Cholesterol-Lowering Intervention Decreases mTOR Complex 2 Signaling and Enhances Antitumor Immunity

Prostate cancer is the most commonly diagnosed cancer and the second leading cause of cancer death among men in North America (1). Yet, there is no widely accepted strategy for prostate cancer chemoprevention. Over the past 10 years a large body of evidence has emerged from preclinical (2–5) and observational studies that cholesterol-lowering HMG-CoA-reductase inhibitors (i.e., statins) exert a protective effect against prostate cancer (6, 7). Cholesterol-lowering drugs have been proposed to directly alter tumor growth (3, 5, 8) as well as T-cell immunity (9–11). However, previous reports that examined the immune effects of cholesterol lowering (CL) at the subcellular level have reached conflicting conclusions about the potential benefits of CL. Some reports suggest that membrane cholesterol promotes T-cell function. Yang and colleagues reported that ACAT1-deficient CD8⁺ T cells, which have increased membrane cholesterol, showed enhanced effector function and improved control of melanoma growth and metastasis in mice (12). Consistent with this model, Wang and colleagues reported that cholesterol sulfate, a naturally occurring analog of cholesterol, inhibits T cells by disrupting cholesterol needed for activation of T-cell receptors (13). However, a report by Swamy and colleagues suggests that membrane cholesterol inhibits T cells because only T-cell receptors that spontaneously detach from cholesterol could activate and signal (14).

These studies relied on in vitro methods and genetic strategies that manipulated cholesterol levels in specific immune cells. However, cholesterol is an essential structural component of all cell membranes, and lowering serum cholesterol is expected to have varied, and even opposing, antitumor effects that depend on the immune context (15, 16). Therefore, we report the first study using preclinical models where it was possible to selectively lower serum cholesterol and assess the final antitumor immunity resulting from the varied effects of CL in mice. Prior attempts to do this in mouse models have relied on statins. In patients, statins' primary function is to inhibit cholesterol synthesis in the liver. However, statins have only a modest effect on mouse serum cholesterol and may increase cholesterol synthesis in the mouse liver (17, 18). Furthermore, statins have the potential to modulate immunity by altering prenylation, which is an effect unrelated to changes in serum cholesterol (17, 19). Therefore, we used oral ezetimibe in mice to inhibit intestinal cholesterol absorption and reduce serum cholesterol by levels analogous to statin-treated patients (3).

In this work we characterize the adaptive immunity in a preclinical model where serum CL is verified. We show that CL in mice decreases the growth of melanoma, prostate cancer, and kidney cancer cells. To establish the role of the immune system, we show that inhibition of tumor growth is CD8⁺ lymphocyte dependent. Additional evidence of immune-mediated antitumor effects is provided by adoptively transferring lymphocytes from mice subjected to CL. These lymphocytes are able to prevent growth of metastatic tumor deposits in mice that did not receive any other treatment. By monitoring Pmel-1 cells adoptively transferred to B6 mice, we show that CL enhanced central CD8⁺ memory cells, providing a mechanism of action. We show in mice and in a window-of-opportunity clinical trial that lowering serum cholesterol inhibits the AKT pathway in lymphocytes, increases infiltration of CD8⁺ lymphocytes in the tumor, and decreases proliferation of tumor cells. We show that these effects are mediated through the mTORC2 pathway in lymphocytes, which is a novel observation that provides the only currently available strategy to inhibit mTORC2 in patients. CL interventions are widely and safely used to reduce the risk of cardiovascular disease in patients and the potential to modulate immunity by simply lowering serum cholesterol has large public health implications.

The following mAbs for flow cytometry, with or without a fluorescent conjugate, were obtained from BioLegend: αCD4 (GK 1.5 and RM4–5), αCD8 (53–6.7), αT-bet (4B10), αCD62 L (MEL-14), αFoxP3 (FJK-16s), αIFN-γ (XMG1.2), αTNFα (TN3–19.12), Rictor (H-11) αCD4 (GK1.5), and αCD8 (2.43) antibodies for T-cell depletion were purchased from BioXcel. Ezetimibe was purchased from Par Phamaceutical. Antibodies for Western blotting against the following targets were purchased from Cell Signaling Technology: p-p70S6K (T389), p70S6K (49D7), pAKT (S473), pAKT (T308), AKT(C67E7), P-4E-BP1 (T37/46), 4E-BP1 (53H1), CPTA (5D8G9), pSGK (D36D11), SGK (D27C11), pPKCα (Thr638), PKCα (Polyclonal ab), pPKCζ (T410/403), and PKCζ (D10E2). Antibody against β-actin (AC-15) was purchased from Sigma.

Cells were washed with ice-cold PBS and lysed with RIPA buffer (50 mmol/L Tris-HCl pH 7.5, 150 mmol/L NaCl, 1 mmol/L EDTA, 10 mmol/L DL-dithiothreitol, 10% glycerol, 0.5% sodium deoxycholate, 1% Triton-X100) with 10 mmol/L NaF, protease inhibitors (1:100 dilution of protease inhibitor cocktail from Sigma). Protein concentrations were measured with the Bio-Rad Protein Assay. Forty micrograms of lysate was loaded for Western blot analysis. Nitrocellulose membranes were used for the protein transfer and Western blots were performed according to recommendations of the antibody manufacturer.

RM1, RENCA, and B16 cancer cell lines were purchased from ATCC. The cells were maintained in DMEM or RPMI1640 medium supplemented with 10% heat-inactivated FBS (Gemcell), 2 mmol/L of l-glutamine, 100 units/mL of penicillin, 100 μg/mL of streptomycin, and 250 ng/mL of Amphotericin B (Thermo Fisher Scientific). All cells were periodically authenticated by morphologic and histologic inspection and animal grafting for assessing their ability to grow and metastasize. The cells were annually tested for mycoplasma using Myco Alert Kit (Lonza).

C57BL/6J, Balb/C, and Pmel-1 mice, 5 to 8 weeks old were purchased from The Jackson Laboratory. All experiments involving animals were in compliance with federal and state standards and were IACUC approved (IACUC006360). Tumors were generated by subcutaneously injecting 2 × 10⁵ tumor cells into the frank. To lower serum cholesterol, mice were administrated daily ezetimibe (30 mg/Kg) by oral gavage for 2 weeks before injection of tumor cells and continuing throughout the experiment. Saline gavage was used in the control group. To deplete CD8⁺ lymphocytes, mice were injected intraperitoneally with 200 μg of CD8 mAb (clone 2.43). Tumor volume was calculated (shortest diameter² × longest diameter/2). After blood draw per tail vein, serum cholesterol was determined using Amplex Red Cholesterol Assay Kit (Invitrogen). For the lung metastasis model, 2 × 10⁵ B16 or RM1 cells were intravenously injected through the tail vein. To monitor CD8⁺ T-cell memory formation in vivo, Pmel-1 lymphocytes were harvested from Thy1.1⁺ mice and 2 × 10⁶ CD8⁺ lymphocytes were adoptively transferred by tail vein injection into Thy1.2+ mice.

Conditional knockout (KO) mice for mTORC2 were generated by crossing B6.Cg-Tg(Lck-cre)548Jxm/J with Rictor^tm1.1Klg/SjmJ (both from The Jackson Laboratory). Resulting mice had T-lymphocyte-specific Rictor KO, Lck-Cre-Rictor^flox (Rictor-KO). Homozygous mice were obtained by interbreeding heterozygotes. Genotyping was performed by PCR. The primer sequences were: Rictor forward, CAAGCATCATGCAGCTCTTC; Rictor reverse TCCCAGAATTTCCAGGCTTA; Cre forward, TGTGAACTTGGTGCTTGAGG; Cre reverse, CAGGTTCTTGCGAACCTCAT.

To prepare dendritic cell (DC) vaccine, bone marrow was harvested from C57BL/6 mice and suspended in RPMI supplemented with 10% FBS. GM-CSF (10 ng/mL) was added to the medium and placed at 37°C with 5% CO₂ for 7 days. DCs were pulsed with tumor cell lysate for 16 hours and activated with 10 μg/mL CpG for 4 hours. Alternatively, DCs were pulsed with 10 ng/mL mouse gp100 peptide (amino acids 25–33, which is presented by H2-D^b class I molecules; Alpha Diagnostic International) for 1 hour and activated with 10 μg/mL CpG for 2 hours. The resulting DC vaccine was injected subcutaneously. For in vitro activation of Pmel-1 cells, DCs were washed with PBS and cocultured with splenic Pmel-1 cells. IFNγ, IL9, and TNFα were analyzed by flow cytometry (LSR II; BD Biosciences).

To monitor CD8⁺ T-cell memory in vivo, spleens and lymph nodes were harvested from Thy1.1⁺ Pmel-1 mice. The spleen and lymph nodes were transferred to a 70 μm cell strainer on a 50 mL tube. A 1 mL syringe plunger was used to gently press the tissue through the strainer using 10 mL red cells lysis buffer to wash the strainer. Cells were then spun at 1,500 rpm for 5 minutes to collect dissociated lymphocytes. CD8⁺ cells were isolated using the Mouse CD8 Cell Recovery Column Kit from Cedarlane Laboratories Ltd. A total of 1 × 10⁵ CD8⁺-enriched cells were stained with CD8-PE-Cy5 and thy1.1-FITC antibodies to confirm presence of CD8 and Thy1.1 by flow cytometry. A total of 2 × 10⁶ Pmel-1 lymphocytes were adoptively transferred by tail vein injection into Thy1.2+ C57BL/6 mice.

Mouse tumors were cut into small fragments ∼2 to 3 mm in length and put in an enzyme digest mix consisting of 10 μg/mL DNAse, 1 mg/mL collagenase, and 25 μg/mL hyaluronidase (all from Sigma), and incubated for 3 hours at 37°C under continuous rotation. TILs were isolated by Ficoll-Paque density gradient centrifugation. CD4 and CD8 cells were analyzed by flow cytometry (LSR II).

Patient prostate tissues or mouse tumors were fixed in 10% formalin in PBS and embedded in paraffin. Four μm sections were baked at 60°C for 1 hour and subjected to immunofluorescence staining using the Ventana Discovery Ultra autostainer with buffers and antibodies from Ventana Medical Systems. Briefly slides were deparaffinized using EZ solution, followed by antigen retrieval with Cell Conditioning 1 (CC1) solution (prediluted Tris solution, pH 8.0) for 64 minutes at 95°C and blocking with Discovery inhibitor (Ventana) for 12 minutes at RT. The slides were incubated with αCD8 (Ventana) or αKi-67 (Ventana) at 37°C for 28 minutes followed by detection with anti-rabbit-HQ/αHQ-HRP/DAB (Ventana) or Red610 Kits (Ventana). DNA dye DAPI (Invitrogen) was used to stain the nucleus. Slides was scanned using TissueFAXS 200 system (TissueGnostics). Image was analyzed using QuPath with default settings. For all summarized data, three random high-power fields were evaluated by automated image analysis and averaged prior to generating a mean for each experimental group.

A prospective, window-of-opportunity trial was conducted (ClinicalTrials.gov identifier: NCT02534376) and provided tissue and serum. The study was conducted in accordance with recognized ethical guidelines (e.g., Declaration of Helsinki, CIOMS, Belmont Report, U.S. Common Rule). In the trial, patients with clinically significant prostate cancer scheduled to undergo prostatectomy provided written informed consent and were enrolled in an IRB-approved study (Pro40945) and received 2 to 6 weeks of maximal CL intervention with Vytorin, which contains ezetimibe (10 mg) and simvastatin (40 mg). For a preplanned exploratory analysis, Ki67 staining in pretreatment biopsies and posttreatment prostatectomies were compared. The Alliance for Clinical Trials in Oncology funded the trial, and their publication committee reviewed and approved this manuscript prior to submission for publication.

Whole blood and prostate biopsy tissue were collected prior to starting Vytorin and posttreatment blood and prostate tissue were collected at the time of prostatectomy. Prostate specimens were processed according to the International Society of Urological Pathology Consensus Guidelines on prostatectomy handling (20). H&E sections were examined by an expert genitourinary pathologist. These annotations were used for IHC image analysis and for macrodissection of tissue for RNA extraction.

As described previously (21), tumor sections were placed in 2.0 mL RNase-free Eppendorf tubes. Sections were treated twice with 1.9 mL xylene for 5 minutes at 55°C while rocking. The sections were washed twice with 100% ethanol. RNA was extracted from the paraffin samples using the MasterPure RNA Purification Kit (Epicenter Biotechnologies). RNA was then treated with 20 units DNase I (Ambion) for 30 minutes and checked for residual genomic DNA by TaqMan RT-PCR targeting ACTB. The final RNA concentration (A260:0.025) and purity (A260:A280 ratio) were measured using a NanoDrop ND-2000 spectrophotometer (NanoDrop Technologies).

For RNA-seq, 300 ng total RNA were sent to MedGenome Inc., where RNA-seq was performed on the HiSeq-2500 platform (Illumina) after using the TruSeq RNA Access Library Prep Kit (Illumina). Reads mapping to ribosomal and mitochondrial genome were removed before performing alignment with STAR (2.4.1; ref. 22). Overall 96% of the total pre-processed reads mapped to the reference gene model/genome. The raw read counts were estimated using RSEM (23). Read count data were normalized by trimmed mean of M-values (TMM) method using DESeq2. Tumor microenvironment analysis was performed using ESTIMATE software (24) for all the protein coding genes at multiple levels, that is, tumor content, innate immune cells, and adaptive immune cells. CD8⁺ T-cell signature was obtained from CIBERSORT (25) and cholesterol biosynthesis signature was from the molecular signature database (MSigDB; ref. 26). Enrichment scores for the signatures were computed using weighted z-score method (27) with intrinsic functionality of the PCTA. Mean values and SEM are provided throughout the manuscript and P values were generated using t tests. Mouse tumor growth curves were compared using the repeated measures ANOVA.

The RNA-seq data are available at GEO (accession # GSE159529).

We have previously reported that cholesterol-lowering interventions reduce the growth of murine prostate cancers by inhibiting mTOR signaling in the tumor (5). We and others have reported that mTOR inhibition can promote the formation of CD8⁺ lymphocyte memory formation (28–30). Therefore, we hypothesized that CL may lead to antitumor immunity mediated by CD8⁺ lymphocytes in mice treated with ezetimibe, which lowers cholesterol by binding to the intestinal cholesterol transporter NPC1L1 and preventing cholesterol absorption (31). In three different syngeneic mouse tumor models (melanoma, prostate cancer, and renal cell carcinoma), ezetimibe decreased tumor growth (Fig. 1A–D; Supplementary Fig. S1A). This was at least in part immune mediated, since depletion of CD8⁺ lymphocytes negated the majority of the antitumor effects of ezetimibe (Fig. 1A–D; Supplementary Figs. S1B and S1C). Some ezetimibe may be absorbed from the gut into the blood stream. Therefore, to rule out the possibility of a direct effect on tumor growth, all tumor cell lines were treated in vitro with ezetimibe or its metabolite, ezetimibe-glucuronide, at various concentrations and treatment durations; there was no effect on proliferation of tumor cells (Supplementary Figs. S2 and S3). We also ruled out the possibility that oral gavage used to administer ezetimibe may produce a stress response that affects tumor growth (Supplementary Fig. S4).

To provide definitive evidence that the antitumor effect associated with CL was mediated by adaptive immunity, we performed a lymphocyte transfer study (Fig. 1E). Mice treated with ezetimibe to lower serum cholesterol were administered a cancer vaccine consisting of DCs pulsed with B16 tumor lysate. Lymphocytes were harvested from these mice 7 days later and adoptively transferred into naive mice (i.e., mice never subjected to CL), which then were challenged intravenously with B16 tumor cells. The tumor cells were administered directly into the tail vein; this model represents a very aggressive cancer scenario where only the most potent treatments are expected to show benefit. Compared with control mice that received lymphocytes from untreated mice, mice that received lymphocytes from vaccine-treated donor mice had fewer lung tumor deposits, and those that received lymphocytes from mice treated with vaccine plus ezetimibe had the fewest lung deposits (Fig. 1F). Consistent with this observation, mice in the latter group with smallest tumor burden had the lowest lung weight. In addition, the CD8⁺ lymphocytes retrieved from these mice tended to have a greater IFNγ response upon restimulation with DC pulsed with B16 tumor lysate (Fig. 1G).

These findings were reproduced using RM1 prostate cancer cells. Mice with lymphocytes transferred from mice treated with vaccine plus ezetimibe had the fewest tumor deposits in the lung (Fig. 1H and I) and the lowest lung weight (Fig. 1J). In another version of this experiment, stimulated lymphocytes were transferred to naive mice before they were challenged with subcutaneous RM1 tumor. Mice that received lymphocytes from donor mice subjected to CL had the slowest tumor growth (Supplementary Fig. S5). These results clearly showed that lowering serum cholesterol stimulated a lymphocyte-mediated antitumor immune response that was independent of any direct effects on the tumor cells. Use of ezetimibe instead of a statin, which can have cholesterol-independent effects, allowed us to attribute all immune effects to CL. In vivo tumor growth was reduced by the sum of all immune effects resulting from serum CL.

We confirmed that ezetimibe decreased serum cholesterol in mice (Fig. 2A). The resulting lymphocytes had decreased mTOR signaling in splenocytes (Fig. 2B), including CD8⁺ lymphocytes (Supplementary Fig. S6). The largest decrease in phosphorylation was seen for AKT, at Ser473, which is associated with mTORC2 signaling. This effect was also observed in patients; pAKT (Ser473) was decreased when comparing CD8⁺ lymphocytes before and after 2 to 6 weeks of CL treatment (Fig. 2C and D). The observation that the mTORC2 pathway may be mediating immune stimulation was surprising because our previous work using first-generation mTOR inhibitors showed that antitumor immunity was associated with inhibition of the mTORC1 pathway (32), and no prior work has linked CL with a decrease in TORC2 signaling.

The experiments in Fig. 1 clearly demonstrated that antitumor immunity resulting from CL was lymphocyte dependant. In a slight variation of the experiment in Fig. 1A–D, CL was initiated after injection of tumor cells to simulate a clinical scenario where cancer is treated with CL after diagnosis. The ezetimibe-treated mice had decreased tumor growth when compared with control mice (Fig. 2E). In ezetimibe-treated mice, the CD8⁺ lymphocytes had decreased mTOR2 signaling as measured by decrease in pAKT (Ser473) and pAKT (Thr308; Fig. 2F). Ezetimibe treatment was associated with an increase in central memory markers. CD8⁺ lymphocytes that were CD44⁺CD62L⁺ increased (Fig. 2G) as well as CD8⁺ lymphocytes expressing Eomes (Fig. 2H). The mTOR pathway and central CD8⁺ memory cells have been linked to fatty acid oxidation, where the rate-limiting step is mediated by CPT1A (33). CL with ezetimibe was associated with increased CPT1A expression in CD8⁺ lymphocytes (Fig. 2I).

Memory CD8⁺ T cells mediate long-term antitumor immunity. Decreased mTOR/mTORC1 signaling can trigger the formation of memory CD8⁺ T cells (30, 34, 35). We considered the possibility that CL, with its inhibition of mTORC2, could enhance the formation of memory CD8⁺ T cells. Thy1.1⁺ Pmel-1 lymphocytes were adoptively transferred into congenic Thy1.2+ B6 mice (Fig. 3A). Pmel-1 lymphocytes recognize gp100 and can be tracked using Thy1.1. Ezetimibe was administered to lower serum cholesterol, and the mice were injected with bone marrow–derived DCs pulsed with gp100 peptide (Supplementary Fig. S7). In the mice treated with ezetimibe, the percent of donor-derived Pmel-1 lymphocytes on day 10 increased significantly, which indicated an increase in gp100-specific cells (Fig. 3B; Supplementary Fig. S8). Central memory cells have a greater recall response and are able to provide greater protection than peripheral memory cells. Both Thy1.1⁺ and Thy1.1⁻ CD8⁺ cells were more likely to have a central memory phenotype (CD62L⁺CD44⁺; Eomes⁺; Fig. 3B; Supplementary Fig. S9). Consistent with this observation, the Thy1.1⁺ cells had higher intensity of CPT1A expression (Fig. 3C).

The most important function of CD8⁺ memory cells is to expand rapidly when rechallenged. When Thy1.1⁺ lymphocytes were rechallenged with DCs pulsed with gp100 peptide on day 20, greatest expansion occurred in lymphocytes previously exposed to ezetimibe where approximately 40% of all CD8⁺ cells were Thy1.1⁺ lymphocytes (Fig. 3D). The CD8⁺ lymphocytes maintained a central memory phenotype on day 23; they expressed CD62 L and CD44. With CL intervention, both Thy1.1⁺ and Thy1.1⁻ CD8 cells had increased PD1 expression, which has been shown to be required for both development and function of CD8⁺ memory T cells (Fig. 3E; ref. 36). Finally, consistent with the increased CD8⁺ lymphocyte expansion upon stimulation after CL, both Thy1.1⁺ and Thy1.1⁻ lymphocytes had an activated phenotype, expressing granzyme B, IFNγ, IL9, or T-bet after CL (Fig. 3F; Supplementary Figs. S10 and S11). These results suggested that lymphocytes stimulated by tumor in the presence of lower serum cholesterol exhibit enhanced CD8⁺ memory lymphocytes with a central memory phenotype.

A genetic approach was used to confirm that TORC2 is the target for CL that leads to enhanced antitumor immunity. mTOR integrates cellular signals from nutrients, growth factors, hormones, stresses, and hypoxia. mTOR can complex either with raptor to form mTORC1 or Rictor to form mTORC2. Conditional KO mice for mTORC2 were generated by crossing B6.Cg-Tg(Lck-cre)548Jxm/J with Rictor^tm1.1Klg/SjmJ. Resulting mice had T lymphocyte–specific Rictor KO, Lck-Cre-Rictor^flox (Rictor-KO; Fig. 4A; Supplementary Fig. S11). As expected, Rictor KO mice had decreased pAKT (Ser473; Fig. 4B), which is a target of TORC2. The flow cytometry findings were confirmed by immunoblot, where it was also noted that total AKT was unchanged by Rictor KO or CL (Supplementary Fig. S12). However, CL decreased Rictor protein expression (Supplementary Figs. S13 and S14), which was unexpected and suggests this as the mechanism for decreased mTORC2 signaling for CL. As expected, pS6K (Thr389), which is a target for mTORC1, was unchanged by KO of Rictor or β-CD (Fig. 4C; Supplementary Fig. S12). However, phosphorylation of other downstream targets of TORC2 such as SGK, PKCα, and PKCζ were decreased in the Rictor-KO and in wild-type lymphocytes treated with β-cyclodextrin (β-CD), which binds and lowers cholesterol in the culture medium (Supplementary Fig. S15). Unlike the control (WT) lymphocytes, Rictor-KO lymphocytes did not respond to β-CD, suggesting that CL decreases TORC2 signaling exclusively through Rictor. As expected, Rictor KO lymphocytes had enhanced activation as indicated by TNFα or IFNγ expression in response to αCD3 and αCD28 (Fig. 4D).

To confirm in vivo that decreased signaling through mTORC2 in lymphocytes results in improved antitumor immunity, B16 or RM1 tumors were grown in Rictor-KO and WT control mice (Rictor^flox). B16 tumor growth was suppressed in Rictor-KO mice compared with WT control mice (Rictor^flox), and tumor suppression was similar to that with CL (Fig. 4E and F). Both Rictor-KO and CL provided similar increase in tumor gp100-specific CD8⁺ lymphocytes (Fig. 4G), which had increased IFNγ responses compared with lymphocytes from WT mice (Fig. 4H). In Rictor-KO mice, CL provided no additional suppression of tumor growth, increase in gp100-specific CD8⁺ lymphocytes, or increase in gp100-specific IFNγ response (Fig. 4E–H). These observations indicate that CL enhances lymphocyte function solely through suppression of Rictor and mTORC2.

To confirm these observations in another model, growth of RM1 tumors were compared in Rictor-KO and control mice. Tumor growth was significantly slower in Rictor-KO mice (Fig. 4I; Supplementary Fig. S16). The CD8⁺ lymphocytes from Rictor-KO mice had decreased mTOR2 signaling as indicated by decrease in pAKT (Ser473) and pAKT (Thr308; Fig. 4J). The lymphocytes from Rictor KO mice had increased expression of Granzyme B and IFNγ (Supplementary Fig. S17). The Rictor-KO mice had increased in markers of central memory, CD8⁺ cells with increased CD62⁺CD44⁺ cells, CD127⁺KLRG11⁻ cells, Eomes, and PD1 (Fig. 4K; Supplementary Fig. S18). However, LAG3, which is a marker of T-cell exhaustion, was not changed by Rictor KO (Fig. 4L). Fatty acid oxidation (FAO) has been linked to increased central T-cell memory formation. The rate-limiting step of FAO is mediation by CPT1A. CPT1A expression increased with CL but not with Rictor KO (Fig. 4M). These suggest that CL decreases FAO, which in-turn decreases mTOR2 signaling.

The mTOR/TORC1 pathway has been shown to regulate CD8⁺ T-cell memory formation (30). Rapamycin and rapamycin analogs, which inhibit TORC1, have been proposed as immunotherapy to enhance CD8⁺ T-cell memory formation (29). However, serum CL had the greatest effect on signaling through the TORC2 pathway.

Another consequence of enhanced antitumor immunity is increased infiltration of lymphocytes into the tumor. Our animal studies described in Fig. 2E revealed that CL increased CD8⁺ lymphocytes in prostate tumor (Fig. 5A). In an exploratory human study, RNA-seq was performed to quantify immune cells present in prostate cancer (Fig. 5B). Two patients were examined, and both showed increased CD8⁺ lymphocytes and M1 macrophages, but no change in other immune populations (Supplementary Fig. S19), following CL intervention with the combination of ezetimibe and simvastatin. This observation was confirmed in 6 additional patients scheduled to undergo radical prostatectomy who were treated with 2 to 6 weeks of CL. Matched pretreatment prostate biopsy and posttreatment prostatectomy tissue showed that CL therapy increased CD8⁺ lymphocytes in the tumor (Fig. 5C).

To provide additional evidence of a relationship between cholesterol and CD8⁺ lymphocytes, we analyzed gene expression from the Prostate Cancer Transcriptome Atlas, which is a virtually organized database of 1,321 clinical specimens from 38 prostate cancer cohorts (Fig. 5D; ref. 37). There was a strong, significant inverse correlation between genes specific for CD8⁺ T cells and genes involved in cholesterol biosynthesis and metabolism. There was a significant positive correlation between genes specific for CD8⁺ T cells and genes involved in cholesterol efflux, which would be involved in removing intracellular cholesterol, but not with genes involved in cholesterol import. Taken together with our animal study and our window-of-opportunity trial, these results suggest that low cholesterol increases CD8⁺ lymphocyte infiltration in the tumor.

If CL produces antitumor immunity, the end result is expected to be a decrease in tumor proliferation. In the murine experiment described in Fig. 1H, lymphocytes from mice treated with a DC vaccine and CL were transferred to naive mice challenged with intravenous tumor cells. The resulting decrease in cellular proliferation measured using Ki-67 staining of lung metastases was due to the transferred lymphocytes since the tumor-bearing mice were never exposed to the DC vaccine or CL intervention (Fig. 5E).

To see if CL decreases Ki-67 in human prostate cancers, we used archival prostate tissue from a previously conducted, prospective clinical trial where men received 2 to 6 weeks of an aggressive cholesterol-lowering intervention prior to radical prostatectomy (Fig. 6A). The study targeted two sources for serum cholesterol in humans: simvastatin inhibited cholesterol synthesis in the liver and ezetimibe decreased cholesterol absorption in the intestine. Patient characteristics are summarized in Fig. 6B. There was a large, significant decrease in serum cholesterol and modest decrease in triglyceride (nonsignificant) and HDL on the trial (Fig. 6C). When comparing pre- and posttreatment prostate tissue, apoptosis assessed using Caspase 3 staining did not change (data not shown). However, proliferation decreased in normal prostate tissue and low-grade prostate cancers (Fig. 6D and E). There was no significant correlation between Ki67 staining and serum cholesterol, however the study was not powered to evaluate this correlation. There was no significant change in Ki-67 staining in high-grade prostate cancers. These results suggest that currently available therapies for CL may decrease growth in both benign prostate that produces voiding symptoms in older men and low-grade prostate cancer, which is managed with active surveillance because of the risk of progression to aggressive cancer.

The mTOR/TORC1 pathway has been shown to regulate CD8⁺ T-cell memory formation (30). In fact, rapamycin and rapamycin analogs, which inhibit TORC1, have been proposed as immunotherapies that can enhance CD8⁺ T-cell memory formation (29). Therefore we assessed the mTOR pathway in lymphocytes from mice treated with ezetimibe. We were surprised to find that lowering serum cholesterol had the greatest effect on signaling through the TORC2 pathway. Therefore, T-lymphocyte-specific TORC2 knockouts were generated and shown to have enhanced lymphocyte function and antitumor capacity. It is not clear if FAO is necessary for CD8⁺ T-cell memory formation (38, 39), however, the rate-limiting step of FAO mediated by CPT1A has been associated with CD8⁺ T-cell memory formation (33). We found that increased CPT1A was associated with CL but not with Rictor KO, suggesting CL first increases FAO, which then inhibits mTOR. Both CL and TORC2 KO increased central memory CD8⁺ lymphocytes. This is in line with a report by Pollizzi and colleagues, who examined the role of mTORC1 and mTORC2 on CD8⁺ T-cell differentiation (40). They found that mTORC1 is critical to CD8⁺ T-cell activation; however, mTORC2 regulates CD8⁺ T-cell memory. There are currently no mTORC2-specific inhibitors in clinical testing and we now describe the only clinical intervention to specifically inhibit mTORC2, and the intervention is safe and already clinically available.

We confirm that our observations extend to patients. We used tissue from a prospective window-of-opportunity clinical trial where patients scheduled to undergo prostatectomy for prostate cancer were treated with 2 to 6 weeks of both ezetimibe and simvastatin to achieve maximal CL. In mice we reduced serum cholesterol by approximately 10% to see the immune response. In patients, we achieved greater CL, which was approximately 25% of baseline. We showed that posttreatment lymphocytes had a decrease in mTORC2 signaling. In our mouse model, CL enhanced CD8⁺ T-cell functions and increased CD8⁺ T cells in the tumor. Using RNA-seq and immunofluorescence in clinical prostate cancers, we showed that posttreatment tumors exhibited an increase in CD8⁺ T cells when compared with matched pretreatment tumors. We used a large, publicly available prostate cancer transcriptome dataset with over 1,000 patient prostate cancers to show an inverse relationship between CD8⁺ lymphocyte infiltration and genes involved in cholesterol biosynthesis and metabolism. Also consistent with the observation that lower cholesterol in the tumor promotes CD8 antitumor immunity, there was a positive correlation between CD8⁺ T-cell infiltration and cholesterol efflux genes. Finally, in both mice and humans, CL decreased tumor proliferation in low-grade cancer. We had too few high-grade cancers to draw any conclusions. However, decrease in cellular proliferation in normal tissue and low-grade cancers suggest CL may be effective for preventing cancer progression in men with low-risk disease on active surveillance.

The combination of our mechanistic animal studies and correlative studies from our window-of-opportunity clinical trial suggest that clinical interventions to lower serum cholesterol may stimulate antitumor immunity by enhancing CD8⁺ memory formation. Localized prostate cancer is a slow growing cancer where only a small minority of patients will eventually die from their disease, even when treatment is not provided. Therefore, any therapy targeting this population should have minimal morbidity. In the case of statins, not only is the therapy safe but there are other potential cardiovascular health benefits, further strengthening the case for CL intervention. Our results provide justification for formal clinical testing of CL intervention for primary or secondary chemoprevention and as a strategy to enhance the effects of immunotherapies for advanced solid tumors.

In clinically relevant murine models for assessing antitumor immunity, we showed that CL inhibits tumor proliferation by enhancing CD8⁺ lymphocytes, which have enhanced memory function and infiltrate tumor tissue. Serum CL inhibited mTOR2 signaling in mouse lymphocytes. We confirmed that our observations extend to patients by examining lymphocytes and prostate tissue from a prospective window-of-opportunity clinical trial where patients scheduled to undergo prostatectomy for prostate cancer were treated with 2 to 6 weeks of ezetimibe and simvastatin to achieve maximal CL. Future clinical trials will define the role of CL in patient care.

No disclosures were reported.

Y. Wang: Conceptualization, data curation, formal analysis, supervision, investigation, visualization, methodology, writing–original draft, writing–review and editing. S. You: Software, formal analysis, supervision, writing–review and editing. S. Su: Formal analysis, investigation, methodology, writing–review and editing. A. Yeon: Formal analysis, investigation, visualization, methodology, writing–review and editing. E.M. Lo: Formal analysis, visualization, methodology. S. Kim: Data curation, formal analysis, methodology. J.L. Mohler: Conceptualization, supervision, funding acquisition, project administration, writing–review and editing. M.R. Freeman: Conceptualization, supervision, writing–review and editing. H.L. Kim: Conceptualization, resources, supervision, funding acquisition, investigation, visualization, writing–original draft, project administration, writing–review and editing.

This work was supported by grants from the Samuel Oschin Comprehensive Cancer Center, Winnick Family Foundation, Homer and Gloria Harvey Family Endowed Chair, Teshinsky Family Foundation, and the NIH (R01CA182438). The clinical trial reported in this publication was supported by the NCI of the NIH under the Award Number UG1CA189823 to the Alliance for Clinical Trials in Oncology NCORP Research Base (Jan C. Buckner, M.D., contact PI). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.