PLT012, a Humanized CD36-Blocking Antibody, ls Effective for Unleashing Antitumor Immunity Against Liver Cancer and Liver Metastasis ?

Sheue-Fen Tzengl.2,Yi-Ru Yu3.45 Jaeoh Park45, Janusz von Renesse45,Huey-Wen Hsiao, Chen-Hsuan Hsu2. Josep Garnica4.57, Jintian Chen8, Lu-Ting Chiu3, Jonas Santol9.10.11, Tse-Yu Chen1213 Pei-Han Chung,( Lana E.Kandalaftl4 Patrick Starlingerl0.15.16 Rodney Cheng-En Hsieh12.13 Ming-Chin Yu17, PeiWen Hsial SantiagoJ,Carmona457, Hung-Kai Chen, Zhen Meng19, Yun-Han Lin3, Jingying Zhou8 Chin-Hsien Tsai220,21 and Ping-Chih Ho4.5

ABSTRACT

Tumor cells develop various strategies to evade immune surveillance,one of which involves altering the metabolic state of the tumor microenvironment. In response to metabolic stress in the tumor microenvironment, several tumor-infiltrating immune subsets upregulate CD36 to take up lipids.This leads to impaired antitumor immunity,as intratumoral regulatoryTcells exhibit increased survival and suppressive activity, whereas C_{ Ḋ }\mathsf{ Ḋ }8^{+} Ḍ T cells become more susceptible to ferroptosis and exhaustion. In this study, we develop a humanized anti-CD36 IgG4 antibody, PLT012, against the lipid-binding domain of CD36 with excellent safety and favorable pharmacokinetic features in mice and cynomolgus monkeys. PLT012 alone or in combination with PD-L1 blockade or standardof-care immunotherapy results in robust antitumor immunity in both immunotherapy-sensitive and -resistant hepatocellular carcinomas (HCC). Notably, PLT012 also reprograms the immune landscape of human HCCex vivo.Ourfindings provide proof-of-concept evidence that PLT012reprograms antitumor immunity in HCC, positioning it as a first-in-class immunotherapy targeting CD36.

SIGNIFICANCE: Despite the success of cancer immunotherapies,like immune checkpoint inhibitors, many patients stillfail to demonstrate significant responses because of metabolic constraints in tumors. PLT012 rejuvenates antitumor immunity by targeting metabolic pathways to reprogram the immune landscape of liver cancer and liver metastasis, with potential to influence future HCC immunotherapy.

INTRODUCTION

Metabolic transformations in cancer cells are controlled by oncogenic driver mutations and immunometabolic editing (1, 2) and are associated with tumor cell survival in the tumor microenvironment (TME). Metabolic reprogramming has been shown to modulate tumor growth, antiapoptotic mechanisms, and immune evasion (3, 4). Interestingly, lipid metabolism reprogramming is an emerging hallmark of cancer. The de novo lipogenesis, together with the recruitment ofadipocytes and adipocyte-likefibroblastsin the TME,can perturb lipid composition and abundance (5-7),which in turn facilitates cancer progression and immune evasion (1,2) Lipid enrichment in combination with other metabolic challenges, including glucose deprivation and hypoxia, imposes considerablemetabolicstress onstromalcellswithin the TME. Tumor-infiltrating lymphocytes (TIL) failing to adapt tometabolichurdles decrease their survivalrates andlose their antitumor capacity (8-11). However, immunosuppressive cells, including regulatory T cells (Treg), pro-tumorigenic tumor-associated macrophages (TAM), and myeloid-derived suppressor cells (MDSC) seem to employ alternative metabolic adaptation strategies to survive and impair antitumor immunity in the TME (4, 12). Therefore, finding a way to block the ability of immunosuppressive cells to escape metabolic stress and, at the same time, to improve the adaptation responses of tumoricidal immune cells is key to reprogramming the immune state in tumors and releasing host antitumor responses.

CD36, a fatty acid (FA) transporter and scavenger receptor, contains two hydrophobic clefts that effectively bind longchain FAs and oxidized low-density lipoprotein (oxLDL） to promote lipid uptake and satisfy the metabolic demands of rapidly proliferating cells (13, 14). In response to the lipidand lactic acid-enriched TME, intratumoral Tregs have been shown to increase CD36 expression, which in turn boosts mitochondrial functions via a peroxisome proliferator-activated receptor \upbeta/\updelta -mediated mechanism (12). By increasing mitochondrial activity, intratumoral Tregs survive in the lactate-rich TME and maintain their immunosuppressive activity. Intriguingly, it has recently been shown that CD36 expression is upregulated in metastasis-associated macrophages isolated from metastatic liver tumors. This increase in CD36 expression allows macrophages to use lipids from tumor cells and stimulate tumor-promoting activities (15). In contrast, a high CD36 expression in tumor-infiltrating CD8^{+} T cells leads to the accumulation of oxidized lipids and subsequent T-cell dysfunction and ferroptosis (10, 11). Similarly, NK cells that show increased CD36 expression in lipid-enriched contexts have impaired in vitro tumoricidal activity (16); also, genetic ablation of CD36 in NK cells to prevent lipid uptake helps maintain NK effector functions in lipid-rich environments (17, 18). These studies suggest that blocking CD36-mediated lipid uptake could be a promising approach to restore immune responses in tumors. Interfering with lipid uptake would simultaneously destroy intratumoral Tregs and macrophages and restore the survival and tumoricidal activities of tumor-infiltrating CD8^{+} T cells and NK cells.

Targeting CD36-mediated metabolic adaptations may elicit antitumor responses while preserving immunehomeostasis as CD36-mediated immune regulation is selectively employed by tumor-infiltrating immune cells rather than by their peripheral counterparts. Despite the fact that CD36 absence can be observed in 5% 0 8% of Asian and African populations without any notable morbidity and that physiologic parameters of CD36- deficient mice are within the norm (19, 20), the broad expression profile of CD36 in the human body, including endothelial cells, cardiomyocytes, and red blood cells (RBC; refs. 21-23), poses a problem. Indeed, using antibodies against broadly expressed proteins may cause a strong antibody-dependent immune activation which may lead to severe complications. Furthermore, small molecules that inhibit CD36-mediated lipid uptake have shown unsatisfying pharmacokinetic and pharmacodynamic properties in vivo (24). It has also been reported that these molecules primarily affect lipid metabolism rather than CD36- mediated lipid transport (24). Thus, it remains a challenge to establish whether we can use classic CD36-targeting approaches for cancer treatment in preclinical models and humans.

Here, we developed PLT012, a humanized IgG4 antibody targeting the lipid-binding pockets of CD36 with reactivity against multiple species and a superior safety profile in monkeys.We used abioinformatic analysis tofind human tumor types characterized by CD36-mediated immune regulation, and we showed that PLT012 antibody effectively increased antihepatocellular carcinoma (HCC) responses, including reduction ofintratumoral Tregs,enhancement of CD8^{+} T-cell infiltration,and improvement of cytotoxic functions in CD8^{+} T cells, in both immune hot and cold HCC types. Notably, we also found that PLT012 can increase the abundance of progenitorexhausted T cells (Prog Tex), a critical cell subset responsible for the responsiveness to immune checkpoint blockade treatment in tumors.In line with this,we demonstrated that PLT012 together with the PD-L1 antibody or in combination with standard liver cancer therapies (anti-VEGF plus antiPD-L1 agents) can elicit strong antitumor responses in mice with cold HCCs or in mice fed a high-fat diet which typically shows reduced responsiveness to anti-PD-L1 therapy. Additionally,we showed that PLT012 can beused toreduce colorectal liver metastases and to restore responsiveness to anti-PD-1 treatment in mice with hepatic metastasis from primary colon cancer. The mechanism of action of PLT012 was further validated in human HCC samples using an ex vivo culture platform.Our results show that PLT012 blocks CD36-mediated metabolic adaptation in Tregs and CD8^{+} TILs, thus leading to robust tumorgrowth inhibition and a shift in the nature of the TME: from immunosuppressive to immunosupportive.

RESULTS

Generation and Characterization of PLT012,a Cross-Reactive anti-CD36 Antibody

The development of antihuman CD36 mAbs that can block CD36-mediated lipid uptake is largely hampered by (i) the hydrophobic nature of the CD36 lipid-binding pockets, which can lead to the identification of candidates with weak and/or non-specific binding and (i) the high frequency of false-negative results in standard antigen capture assays for candidate identification (25). To overcome these obstacles, we first screened synthetic human single-chain variable fragment phage display libraries to identify a primary hit antibody, followed by panning with light-chain shuffling libraries to enhance binding affnity against CD36. To further improve drug properties,two additional phage display libraries harboring complementary-determining region (CDR) mutations were used for panning, resulting in an optimized CDR sequence for the lead antibody. Moreover, site-specific mutagenesis was performed on the CDR, guided by artificial intelligence modeling and CamSol predictions,to improve retention time, which may be associated with non-specific hydrophobic interactions. Considering that CD36 can be expressed in many cell types, we engineered antibody candidates into an IgG4 form to reduce antibody-dependent immune responses (26) and optimized the Fc domain to minimize aggregation during antibody production. By using these approaches, we identified PLT012, a full-length human monoclonal IgG4 antibody, displaying reactivity against CD36 across various species, including rodents, monkeys, and humans. Of note, because of its cross-reactivity, we could examine PLT012 biochemical properties and antitumor functions without using murine surrogate antibodies. Next, we evaluated whether PLT012 could compete with D2712, a commercially available antibody against the murine CD36 lipid-binding pocket. Our results showed that PLT012 can compete with D2712 for binding to CD36 (Fig. 1A), indicating that PLT012 may recognize a similar epitope and inhibit CD36-mediated uptake of FAs and oxLDL. Indeed, we found that PLT012 strongly inhibited the binding (IC_{50}; 1.798nmol/L and uptake (IC_{50} 1.357\ nmol/L ）of fluorochrome-labeled oxLDL in F293 cells that overexpress human CD36 (Fig. 1B). In addition, we confirmed that PLT012 treatment significantly repressed oxLDL uptake compared with control IgG in primary tumor-infiltrating immune cells, including myeloid cells, CD3^{+} T lymphocytes, and intratumoral Tregs isolated from MC38 tumor-bearing mice (Fig. 1C). Then, we examined whether

FigurAsenfthiinpifPanhenfAtDxresincel incubated with PLT012 at concentrations ranging from 0.002 to 10\upmug/mL followed by FACS analysis using commercial anti-CD36 antibodies, D2712. B,HumanCD36-expressingF293cellswerepre-incubatedwithPLT012atconcentrationsrangingfrom0.01t 30\upmug/mL followedby oxLDL-Dil staining for oxLDL binding with an 1mathsf{C}_{50} at 0.25 (left) and uptake with an 1C_{50} at 0.19 (right）MFI,mean fluorescence intensity.C,oxLDL-Diluptake in the indicatdelltt compriinfntfsicrytinieei corp FLAG-ta3ieifiinttiomnmntreefneratiof octitist was conductedusinga two-tailed,unpairedStudentttest.

PLT012 treatment could suppress tumor growth in syngeneic mouse melanoma and colorectal cancer models. In agreement with previous reports showing that CD36-deficient mice displayed reduced tumor growth (10-12), PLT012 promisingly exhibited antitumor responses in both Yumm1.7 melanoma and MC38 colorectal cancer models (Supplementary Fig. S1A and S1B).Altogether, our results show that PLT012, an anti-CD36 mAb with reactivity against multiple species, can effectively inhibit CD36-mediated lipid uptake and induce effective antitumor responses.

Identification of PLT012-BindingSites on CD36

PLT012 exhibited affinity for CD36 epitopes that are analogous to those recognized by the commercially available antibody against the murine CD36 and also possessed the ability toinhibit CD36-mediatedlipid uptake;however,the specific regions involved in the CD36-PLT012 interaction remain unknown. To characterize the binding properties of PLT012 and to examine whether the interaction interface is within thelipid-binding pocket of CD36, we used cryogenic electron microscopy (cryo-EM). Our cryo-EM analysis identified two interaction domains of PLT012 on the extracellular domain (ECD) of CD36: domain 1,located within amino acids 153 to 160,and domain 2,located in a region between amino acids 191 to 197 (Fig. 1D). These domains are positioned within the FA-binding and oxLDL-binding regions of CD36 as confirmed by prior domain-mapping studies (27). To investigate the impact of these domains in the PLT012-CD36 interaction, we first introduced point mutations to alter the charge and hydrophobicity of CD36 and then measured the binding affinity with PLT012. Flow cytometry-based binding assays showed that point mutations in domain 1 reduced the bindingof PLT012 to theECDof CD36by almost 50% whereas mutations in domain 2 did not affect the interaction between the protein pair.Notably, the interaction between PLT012 and CD36 was almost completely abolished when both domains were mutated (Fig. 1E). These results suggest that PLT012 preferentially interacts with domain 1 to limit lipid entry into the cell, whereas domain 2 only provides supplementary support.

PLTO12 Targets Immune Infiltrates in HCCs

CD36-mediated lipid uptake influences the immunosuppressive TME by modulating the survival and functions of intratumoral Tregs and CD8^{+} T cells (10-12). Considering this, we hypothesized that PLT012 could elicit robust antitumor responses in lipid-rich tumors. By using data from The Cancer Genome Atlas (TCGA), we performed an in silico analysis to examine the tumor types that displayed both high expression of CD36 and an enrichment in genes relevant to FA metabolism (Fig. 2A). The transcriptional profiles of the top nine tumor types retrieved from the in silico analysis were further analyzed with CIBERSORT - a deconvolution method that characterizes the cell composition of complex tissues from their gene expression profiles - to determine the relative abundance of CD8^{+} T cells and Treg cell subsets (28). The results from the sequential analysis suggest that liver HCC may be a tumor type that can be sensitive to anti-CD36 antibody treatment (Fig. 2B). We did not include cholangiocarcinoma in our study because of the potential biases stemming from a limited sample size. To confirm whether TILs increase their CD36 expression in the context of liver cancer, we measured CD36 expression in Tregs and CD8^{+} T cells from peripheral blood mononuclear cells, spleens, and liver of HCC-bearing mice. To replicate the pathologic progression of HCCs, we induced liver cancer onset by overexpressing MYC and a constitutively active mutant of \upbeta -catenin in hepatocytes (M Y C^{oE}/ CTNNB1N90 HCC) with a hydrodynamic injection system (Supplementary Fig. S2A).By staining single-cell suspensions with Alexa Fluor 647-conjugated PLT012, we found that both Tregs (CD45^{+C D3^{+}C D4^{+}F o x P3^{+})} and CD8^{+} T cells (CD45^{+C D3^{+}C D8^{+})} isolated from HCC-bearing mice expressed higher levels of CD36 compared with their counterparts obtained from liver, spleen, or peripheral blood of normal healthy mice (Fig. 2C and D). Of note, we also observed that NK cells (CD45^{+}CD3^{-}NK1.1^{+}) in tumors have increased CD36 expression, which has been reported to impair NK cell cytotoxicity (16, 17). These findings are consistent with previous studies reporting that the TME promotes CD36 expression in tumor-infiltrating immune cells (10, 12). To further examine PLT012 distribution within tissues and to determine the kinetics of the treatment with PLT012, we intraperitoneally injected Alexa Fluor 647-conjugated PLT012 ( 10mg per kg of body weight) into control or CTNNB ^{\primeN90}/M Y C^{OE} HCC-bearing mice and collected samples from various organs at 4, 24, and 48 hours after injection. We used in vivo bioluminescent imaging (IVIS) to quantify the fluorescence intensity in the samples and showed that PLT012 mainly accumulates in the liver, lung, intestine, epididymal white adipose tissue, and pancreas (Supplementary Fig. S2B and S2C).Healthy and tumor-bearing mice showed similar fluorescence intensity levels which maybe attributed to the presence of adipocytes and endothelial cells, which express high levels of CD36 (5, 21, 29). Importantly, 24 hours after injection, PLT012 signal was significantly stronger in livers from HCC-bearing mice compared with livers from normal mice, suggesting that the HCC microenvironment may facilitate the recruitment of CD36-expressing cells (Fig. 2E). To verify that PLT012 treatment can target TILs,we quantified theAlexa Fluor647signal in Tregs, CD8^{+} T cells, and NKs from livers of normal and HCC-bearing mice intraperitoneally treated with one dose of labeled PLT012. We observed that both Tregs and CD8^{+} T cells from HCC-bearing mice had a higher percentage oflabeled cells compared with their counterparts from normal mice. Notably, these differences persisted for 48 hours following mAb administration (Fig. 2F and G). In addition, NK cells from tumor-bearing mice exhibited higher levels of PLT012 signal at the 24-hour time point, but not at the 48-hour time point, compared with NK cells from normal livers (Fig. 2H). Taken together, these results reveal that the HCC microenvironment could increase CD36 expression in infiltrating immune cells, including Tregs, CD8^{+} T cells, and NK cells, and that PLT012 may have the potential to restore antitumor immunity against HCC by targeting CD36-mediated immune regulations.

PLT012 Elicits Strong Antitumor Responses in Murine HCCs

We next examined the ability of PLT012 to stimulate antitumor responses in HCCs. To do so, we first assessed the impact of PLT012 treatment on T-cell activity and tumor growth in two murine HCC models:(i) the MYCOE/ T r p53^{KO} HCC model - driven by MYC overexpression and theloss of p53 function in hepatocytes - and (i) the C T N N B I^{N90}/M Y C^{OE} HCC model - driven by MYC overexpression and characterized also by the overexpression of a mutated \upbeta -catenin (C T N N B I^{N90} ;Supplementary Fig. S2A). M Y C^{oE}/T r p53^{\tinyKO} HCC is known to have an inflamed (hot) tumor microenvironment, characterized by high infiltration of CD103^{+} dendritic cells and CD8^{+} T cells,and exhibits a slower progression compared with CTN N B I^{N90}/M Y C^{OE} HCCs (Supplementary Fig. S3A and S3B; ref. 30). We found that PLT012 treatment strongly limited M Y C^{OE}/T r p53^{KO} HCC growth compared with control treatment and resulted in 32% of the HCC-bearing mice being tumor-free (Fig. 3A; Supplementary Fig.S3C). Measurements of tumor weight and ex vivo liver bioluminescence imaging (BLI) further corroborated these findings (Fig. 3B and C). Of note, PLT012 treatment also mitigated liver injury associated with tumor progression as indicated by the reduced serum alanine transaminase (ALT) activity (Supplementary Fig. S3D). Furthermore, PLT012 treatment led to a reduction of Tregs and an increase of CD8^{+} T cells within the TME (Fig. 3D). Consequently, the CD8^{+} T/Treg ratios in HCC from PLT012- treated mice were higher compared with the control group (Supplementary Fig.S3E).As targeting CD36 has been shown tomitigate exhaustion and maintain the effector function of CD8^{+} T cells (10, 11), we next examined whether PLT012 could alleviate T-cell exhaustion by controlling the expression of exhaustion markers as well as granzyme B (GrB, a key molecule for cytolytic activity). Strikingly,PLT012 enhanced the abundance of CD8^{+G r B^{+}} T cells (Fig. 3E). Moreover, PLT012 treatment resulted in an increase in both Prog Tex (characterized by CD44^{+P D-1^{+}T C F1^{+}T i m}3^{-}) and terminally exhausted T cells (Term Tex, characterized by CD44^{+P D^{-1^{+}T C F1^{-}}} Tim3^{+}; Fig.3F).Ourresults,together with theobservation that Prog Tex abundance can influence the responsiveness to PD-1 blockade treatment (31-33), suggest that PLT012 is capable of restoring antitumor responses in HCC and might also improve responsiveness to PD-1 blockade.

More than 35% of patients with HCC harbor genetic mutations that lead to aberrant activation of WNT/ ^{\primeβ} -catenin signaling. These types of HCC have a cold TME with extremely low responsiveness to immune checkpoint inhibitors, in part due to a scarcity of TILs (30). To verify that CD36 blockade with PLT012 can overcome \upbeta -catenin-driven immune tolerance in HCC, we treated C T N N B I^{N90}/M Y C^{OE} HCC-bearing mice with PLT012.Our results showed that, like in the M Y C^{oE}/T r p53^{\tinyKO} HCC model, PLT012 significantly inhibited tumor growth and markedly decreased circulating ALT levels compared with the control treatment (Fig. 3G-I; Supplementary S3F). PLT012 treatment also increased the abundance of CD8^{+} T cells and decreased the proportion of Tregs in the tumors, thereby resulting in increased ratios of CD8^{+} T cells to Tregs within individual tumors (Supplementary Fig. S3G and S3H). In a similar way, PLT012 treatment of CTN N B I^{N90}/M Y C^{OE} HCC mice also increased the abundance of GrB-expressing CD8^{+} T cells as well as the population of Prog Tex and Term Tex CD8^{+} T cells (Supplementary Fig. S3I and S3J). To confirm the compositional changes in the immune subsets, the activation of T cells, and tumor cell death occurrence in the tumors, we performed IHC experiments to visualize FOXP3, CD8, GrB, and cleaved caspase-3. Our results confirm that PLT012 treatment can transform the TME from an immunosuppressive milieu to an immunosupportive environment characterized by high infiltration of CD8^{+} T cells into the tumor and a low abundance of intratumoral Tregs (Fig. 3J and K). We also observed a stronger signal for GrB and cleaved caspase-3 in tumors from PLT012-treated mice compared with controls, suggesting that CD8^{+} T cells exhibited enhanced cytotoxic activity that promotes tumor cell death. PLT012 significantly enhanced the antigen-specific T-cell response as assessed by ovalbumin (OVA)-specific tetramer staining (Supplementary Fig. S3K).In tetramerpositive CD8^{+} T cells, there were increased frequencies of antigen-specific Prog Tex and Term Tex in the PLT012 group, indicating that PLT012 boosts the adaptive immune system to control tumor progression (SupplementaryFig.S3K).

To explore the global immune response following treatment with PLT012, we isolated CD45^{+} tumor-infiltrating cells and subjected them to single-cell RNA sequencing (scRNA-seq).We next studied the changes in T-cellpopulations and classified the different cell types with scGate (34). We identified various CD8^{+} T-cell subsets, Tregs, Th1 cells, T follicular helper cells, and unconventional T cells based on their lineage-specific markers (Fig. 3L; Supplementary Fig. S3L) In line with the flow cytometry results, the scRNA-seq compositional analysis showed a reduction in Tregs and an increase in both total Term Tex and Prog Tex cells (Fig. 3M; Supplementary Fig. S3M). Noteworthy,PLT012 treatment induced robust changes in the transcriptome of Term Tex cells compared with their counterpart from the control group (Supplementary Fig.S3N;Supplementary Table S1).The analysisofdifferentiallyexpressedgenesshowedthatgene signatures associated with T-cell activation and effector function were significantly enriched in Term Tex cells from PLT012-treated tumors (Fig. 3N; Supplementary Fig. S3O; Supplementary Table S2), suggesting that PLT012 treatment can restore effector functions in exhausted T cells. Additionally,we observed that PLT012 treatment alters the expression of genes involved in lipid metabolism pathways in Term Tex cells (Supplementary Fig. S3P; Supplementary Table S2). Furthermore, the depletion of CD8^{+} T cells significantly undermines the therapeutic effects of PLT012 (Supplementary Fig.S3Q),indicating that PLT012 treatment can induce CD8 T cell-mediated anti-HCC responses. Together, our results demonstrate that PLT012 treatment can induce robust anti-HCC immune responses in both hot and cold TME, increase CD8^{+} TIL cell number, restore CD8^{+T I L} effectorfunctions, as well as reduce the intratumoral Treg populations.

PLT012EnhancestheEfficacyoftheStandard-ofCare Immune Therapy for HCCs

Considering thatProg Tex cells are responsiblefor the antitumor responses induced by PD-1/PD-L1-blocking therapy (31-33) and that PLT012 increases the amount of Prog Tex populations in HCCs, we hypothesized that PLT012 may enhance the therapeutic effcacy of the PD-1/PD-L1-blocking therapy. To verify this, we treated CTNNB1N90/MYCOE HCCbearing mice with either anti-PD-L1 mAb alone, PLT012 alone, or a combination of PLT012 and anti-PD-L1 mAbs. As expected, anti-PD-L1 mAbs induced mild antitumor responses in cold HCCs; however, the combination of PLT012 and anti-PD-L1 mAbs in a single treatment exhibited more profound antitumor effects than monotherapy (Fig. 4A-C). Interestingly, the combinatorial therapy further boosted the ability of PLT012 to mediate the reprogramming of the immune state in tumors, including reducing Treg percentages, increasing the total number of CD8^{+} T cells, Term Tex cells, and Prog Tex cells, and concurrently elevating the amount of GrB-expressing CD8^{+} T cells (Fig. 4D-H). Next, we wondered whether the incorporation of PLT012 into the standard-ofcare (SoC) regimen for the treatment of human HCC, which uses anti-VEGF and anti-PD-L1 mAbs (35), could induce superior outcomes and therapeutic benefits.We found that including PLT012 in the SoC regimen resulted in a significant improvement in tumor suppression (Fig. 4I-K). Notably, more than 70% of the treated mice showed an overall positive response to the triple therapy (PLT012 ^+ anti-PD-L1 ^+ anti-VEGF), whereas the therapeutic effects following SoC therapy alone were only observed in 27% of the subjects (Fig. 4L). Importantly, although no subjects achieved a tumorfree status with the control treatment alone,anti-VEGF alone, or SoC alone, the combination of SoC and PLT012 led to a tumor-free status in 45% of the C T N N B I^{N90}/M Y C^{OE} HCCbearing mice. In conclusion, targeting CD36 with PLT012 presents a promising strategy tobe used in combination with immune checkpoint inhibitors to reprogram the tumor immune microenvironment in the management ofliver cancers.

FiueCnnhfthlitnnJnentatihis maesqtittittnfiarl counterstained withhematoxylin.Scalebars, 50\upmum L,UniformManifoldApproximationand Projection(UMAP)visualizationof Tcellsfilteredinsilico fromwhleivceliingbtanCtaleheulatodettedwaatateinPrjlngsi maalit subsets,respectively (See Supplementary Fig. 3MandSupplementaryTableS1).N,GSEAnormalizedenrichment scoresforT-cellactivationpathways, with significant enrichment(adjusted P value <0.05 )based on differential gene expression in C D8^{+} Tex cells from PLT012 versus Ctrl conditions sufilintsastiaththat Data are cumulative results of three independent experiments,represented as means ± SEM A-K and as a box plot with Tukey whiskers indicating the minimum to maximum range in K. The P value is determined by a two-tailed, unpaired Student t test.

PLT012 Restrains HCC Progression under High Dietary Lipid Uptake Conditions

Excessive dietaryfatintake andaderegulated lipidmetabolism in livers havebeen reported tomodulate FAbiosynthesis in HCC and promote aggressiveness in human and murine HCC (36-38). HCC subtypes characterized by increased FA biosynthesis are also associated with more aggressive phenotypes and reduced survival rates (39); to date, an effective treatment to restrict HCC progression under high dietary fat intake is still missing because of the complications due to auto-reactive T cells and changes in other immune parameters (40-42). As, in Tregs and CD8^{+} T cells, CD36 expression can be induced by a lipid-rich environment (10,12),we reasoned that feeding HCC-bearing mice a high-fat diet might promote CD36 expression in these cell types and enhance the immunosuppressive state in tumors. To test this hypothesis, 20 days after hydrodynamic injection, CTNNB1N90/ M Y C^{{oE}} HCC-bearing mice were fed either a chow diet (CD) or a Western diet (WD) rich in sugars and fats (Fig. 5A). The WDincreased mice body weight, accelerated the progression of liver cancer, and promoted an immunosuppressive TME as evidenced by the increase in intratumoral Tregs and the reduction in CD8^{+} T cells (Supplementary Fig. S4A-F). Notably, by using both PLT012 and HM36 -a commercially available anti-CD36 antibody -we found that feeding mice a WDpromoted CD36expression in intratumoral Tregs and tumor-infiltrating CD8^{+} T cells,including both Prog Tex and Term Tex cells (Supplementary Fig. S4G and S4H),indicating that an increase in lipid uptake may reinforce CD36-mediated immune responses in the TME. Next, we tested the antitumor responses of PLT012 and anti-PD-L1 mAbs in these experimental settings. In alignment with a previous report, the administration of the anti-PD-L1 antibody only limitedly curtailed tumor growth (43); however, PLT012 consistently showedagreat ability toreducetumorgrowthwhen themice were maintained on a CD (Fig. 5B-D). Notably,PLT012 treatment remained effective in restricting tumor growth even in mice fed a WD. Differently from the observations with the anti-PD-L1 mAb treatment, PLT012 treatment resulted in less intratumoral Tregs in HCC from both CD- and WD-fed groups (Fig. 5E). Furthermore, PLT012 treatment strongly promoted the infiltration of CD8^{+} T cells, including Prog Tex and Term Tex cells,and increased the amount of {GrB^{+}} populations (Fig. 5F-I), which led to a more robust antitumor immune response in liver cancers with an aberrant dietary lipid uptake. Collectively,these results support the hypothesis that targeting CD36with PLT012 represents a promising strategy to restore immune responses in HCCs with or without excessive metabolic challenges due to dietary lipid uptake.

PLT012 Treatment Restricts Liver Metastasis and Restores Systemic Antitumor Immunity

The immune tolerance of the liver, in particular the development of Tregs, significantly impairs the immunosurveillance mechanisms (44-46), rendering the liver a frequent site for metastases from various malignancies.The presence of liver metastases in patients with cancer is associated with a reduced response to immunotherapy, which represents a critical unmet medical need that requires additional attention (47). Taking advantage of the unique attributes of the POG570 cohort (48), which includes a comprehensive transcriptome dataset of biopsy samples from various metastatic sites in patients with advanced malignancies, we initially assessed the metabolic properties of the TME (Supplementary Fig. S5A). Following the exclusion of metastatic samples with inadequate sample sizes, our analysis revealed that liver metastases across multiple cancer types exhibited a TME characterized by high CD36 expression,increased FA metabolism scores, Treg abundance, and reduced infiltration of CD8^{+} T cells when compared with other metastatic locations (Fig. 6A). To explore the primary tumor types that exhibited high FA scores and high CD36 expression in liver colonization, we applied the same analytic methodology to samples derived from liver metastasis of various cancer types in the same cohort. Our analysis showed that liver metastasis from breast cancer, colon adenocarcinoma, pancreatic cancer, and stomach cancer exhibited significant lipid metabolism alterations with respect to other metastatic diseases or other metastatic sites within the same cancer type (Fig. 6B). Notably, the tissue microenvironment is not the sole determinant of FAmetabolism activation.For instance, liver metastases from non-small cell lung carcinoma did not display significant FA metabolismrelated changes in comparison with other metastatic sites of non-small cell lung carcinoma, suggesting that both intrinsic factors and the surrounding environment collectively influence metabolic features (Fig. 6B; Supplementary Fig. S5B). Therefore, we further assessed the correlation among CD36 expression, FA score, and exhaustion score to identify tumor types with liver metastasis which could be used for CD36 blockade treatment with mAbs (Supplementary Fig. S5C). Our data showed that, in liver metastases originating from colon adenocarcinoma, CD36 expression strongly correlates with the FA score and the exhaustion score. This suggests that the growth of liver metastases may be significantly influenced by CD36-mediated immune regulation, highlighting the potential therapeutic implications of targeting CD36 in this specific context.

A previously published work reports that, in colon cancer liver metastasis, antigen-specific CD8^{+T} cells undergo apoptosis following interaction with CD11b^{+F4/80^{+}} monocytederived macrophages. Consequently,liver metastases eliminate tumor-specificT cells within thebody,leadingto acquired resistance to immune checkpoint blockade (47). Indeed,by using a tumor engraftment model that allows establishing subcutaneous and liver metastases within the same mouse, we found that PD-1 blockade partiallyloses its ability to limit subcutaneous MC38colorectal cancer growth whenliver metastases were present (Supplementary Fig.S6A).As we found that most liver metastases of human cancers have elevated

CD36 expression and display several tumor-immune microenvironmental features controlled by CD36, we reasoned that PLT012 could also target liver metastases and re-instate systemic antitumor immunity.To verify this, we established a preclinical model of liver metastasis by inoculating syngeneic MC38 colorectal cancer cells both subcutaneously and intrahepatically. We treated mice engrafted with MC38 cells with either control vehicle, anti-PD-1 mAb alone, PLT012 alone, or anti-PD- 1+{PLT}012\ (Fig. 6C). PLT012 exhibited superior antitumor effects compared with control treatment or antiPD-1 monotherapy on both liver metastasis (Fig. 6D) and local tumors (Fig. 6E) as assessed by BLI and endpoint tumor weight measurements (Supplementary Fig.S6B-S6E).Most importantly, the combination of PLT012 with the anti-PD-1 therapy significantly improved tumor suppression in primary tumors and the abundance of CD8^{+} TILs in both primary and metastatic tumors compared with monotherapy (Fig. 6D-G; Supplementary Fig. S6F). In addition, PLT012 treatment alone and in combination with anti-PD-1 effectively reduced intratumoral Treg abundance in both primary and metastatic tumors (Fig. 6F and G). Next, we examined whether PLT012 treatment could influence TAM phenotypes in subcutaneous tumors and liver metastases (Supplementary Fig. S6G). Our findings show that PLT012 alone is ableto promote an increase in the M1-like macrophage population (CD45+Ly6G-CD11b+F4/80+Tim4-Clec4fMHCII+CD206-) and a decrease in the M2-like population (CD45+Ly6G-CD11b+F4/ 80^{+}Tim4^{-} Clec4fMHCII-CD206+) in liver metastases (Fig. 6H and I). Similarly, PLT012 alone induced a remarkable increase in M1-like TAMs (CD45+Ly6G-CD11b+F4/80+MHCII+CD206-) but a reduction in M2-like TAMs (CD45+Ly6G-CD11b+F4/ 80+MHCIl-CD206t) in primary tumors (Supplementary Fig. S6H and S61). Notably, the combined treatment with PLTO12 and anti-PD-1 further amplified these differences in subcutaneous tumors (Supplementary Fig. S6H and S6I). Surface CD206 staining confirmed that PLT012 treatment, as well as the combined therapy treatment, reduced the {CD}206^{+} population in tumors, whereas treatment with anti-PD-1 alone failed to produce the same beneficial outcome (Supplementary Fig. S6J and S6K). Taken together, our results indicate that PLT012 can reprogram the tumor immune microenvironment, thus restricting colon cancer liver metastasis and improving sensitivity to PD-1-blocking therapy in mice with liver metastasis.

PLT012isWellToleratedinNonhumanPrimates

To get approval for new treatments in human patients, it is key to assess and prove the safety of the molecule under study. To this aim,we first confirmed CD36 expression in RBCs and platelets from mice, monkeys, and humans. Our results showed that among the three models, monkeys and humans have the most comparable CD36 expression patterns, characterized by low and high CD36 expression levels in RBCs and platelets, respectively (Fig. 7A). Also, CD36 expression levels in peripheral immune cell populations, including T cells, B cells, and monocytes, were comparable in monkeys and humans (Fig. 7B). Therefore, cynomolgus monkeys were chosen for the safety testing of PLT012 in a repeated-dose study (Fig. 7C). The administration of PLT012 was well tolerated in monkeys receiving three different dose levels (10,60,and 200\mg/kg) ，with no treatment-related mortality, clinical signs, or alterations in body weight, food intake, hematologic parameters, or coagulation metrics.Furthermore, alanine transaminase and aspartate aminotransferase levels were comparable between the three dosing groups and control groups, apart from a transient increase in the 200~mg/kg group at 24 hours after injection,indicating that PLT012 treatment did not affect liver functions throughout the study (Fig. 7D). Moreover, PLT012 treatment did not affect RBC and platelet counts, regardless of the different CD36 expression patterns in monkey RBCs and platelets (Fig. 7E). Immunophenotyping with FACS further indicated a good tolerance to PLT012 treatmentcharacterizedby astableimmuneprofile,including CD4^{+} T cells, CD8^{+} T cells, NK cells, and monocytes, maintained throughout the study (Fig. 7F). Consistently, a previously published phenotypic assay also indicated limited effects on hematopoiesis and similar T-cell phenotypes in Cd36-KO and wild-type mice (49). Altogether, these results demonstrate that PLT012 is exceptionally safe with no observed adverse effects,even at the highest dose (200mg/kg) ：

PLT012 Triggers Desirable Immune Responses in Human HCC Tumors

The pharmacodynamic effects, antitumor activity, and safety profile observed in the preclinical studies emphasize the potential of using PLT012 in cancer therapy. Interestingly, a newly established ex vivo culture model-developed from patient-derived tumor fragments - has been shown to maintain stromal compartments,including immune infiltrates and tumor micro-structures, and can be used to follow the immunologic responses ofhuman tumor tissues in response to ex vivo PD-1 blockade (50). To translate our preclinical findings into a clinically relevant context, we tested the effcacy ofPLT012 in modulating the abundance and functions ofintratumoral CD8^{+T} and Treg cells in ex vivo cultures obtained from HCC tumor samples. We obtained HCC samples from 11 patients at different stages of disease progression (Supplementary Table S3), generated ex vivo cultures, and treated them with control IgG or PLT012 for 2 days. We used flow cytometry to identify intratumoral CD8^{+} T cells and Tregs in the samples (Supplementary Fig.S7).We first assessed CD36 expression in intratumoral Tregs and CD8^{+} T cells before treatment; then,we characterized the immune profile following treatment with either control IgG or PLT012.Despite the high variability in the percentage of CD36-positive intratumoral Tregs and CD8^{+} T cells (range, 10%{-}80% ;Fig.7G), nearly all cultures positively responded to PLT012 treatment with an increase in the proportion of CD8^{+} T cells, a reduction in the proportion of Tregs, an expansion of the CD8^{+} {GrB^{+}} T population, and an increase of the CD8/Treg ratio (Fig. 7H). Importantly, assessments of Treg-related pharmacodynamic markers (i.e., Treg percentages and CD8/Treg ratios) indicated an 81.8% response rate following PLT012 treatment. Furthermore, PLT012-induced increase in CD8^{+} T-cell percentage and GrB expression occurred in 45.4% of the patients (Fig. 7I). Of note, no correlation between the expression of CD36 in Tregs and CD8^{+} T cells and the level of response to PLT012 treatment was found, suggesting a feed-forward effect of PLT012 in promoting an immunosupportive microenvironment.These findings point out that PLT012 effectively modulates the TME in HCCs by targeting CD36. The efficacy of PLT012 in modulating the immune profiles of cells ex vivo from samples of patients with HCC and its good safety profile in nonhuman primates make it a promising therapeutic agent to improve outcomes of patients with HCC.

DISCUSSION

Here,we demonstrate that modulationofCD36-mediated lipid signaling with the PLT012 antibody leads to better therapeutic outcomes in liver cancer and patients with colon cancer with metastasis in the liver.PLT012 has the potential to enhance the tumoricidalefficacy ofimmune cells and to modifythe TME in a way that promotes a shift toward a more immunosupportive milieu characterized by an increased M1/ M2 macrophage ratio and a higher CD8^{+} T-cell to Treg ratio. Moreover, PLT012 can provide superior anti-HCC immune responses in mice under a high-fat diet and elicit anti-liver metastasis responses in murine models. Also, PLT012 can be used to restore full responsiveness to PD-1 blocking agents in primary colon cancer in mice with liver metastasis. Importantly, PLT012 targets the FA-binding region of CD36 with high affinity and, thanks to its isotype, exhibits limited antibody-dependent cellular cytotoxicity and complementdependent cytotoxicity. The rationale design renders PLT012 a superior molecule for tumor suppression with a negligible impact on normal immune homeostasis and vascular balance. Notably, our findings can facilitate the translation of fundamental research into clinical oncology as evidenced by the results obtained in human HCC tumor ex vivo cultures and cynomolgus monkeys for product safety and pharmacokinetic studies. In summary, our work introduces a novel perspective in tumor immunology, providing valuable insights for the development of metabolism-targeting therapies that could significantly influence the management of liver cancer and other metastatic conditions affecting the liver.

Treatment with checkpoint inhibitors has resulted in improved survival rates in patients with liver cancer when compared with sorafenib (51). The combination of atezolizumab and bevacizumab has revealed an objective response rate of approximately 30% with amedian overall survival of19.2 months. Additionally, another SoC therapy combining tremelimumab and durvalumab demonstrated an objective response rate of 20% with a median overall survival of16.43months,which is comparatively lower (43).Nonetheless, the mortality rate in patients with HCC still remains too high. Given the unmet medical needs and the growing market demand for treatments that can address a wider range of patients, it is crucial to evaluate new therapies against atezolizumab plus bevacizumab. Additionally, further research defining HCC subtypes based on immune profles helps predict optimal combinations and innovative agents that promise better outcomes. Our findings reveal that the use of PLT012 with either anti-PD-L1 or in SoC therapy has remarkable therapeutic effects and leads to 30% to 40% of HCC-bearing mice becoming tumor-free, which demonstrates the importance of exploring CD36 targeting as an option for the treatment of HCC and metastatic HCC. Of note, in addition to its pro-angiogenic effects, VEGF-A enhances theinfiltration and activation ofMDSCs and induces M2 macrophage polarization (52-54),ultimately contributing to the development of an immunosuppressive TME. Furthermore,CD36-mediated FA uptakehasbeen reported to activate oxidative metabolism in both MDSCs and M2 macrophages (55, 56). Given the distinct mechanisms, combining bevacizumab with PLT012 shows promise for overcoming VEGFA-dependent immunosuppression and disrupting the metabolic advantages,which could enhance T-cell infiltration and activation within the TME. Additionally, it has been reported that CD36 promotes ferroptosis and increases exhaustion severity in TILs, suggesting that CD36 antagonizes the effect of PD-1/PD-L1 blockade which promotes T-cell reinvigoration and enhances antitumor immunity of TILs (57). Moreover, genetic ablation of CD36 leads to increased production of IFN{*}γ and TNF in intratumoral CD8^{+} T cells and a higher frequency of Prog Tex cells (10). However, our results show that PLT012 does not prevent the formation of Term Tex cells, but PLT012 treatment promotes effector functions in those Term Tex cells. Of note, PLT012 treatment promotes the abundance of Prog Tex cells. These findings suggest that the expression of CD36 may contribute to the functional and survival decline of both Term Tex and Prog Tex cells. Given that T cells engage various metabolic pathways to finetune their differentiation and functional status (58-61), our results suggest that CD36-mediated lipid uptake may modulate effector function in Term Tex cells and stemness in Prog Tex cells.However,it remains unclear whether the inhibition of CD36 differentially regulates T-cell differentiation programs such as memory formation and immune rejuvenation by influencing lipid compositions. Studying the influence of CD36-driven changes in lipid composition on transcriptional and functional changes may provide critical information for this yet unexplored area.

In addition tofacilitating FA uptake,CD36 hasbeen shown to interact with protein-related ligands such as thrombospondin-1 via the CLESH (CD36, LIMP-2, Emp sequence homologous) domain, thereby regulating angiogenesis and platelet aggregation (62, 63). As PLT012 is designed to specifically target the lipid-binding pocket of CD36, we did not observe changes in platelet number and aggregation in any of our preclinical models. This emphasizes the importance of targeting the lipid-binding pocket of CD36 to minimize side effects.

Interestingly, it is well documented that tumor cells with higher metastatic potential, including oral squamous cell carcinomas,melanoma, and breast cancer,have increased CD36 expression(64-66).In gastric cancer cells,CD36-mediated FA uptake leads to the phosphorylation of AKT that inhibits the degradation of glycogen synthase kinase 3\upbeta/\upbeta catenin,thereby facilitating metastasis formation (67). In addition, CD36- mediated lipid uptake can initiate epithelial-mesenchymal transition by modulating Wnt and TGF- \boldsymbol{β} signaling pathways (68). In the context of oral squamous cell carcinoma, short hairpin RNA-mediated knockdown of CD36 or the application of an antibody targeting the lipid-binding pocket of CD36 has been shown to suppress metastasis formation without causing significant adverse effects in preclinical models (65). These findings point out that lipid metabolism regulationby CD36isimportantforimmune surveillance regulation within the TME and metastatic colonization promotion at distant sites.For this reason,it is worth investigating whether PLT012-mediated antimetastatic effects are restricted to liver metastasis or can be exploited to restrict metastasis formation in other organs.

Furthermore,as the inhibition of CD36-mediated FA uptakemay altermetastaticbehaviorsofcancer cells and increase the susceptibility of drug-resistant cells to tyrosine kinase inhibitors (69, 70),CD36 can be considered a viable targetforthe prevention and treatment of cancer metastasis.

MATERIALSANDMETHODS

Assays for Affinity Binding of Antibody, oxLDL Binding. and oxLDL Uptake

For the evaluation of PLT012 binding affinity, human CD36- expressing F293 cell lines were incubated with PLT012 at concentrations ranging from 0.002 to 10~\upmug/mL at 4°C for 30 minutes, followed by staining with commercial PE-conjugated anti-CD36 antibodies (0.4~\upmug/mL) ,clones D-2712 (BD Biosciences, cat.#552544, RRID: AB_2072646), at 4°C for 20 minutes.For the assessment of oxLDL binding in TILs, CD45^{+} TILs were isolated from MC38 tumors 2 weeks after inoculation using the mouse Tumor DissociationKit and CD45 microbeads (Miltenyi Biotec) in accordance with the manufacturer's protocols.To evaluate the inhibitory effect of PLT012 on oxLDL binding and uptake, human CD36-expressing F293 cells or indicated TILs were initially pre-incubated with either control IgG or anti-CD36 antibodies (5~mg/mL) at 4°C for 30 minutes. Subsequently, the cells were treated with oxLDL-Dil ⟨{5\mg/mL} Kalen Biomedical, #770232-9) in RPMI medium containing 1% FA-free BSA at 4°C for 2 hours to measure oxLDL binding or at 37°C for 5 to 15 minutes to measure oxLDL uptake,respectively. Data collection was done using an Attune NxT Flow Cytometer (Thermo Fisher Scientific). The uptake of oxLDL was analyzed by flow cytometry and expressed as a percentage (%) of positive cells based on a histogram gate placed on negative control cells.

Cryo-EMStructuralAnalysisoftheCD36(ECD)-PLT012 (Fab) Complex

The CD36 ECD complexed with the Fab domain ofa specific protein was reconstituted and then loaded into a Chameleon system (SPT Labtech) by spraying a nanoliter onto a self-wicking 1.2/1.3 grid (SPT Labtech). The single-particle cryo-EM data were collected using an automated process on a Titan Krios G4 TEM operating at 300kV (Thermo Fisher Scientific). Raw Electron Event Representation (EER) movies were imported to cryoSPARCv4.3 (71), drift-corrected and dose-weighted with Patch Motion Correction, and estimated for contrast transfer function (CTF) with Patch CTF Estimation.After initial processing and selection, a total of 21,008 images were used for further analysis.Particles were picked and further processed to obtain a three-dimensional (3D) reconstruction. The resulting 3D reconstruction was used to generate templates for particle picking,and a total of 16,529,240 particles were extracted.Following three rounds of two-dimensional classification and refinement, the final 3D class with 84,937particles reached a resolution of 3.75~\AA~ To account for flexibility in the Fab region, a flexible 3D refinement was performed, resulting in a 3.45 A reconstruction (B-factor -127).

Tocompensatefor artifacts andimprove theinterpretability ofthe model building, the quality of 3Dreconstruction was enhanced with EMReady (72). The initial model was obtained from AlphaFold (73), fitting into the cryo-EM density using a rigid body fit in Chimera (74). The atomic model of the Fab region was manually inspected and adjusted in Coot (75), refined with Servalcat (76), and the quality was assessed with MolProbity(77).The atomic modelofthe CD36 region was fit into the cryo-EM density using rigid body fit in Chimera and manually inspected and real-space refined with Phenix (78). Raw cryo-EM images were deposited to the Electron Microscopy Public Image Archive (EMPIAR) with accession code EMPIAR-X. The cryo-EM map of CD36-Fab was deposited to the Electron Microscopy Data Bank with accession code EMD-X.The built atomic model of the Fab region was deposited to the Protein Data Bank with accession code PDB-X.

Cell Lines

The YUMM1.7 melanoma cell line (RRID: CVCL_JK16) and MC38 celline (RRID: CVCL_B288) were kindly provided by Marcus Bosenberg (79) and Pedro Romero, respectively.YUMM1.7 and MC38 cells were cultured in DMEM with 10% FBS and 1% penicillin-streptomycin and used for experiments when they were in the exponential growth phase. Cellines were regularly authenticated by short tandem repeat profiling. All the cell lines were verified tobe free from Mycoplasma contamination.

InVivoExperiment

In subcutaneous engraftment tumor models,Yumm1.7-gp33 (8x10^{5} cells) and MC38-OVA cells (5x10^{5} cells) were administeredvia subcutaneous injection into 6-week-old C57BL/6 mice. Tumor dimensions were assessed every 2 to 3 days after engraftment, with tumor volume calculated using theformula volume =(lengthx width2)/2. MC38 cells were injected subcutaneously (1x10^{6}) and intrahepatically (5x10^{5}) for experimental liver metastasis models as previously described (47).

HCC mouse models were generated by hydrodynamic delivery of endotoxin-free plasmid DNA through the lateral tail vein within 5 to 7 seconds as previously described (80). The pT3-Myc_LucOVA plasmid was created by replacing the epitope OVA323-339/SIYRYYGL (OS) sequence in the pT3-EF1A-MYC-IRES-LucOS plasmid (RRID: Addgene_129776) with the OVA323-339 sequence. In the M Y C^{OE}/T r p53^{KO} HCC model, a total of 12\upmug of the pT3-Myc_LucOVA plasmid, 13.2\upmug of the Trp53 gRNA plasmid (RRID: Addgene_59910), and 4~\upmug of SB100x (RRID:Addgene_34879) were injected into the mice.For the CTNNB1N90/MYCOE HCC model, each mouse received 12~\upmu\up g of pT3- b-catenin (RRID: Addgene_31785), 12~{\upmug} of the pT3-Myc_LucOVA plasmid, and {8\upmug} of S_{B100x.} Tumor growth was monitored weekly through BLI starting from week 2. Mice exhibiting persistent tumor growth were randomly assigned to specific treatment groups to ensure objectivity. PLT012, anti-mouse VEGF-A (2G11-2A05, Bio X Cell), and anti-mouse PD-L1 (10F.9G2, Bio X cell) were administered intraperitoneally at a dosage of 10mg/kg Tumor samples were collected 35 to 38 days after hydrodynamic injection for subsequent weight measurement and immune profiling via flow cytometry. The bioluminescence activity of the mice was evaluated using an IVIS imaging system following luciferin administration (150mg/Kg) All procedures were approved by the University of Lausanne's and the National Defense Medical Center's Institutional Animal Care and Use protocols.

Flow Cytometric Analysis

The entire liver with tumor lesions was finely minced and digested in RPMI medium with 2% FBS, 1% penicillin-streptomycin, DNase I (1\upmugmL^{-1} Sigma-Aldrich),and collagenase (0.5~mgmL^{-1} Sigma-Aldrich) at 37°C for 40 minutes. The resulting mixture was filtered through a {70}{*}{\upmu{m}} cell strainer, and the filtered cells were treated with RBC Lysing Buffer (BioLegend), followed by washing with FACS buffer (PBS with heat-inactivated 2% FBS and 0.1% sodium azide). Tumor-infiltrating leukocytes were then enriched usingPercoll density gradient centrifugation 800\ g for 20 minutes) at room temperature. Single-cell suspensions were incubated with anti-CD16/32 Fc receptor-blocker (BioLegend, cat. \#101320 RRID: AB_1574975) on ice for 10 minutes before staining. Viable cells were identified by stainingwith BDHorizon FixableViabilityStain 450 (BD Biosciences, cat. \#562247 RRID:AB_2869405) at room temperature for 15 minutes. The cells were processed for surface marker staining for 1 hour at 4°C and then for intracellular staining. Intracellular staining was performed according to the instructions of Transcription Factor Fixation/PermeabilizationBufferSet(BioLegend).Briefly, the cells were fixed and permeabilized in TrueNuclear Fix solutionfor 20 minutes and then stained byindicated antibody for 1 hour. Samples were analyzed on Cytek Northern Lights flow cytometers, and data were analyzed with FlowJo v10 (RRID: SCR_008520). The following antibodies were used for flow cytometry: anti-CD45 (BioLegend, cat. \#103136 RRID: AB_2562612), anti-CD19 (BioLegend, cat.#115530,RRID:AB_830707), anti-CD38 (BD Biosciences, cat. #564378, RRID: AB_2738779), anti-CD4 (BD Biosciences,cat.#566939,RRID:AB_2869957),anti-FOXP3(BD Biosciences, cat. #560401, RRID: AB_1645201), anti-CD25 (BioLegend, cat. #102048, RRID: AB_2564124), anti-CD8a (BD Biosciences, cat. #747134, RRID: AB_2871881), anti-CD44 (BD Biosciences, cat. #566506, RRID: AB_2744396), anti-granzyme B (BD Biosciences, cat. #563388, RRID: AB_2738174), anti-TIM-3 (BD Biosciences, cat.#747622,RRID: AB_2744188), anti-PD-1 (BioLegend, cat. #109110, RRID: AB_572017), anti-CD11b (BioLegend, cat. #101217, RRID:AB_389305),anti-CD11c (N418),anti-CD206 (BioLegend, cat.#141732,RRID:AB_2565932),anti-CD103 (BioLegend,cat. #156915, RRID: AB_2904296), anti-MHC class II I-Ab/I-E (BioLegend, cat. #107639, RRID: AB_2565894), anti-F4/80 (BD Biosciences, cat.#746070,RRID: AB_2743450),anti-Ly6C (BD Biosciences, cat. #566987,RRID: AB_2869991), anti-Ly6G (BD Biosciences, cat. #740554,RRID:AB_2740255),anti-Tim-4 (BioLegend,cat.#130010, RRID: AB_2565719), and anti-NK1.1 (BioLegend, cat.#108724, RRID: AB_830871). The OVA tetramer staining was performed following the provided instructions (MBL International, cat.#TS-5001-2C, RRID: AB_3090188).

IHC Analysis

Paraffin sections at a thickness of 4~{\upmum} were deparaffinized and stained with antibodies against CD8α (Abcam, cat. #ab217344, RRID: AB_2890649), FOXP3 (CellSignaling Technology, cat.#12653, RRID: AB_2797979), cleavage caspase-3 (Cell Signaling Technology, cat. #9661, RRID: AB_2341188),and granzyme B (Cell Signaling Technology, cat. #44153, RRID: AB_2857976),followed by imaging with the PolyTek Polymerized Imaging System (ScyTek Laboratories). Chromogenic staining was determined in the Zeiss Axioscan 7 slide scanner.Six random regions were selected to quantify positive levels for each tumor tissue section, and the value for each image was quantified using ImageJ (RRID: SCR_003070).

InSilicoAnalysisonTCGAandPOG570Cohorts

Data from TCGA cohort were acquired from UCSC Xena (http:/ xena.ucsc.edu/).The gene expression quantification was conducted using the UCSC Xena Toil RNA-seq pipeline.Furthermore, information from the POG570 cohort was obtained from cBioPortal. The Reads Per Kilobase per Million mapped reads (RPKM) gene expression values were transformed to Transcripts per million (TPM) using the method suggested by Wagner and colleagues (81).

TheFatty Acid Metabolism signature was obtained from the Molecular Signatures Database (HALLMARK_FATTY_ACID_ METABOLISM), and the T Cell Exhaustion signature was sourced from Bao and colleagues (82). The enrichment levels of these signatures in each sample were quantified using the“GSVA"R package (version 1.48.3) and the single-sample gene set enrichment analysis (GSEA) method was used.

CIBERSORT, a deconvolution algorithm, was used to estimate cell type proportions within bulk cancer samples based on gene expression data.We estimated the proportions of 22 types of infiltrating immune cells using the CIBERSORT method (83,84).For TCGA cohort,theimmune infiltration proportions were obtained from TIMER2.0 (http://timer.cistrome.org/).For the POG570 cohort, the “immunedeconv"R package(version 2.1.0) was used with CIBERSORT (absolute mode), and the CIBERSORT source code was downloaded under a licensed agreement. We extracted the proportions of Treg cells and CD8 T cells with the default parameters.

Cancer types with a mean of FA enrichment score greater than zero and a CD36 expression level higher than the median of the mean CD36 expression among 33 cancer types,minus one SD, were selected.A higher gene set enrichment score indicates ahigher activity level of the associated biological process or pathway.

scRNA-seq Analysis

The cell multiplexing (CellPlex reagents, 10x Genomics) and single-cellibrary preparation were conducted using the Single Cell 3^{\prime} v3.1 kit ( 10x Genomics) according to the manufacturer's protocol. The libraries were sequenced on the Illumina NovaSeq 60o0 platform with the NovaSeq 6000 S4 Reagent Kit (300 cycles). The scRNA-seq data were processed, and the matrix data containing gene counts for each cell per sample were generated using Cell Ranger (V3.0.2, https://support.10xgenomics.com/).

CellBender (85) was used to minimize ambient RNA content. Subsequent downstream analyses and data normalization were performed using the Seurat v5 pipeline (86).For quality control, cells were filtered based on specific criteria: cells with fewer than 400 or more than 10,000 detected genes, fewer than 1,000 or more than 40,000 counts, a ribosomal gene proportion below 0.04 or above 0.50, a mitochondrial gene proportion above 0.2, log1o genes per Unique Molecular Identifier (UMI) ratio below 0.65, or a hemoglobin-related gene proportion exceeding 0.03 were excluded.After this step,9,422 cells from the condition PLT012 and 12,322 cells from the control (Ctrl) condition were retained for further analysis.

Cell types were annotated using scGate (34), and doublets were removed using DoubletFinder (87), assuming a doublet rate of 10% The doublet-free dataset was re-annotated with scGate, and cell type classifications were further refined through unsupervised clustering and marker definition. This process allowed removal of additional doublets missed by DoubletFinder and better defined clusters of unconventional T cells, such as γδT and iNKT cells. Subsequent analyses were performed on 2,227T cells from the conditionPLT012 and 4,385 T cells from the Ctrl condition.

Conventional T cells( CD4^{+} and \scriptstyle\mathbf{CD8^{+}} T cells) were further subclassified using ProjecTILs (88), with a murine reference atlas for tumor-infiltrating conventional T cells, including CD8^{+} exhausted T cells (Tex), CD8^{+} progenitor exhausted T cells (Tpex), and Treg (https://doi.org/10.6084/m9.figshare.12478571.v3). Differential gene expression was performed for each T-cell subtype between PLT012 and Ctrl using the“FindMarkers”function from Seurat, with default parameters (Wilcoxon rank-sum test). Differential expression analysis was performed only when at least 10 cells were present in at least one condition.

GSEA focused on T-cell activation,effectorfunctions, and cellular metabolism pathways.These pathways were derived by filtering the mouse Molecular Signatures Database (89).The custom combined list of pathways was used for GSEA using ClusterProfler (90)using 2,000 permutations. The enrichment analysis was based on genes ranked by -log_{10} of the adjusted P value plus the absolute value of \log_{2} fold change and its sign, obtained from the differential expression analysis between conditions.

Cynomolgus Monkey Toxicity Study

Cynomolgus monkeys(Macacafascicularis)received five onceweekly intravenous doses of PLT012 at concentrations of 0, 10, 60,or 200~mg/kg on days 1, 8,15,22, and 29.Each group was comprised of two males and two females, with one animal of each sex exhibiting low expression and one exhibiting high expression of CD36 in RBC. The animals were monitored for clinical signs, including injection site reactions,body weight, and ophthalmoscopy.Clinical pathology (hematology, serum chemistry, and coagulation) was evaluated before the study commenced and on days 2, 8, 16, and 30 (before the scheduled necropsy). Immunotoxicology (immunophenotyping) was conducted on blood samples collected before the study began and 4, 24, and 168 hours after the first and fourth administrations. Upon completion of the 30-day study, all cynomolgus monkeys underwent necropsy with comprehensive macroscopic postmortem examination and organ weight recording.A complete list of tissues was obtained for histopathologic examination.

The experiments involving cynomolgus monkeys were carried out at WuXi AppTec. The protocol, along with any amendments or procedures with regard to the care or use of animals in this study, was reviewed and approved by the WuXi AppTecInstitutional Animal Care and Use Committee before the study commenced.

Human Sample Ex Vivo CulturePlatform

Human HCC samples were collected from Centre Hospitalier Universitaire Vaudois (Switzerland), Clinic Favoriten (Vienna), Chang Gung Memorial Hospital (Taiwan), and New Taipei City Tucheng Hospital (Taiwan) in accordance with the approval of the institutional review board at each institution. The detailed culture procedures were described previously (50, 91). In short, cryopreserved vials containing minced tumor fragments were thawed,embedded in Matrigel-containing matrix (Matrix High Concentration, phenol red-free, 4mg/mL final concentration; BD Biosciences), and cultured for 2 days with either 10~\upmug/mL of control human anti *\ B Gal-hIgG4 (S228P;InvivoGen） or PLT012. Tumor-associated cell suspensions were then collected from pooled tumor fragments. These suspensions underwent FACS analysis of the immune profile using Aurora spectral flow cytometry (Cytek)with the following antibodies: anti-CD3 APC-F750 (BioLegend, cat.#981006, RRID: AB_2894549),anti-CD4 NovaFB585 (Thermo Fisher Scientific, cat. #H001T03B04-A,RRID: AB_3097881),anti-CD8 PerCP (BioLegend, cat.#980916,RRID: AB_2890877),anti-CD14 PerCP-eF710 (Thermo Fisher Scientific, cat.#46-0149-42,RRID:AB_10671405),anti-CD19 PerCP-eF710 (Thermo Fisher Scientific, cat.#46-0199-42, RRID: AB_2866432), anti-CD36 BUV805 (BD Biosciences,cat.#748645,RRID:AB_2873052), anti-CD45RA BV785 (BioLegend, cat. #304140, RRID: AB_2563816), anti-CCR7 PE-Cy7 (BioLegend, cat.#353226, RRID:AB_11126145), anti-FOXP3 PE-Cy5 (Thermo Fisher Scientific,cat.#15-4777-42,RRID: AB_2811750),anti-GzmB BV510(BD Biosciences,cat.#563388,RRID: AB_2738174), and ZombieNIR (BioLegend, cat. #423106).

Quantification and Statistical Analysis

Statistical analysesfor multiplegroups wereperformedusingoneway ANOVA, and two-tailed, unpaired Student t tests were conducted for two-group comparisons.The data are presented as means ± SEM, and each point represents a biological replicate. The box and whisker plot displayed the distribution of immune cells, with the whisker indicating minimum to maximum values.

Data Availability

Data generated in this study are publicly available in the Gene Expression Omnibus (RRID: SCR_005012) at GSE291964.Raw cryo-EM images were deposited to the EMPIAR with accession code EMPIAR-12599. The cryo-EM map of CD36-Fab was deposited to the Electron Microscopy Data Bank with accession code EMD-52203. No customized code has been used to produce analytic results for this study. Correspondence and requests for materials should be addressed to Chin-Hsien Tsai and Ping-Chih Ho.

Authors'Disclosures

Y-R. Yu reports personal fees, non-financial support, and other support from Pilatus Biosciences during the conduct of the study; personal fees, non-financial support, and other support from Pilatus Biosciences outside the submitted work; and a patent for 385670-P pending. H.-W. Hsiao reports other support from Pilatus Biosciences during the conduct of the study, as well as a patent for PCT/ US2023/063766 pending and licensed to Pilatus Biosciences. L.-T. Chiu reports personal fees, non-financial support, and other support from Pilatus Biosciences during the conduct of the study, as well as personal fees, non-financial support, and other support from Pilatus Biosciences outside the submitted work. P.-H. Chung reports other support from Pilatus Biosciences during the conduct of the study,as well as a patent for PCT/US2023/063766 pending and licensed to Pilatus Biosciences. H.-K. Chen reports other support from Pilatus Biosciences during the conduct of the study, as well as a patent for PCT/US2023/063766 pending and licensed to Pilatus Biosciences. Y.-H. Lin reports personal fees, non-financial support, and other support from Pilatus Biosciences during the conduct of the study; personal fees, non-financial support, and other support from Pilatus Biosciences outside the submitted work;and a patent for 385670-P pending. P.-C. Ho reports other support from Pilatus Biosciences during the conduct of the study; other support from Elixiron Immunotherapeutics outside the submitted work; and a patent for CD36 targeting pending. No disclosures were reported by the other authors.

Authors'Contributions

S.-F.Tzeng:Data curation, methodology,writing-original draft. Y.-R. Yu: Conceptualization, data curation, methodology, writingoriginal draft. J. Park: Data curation. J. von Renesse: Data curation, formal analysis.H.-W. Hsiao: Data curation,methodology.C.-H. Hsu: Data curation. J. Garnica:Data curation, formal analysis, methodology. J. Chen: Data curation. L.-T. Chiu: Data curation, formal analysis, methodology. J. Santol: Resources, methodology, project administration.T.-Y. Chen: Resources, project administration. P.-H. Chung: Data curation. L.E. Kandalaft: Resources. P. Starlinger: Resources,project administration.R.C.-E.Hsieh:Resources, project administration.M.-C.Yu:Data curation,project administration. P.-W. Hsiao: Resources. S.J. Carmona: Data curation, methodology. H.-K. Chen: Data curation, supervision. Z. Meng: Conceptualization,methodology.Y.-H.Lin: Conceptualization, data curation. J. Zhou: Resources, data curation. C.-H. Tsai: Conceptualization, resources, data curation,formal analysis, supervision,funding acqui sition. P-C. Ho: Conceptualization, resources, supervision, funding acquisition, writing-original draft, writing-review and editing.

Acknowledgments

This work was supported in part by research funding of the National Science and Technology Council (NSTC 112-2628-B-016-002 and NSTC 113-2628-B-016-005), Ministry of National DefenseMedical Affairs Bureau (MND-MAB-D-112095), and Pilatus Biosciences SA research grant to C.-H. Tsai. P.-C. Ho was supported in part by the Cancer Research Institute (Lloyd J.Old STAR award),the Swiss Science National Foundation (310030L_208130, IZLCZ0_206083,and CRSII5_205930)and Swiss Science National Foundation Consolidator Grant (TMCG-3_213736),the Cancer Research Institute (Lloyd J. Old STAR award), Helmut Horten Stifung, and Swiss Cancer Foundation.J. Garnica was supported by SNF fellowship (205930). R.C.-E. Hsieh was supported by National Science and Technology Council (NSTC111-2314-B-182A-160- MY2 and NSTC112-2628-B-182A-007-MY3) and the Ministry of Education in Taiwan (MOE-113-YSFMN-1009-001-P1). M.-C. Yu was supported by the National Science and Technology Council (NSTC113-2314-B-182A-068). This study was in part funded by the

NIH (R01DK122813) received by P. Starlinger and by the SFU MED Research Promotion Fund (FFF 12/22 and FFF 12/23) received by J. Santol and P. Starlinger. J. Zhou was supported by General Research Fund (14104820)from Hong Kong SAR We thank Laboratory Animal Core Facility (Agricultural Biotechnology Research Center, ABRC) for their services, ABRC Instrument Core Facilities for providing the Cytek Northern Lights and PerkinElmer IVIS Lumina XRMS,and Instrument Center of National Defense Medical Center for their assistance with the Axioscan 7 Slide Scanner.We thank ProteinProduction and Structure Core Facility(PTPSP and EPFL) and DCI-Laussane (EPFL)for structural analyses.We also thank Yu-Chih Yang for help with experiments.

Note

Supplementary data for this article are available at Cancer Discovery Online (http://cancerdiscovery.aacrjournals.org/).

Received October 2,2024; revised March 7,2025;accepted April 11, 2025; posted first April 28, 2025.

REFERENCE:

1. Chuang YM, Tzeng SF, Ho PC, Tsai CH. Immunosurveillance encounters cancer metabolism. EMBO Rep 2024;25:471-88.
2. Zhang S, Lv K, Liu Z, Zhao R, Li F. Fatty acid metabolism of immune cells: a new target of tumour immunotherapy. Cell Death Discov 2024;10:39.
3.Hanahan D. Hallmarks of cancer: new dimensions. Cancer Discov 2022;12:31-46.
4. Li X, Wenes M, Romero P, Huang SC, Fendt SM, Ho PC.Navigating metabolic pathways to enhance antitumour immunity and immunotherapy. Nat Rev Clin Oncol 2019;16:425-41.
5. Daquinag AC, Gao Z, Fussell C, Immaraj L, Pasqualini \operatorname{R}, Arap W, et al.Fatty acid mobilization from adipose tissue is mediated by CD36 posttranslational modifications and intracellular trafficking. JCI Insight 2021;6:e147057.
6.Zhu GQ, Tang Z, Huang R, Qu WF, Fang Y, Yang R, et al. CD36+ cancer-associated fibroblasts provide immunosuppressive microenvironment for hepatocellular carcinoma via secretion of macrophage migration inhibitory factor. Cell Discov 2023;9:25.
7. Lee-Rueckert M, Canyelles M, Tondo M, Rotllan N, Kovanen PT, Llorente-Cortes V, et al. Obesity-induced changes in cancer cells and their microenvironment:mechanisms and therapeutic perspectives to manage dysregulated lipid metabolism. Semin Cancer Biol 2023; 93:36-51.
8.Yu YR, Imrichova H, Wang H, Chao T, Xiao Z, Gao M, et al. Disturbed mitochondrial dynamics in CD8^{+} TILs reinforce T cell exhaustion. Nat Immunol 2020;21:1540-51.
9. Ho PC, Bihuniak JD, Macintyre AN, Staron M, Liu X, Amezquita R, et al.Phosphoenolpyruvate is a metabolic checkpoint of anti-tumor T cell responses. Cell 2015;162:1217-28.
10. Xu S, Chaudhary O, Rodriguez-Morales P, Sun X, Chen D, Zappasodi R,et al.Uptake of oxidized lipids by the scavenger receptor CD36 promotes lipid peroxidation and dysfunction in CD8^{+} T cells in tumors. Immunity 2021;54:1561-77.e7.
11. Ma X, Xiao L, Liu L, Ye L, Su P, Bi E, et al. CD36-mediated ferroptosis dampens intratumoral CD8+ T cell effector function and impairs their antitumor ability. Cell Metab 2021;33:1001-12.e5.
12. Wang H, Franco F, Tsui YC, Xie X, Trefny MP, Zappasodi R, et al. CD36-mediated metabolic adaptation supports regulatoryT cellsurvival and function in tumors. Nat Immunol 2020;21:298-308.
13. Koundouros N, Poulogiannis G. Reprogramming of fatty acid metabolism in cancer. Br J Cancer 2020;122:4-22.
14. Zhang Y, Meng \scriptstyleQ, Sun \scriptstyleQ, Xu ZX, Zhou H, Wang Y. LKB1 deficiencyinduced metabolic reprogramming in tumorigenesis and non-neoplastic diseases. Mol Metab 2021;44:101131. metabolic crosstalk between tumor cells and macrophages affects liver metastasis. Nat Commun 2022;13:5782.
16. Niavarani SR, Lawson C, Bakos O, Boudaud M, Batenchuk C, Rouleau S, et al. Lipid accumulation impairs natural killer cell cytotoxicity and tumor control in the postoperative period. BMC Cancer 2019;19:823.
17. Michelet X, Dyck L, Hogan A, Loftus RM, Duquette D, Wei K, et al. Metabolic reprogramming of natural killer cells in obesity limits antitumor responses. Nat Immunol 2018;19:1330-40.
18. Kobayashi T, Lam PY, Jiang H, Bednarska K, Gloury R, Murigneux V, et al. Increased lipid metabolism impairs NK cell function and mediates adaptation to the lymphoma environment. Blood 2020; 136:3004-17.
19. Hirano K, Kuwasako T, Nakagawa-Toyama Y, Janabi M, Yamashita S, Matsuzawa Y. Pathophysiology of human genetic CD36 deficiency. Trends Cardiovasc Med 2003;13:136-41.
20. Febbraio M, Abumrad NA, Hajar DP, Sharma K, Cheng W, Pearce SF, et al. A null mutation in murine CD36 reveals an important role in fatty acid and lipoprotein metabolism.J Biol Chem 1999;274:19055-62.
21. Son NH, Basu D, Samovski D, Pietka TA, Peche VS, Willecke F, et al. Endothelial cell CD36 optimizes tissue fatty acid uptake.JClin Invest 2018;128:4329-42.
22. Tanaka T, Nakata T, Oka T, Ogawa T, Okamoto F, Kusaka Y, et al. Defect in human myocardial long-chain fatty acid uptake is caused by FAT/CD36 mutations. J Lipid Res 2001;42:751-9.
23. van Schravendijk MR, Handunnetti SM, Barnwell JW, Howard RJ. Normal human erythrocytes express CD36, an adhesion molecule of monocytes, platelets, and endothelial cels. Blood 1992;80:2105-14.
24. Jay AG imard JR,Huang N, Hamilton JA.SO and other putative inhibitors of FA transport across membranes by CD36 disrupt intracellular metabolism, but do not affect FA translocation.J Lipid Res 2020;61:790-807.
25. Xu X, Chen D, Ye X, Xia W, Shao Y, Deng J, et al. Improvement of anti-CD36 antibody detection via monoclonal antibody immobilization of platelet antigens assay by using selected monoclonal antibodies. Ann Lab Med 2023;43:86-91.
26. Rispens T,Huijbers MG.The unique properties of IgG4 and its roles in health and disease. Nat Rev Immunol 2023;23:763-78.
27. Nergiz-Unal R, Rademakers T, Cosemans JM, Heemskerk JW. CD36 as a multiple-ligand signaling receptor in atherothrombosis.Cardiovasc Hematol Agents Med Chem 2011;9:42-55.
28. LiT,FuJ,Zeng Z, Cohen D,LiJ,Chen Q,et al. TIMER2.0 for analysis of tumor-infiltrating immune cells. Nucleic Acids Res 2020;48:W509-14.
29. Peche VS, Pietka TA, Jacome-Sosa M, Samovski D, Palacios H, Chatterjee-Basu G, et al. Endothelial cell CD36 regulates membrane ceramide formation, exosome fatty acid transfer and circulating fatty acid levels. Nat Commun 2023;14:4029.
30. Ruiz de Galarreta M, Bresnahan E, Molina-Sanchez P, Lindblad KE, Maier B, Sia D, et al. Beta-catenin activation promotes immune escape and resistance to anti-PD-1 therapy in hepatocellular carcinoma. Cancer Discov 2019;9:1124-41.
31. Kurtulus S, Madi A, Escobar G, Klapholz M, Nyman J, Christian E, et al.Checkpoint blockade immunotherapy induces dynamic changes in PD-1-CD8+ tumor-infiltrating T cells. Immunity 2019;50:181-94.e6.
32. Siddiqui I, Schaeuble K, Chennupati V, Fuertes Marraco SA, Calderon-Copete S, Pais Ferreira D, et al. Intratumoral Tef1+PD-1+CD8+ Tcells with stem-like properties promote tumor control in response to vacination and checkpoint blockade immunotherapy.Immunity 2019;50:195-211.e10.
33. Miller BC, Sen DR, Al Abosy R, Bi K, Virkud YV, LaFleur MW, et al. Subsets of exhausted CD8+T cells differentially mediate tumor control and respond to checkpoint blockade.Nat Immunol 2019;20:326-36.
34.Andreatta M, Berenstein AJ, Carmona SJ. scGate: marker-based purifcation of cell types from heterogeneous single-cell RNA-seq datasets. Bioinformatics 2022;38:2642-4.
35. Finn RS, Qin S,Ikeda M, Galle PR, Ducreux M, Kim TY, et al. Atezolizumab plus bevacizumab in unresectable hepatocellular carcinoma. N Fnal I Med 2020-382-1894_905 hepatocellular carcinoma and chronic liver disease mortality in the UK Biobank. Nat Commun 2024;15:3707.
37. Moussa I, Day RS, Li R, Kaseb A, Jalal PK, Daniel-MacDougall C, et al. Association of dietary fat intake and hepatocelular carcinoma among US adults. Cancer Med 2021;10:7308-19.
38. Barupal DK, Ramos ML, Florio AA, Wheeler WA, Weinstein SJ, Albanes D,et al. Identification of pre-diagnostic lipid sets associated with liver cancer risk using untargeted lipidomics and chemical set analysis: a nested case-control study within the ATBC cohort. Int J Cancer 2024;154:454-64.
39. Li B, Li Y, Zhou H, Xu Y, Cao Y, Cheng C, et al. Multiomics identifies metabolic subtypes based on fatty acid degradation allocating personalized treatment in hepatocellular carcinoma. Hepatology 2024;79:289-306.
40. Dudek M, Pfister D, Donakonda S, Filpe P, Schneider A, Laschinger M, et al.Auto-aggressive CXCR6+ CD8 T cells cause liver immune pathology in NASH. Nature 2021;592:444-9.
41. Yahoo N, Dudek M, Knolle P, Heikenwalder M. Role of immune responses in the development of NAFLD-associated liver cancer and prospects for therapeutic modulation.JHepatol 2023;79:538-51.
42. Piening A, Ebert E, Gottlieb C,Khojandi N, Kuehm LM, Hoft SG, et al. Obesity-related T cell dysfunction impairs immunosurveillance and increases cancer risk.Nat Commun 2024;15:2835.
43.Llovet JM, Castet F, HeikenwalderM, Maini MK,Mazzaferro V, Pinat DJ, et al. Immunotherapies for hepatocellular carcinoma.Nat Rev Clin Oncol 2022;19:151-72.
44. Crispe IN. Hepatic T cells and liver tolerance. Nat Rev Immunol 2003;3:51-62.
45. Horst AK, Neumann K, Diehl L, Tiegs G. Modulation of liver tolerance by conventional and nonconventional antigen-presenting cells and regulatory immune cells. Cell Mol Immunol 2016;13:277-92.
46. Wang H, Tsung A, Mishra L, Huang H. Regulatory T cell: a doubleedged sword from metabolic-dysfunction-associated steatohepatitis to hepatocellular carcinoma. EBioMedicine 2024;101:105031.
47. YuJ, Green MD,LiS, Sun Y, Journey SN, Choi JE,et al.Liver metastasis restrains immunotherapy efficacy via macrophage-mediated T cell elimination. Nat Med 2021;27:152-64.
48. Pleasance E, Titmuss E, Williamson L, Kwan H, Culibrk L, Zhao EY,et al. Pan-cancer analysis of advanced patient tumors reveals interactions between therapy and genomic landscapes. Nat Cancer 2020;1:452-68.
49. Meng Y, Pospiech M, Ali A, Chandwani R, Vergel M, Onyemaechi S, et al. Deletion of CD36 exhibits limited impact on normal hematopoiesis and the leukemia microenvironment. Cell Mol Biol Lett 2023;28:45.
50. Voabil P, de Bruijn M, Roelofsen LM, Hendriks SH, Brokamp S, van den Braber M, et al. An ex vivo tumor fragment platform to dissect response to PD-1 blockade in cancer. Nat Med 2021;27:1250-61.
51. Rimassa L, Finn RS, Sangro B. Combination immunotherapy for hepatocellular carcinoma.J Hepatol 2023;79:506-15.
52. Huang Y, Chen X, Dikov MM, Novitskiy SV, Mosse CA, Yang L, et al. Distinct roles of VEGFR-1 and VEGFR-2 in the aberrant hematopoiesis associated with elevated levels of VEGF. Blood 2007;110:624-31.
53.Linde N,Lederle W,DepnerS,van Rooijen N,Gutschalk CM,Mueller MM.Vascular endothelial growth factor-induced skin carcinogenesis depends on recruitment and alternative activation of macrophages. JPathol 2012;227:17-28.
54.Movahedi K, Laoui D, Gysemans C, Baeten M, Stange G, Van den Bossche J, et al. Different tumor microenvironments contain functionally distinct subsets of macrophages derived from Ly6C(high) monocytes.Cancer Res 2010;70:5728-3.
55. Al-Khami AA, Zheng L, Del Valle L, Hossain F, Wyczechowska D, Zabaleta J,et al. Exogenous lipid uptake induces metabolic and functional reprogramming of tumor-associated myeloid-derived suppressor cells. Oncoimmunology 2017;6:e1344804.
56. Rios FJ, Koga MM, Pecenin M, Ferracini M, Gidlund M, Jancar S. Oxidized LDL induces alternative macrophage phenotype through activation of CD36 and PAFR. Mediators Inflamm 2013;2013:198193.
57. Thommen DS, Koelzer VH, Herzig P, Roller A, Trefny M, Dimeloe S, et al. A transcrinrionallv and fincrionallv disrincr PD *1^{+} CD8+ T cell pool with predictive potential in non-small-cell lung cancer treated with PD-1 blockade. Nat Med 2018;24:994-1004.
58. Buck MD, O'Sullivan D, Klein Geltink RI, Curtis JD, Chang CH, Sanin DE, et al. Mitochondrial dynamics controls T celfate through metabolic programming.Cell 2016;166:63-76.
59. Franco F, Bevilacqua A, Wu RM, Kao KC, Lin CP, Rousseau L, et al. Regulatory circuits of mitophagy restrict distinct modes of cell death during memory CD8+Tcell formation.Sci Immunol 2023;8:eadf7579.
60. Bevilacqua A, Franco F, Lu YT, Rahiman N, Kao KC, Chuang YM, et al. PPARB/8-orchestrated metabolic reprogramming supports the formation and maintenance of memory CD8+ T cells. Sci Immunol 2024;9:eadn2717.
61.Franco F,Jaccard A,Romero P, Yu YR,Ho PC.Metabolic and epigenetic regulation of T-cell exhaustion. Nat Metab 2020;2:1001-12.
62. Nergiz-Unal R, Lamers MM, Van Kruchten R, Luiken J, Cosemans JM, Glatz JF, et al. Signaling role of CD36 in platelet activation and thrombus formation on immobilized thrombospondin or oxidized low-density lipoprotein.J Thromb Haemost 2011;9:1835-46.
63.Silverstein RL,Febbraio M.CD36, a scavenger receptor involved in immunity, metabolism, angiogenesis, and behavior.Sci Signal 2009;2:re3.
64. Terry AR, Nogueira V, Rho H, Ramakrishnan G, Li J, Kang S, et al. CD36 maintains lipid homeostasis via selective uptake of monounsaturated fatty acids during matrix detachment and tumor progression. Cell Metab 2023;35:2060-76.e9.
65. Pascual G, Avgustinova A, Mejetta S, Martin M, Castellanos A, Attolini CS,et al.Targeting metastasis-initiating cells through the fatty acid receptor CD36.Nature 2017;541:41-5.
66. Pascual G, Dominguez D, Elosua-Bayes M, BeckedorffF, Laudanna C, Bigas C,et al. Dietary palmitic acid promotes a prometastatic memory via Schwann cells. Nature 2021;599:485-90.
67. Pan J, Fan Z, Wang Z, Dai \scriptstyleQ, Xiang Z, Yuan F, et al. CD36 mediates palmitate acid-induced metastasis of gastric cancer via AKT/GSK3\upbeta/\upbeta -catenin pathway.JExp Clin Cancer Res 2019;38:52.
68.Nath A, Li I, Roberts LR, Chan C. Elevated free fatty acid uptake via CD36 promotes epithelial-mesenchymal transition in hepatocellular carcinoma.Sci Rep 2015;5:14752.
69. Landberg N, von Palffy S, Askmyr M, Lilljebjorn H, Sanden C, Rissler M, et al. CD36 defines primitive chronic myeloid leukemia cells less responsive to imatinib but vulnerable to antibody-based therapeutic targeting.Haematologica 2018;103:447-55.
70. Feng WW, Wilkins O, Bang S, Ung M, LiJ, An J, et al. CD36-Mediated metabolic rewiring of breast cancer cells promotes resistance to HER2-targeted therapies. Cell Rep 2019;29:3405-20.e5.
71. Punjani A, Rubinstein JL, Fleet DJ, Brubaker MA. cryoSPARC: algorithms for rapid unsupervised cryo-EM structure determination. Nat Methods 2017;14:290-6.
72. He J, Li T, Huang SY. Improvement of cryo-EM maps by simultaneous local and non-local deep learning. Nat Commun 2023;14:3217.
73. Jumper J, Evans R, Pritzel A, Green T, Figurnov M, Ronneberger O, et al. Highly accurate protein structure prediction with AlphaFold. Nature 2021;596:583-9.
74. Pettersen EF, Goddard TD, Huang CC, Couch GS, Greenblatt DM, Meng EC,et al.UCSFChimera-avisualization system forexploratory research and analysis.J Comput Chem 2004;25:1605-12.
75. Emsley P, Lohkamp B, Scott WG, Cowtan K. Features and development of Coot. Acta Crystallogr D Biol Crystallogr 2010;66: 486-501.
76. Yamashita K, Palmer CM, Burnley T, Murshudov GN. Cryo-EM single-particle structure refinement and map calculation using Servalcat.Acta Crystallogr D Struct Biol 2021;77:1282-91.
77. Williams CJ, Headd JJ, Moriarty NW, Prisant MG, Videau LL, Deis LN, et al. MolProbity: more and better reference data for improved all-atom structure validation.Protein Sci 2018;27:293-315.
78.Afonine PV, Poon BK,Read RJ, Sobolev OV,Terwilliger TC, Urzhumtsev A, et al. Real-space refinement in PHENIX for cryo-EM and crystallography. Acta Crystallogr D Struct Biol 2018;74:531-44.
79.Meeth K, Wang JX, Micevic G, Damsky W, Bosenberg MW. The YUMM lines: a series of congenic mouse melanoma cellines with defined genetic alterations.Pigment Cell Melanoma Res 2016;29:590-7.
80. Molina-Sanchez P, Ruiz de Galarreta M, Yao MA, Lindblad KE, Bresnahan E,Bitterman E, et al. Cooperation between distinct cancer driver genes underlies intertumor heterogeneity in hepatocellular carcinoma.Gastroenterology 2020;159:2203-20.e14.
81.Wagner GP, Kin K, Lynch VJ. Measurement of mRNA abundance using RNA-seq data: RPKM measure is inconsistent among samples. Theory Biosci 2012;131:281-5.
82. Bao X, Zhang H, Wu W, Cheng S, Dai X, Zhu X, et al. Analysis of the molecular nature associated with microsatellite status in colon canceridentifies clinical implications for immunotherapy.J Immunother Cancer 2020;8:e001437.
83.Newman AM, Liu CL, Green MR, Gentles AJ, Feng W, Xu Y, et al. Robust enumeration of cell subsets from tissue expression profiles. Nat Methods 2015;12:453-7.
84. Chen B, Khodadoust MS, Liu CL, Newman AM, Alizadeh AA. Profiling tumor infiltrating immune cells with CIBERSORT. Methods Mol Biol 2018;1711:243-59.
85. Fleming SJ, Chaffn MD, Arduini A, Akkad AD, Banks E, Marioni JC, et al. Unsupervised removal of systematic background noise from droplet-based single-cellexperiments using CellBender. Nat Methods 2023;20:1323-35.
86. Hao Y, Stuart T, Kowalski MH, Choudhary S, Hoffman P, Hartman A,et al. Dictionary learning for integrative,multimodal and scalable single-cell analysis.Nat Biotechnol 2024;42:293-304.
87.McGinnis CS, MurrowLM, Gartner ZJ. DoubletFinder: doublet detection in single-cell RNA sequencing data using artificial nearest neighbors. Cell Syst 2019;8:329-37.e4.
88.Andreatta M, Corria-OsorioJ, MillerS, Cubas R, Coukos G, Carmona SJ. Interpretation of T cell states from single-cell transcriptomics data using reference atlases. Nat Commun 2021;12:2965.
89. Castanza AS, Recla JM, Eby D, Thorvaldsdottir H, Bult CJ, Mesirov JP. Extending support formouse data in the molecular signatures database (MSigDB). Nat Methods 2023;20:1619-20.
90. Wu T, Hu E, Xu S, Chen M, Guo P, Dai Z, et al. clusterProfiler 4.0: a universal enrichment tool for interpreting omics data. Innovation (Camb) 2021;2:100141.
91. Roelofsen LM, Voabil P, de Bruijn M, Herzig P, Zippelius A, Schumacher TN, et al. Protocol for ex vivo culture of patient-derived tumor fragments.STAR Protoc 2023;4:102282.