that rejected the tumor (Fig. S1B). Growth inhibition was observed with a metformin dose as low as 0.2 mg/mL (Fig. S1C). Of note, a previous report identified the achievement of plasma metformin concentrations of 0.45 and 1.7 μg/mL using 1 and 5 mg/mL of metformin, respectively, in drinking water (31); these plasma concentrations are similar to those in patients with diabetes treated using metformin (0.5–2 μg/mL). Administration of metformin beginning on day 0, the time point of tumor inoculation, resulted in more effective rejection than on day 7. Beginning treatment on day 10 and 13 was also effective, although the effect was less than on day 0 (Fig. S1D). Finally, as expected, CD8⁺ but not CD4⁺ T cells were proven to be responsible for the antitumor effect, because their depletion by mAb completely abrogated the response (Fig. 1B). Complete rejection by metformin was also observed with Renca (renal cell carcinoma), although partial but significant growth inhibition was observed with other tumors, 3LL (non small cell lung carcinoma), Colon 26 (intestinal carcinoma), and 4T1 (breast cancer) (Fig. S1 E–H).

Injection of a vaccine consisting of antigen (Ag) and adjuvant primes and generates specific T-cell immunity, mainly in draining lymph nodes near the injection site. However, we did not inject tumor antigens with any kind of adjuvant in Fig. 1. Therefore, it is possible that a unique process occurs at the tumor site and leads to antitumor immunity. Based on this notion, we focused on TILs throughout the experiment to clarify the associated mechanism. We found that total numbers of TILs dramatically increased when metformin administration was started on day 7, and that both CD8⁺ and CD4⁺ T cells were involved in the increment (Fig. 1 C–E). In particular, the number of CD8⁺ TILs increased nearly fourfold. We considered the possibility that metformin may suppress expression of the immune exhaustion markers PD-1 and Tim-3 on CD8⁺ TILs, thus avoiding immune exhaustion. Therefore, we investigated the expression of these markers on CD8⁺ TILs derived from individual tumor-bearing mice (Fig. S1B). The number of PD-1⁻Tim-3⁻ CD8⁺ TILs decreased from day 7–10, irrespective of metformin use (Fig. S2B). The PD-1⁻Tim-3⁺CD8⁺ TIL population increased progressively, whereas PD-1⁺Tim-3⁻ and PD-1⁺Tim-3⁺CD8⁺ TILs remained stable. Metformin did not affect any subset populations (Fig. S2 B–E). However, we surprisingly found that a significant proportion of CD8⁺TILs underwent apoptosis, detected by Annexin V (Fig. 1F and Fig. S3A), and that metformin suppressed apoptosis induction in all subsets, including PD-1⁻Tim-3⁺CD8⁺TILs (Fig. S3 B–E). Of note, the physiologically essential apoptotic process of CD4⁺CD8⁺ thymocytes, which depends on a mitochondrial pathway (32), was not down-regulated by metformin (Fig. S4), suggesting that an apoptotic mechanism unique to the tumor microenvironment is metformin-sensitive.

We next examined the metformin effects in another tumor system. MO5 is a subclone of B16 melanoma cells expressing ovalbumin (OVA) (33). Metformin administration induced significant antitumor activity (Fig. 2A). OVA- and TRP2-specific CD8⁺ TILs were identified by specific tetramers. Both TIL populations in untreated mice decreased gradually from day 7–13; in contrast, metformin administration maintained or increased these populations (Fig. 2B). CD8⁺ TILs again underwent apoptosis, which was suppressed by metformin administration (Fig. S5 A and B). The Annexin V-positive populations among OVA tetramer-positive and -negative (includes TRP-2–positive population) CD8⁺ TILs were near 80% at day 10; however, metformin suppressed this rate to <20–40% (Fig. S5 C and D). These results are consistent with those observed in the RLmale1 model. Next, to examine the functional state of antigen-specific TILs, magnet-purified CD8⁺ TILs isolated from tumor tissues were incubated with DC-like DC2.4 cells that had been pulsed with an epitope peptide (OVA_257–264); TILs were later examined for their cytokine production capacity. Only IFNγ-producing cells or very small populations producing both IFNγ and TNFα or IL-2 could be identified in untreated mice, whereas a marked increase in the population producing both IFNγ and TNFα was observed with metformin (Fig. 2C).

CD8⁺TILs in the context of memory T cells are poorly understood. Elegant studies with an acute viral infection model have proposed classification of memory T cells into central memory (TCM; CD44⁺, CD62L^high) and effector memory (TEM; CD44⁺,CD62L^low) (34, 35). TCM were shown to mediate viral-specific recall responses. Based on this model, we investigated TCM and TEM CD8⁺ TILs. Without metformin, the staining of CD8⁺ TILs from an RLmale1 tumor using antibodies against CD62L and CD44 revealed that proportions of TCM and TEM were nearly equal on day 7 and 10 but shifted to TCM dominance on day 13. In contrast, metformin maintained TEM dominance from day 10 to day 13 (Fig. 3A). Further dissection of the TIL compartment based on CD62L and KLRG1 expression revealed that short-lived effector T cells (TE; CD62L^lowKLRG1^high) were visible on day 7 but gradually decreased by day 13. In contrast, metformin yielded increases in both TEM and TE populations on day 13 (Fig. 3B), coinciding with tumor regression (Fig. 1A). In the MO5 model, metformin again caused TEM dominant over TCM (Fig. 3 C and D). At this stage, we concluded that TEM and/or TE are more responsible than TCM for tumor rejection.

We next investigated the capacity for triple cytokine (IL-2, TNFα, IFNγ) production or the multifunctionality of CD8⁺ TILs in the context of TCM/TEM classification. CD8⁺ TILs recovered from RLmale1 tumor masses were stimulated with PMA/ionomycin for 6 h in vitro and monitored for cytokine production. Without metformin, the cytokine-producing cells on day 10 were mainly identified as TCM (Fig. 4A). In contrast, with metformin, triple cytokine-producing cells appeared in correlation with the increased population of TEM (Fig. 4A). The populations with various cytokine producing patterns in the presence and absence of metformin are summarized in Fig. 4B. Metformin markedly changed the multifunctionality of CD8⁺ TILs. Taking these results together, we concluded that metformin-induced TEM capable of producing multiple (triple and double) cytokines are most important for tumor rejection. We next classified CD8⁺ TILs on the basis of their expression of PD-1 and Tim-3, followed by intracellular cytokine staining. We found that CD8⁺ TILs with triple cytokine-producing abilities belonged exclusively to the PD-1⁻Tim-3⁺ subset, which was the supposedly exhausted population in the RLmale1 tumor model (Fig. S6). We further confirmed this notion using adoptive transfer experiments. MO5-inoculated mice were adoptively transferred with OT-I CD8⁺ T cells. The transferred T cells had been previously shown to undergo vigorous division and were thus cross-primed in vivo via the adjuvant-free administration of a fusion protein comprising OVA and Mycobacterium heat shock protein 70 (OVA-mHSP70) as a vaccine (36, 37). OVA-mHSP70 injection significantly enhanced the migration of the transferred CD45.1⁺OT-I CD8⁺ T cells into the tumor tissues; however, the cytokine-producing abilities of these cells were poor (Fig. 5A). In contrast, injection of the fusion protein together with oral metformin administration apparently improved the multifunctionality of the migrated T cells, which were classified as the Tim-3⁺ population (Fig. 5A).

It is unknown whether plasma metformin concentrations as low as 10 μM (1.6 μg/mL) would directly influence the fate of T cells. To address this important question, we incubated CD8⁺ T cells isolated from naïve OT-I mice with 10 μM metformin for 6 h in the presence or absence of different doses of the AMPK inhibitor compound C (38) as indicated (Fig. 5B). After extensive washing, the cells were transferred into MO5-bearing mice. Two days later, splenic T cells and TILs were recovered and investigated for the presence and multifunctionality of donor-derived CD8⁺ T cells. Metformin-treated CD8⁺ TILs comprised up to 9.9% of all CD8⁺ T cells and were identified as triple cytokine-producing cells (Fig. 5B). However, compound C treatment abrogated the migration, although donor CD8⁺ T cells were present in the spleens of all groups (Fig. 5B). Accordingly, tumor growth inhibition was apparent in the metformin-treated group, although this effect was blocked by compound C (Fig. 5C). The weak but significant metformin-mediated increase in the phosphorylation of AMPK and its downstream target acetyl-CoA carboxylase (ACC) and the abrogation of this effect by compound C were observed by Western blot analysis (Fig. 5D). The results led us to conclude that the direct action of metformin on CD8⁺ T cells, at least partly, reduced their exhaustion within the tumor microenvironment in a manner sensitive to the AMPK inhibitor compound C.

Finally, we examined the expression of CD8⁺ TIL molecules that may possibly be influenced by metformin administration. After CD8⁺ TIL purification on day 10, cell lysates were immediately prepared for candidate molecule detection via Western blot analysis and for caspase-3 activity measurement using a fluorescent substrate. The levels of phosphorylated AMPKα and β were increased; a twofold increase in Bat3 expression was also observed, whereas Bcl2 and Bax expression were unaltered (Fig. S7A). As expected, caspase-3 activity was prominent without metformin but was completely abrogated in CD8⁺ TILs from metformin-treated mice (Fig. S7B), which offers a plausible explanation for apoptosis inhibition. To further examine the apoptotic cell populations, we evaluated the expression of active caspase-3 in TCM, TEM, and TE. Without metformin, TCM, TEM, and TE all expressed active caspase-3 whereas with metformin, primarily TCM expressed this activated enzyme (Fig. S7C). These results may explain the dominance of TCM over TEM in the absence of metformin and the dominance of TEM and TE in the presence of metformin. pS6, a downstream target of mTOR, was positive in TCM, TEM, and TE without metformin but negative with metformin (Fig. S7D), indicating that metformin inhibits mTOR, possibly via AMPK activation.

BALB/c and C57BL/6 (B6) mice were purchased from CLEA Japan and SLC. Breeding pairs of CB-17 SCID mice were provided by K. Kuribayashi, Mie University School of Medicine, Mie, Japan.

BALB/c radiation leukemia RLmale1, B6 OVA-gene introduced B16 melanoma MO5, B6 nonsmall cell lung carcinoma 3LL, BALB/c intestinal carcinoma Colon 26, BALB/c renal cell carcinoma Renca, and BALB/c breast cancer cell 4T1 were used for the tumor assay. 3LL, Colon 26, Renca, and 4T1 were kindly provided by H. Yagita, Juntendo University School of Medicine, Tokyo, Japan.

Mice were intradermally inoculated with 2 × 10⁵ tumor cells (in 0.2 mL) on the right back with a 27-gauge needle. Before inoculation of tumor cells, the hair was cut with clippers. Mice were orally administrated metformin hydrochloride (Wako) (5 mg/mL) or as indicated dissolved the drinking water. The diameter of the tumors was measured with Vernier calipers twice at right angles to calculate the mean diameter.