AMPK promotes antitumor immunity by downregulating PD-1 in regulatory T cells via the HMGCR/p38 signaling pathway

AMP-activated protein kinase (AMPK) is a metabolic sensor that maintains energy homeostasis. AMPK functions as a tumor suppressor in different cancers; however, its role in regulating antitumor immunity, particularly the function of regulatory T cells (Tregs), is poorly defined. AMPKα1fl/flFoxp3YFP-Cre, Foxp3YFP-Cre, Rag1−/−, and C57BL/6 J mice were used for our research. Flow cytometry and cell sorting, western blotting, immuno-precipitation, immuno-fluorescence, glycolysis assay, and qRT-PCR were used to investigate the role of AMPK in suppressing programmed cell death 1 (PD-1) expression and for mechanistic investigation. The deletion of the AMPKα1 subunit in Tregs accelerates tumor growth by increasing the expression of PD-1. Metabolically, loss of AMPK in Tregs promotes glycolysis and the expression of 3-hydroxy-3-methylglutaryl-CoA reductase (HMGCR), a key enzyme of the mevalonate pathway. Mechanistically, AMPK activates the p38 mitogen-activated protein kinase (MAPK) that phosphorylates glycogen synthase kinase-3β (GSK-3β), inhibiting the expression of PD-1 in Tregs. Our study identified an AMPK regulatory mechanism of PD-1 expression via the HMGCR/p38 MAPK/GSK3β signaling pathway. We propose that the AMPK activator can display synergic antitumor effect in murine tumor models, supporting their potential clinical use when combined with anti-PD-1 antibody, anti-CTLA-4 antibody, or a HMGCR inhibitor.


Background
Tumor cells reprogram several metabolic pathways to meet their bioenergetics/biosynthetic demands [1]; this leads to changes in the tumor microenvironment (TME), affecting tumor-infiltrating cells [2]. Regulatory T cells (Tregs) are a major barrier to antitumor immunity [3]. In the TME, Tregs also undergo metabolic reprogramming, where glycolysis is inhibited, but fatty acid oxidation (FAO) and oxidative phosphorylation are promoted, thereby enhancing Treg-mediated immunosuppression and promoting tumor progression [4]. However, contrary to Tregs, cytotoxic CD8 + T cells play an important role in antitumor immunity through the production of interferon (IFN)-γ and granzyme B (GZB) [5]. Notably, T cell activation also depends on metabolic pathways, including aerobic glycolysis, amino acid metabolism, glutaminolysis, and de novo fatty acid synthesis [6]. Since these pathways are equally important for the proliferation and survival of tumor cells, T cells and tumor cells compete for nutrients. Therefore, cellular metabolism actively regulates tumorigenesis.
AMP-activated protein kinase (AMPK) is an evolutionarily conserved serine/threonine kinase acting as an energy sensor for the maintenance of energy homeostasis [7]. AMPK activation by the tumor suppressor Liver kinase B1 (LKB1) supports the hypothesis that AMPK is also a tumor suppressor [8]. Recent studies have shown that AMPK mediates the inhibition of cell proliferation and growth of tumor cells [9]. Notably, studies have shown that the AMPK activators metformin and 5-aminoimidazole-4-carboxamide 1-β-D-ribofuranoside (AICAR) inhibit tumor progression [10,11]. AMPK also inhibits the expression of Glut1 and glycolysis in Tregs by inhibiting mTORC1 signaling [12]. As glycolysis and lipid metabolism are the primary pathways underpinning the survival of tumor cells, we can speculate that AMPK and Tregs are metabolically linked in cancer. However, the potential role of AMPK in Tregs in cancer has not been studied yet.
Programmed cell death 1 (PD-1) is an inhibitory molecule expressed on the surface of Tregs and effector T cells that serves as a major immune checkpoint in antitumor responses [13]. In fact, anti-PD-1-mediated immune checkpoint blockade is used as first-line therapy against lung cancer [14]. Interestingly, a recent study showed that the inhibition of glycogen synthase kinase-3 beta (GSK3β) decreases the expression of PD-1 on CD8 + T cells and is as effective as anti-PD-1/anti-PD-L1 antibodies for controlling B16F10 melanoma [15]. Thus, although the regulatory mechanism of PD-1 expression in Tregs is unclear, GSK3β could serve as a potential target for the modulation of PD-1 in cancer.
In the present study, we addressed the role of AMPK in Tregs in cancer by generating Treg-specific AMPKα1knockout mice. We showed that AMPK loss in Treg cells accelerated tumor growth by increasing the expression of PD-1. Our results indicate that AMPK regulates the expression of PD-1 via the HMGCR/P38 MAPK/GSK3β signaling axis, suppressing tumor progression.

Materials and methods
Mice C57BL/6 J, Rag1 −/− , AMPKα1 fl/fl , CD4 Cre and Foxp3 YFP-Cre mice were purchased from the Jackson Laboratory (Bar Harbor, ME, USA). AMPKα1 fl/fl mice were crossed with Foxp3 YFP-Cre mice or CD4 Cre to generate AMPKα1 fl/fl Fox-p3 YFP-Cre offspring (referred to as AMPK fl/fl Foxp3-Cre mice hereafter or also denoted as AMPK-KO in figure) or AMPKα1 fl/fl CD4 Cre offspring (referred to as AMPK fl/ fl CD4-Cre mice hereafter) respectively. AMPKα1 fl/fl Fox-p3 YFP-Cre mice were used at 6-10 weeks of age unless otherwise specified. Age-and sex-matched littermate Foxp3 Cre control mice were used as controls and referred to as wild type (WT) hereafter. All mice were maintained under specific pathogen-free conditions in the animal facilities of the Kangwon National University and the Yeungnam University. All animal experiments were approved by the Institutional Animal Care and Use Committees (IACUC) of the Kangwon National University (Permit Number: KW-190729-1) and the Yeungnam University (Permit Number: 2017-034).

Tumor models
AMPKα1 fl/fl Foxp3 YFP-Cre and Foxp3 YFP-Cre , Rag1 −/− , and C57BL/6 J mice were injected subcutaneously with 2.5 × 10 5 B16F10 melanoma cells. WT and AMPK fl/ fl Foxp3-Cre mice were injected subcutaneously with 1 × 10 6 TC-1 cervical cancer cells and 2.5 × 10 5 MC38 colon cancer cells. Tumors were measured every day with a digital caliper in two dimensions (length and width); tumor volume (mm 3 ) was determined using the formula V = W 2 × L/2, where W and L are the shortest and longest diameters in mm, respectively. Checkpoint blockade monoclonal antibodies were administered every 3 days starting from day 8 post-tumor challenge until the end of the experiment. Anti-mouse PD-1 (Clone#RPM1-14; BioXcell, Lebanon, NH, USA) and anti-mouse cytotoxic T-lymphocyte associated protein 4 (CTLA4) (Clone#9D9; BioXcell) were administered intraperitoneally (i.p.; 200 μg/mouse and 100 μg/mouse, respectively). AICAR (500 mg/kg) was injected i.p. daily either alone or in combination with anti-PD-1 or anti-CTLA4. Statin (15 mg/kg) was injected i.p. daily into C57BL/6 J mice either alone or in combination with AICAR. Mice were assigned into different groups in a randomized fashion based on their ear tag number.
For the isolation of tumor-infiltrating lymphocytes, tumor tissues were harvested and minced using sterile razor blades. Cleared tumor pieces were digested using an enzyme mixture containing 0.5 mg/mL Collagenase D (Cat#11088866001, Roche Dagnostics GmBH, Mannheim, Germany) and 0.02 mg/mL DNase I (Cat#10104159001, Sigma Aldrich, St. Louis, MO, USA) in RPMI 1640 at 37 °C for 45 min and passed through 70 μM cell strainers (BD Biosciences, Franklin Lakes, NJ, USA). A Percoll gradient (Cat#17-0891-01, GE Healthcare, Chicago, IL, USA) was then used to separate cancer cells and enrich lymphocytes as described previously [16].

Flow cytometry and cell sorting
Single cell suspensions were prepared from spleens, peripheral lymph nodes, and mesenteric lymph nodes.

Immunoprecipitation
Tregs were harvested from WT mice and lysed in IP lysis buffer (Cat#87787, Pierce, Appleton, WI, USA). Immunoprecipitation was performed using anti-AMPK, anti-p38, anti-GSK3β, and anti-HMGCR antibodies (Cell Signaling Technology) in 200 μL of total cell lysate mixed with 25 μL protein G agarose beads (Cat#22851, Pierce), followed by overnight incubation at 4 °C. The immunoprecipitated proteins were then washed 3-5 times with lysis buffer and analyzed by western blotting.

Glycolysis assay
YFP + CD25 + Tregs sorted from WT and AMPK-KO mice were seeded into 96-well plates at a density of 5 × 10 5 cells/well, followed by incubation overnight in a 5% CO 2 atmosphere at 37 °C. On the next day, a CO 2 purge was performed by incubating cells in a CO 2 -free incubator at 37 °C for 3 h. Cell suspensions were then harvested and washed with respiration buffer. The glycolysis assay reagent (Cat#ab197244, Abcam, Cambridge, UK) was then added. Fluorescence intensity was measured at the excitation and emission wavelengths of 380 nm and 615 nm, respectively.

In vitro Tregs suppression assay
Naïve CD4 + CD25 − T cells were labeled with carboxyfluorescein diacetate succinimidyl ester (CFSE, Ther-moFisher Scientific) and co-cultured with an increasing ratio of sorted Tregs for 4 days in the presence of anti-CD3 (1 μg/mL) and irradiated splenocytes. The suppressive activity of Tregs was analyzed by measuring the proliferation of activated effector T cells based on CFSE dilution as described previously [19].

RNA isolation and real-time PCR
Total RNA was extracted from sorted YFP + CD25 + Tregs using the ReliPrep ™ RNA Cell Miniprep System (Cat#Z6011, Promega Corporation, Madison, WI, USA), and cDNA was synthesized using the Goscript Reverse Transcription system (Cat#A5001, Promega Corporation). mRNA expression was measured by real-time PCR using the QuantiTect SYBR Green PCR kit (QIAGEN). 25 μL of the mixture, containing 10 ng of the total RNA sample, was used for real-time PCR which was performed on a CFX96 thermal cycler (Bio-Rad, Hercules, CA, USA) using the following thermal cycle: 60 min at 42 °C (for reverse transcription), 15 min at 95 °C (for heat inactivation or pre-denaturation), and 40 cycles for 15 s at 95 °C, 30 s at 58 °C, and 30 s at 72 °C, each. All data analyses were performed using the comparative C T method, and the fold change was calculated using the 2 -△△C(T) equation, as previously described [20].
Melting curve analysis was performed to check for non-specific amplification and to confirm that a single amplicon was generated by qPCR. The PCR efficiency was > 90%. The PCR target genes and primer sequences are listed in supplementary table.

Statistical analysis
Statistical analysis was performed using GraphPad Prism 9 (GraphPad Software Inc., San Diego, CA, USA). The unpaired two-tailed student's t-test was used for comparisons between two groups. Tukey's multiple comparisons test was used for multiple comparisons. Data are presented as the mean ± standard deviation (SD). Statistical significance was defined as p < 0.05.

AMPK deficiency in Tregs promotes tumor growth
Although the T cell-specific deletion of AMPK promotes tumor growth in mice [21], the specific role of AMPK in Tregs in antitumor immunity is still controversial. Interestingly, we found that the AMPK expression levels were markedly reduced in Tregs from tumor tissues and draining lymph nodes in tumor-bearing mice, compared to those of peripheral Treg cells in tumor-free mice, whereas the levels of LKB1, an upstream regulator of AMPK, remained unchanged (Fig. 1A). Likewise, the transcription of AMPK also decreased in Tregs in tumorbearing mice, compared to that in WT tumor-free mice (Fig. 1B). To determine whether the correlation between AMPK and PD-1 expression was also evident in cancer patients, we analyzed gene expression profile from public database such as TCGA and GEO [22,23]. In single-cell RNA-seq dataset which was originated from melanoma tissue (GSE72056), we downloaded the TPM dataset file and sorted any values more than 3 TPM for cells that highly expressed Foxp3 transcripts and excluded invalid TPM values for Pdcd1 and Prkaa1 expression. There was a significant negative correlation between the transcripts of Prkaa1 and Pdcd1 at the single-cell level (r = − 0.627, p = 0.012; Fig. S1A). In addition, the gene expression Pokhrel et al. Mol Cancer (2021) 20:133 profiles of Prkag2 and Pdcd1 were analyzed in resident Tregs in patient cancer specimens using the available datasets (GSE89225) (r = − 0.621, p = 0.0034; Fig. S1B). Further analysis also revealed that Pdcd1 was negatively correlated with AMPK subunit mRNAs including Prkaa1, Prkaa2, and Prkag1 in TCGA database obtained from prostate, melanoma, and breast cancer patients (Fig. S1C). These results suggest that Tregs with lower levels of AMPK expression exhibited higher levels of PD-1 expression in patients with cancer, indicating the importance of AMPK in the regulation of PD-1 expression in Tregs.
To address the specific role of AMPK in tumor-associated Tregs, we crossed mice carrying loxP-flanked Prkaa1 alleles (AMPK fl/fl ) with Foxp3 YFP-Cre (referred to as Foxp3-Cre) mice [24] to generate progeny in which AMPK alleles are conditionally deleted in Treg cells, but not in other T cells (hereafter referred to as AMPK fl/ fl Foxp3-Cre mice; AMPK +/+ Foxp3-Cre mice were used as control (WT)). AMPK fl/fl Foxp3-Cre mice (also denoted as AMPK-KO in the figure) were born at the expected Mendelian ratios and seemed grossly normal. There was no differences between the size of the spleen and lymph nodes of WT and AMPK fl/fl Foxp3-Cre mice (data not shown). AMPK fl/fl Foxp3-Cre mice exhibited no significant change in the number and percentage of CD4 + , CD8 + T cells, effector T, Th1, Th17 and Treg cells in spleen, MLN and PLN (Fig. S2A-E). Next, we inoculated several kinds of syngeneic tumor cells into AMPK fl/fl Foxp3-Cre mice and into WT control mice. Interestingly, we found that tumors formed by B16F10 melanoma, TC-1 cervical cancer, and MC38 colon cancer cells grew more rapidly in AMPK fl/fl Foxp3-Cre mice than in WT mice (Fig. 1C, D and Fig. S3A-C).
Moreover, we found that the levels of effector T cells (CD4 + and CD8 + ) in the tumors and draining lymph nodes (dLNs) of AMPK fl/fl Foxp3-Cre mice were significantly reduced ( Fig. 1E and Fig. S4A). Additionally, although the frequencies of Foxp3 + Tregs in the tumors and dLNs were unchanged ( Fig. 1E and Fig. S4B), cellularity and proliferation were higher in Tregs from AMPK fl/fl Foxp3-Cre mice ( Fig. 1E and Fig. S4C), suggesting that the proliferation and survival of Tregs might be regulated by AMPK. In addition, IFN-γ-producing CD4 + and CD8 + T cells and GZB-producing CD8 + T cells were significantly decreased in tumors and dLNs from AMPK fl/ fl Foxp3-Cre mice (Fig. 1F, G, and Fig. S4D, E). In addition, we confirmed that GZB-producing CD8 + T cells The data are presented as the mean ± standard deviation (SD); n = 5 mice per group. *P < 0.05; **P < 0.01; ***P < 0.001 in tumors were found to be reduced in AMPK fl/fl Foxp3-Cre mice (Fig. S5). Next, we assessed the tumor-antigenspecific cytotoxic T cell response using an in vivo CTL assay, and we observed that cytolytic activity of CD8 + T cells in the spleen and lymph nodes was significantly decreased in AMPK fl/fl Foxp3-Cre mice (Fig. 1H), suggesting that AMPK supports the cytotoxic effect exerted by CD8 + T cells. Thus, these findings suggest that the loss of AMPK in tumor-associated Tregs promoted tumor growth in mice via the impact on antitumor T cells and their functions.

AMPK-deficient Tregs show high expression of PD-1
Tregs express various inhibitory receptors on their surface [25]. We found that AMPK-KO Tregs expressed higher levels of PD-1, Nrp1, and ICOS than Tregs from WT mice in both tumor-free and tumor-bearing mice ( Fig. 2A, B). However, other receptors associated with the suppressive function of Tregs, including CTLA4, GITR, and OX40, were not differentially expressed ( Fig. 2A, B). The expression of CD39 and CD73, which contribute to energy depletion in the TME, was also not changed ( Fig. 2A, B).
To confirm whether the expression of PD-1, Nrp1, and ICOS in Tregs was regulated by AMPK, we treated WT mice with compound C, a pharmacological inhibitor of AMPK. We found that the expression of PD-1 was significantly upregulated in Tregs after treatment with compound C, whereas no significant alterations were detected in the levels of ICOS and Nrp1 (Fig. 2C). Conversely, treating WT mice with a pharmacological activator of AMPK, AICAR, decreased the expression of PD-1 in Tregs (Fig. 2D). Similar results were obtained with another AMPK activator, metformin (Fig. S6), suggesting that AMPK could restrict the expression of PD-1 in Tregs. Western blot analysis further confirmed that the expression of PD-1 in WT Tregs was reduced after AICAR treatment and increased after compound C treatment in a time-dependent manner (Fig. 2E). We also found that AMPK-KO Tregs inhibited the proliferation of T cells more potently than WT Tregs in a co-culture system (Fig. 2F). Notably, AMPK-KO Tregs expressed higher levels of IL-10 and TGF-β at both the mRNA and protein levels (Fig. 2G, H) than WT Tregs. Collectively, these findings suggest that AMPK suppresses the expression of PD-1 in Tregs, attenuating their immunosuppressive potential.
We also observed that PD-1 expression was higher in CD4 + T cells from tumor-bearing mice than tumorfree mice (Fig. S7A). In tumor-free mice, PD-1 expression was comparable between AMPK-deficient CD4 + T cells and WT CD4 + T cells (Fig. S7B, C). In contrast, in tumor-bearing mice, PD-1 expression was higher in AMPK-deficient CD4 + T cells, but not in CD8 + T cells (Fig. S7D, E). Therefore, we can presume that AMPK negatively regulates PD-1 expression in CD4 + T cells as well as in Tregs.
Immune checkpoint blockade reduces tumor growth in AMPK fl/fl Foxp3-Cre mice and synergizes with AMPK activation for antitumor immunity We reasoned that the increased expression of PD-1 in Tregs might be responsible for the increased tumor growth in AMPK fl/fl Foxp3-Cre mice; therefore, we treated tumor-bearing WT and AMPK fl/fl Foxp3-Cre mice with an anti-PD-1 antibody. Remarkably, the growth of B16F10 tumors in anti-PD-1-treated AMPK fl/fl Foxp3-Cre mice was significantly reduced, compared to that in anti-PD-1-treated WT mice (Fig. 3A, B). Infiltration by CD4 + and CD8 + T cells was highly increased in AMPK fl/fl Foxp3-Cre mice after anti-PD-1 treatment, whereas no difference was observed for Foxp3 + Tregs (Fig. 3C). Similarly, the counts of GZB-and IFN-γ-producing CD8 + T cells were highly increased in AMPK fl/fl Foxp3-Cre mice after anti-PD-1 treatment (Fig. 3D). Further, we assessed the expression of IL-10 in Tregs, a hallmark of immunosuppression [26], and found that the counts of IL-10 + Foxp3 + T cells were significantly decreased in AMPK fl/fl Foxp3-Cre mice following anti-PD-1 treatment (Fig. 3E). Notably, anti-PD-1 antibody treatment also significantly limited the expression of ICOS and Nrp1 in AMPK-KO Tregs (Fig. 3F), suggesting that this antibody may work by direct binding to Tregs and by rescuing exhausted T cells. Overall, anti-tumor immunity was more enhanced in AMPK fl/fl Foxp3-Cre mice after anti-PD-1 antibody treatment compared to WT mice, indicating that AMPK is expected to have other functions besides regulating PD-1 expression. When we analyzed Tregs in anti-PD-1 antibody-treated mice, the CD25 expression levels decreased in AMPK-KO Tregs compared to WT Tregs (Fig. S8A).
To confirm this, we stimulated anti-CD3/anti-CD28 antibodies-treated Tregs with IL-2 in vitro. Before stimulation, CD25 expression levels were comparable between WT and AMPK-KO Tregs. However, after treatment with anti-PD-1 antibody, CD25 expression levels were slightly reduced in AMPK-KO Tregs (Fig. S8B). Thus, the enhanced anti-tumor immunity in AMPK fl/fl Foxp3-Cre mice after PD-1 treatment may be associated with the regulation of CD25 expression in AMPK-KO Tregs.
Studies have shown that AMPK activators, including AICAR and metformin, can be used to effectively treat several cancers [10,26]. Given that the activation of AMPK results in the downregulation of PD-1 in Tregs, we next assessed the antitumor effect of combination therapy with AICAR and anti-CTLA4 antibodies. AICAR treatment reduced B16F10 tumor growth in immunecompetent mice (Fig. S9A). Additionally, AICAR treatment increased the frequency of CD4 + and CD8 + T cells in B16F10-bearing C57BL/6 J mice without affecting the percentage of Treg cells (Fig. S9B). The downregulation of PD-1 was also observed in Tregs isolated from tumor tissues following AICAR treatment (Fig. S9C), and combination therapy with AICAR with anti-CTLA4 antibody significantly reduced the tumor growth due to increased antitumor T cell activity in comparison with AICAR monotherapy (Fig. S9A-C). To confirm whether the antitumor effect of AICAR is dependent on T cells, we used immune-deficient Rag1 −/− mice. AICAR had no effect on B16F10 tumor growth in immune-deficient mice (Fig. S9D), which confirmed that the antitumor effect of AICAR was mainly dependent on T cells. Next, we assessed the combination therapy of AICAR with anti-PD-1 antibody. Similar to the treatment with AICAR and anti-CTLA-4 antibody, the combination therapy with AICAR and anti-PD-1 antibody significantly reduced tumor growth due to increased anti-tumor T cell activity in comparison with AICAR monotherapy or

Loss of AMPK in Tregs disrupts metabolic process and enhances the expression of HMGCR
As AMPK is essential for energy homeostasis, AMPK deficiency in Tregs may contribute to the alteration of energy metabolism. We found that the mitochondrial potential and content were significantly reduced in AMPK-KO Tregs (Fig. S11A), suggesting a mitochondrial defect; therefore, we measured the ATP levels in these cells. Although the ATP levels were significantly reduced in AMPK-KO Tregs (Fig. S11B), glycolysis was elevated compared to that in WT Tregs (Fig. S11C). As AMPK was reported to inhibit glycolysis in Tregs by inhibiting mTORC1 signaling [27], we examined CD71 and CD98, key nutrient receptors regulated by mTORC1 [28]. The expression levels of CD71 (p = 0.13) and CD98 (p = 0.37) were not significantly different between WT and AMPK-KO Tregs (Fig. S11D). Regarding the activation of p70S6 kinase (S6K) and phosphoinositide 3-kinase (PI3K), the phosphorylation levels of S6K and PI3K were markedly increased in WT Tregs, but not in AMPK-KO Tregs after TCR stimulation (Fig. S11E), suggesting that the upregulation of glycolysis in AMPKdeficient Tregs is not mTORC1-mediated. Further, we analyzed the expression of the mTOR complex and of several enzymes associated with energy metabolism. Interestingly, the expression of HMGCR, one of the primary rate-limiting enzymes of the mevalonate pathway, and glycolytic signature genes, including Glut1 and Ldha, was significantly upregulated in AMPK-KO Tregs (Fig. 4A). Western blot analysis revealed that the accumulation of the dephosphorylated form of HMGCR (the active form) was increased in AMPK-KO Tregs, compared to WT Tregs, whereas the expression of phosphorylated-HMGCR (the inactivated form) was decreased, indicating HMGCR activation (Fig. 4B). However, the expression of ACC, SREBP1, and Foxo3a was not markedly increased in AMPK-KO Tregs compared to WT Tregs after TCR stimulation (Fig. 4B). In addition, the phosphorylated-HMGCR increased, whereas dephosphorylated-HMGCR decreased following AICAR treatment (Fig. 4C), suggesting that HMGCR is a downstream substrate of AMPK in Tregs. Furthermore, we found that AMPK could bind to HMGCR using an immunoprecipitation assay (Fig. 4D). These results suggest that AMPK deficiency induces metabolic changes in Tregs, favoring glycolysis, and that an increase in the HMGCR expression occurred, which may be the underlying reason for the increased tumor growth in AMPK fl/ fl Foxp3-Cre mice. Altogether, these findings suggest that AMPK negatively regulates HMGCR activation in Tregs.
Statin, a specific HMGCR inhibitor, was reported to exert antitumor effects in different cancers [29]. Thus, we assessed the antitumor effect of statin in combination with AICAR; WT mice transplanted with B16F10 tumors were treated with AICAR, statin, or both. We observed that the combination therapy led to a reduction in tumor growth, with increased frequencies of CD4 + and CD8 + T cells and no change in Tregs (Fig. 4E and Fig. S12A). However, reduced tumor growth was not observed in Rag1 −/− mice (Fig. 4F), suggesting that the antitumor effect of this combination therapy depends on T cells and B cells. Indeed, GZB-and IFN-γ-producing CD8 + T cells were highly increased after AICAR and statin co-treatment in tumor-bearing C57BL/6 J mice (Fig. S12B). Also, . D Immunoprecipitation of the lysates of WT Tregs using anti-AMPK and anti-HMGCR antibodies. E The volume and weight of B16F10 tumors from WT mice after treatment with PBS, AICAR, statin, and AICAR + statin combination therapy. F The volume and weight of B16F10 tumors in Rag1 −/− mice following PBS treatment or co-treatment with AICAR and statin. G MFI of PD-1 in Tregs from tumor-bearing WT mice after the indicated treatments. The data are presented as the mean ± standard deviation (SD); n = 5 mice per group. *P < 0.05; **P < 0.01; ***P < 0.001 the expression of PD-1 in Tregs was significantly reduced after combination therapy (Fig. 4G). These findings suggest that the activation of AMPK synergizes with the inhibition of HMGCR to suppress tumor growth via the downregulation of PD-1.

HMGCR regulates the expression of PD-1 via p38 MAPK
We determined the mechanism underlying HMGCRmediated PD-1 regulation. We hypothesized that HMGCR might regulate the expression of PD-1 through the mevalonate pathway. Therefore, we assessed the expression of PD-1 in vitro in Tregs treated with the byproducts of the mevalonate pathway, including mevalonate, GGPP, and cholesterol. However, there was no significant difference in the expression of PD-1 after treatment with these byproducts in vitro (Fig. 5A) and in vivo (Fig. S13), suggesting that the HMGCR-mediated regulation of the expression of PD-1 occurs in a mevalonate pathway-independent manner.
Since previous findings suggested that p38 MAPK is involved in the regulation of cholesterol metabolism [30], we investigated whether the canonical MAPK pathway was responsible for the HMGCR-mediated regulation of PD-1. We found no changes in the levels of the phosphorylated and total forms of ERK and JNK, whereas the phosphorylation of p38 was decreased in AMPK-KO Tregs (Fig. 5B). In addition, the level of phosphorylated p38 was increased by treating Tregs with AICAR (Fig. 5C) and decreased by treating with compound C (Fig. 5D), suggesting that AMPK positively regulates p38. Further, the phosphorylation of p38 was significantly increased in Tregs after statin treatment in a dose-dependent manner, whereas no changes were observed in the expression of ERK and JNK (Fig. 5E, F), suggesting that HMGCR negatively regulates the phosphorylation of p38.
We also investigated whether p38 regulates the expression of PD-1. The expression of PD-1 in WT Tregs was significantly increased after treatment with a p38 inhibitor (SB203580) in vitro (Fig. 5G) and in vivo (Fig. 5H). Furthermore, the p38 inhibitor led to increased B16F10 melanoma tumor growth in WT mice in vivo (Fig. 5I,  J). Remarkably, the expression of PD-1 in tumor-isolated Tregs was significantly upregulated by SB203580 (Fig. 5K), suggesting that p38 could regulate the expression of PD-1 both in tumor-free and in tumor-bearing mice. Thus, these findings suggest that HMGCR regulates the expression of PD-1 via p38 MAPK.

Degradation of GSK3β by p38 is involved in the regulation of PD-1
We identified the likely downstream targets of p38 MAPK regulating the expression of PD-1. Since p38 MAPK regulates the Wnt-β-catenin signaling via the inactivation of GSK3β [31,32], we first checked the expression of GSK3β and β-catenin in WT and AMPK-KO Tregs. We found that the total form of GSK3β (the active form) was highly increased, whereas phosphorylated GSK3β at Ser9 (the inactivated form) was reduced in AMPK-KO Tregs (Fig. 6A). In accordance, the levels of phosphorylated β-catenin were highly increased, whereas the expression of total β-catenin was decreased in AMPK-KO Tregs, suggesting the degradation of β-catenin and the activation of GSK3β (Fig. 6A). To disclose the potential role of AMPK, we treated WT Tregs with AICAR and compound C. The degradation of GSK3β was enhanced, as evidenced by the increase in the expression of the phosphorylated form after AICAR treatment (Fig. 6B). Conversely, compound C increased the level of total GSK3β and decreased that of phosphorylated GSK3β in Tregs (Fig. 6C). Additionally, the levels of total β-catenin increased after AMPK activation (Fig. 6B) and decreased after AMPK inhibition (Fig. 6C), indicating that AMPK positively regulates the expression of β-catenin and negatively regulates that of GSK3β.
To understand whether p38 MAPK is an upstream regulator of GSK3β, we treated WT Tregs with SB203580. As expected, SB203580 treatment reduced phosphorylated GSK3β, resulting in the accumulation of total GSK3β. Notably, AMPK and HMGCR remained unaffected (Fig. 6D), suggesting that p38 MAPK is an upstream regulator of GSK3β and a downstream effector of AMPK and HMGCR. Furthermore, we observed that p38 MAPK binds to GSK3β (Fig. 6E). Thus, these data suggest that AMPK in Tregs regulates GSK3β via p38 MAPK, thereby inhibiting the expression of PD-1.
To determine whether the increased levels of GSK3β in AMPK-KO Tregs are directly responsible for the upregulation of PD-1 expression, we treated AMPK-KO Tregs with SB216763, a GSK3β inhibitor. Inhibition of GSK3β reduced the expression of PD-1 in WT and AMPK-KO Tregs, compared to their untreated counterparts, albeit AMPK-KO Tregs still exhibit a substantially higher level of PD-1 than WT Tregs, suggesting that AMPK regulates PD-1 expression via both GSK-dependent and -independent mechanisms (Fig. 6F). Previous findings suggested that the inhibition of GSK3β enhanced the expression of T-bet, thereby downregulating the expression of PD-1 in CD8 + T cells [33]. Here, we found that T-bet was downregulated in AMPK-KO Tregs (Fig. 6G). Notably, we confirmed this phenotype depended on GSK3β as the expression of T-bet was significantly increased in WT and AMPK-KO Tregs treated with SB216763 (Fig. 6G). Thus, these findings suggest that GSK3β promotes the expression of PD-1 in Tregs through the inhibition of T-bet.

Discussion
T cell activation is associated with metabolic changes to address the increased energy demand for effector functions. Especially, activated Th1 and Th17 cells mainly depend on glycolysis, controlled by the energy sensor AMPK coordinating energy homeostasis [6,34]. In cancer cells under metabolic stress, lowered expression of AMPK is associated with increased energy demand, probably due to the low oxygen levels in the TME and the consequent shift toward glycolysis. In this respect, T cells lacking AMPKα1 also display reduced mitochondrial bioenergetics and cellular ATP levels in response to glucose deprivation or to pathogens [2]. Likewise, AMPK is also crucial to maintain the function of Tregs [35]. However, Tregs mainly depend on FAO to survive and function [4,35]. In this study, we found that the ablation of AMPK in Tregs increased glycolysis and the expression of PD-1, suggesting that AMPK may be modulated in Tregs to The I tumor volume and J weight of B16F10 melanoma tumors in WT mice treated with PBS or SB203580. K MFI of PD-1 in Tregs isolated from tumor tissues after treatment with PBS or SB203580. The data are presented as the mean ± standard deviation (SD); n = 5 mice per group. *P < 0.05; **P < 0.01; ***P < 0.001 promote tumor suppression. Although the regulatory mechanism underlying the expression of PD-1 under conditions of metabolic stress is not fully understood, we suggest that AMPK regulates the expression of PD-1 via the HMGCR/P38 MAPK/GSK3β axis in Tregs.
Tregs are a major barrier to antitumor immunity in various cancers [36]; for instance, they secrete immunoregulatory cytokines, including IL-10 and TGF-β. In addition, Tregs express several inhibitory surface receptors including PD-1, ICOS, and Nrp1 in tumors, mediating immunosuppression [25]. Notably, PD-1 is also expressed in exhausted T cells [37]. In fact, PD-1 blockade is used to treat cancer patients not responding to classical chemotherapy agents, promoting the restoration of the effector functions of exhausted T cells [38,39]. However, despite the importance of PD-1 in immune regulation, the role of PD-1 in Tregs remains controversial [39,40]. A recent study suggested that PD-1 expression in Tregs was amplified by PD-1 blockade; consequently, PD-1 blockade accelerated tumor growth [40,41]. This contradicts our results which shows that the PD-1 blockade in Tregs suppresses tumor growth. However, since antitumor CD4 + and CD8 + T cells were restored by the PD-1 blockade in our study, it is highly plausible that our phenotype depends mainly on the PD-1 expression in Tregs. In this regard, we presume that the enhanced tumor growth in AMPK fl/fl Foxp3-Cre mice is associated with high levels of PD-1 in Tregs. Notably, the PD-1 blockade exhibited profound anti-tumor effects in murine syngeneic tumor models, particularly in AMPK fl/fl Foxp3-Cre mice. Although anti-PD-1 antibody is expected to act on both Tregs and exhausted T cells, it profoundly exhibited antitumor effects in AMPK fl/fl Foxp3-Cre mice compared to WT mice, suggesting that the increased antitumor immunity is due to the PD-1 blockade in Tregs, which is increased in the absence of AMPK signaling.
Despite the importance of PD-1 as a negative feedback regulator of T cell effector functions, the upstream pathway involved in the downregulation of PD-1 is yet unknown. Here, we discovered that AMPK regulates the expression of PD-1 through the HMGCR/p38 MAPK/ Fig. 6 AMPK suppresses the activity of GSK3β through p38 MAPK activation to regulate the expression of PD-1. A Analysis of phosphorylated and total GSK3β and β-catenin in WT and AMPK-KO Tregs stimulated with or without anti-CD3/CD28 stimulation for 4 h. Analysis of the expression of GSK3β and β-catenin by western blotting in activated Tregs after B AICAR and C compound C treatment at the indicated doses. D Immunoblot analysis of the indicated proteins in activated Tregs treated with the p38 MAPK inhibitor SB203580 at the indicated doses. E Immunoprecipitation assay of the lysates of WT Tregs using anti-p38 and anti-GSK3β antibodies. Flow cytometric analysis of F PD-1 and G T-bet in Tregs from WT and AMPK fl/fl Foxp3-Cre mice after i.p. treatment with or without the GSK3β inhibitor SB216763 daily for 7 days (2 μg/kg). The data are presented as the mean ± standard deviation (SD); n = 5 mice per group. *P < 0.05; **P < 0.01; ***P < 0.001 GSK3β axis. Given that p38 MAPK is regulated by AMPK and by statin, an HMGCR inhibitor [42,43], we found that HMGCR in Tregs could be regulated by AMPK. Consistently, we found that the inhibition of HMGCR by statin activates p38 and suppresses the expression of PD-1 in Tregs, suggesting that HMGCR is an upstream regulator of p38MAPK. However, the detailed mechanism underlying the HMGCR-mediated regulation of p38 MAPK needs to be further explored. Additionally, previous studies suggested that p38 MAPK inactivates GSK3β via phosphorylation [32]. We also found that p38 MAPK interacted with GSK3β and promoted its phosphorylation, suppressing the expression of PD-1; these data suggest that GSK3β is a downstream substrate of p38 in Tregs. Interestingly, our data suggest that GSK3β upregulates the expression of PD-1 in Tregs via the inhibition of T-bet, in line with data reported in CD8 + T cells [36]. Notably, GSK3β-mediated phosphorylation enhanced the proteasomal degradation of β-catenin, a key mediator of Wnt signaling [44]. Activated Tregs increased the expression of β-catenin, but it was attenuated in LKB1deficient Tregs [45]. In this regard, the decreased levels of β-catenin detected in AMPK-KO Tregs may explain the increased expression of PD-1; However, it is still unclear how GSK3β regulates PD-1 expression directly. One hypothesis is that GSK3β-mediated ER stress can induce PD-1 expression. GSK-3β inhibition increases IRE1αdependent XBP1 splicing, which directs the transcription of several genes involved in the functional and structural expansion of the endoplasmic reticulum (ER) and genes associated with the ER-associated degradation (ERAD) pathway in order to reduce ER stress and restore ER homeostasis [46]. In addition, another study reported that activated XBP1 binds to the Pdcd1 promoter and 2B4 promoter to regulate the expression of inhibitory receptors in CD8 + T cells [47] however, the level of PD-1 expression was not changed in XBP1-deficient CD4 + T cells [48]. Our preliminary data suggested that the expression of the spliced form of XBP1 was increased in AMPK-deficient CD4 T cells (data not shown). However, the mechanism of XBP1-mediated PD-1 expression is still unclear and further studies are needed to elucidate the mechanism.
LKB1 is a master kinase that functions upstream of AMPK, along with TAK1 and CaMKKβ, which acts as a potent tumor suppressor that directly phosphorylates and activates AMPK [8]. However, our previous study suggested that the phenotype of Treg-specific AMPK fl/fl Foxp3-Cre mice differs from Treg-specific LKB1-KO mice, as only Treg-specific LKB1-KO mice developed spontaneous autoimmune inflammation [49]. Interestingly, LKB1-KO Tregs showed hyperactivation of mTORC1 signaling [45], while AMPK-KO Tregs did not. Moreover, the loss of LKB1 in Tregs reduced HMGCR expression [45], but the loss of AMPK in Tregs increased HMGCR expression. Thus, we presumed that the role of AMPK is distinct from that of LKB1 in CD4 + T cells. Additionally, AMPK expression was significantly reduced in tumor-infiltrating Tregs of WT mice, whereas there was no significant change in LKB1 levels. This finding suggests that AMPK exerts a tumor-suppressive effect independent of LKB1 in Tregs. Collectively, it can be presumed that AMPK attenuates the immune suppressive function of Tregs in the TME, thereby potentiating antitumor immunity. " Inhibitory immune checkpoint blockade using anti-PD-1 and anti-CTLA4 antibodies has provided substantial benefits to certain cancer patients [50]. However, the monotherapy clinical outcomes are not satisfactory yet. Several approaches are therefore being considered to improve the efficacy of these immune checkpoint inhibitors. For instance, combined therapy with existing anticancer therapies, such as chemotherapy, radiotherapy, and targeted therapy, was proposed [51]. However, such combined therapy could induce severe toxicity and side effects. Our findings showed that the AMPK activator, AICAR, exhibited synergistic antitumor effects when combined with anti-PD-1 and anti-CTLA4 antibodies, suggesting that AMPK activators could be considered to complement anti-PD-1 and anti-CTLA4 therapy to improve the outcomes of cancer patients. In addition, HMGCR inhibitors have been widely used as cholesteroltargeting drugs in clinical studies in cancer patients [29]. In our study, a synergistic effect was observed when statin was combined with AICAR, suggesting that AMPK activation combined with the inhibition of HMGCR could be a potential combination therapy for cancer treatment.

Conclusion
In conclusion, we demonstrated the role of AMPK in Tregs in regulating antitumor immunity: AMPK promotes antitumor immunity due to the downregulated expression of PD-1 via the HMGCR/P38 MAPK/GSK3β axis. The activation of AMPK combined with anti-PD-1 and anti-CTLA4 antibodies or with a HMGCR inhibitor exhibited synergic anti-cancer activity in murine tumor models, supporting their potential clinical use. Altogether, our findings provide support for the notion that AMPK in Tregs acts not only as a crucial regulator maintaining metabolic homeostasis, but also as a potent tumor suppressor.
Additional file 1: Table 1. The real-time PCR primer list, along with the sequences.
Additional file 2: Figure S1. The correlation between AMPK and PD-1 mRNA expression in multiple cancer patient datasets. Figure S2. The phenotype of AMPK fl/fl Foxp3-Cre mice is similar with that of WT mice. Figure S3. Loss of AMPK in Tregs promotes TC-1 and MC38 tumor growth. Figure S4. AMPK deficiency in Tregs reduces antitumor T cell populations. Figure S5.AMPK fl/fl Foxp3-Cre mice show low GZB + CD8 T cell numbers per unit area. Figure S6. Metformin inhibits the expression of PD-1. Figure S7. Loss of AMPK enhances the expression of PD-1 in tumor infiltrated CD4 + T cells. Figure S8. Deficiency of AMPK causes down-regulation of CD25 expression by anti-PD1 antibody treatment. Figure S9. AMPK activation synergizes with CTLA4 checkpoint blockade to limit B16F10 tumor growth. Figure S10. AMPK activation synergizes with PD-1 checkpoint blockade to suppress tumor growth. Figure S11. Deficiency of AMPK in Tregs changes metabolic process with independent of mTORC1 signaling. Figure S12. Combined therapy of AICAR and Statin increases effector T cells. Figure S13. PD-1 expression in splenic Tregs from WT mice treated with mevalonate pathway by products in vivo.