Metformin combined with local irradiation provokes abscopal effects in a murine rectal cancer model

Although preoperative chemoradiation therapy can down-stage locally advanced rectal cancer (LARC), it has little effect on distant metastases. Metformin exerts an anti-cancer effect partly through the activation of host immunity. LuM1, a highly lung metastatic subclone of colon 26, was injected subcutaneously (sc) in BALB/c mice and treated with metformin and/or local radiation (RT). Lung metastases and the primary tumors were evaluated and the phenotypes of immune cells in the spleen and lung metastases were examined with flow cytometry and immunohistochemistry. Local RT, but not metformin, partially delayed the growth of sc tumor which was augmented with metformin. Lung metastases were unchanged in metformin or RT alone, but significantly reduced in the combined therapy. The ratios of splenic T cells tended to be low in the RT group, which were increased by the addition of metformin. IFN-γ production of the splenic CD4(+) and CD8(+) T cells was enhanced and CD49b (+) CD335(+) activated NK cells was increased after combined treatment group. Density of NK cells infiltrating in lung metastases was increased after combination treatment. Metformin effectively enhances local and abscopal effects of RT though the activation of cell-mediated immunity and might be clinically useful for LARC.


Results
Metformin augments the suppressive effects of RT on the growth of sc tumors. Subcutaneous tumors were observed in all mice about 10 days after sc injection and treatment started from day 13. Weight did not show significant differences in all groups ( Fig. 2A). As shown in Fig. 2B,C, local RT (4 Gy × 2) suppressed the growth of sc tumors to a size approximately half that observed in the no treatment group although not statistically significantly (p = 0.067). Metformin alone did not significantly suppress tumor growth. However, when combined with RT, tumor weight at day 28 was significantly reduced compared with no treatment group (p < 0.01, n = 10).
Combined RT and metformin treatment suppressed the growth of non-irradiated lung metastases. The number of macroscopic metastases in non-irradiated lungs was counted after sacrifice on day 28. As shown in Fig. 3, multiple metastases developed in the lungs of mice in the no treatment group (median (M) = 120, 22-188) which was not significantly changed by treatment with metformin alone. The number of metastases tended to be reduced in the RT treated group, although not statistically significantly so (M = 34, 3-78, Figure 1. Treatment with radiation therapy (RT) and/or metformin. LuM-1 cells (1 × 10 6 per mouse) were subcutaneously injected in the right flank of 6-7 week-old female BALB/c mice. Local RT was delivered using MX 160 Labo (mediXtec, Chiba, Japan), as described in Materials and Methods. Metformin was dissolved in the drinking water (1 mg/mL) and administrated continuously from day13. www.nature.com/scientificreports/ p = 0.12). In comparison, when metformin was combined with RT treatment, the number of lung metastases was significantly decreased (M = 13, 0-70) with no metastases detected in 3/10 mice (Fig. 3). We also examined the metastatic lesions including the microscopic tumors inside the lung using microscopic observation in representative coronal tissue sections, and found that lung metastases were also suppressed significantly by combination treatment (Supplementary Fig. 1). Fig. 4A, the ratios of CD3(+) or CD4(+) T cells at day 28 tended to be reduced in mice treated with RT alone, which were recovered in mice treated with both RT and metformin. The difference between RT treated mice and the combined therapy group was statistically significant (CD3:17.8 ± 2.9% vs 26.5 ± 5.9%, p < 0.05; CD4:14.7% ± 1.5% vs 17.1 ± 3.8%, p < 0.05). The ratios of CD8(+) T cells also showed the same trend (4.6% ± 1.2% vs 7.1 ± 1.8%, p = 0.14). The rate of CD19(+) B was not different (Fig. 4B). However, the rate of activated NK cells was significantly increased in the combined therapy group (1.1% ± 0.34% vs 2.6% ± 1.0%, p < 0.05) (Fig. 4C). The ratios of regulatory T cells (T reg ) and granulocytic-or monocytic-myeloid derived suppressor cells in splenocytes did not show significant differences among the 4 groups (Fig. 5).

Combined RT and metformin treatment increased the frequencies of T and natural killer (NK) cells in the spleens of LuM-1 bearing mice. As shown in
Combined radiation and metformin treatment restored T cell exhaustion in spleens of LuM-1 bearing mice. We next examined T cell exhaustion in splenocytes. Although the expression of a representative exhaustive marker, PD-1 did not change significantly in CD4(+) and CD8(+) T cells, the ratios of PD-1 (+) cells tended to be lower in the combined treatment group compared with RT alone treated group (CD4;15.4 ± 6.7% vs 9.4% ± 1.7%, p = 0.11 CD8; 5.4 ± 1.0% vs 4.2% ± 0.80%, p = 0.22) (Fig. 4D). After stimulation with PMA and Ca 2+ ionophore, the ratios of IFN-γ (+) cells were significantly increased in mice treated with both RT and metformin compared with those in the control group (CD4; 6.8 ± 2.4% vs 10.4 ± 1.3%, p < 0.05; CD8; 24.1 ± 3.1% vs 40.4 ± 3.1%, p < 0.01) (Fig. 6).  Combination therapy increased the density of NK cells while decreased the density of MDSC in lung metastases was decreased. The character of infiltrating immune cells in metastatic lesions larger than 40,000 µm 2 in lung tissue was evaluated with immunohistochemistry. The density of CD8a (+) cells was not changed among the 4 groups (Fig. 8B). However, as shown in Fig. 8A,C, the numbers of NKp46/CD335(+) NK cells in metastatic tumors were markedly increased only in mice treated with RT+ metformin compared with the no treatment group (p < 0.0001). In contrast, the density of Gr-1 (+) cells was significantly decreased in combined therapy treated group (p = 0.012) (Fig. 8A,D).

Discussion
Metformin has been shown to reduce the development of various malignancies and cancer-related mortality in patients with type 2 diabetes mellitus [6][7][8][9][10] . In patients with colorectal cancer, metformin has been shown to improve patient outcomes although the underlying mechanisms of how metformin exerts its anti-tumor effects are not fully understood 11,[44][45][46][47] . Based on these observations, many preclinical studies were conducted to determine the clinical usefulness of combined RT and metformin which have had promising results 12,16,17 . However, the results of clinical studies are still inconsistent 21,22,25 .
In previous in vivo studies 12,17 , the anti-tumor effects of metformin on irradiated tumors were evaluated in immune deficient mice. However, recent studies have suggested that the anti-tumor effects of metformin are largely dependent on host immunity [35][36][37][38][39]43 . In this study, therefore, we used a syngeneic murine model of spontaneous lung metastases after sc implantation of colorectal cancer tumor cells and reevaluated the synergistic effects of combined RT and metformin.
LuM-1 is a highly aggressive and non-immunogenic subclone of colon 26 48 , and metformin alone did not show significant anti-tumor effects on sc tumors or lung metastases. However, synergistic effects of RT and  www.nature.com/scientificreports/ metformin were observed. Selective RT (4 Gy × 2) delayed the growth of sc tumors with marginal statistical significance. However, when combined with oral metformin, the tumor suppressive effects of RT were markedly enhanced. More interestingly, combined therapy suppressed the growth of metastatic tumors in non-irradiated lungs. No metastases were detected in 3/10 mice. Metastatic lesions examined with tissue sections also show the same trends. The measurement of metastases with these methodologies may not be enough to assess the accurate total metastatic burden in lung. However, microscopic metastases from the primary LuM-1 tumor were already present in the lungs at the time of starting treatment 49 , these data are highly suggestive that metformin not only enhances the cytotoxic effects on tumor cells directly exposed to RT but also causes tumor regression outside the irradiated field as a so-called "abscopal effect" 30,50 .
Although DNA double-strand breaks in tumor cells are considered to be the primary mechanism of the antitumor effects of RT, recent studies suggest that a reduction in tumor size is also strongly dependent on immunogenic death of tumor cells mediated by T cell-mediated host immune responses after RT 51,52 . Indeed, local RT has been shown to increase the generation of tumor antigen-specific effector T cells in murine tumor models 52,53 .
In data from the present study, the rates of CD3(+), CD4(+) and CD8(+) T cells tended to be decreased in the spleens of mice treated with RT alone presumably due to direct toxic effects of RT 54 . However, the ratios were significantly increased by the additional treatment with metformin. The frequencies of IFN-γ positive cells in CD4(+) and CD8(+) T cells were significantly elevated, and PD-1(+) cells in both T cell populations tended to be reduced in mice receiving combined therapy. The results observed in irradiated mice are consistent with those from previous studies showing that the anti-tumor effects of metformin result mostly from T cell mediated host immunity 35,36,43 . The IFN-γ producing activities of T cells is critical to suppress the growth of metastases since they inversely correlate with the degree of lung metastases in the mice in all 4 groups in the present study. Taken together, it is suggested that RT alone may reduce systemic T-cell mediated immunity associated with a non-immunogenic tumor such as LuM-1, while metformin restore T cell exhaustion which can efficiently induce abscopal effects.
Another interesting finding is that the frequency of activated NK cells defined by the CD49(+) CD335(+) phenotype showed a strong inverse correlation with the number of lung metastases and their ratios were significantly increased in mice treated with combined RT and metformin. Immunohistochemical analysis revealed that the number of CD335(+) activated NK cells, but not CD8(+) T cells, infiltrating metastatic lesions was increased in those mice. Consistent with this result, a recent study has demonstrated that metformin increases the frequency of CD335(+) NK cells in the spleen which suppresses pulmonary metastases from B16F10 melanoma 55 .
However, infiltration of Gr-1(+) MDSC in lung metastases was reduced after combined treatment. Numerous studies have suggested that accumulation of MDSC provide a tumor permissive microenvironment through stimulation of angiogenesis as well as potent suppression of T cell functions 56,57 . MDSC have also been reported to inhibit NK cell functions [58][59][60] , although the mechanistic interactions with NK cells are less clear than with T cells 61 . Considering the results of these studies, it is suggested that the abscopal effects induced by metformin and RT might be partially attributed to augmented NK cell activity caused by a reduced number of MDSC in the lung.
There is growing evidence that RT can result in in situ tumor vaccination by exposing tumor specific neoantigens to the host's innate immune system which then leads to immunogenic cell death of non-immunogenic tumor cells [51][52][53]62 . After successful results were found with immune check point inhibitors (ICI) in the treatment of various malignancies, radioimmunotherapy using the combination of RT and ICI has attracted much attention as an effective novel therapy for advanced malignancies. However, studies have shown conflicting results of the clinical usefulness of combined RT and ICI presumably due to differences in cancer immunogenicity as well as host immunity [31][32][33][34] . The present study shows that metformin efficiently enhances the abscopal effects of RT in non-immunogenic colorectal cancer tumors though the activation of T cell-and NK cell-mediated immunity. The results may not directly apply to humans. However, since metformin activates anti-tumor immunity by different mechanisms from anti-PD-1/PD-L1 or anti-CTLA-4 mAbs, the combination of immune check point inhibitors and metformin may provoke remarkable radiosensitizing effects leading to improved outcomes in patients with LARC. Cell lines. LuM-1, a highly metastatic sub-clone of murine colon cancer, colon26 63  Animal model. The experimental protocol is shown in Fig. 1A. Female BALB/c mice 5-6 weeks old were purchased from CLEA Japan. (Tokyo, Japan) and housed in specific pathogen-free conditions. LuM-1 cells (1 × 10 6 per mouse) were injected subcutaneously in the right flank of 6-7 weeks-old female BALB/c mice. When the primary tumors reached a volume of 100-300 mm 3 at day13, mice were divided into four groups, each group containing 5-6 mice to enable statistical analysis (Group A, no treatment; Group B, metformin treated; Group C, radiation treated; Group D, metformin and radiation treated). Local RT was delivered using MX 160 Labo (mediXtec, Chiba, Japan), as described previously 49 . In short, anesthetized mice were held in the decubitus position, and radiation delivered only to the subcutaneous (sc) tumor with the remainder of the mouse including the lung covered with 3 lead plates with a thickness of 5 mm (Fig. 1B). We confirmed the X-ray was completely blocked by this apparatus. As a control, mice were similarly placed in the same conditions under anesthesia. Mice were administrated metformin hydrochloride (Wako) (1 mg/mL) or as indicated, dissolved in drinking water. Tumor diameter was measured with calipers and tumor volume calculated using tumor volume (mm 3 ) = [tumor length (mm) × tumor width (mm) 2 ]/2 and measured once every 3 days. Mouse weight was measured every 3 days. Mice were sacrificed using deep anesthesia by isoflurane on day 28, and the weight of the sc tumor and numbers of macroscopic metastatic nodules in the lungs were blindly evaluated by 2 investigators and average of the values were adopted. All procedures were approved by the Animal Care Committee of Jichi Medical University (No 19035-01) and performed in accordance with ARRIVE guidelines and the Japanese Guidelines for Animal Research. www.nature.com/scientificreports/ Immunohistochemistry of mice lung samples. Mice were sacrificed at day 28, lungs were internally fixed with 4% formalin, excised, and paraffin-embedded 4 µm sections were prepared for immunohistochemical evaluation. After endogenous peroxidase blocking by methanol and 30% hydrogen peroxide and heat-induced antigen retrieval in citrate buffer with microwaves, the specimens were incubated with 1% BSA for 30 min to block nonspecific antibody binding. Then, the slides were incubated with Abs to CD8a (at a dilution of 1:100), Ly-6G/Ly-6c (1:50), or NKp46/CD335 (1:500) in humid chambers overnight at 4 °C. After washings with PBS, sections were incubated with anti-rat or anti-rabbit secondary antibody conjugated with peroxidase for 30 min at room temperature. After washing, the enzyme substrate 3, 30-diaminobenzidine (Dako REAL EnVision Detection System, DAKO) was used for visualization and counterstained with Meyer's hematoxylin. In each tissue section, 1-5 different metastatic lesions larger than 40,000 µm 2 were randomly selected and positively stained cells were counted in those areas under the microscope. Thereafter, the densities of each cell type (/mm 2 ) were calculated using ImageJ software (NIH, Bethesda, MD).

Statistical analysis. For data on tumor volume, lung metastases and immunohistochemistry, p-values
were evaluated with Kruskal-Wallis analysis followed by Dunn's multiple comparison tests. For splenocyte data, p-values were evaluated with one-way ANOVA followed by Tukey's honestly significant difference test. Correlation was examined with simple linear regression analysis. All analyses were performed with Graph Pad Prism 8 Software (San Diego, CA, USA), and p-values < 0.05 were considered statistically significant. www.nature.com/scientificreports/

Data availability
All data generated or analysed during this study were included in this published article and its supplementary information files.