RORγ Bridges Cancer-Driven Lipid Dysmetabolism and Myeloid Immunosuppression

Tumor-related IL6/IL1β-dependent induction of hepatic PCSK9 downregulates hepatic LDL receptor (LDLR) expression, activating RORγ-dependent expansion of suppressive protumor myeloid populations and favoring tumor progression.

Abstract

Despite well-documented metabolic and hematopoietic alterations during tumor development, the mechanisms underlying this crucial immunometabolic intersection remain elusive. Of particular interest is the connection between lipid metabolism and the retinoic acid–related orphan receptor (RORC1/RORγ), whose transcriptional activity modulates cancer-related emergency myelopoiesis and is boosted by cholesterol metabolites, whereas hypercholesterolemia itself is associated with dysregulated myelopoiesis. In this study, we show that cancer and hypercholesterolemic diet independently or cooperatively activate RORγ-dependent expansion of myeloid-derived suppressor cells (MDSC) and M2-polarized tumor-associated macrophages (TAM), supporting cancer spread. Moreover, we report that tumor-induced expression of IL1β and IL6 promotes hepatic expression of proprotein convertase subtilisin/kexin type 9 in preclinical models and patients. Importantly, lowering cholesterol levels, by genetic or pharmacologic inhibition of proprotein convertase subtilisin/kexin type 9, prevents MDSC expansion, M2 TAM accumulation, and tumor progression in a RORγ-dependent manner, unleashing specific antitumor immunity. Overall, we identify RORγ as a key sensor of lipid disorders, bridging hypercholesterolemia and protumor myelopoiesis.

Significance:

Cancer and a hypercholesterolemic diet independently or collaboratively increase blood cholesterol, which in turn triggers RORγ-dependent expansion of suppressive monocytic MDSCs and M2-like TAMs, thus inhibiting specific antitumor immunity and facilitating disease progression.

Introduction

Cancers generate complex immunologic stressors that can alter the myelopoietic output, leading to the formation of heterogeneous myeloid cell populations [i.e., tumor-associated macrophages (TAM) and myeloid-derived suppressor cells (MDSC)] endowed with immunosuppressive and tumor-promoting activities (1–4). The concerted action of selected transcription factors, such as c/EBPβ, retinoic acid–related orphan receptor (RORC1/RORγ), STAT3, IRF8, and p50 NF-κB (5–7), translates tumor-derived signals into altered myelopoiesis, causing changes in the differentiation, metabolism, and activation state of myeloid cells (8, 9). Adipose tissue signaling has also been shown to influence the composition of myeloid cells in various tissues (10–12) and contribute to cancer-related immune dysfunction (13). Whereas the correlation between cholesterol levels and cancer risk or progression remains controversial (14), compelling evidence links dysregulated lipid and cholesterol pathways to cancer invasion and metastasis through transcriptional and epigenetic changes that reprogram tumor metabolism, favoring inflammation in the tumor microenvironment (TME) while evading immune destruction (13, 15–22). Indeed, recent studies have brought attention to the association between cholesterol dysmetabolism and tumor progression (23). Furthermore, hepatic proprotein convertase subtilisin/kexin type 9 (PCSK9), which interferes with cholesterol absorption and clearance by promoting endo-lysosomal degradation of hepatic low-density lipoprotein receptor (LDLR; 24), has been found to reduce the recycling of MHC-I on the surface of tumor cells regardless of its cholesterol-regulating functions, potentially affecting immune checkpoint therapy outcomes (25). A further correlation between hypercholesterolemia and myeloid cell expansion, mobilization, and inflammatory activities comes from cardiovascular studies (26, 27), suggesting that the identification of molecular sensors of lipid dysregulation may also be relevant for diseases associated with chronic metabolic inflammation, such as atherosclerosis. Given the critical role of RORC1/RORγ in controlling emergency myelopoiesis in patients with cancer (5) and its activation by cholesterol and its metabolites (28), in this study we sought to determine the potential involvement of this cholesterol sensor in linking lipid dysmetabolism to protumoral myelopoiesis.

Results

Cancers Alter Cholesterol Metabolism

We initially conducted a comprehensive evaluation of blood cholesterol levels in mice bearing various tumor types [i.e., MN/MCA1 fibrosarcoma, B16-F10 melanoma, MC38 colon adenocarcinoma, and Lewis lung carcinoma (LLC)] at both early disease (ED) and advanced disease (AD) stages (see “Methods”; ref. 5). Our findings revealed a consistent increase in total and LDL cholesterol levels, particularly in AD mice (Fig. 1A–D). We also observed alterations in total, LDL, and high-density lipoprotein (HDL) cholesterol levels in patients with stages I-II (early) and stage III-IV (advanced) cancer (Fig. 1E), including non–small cell lung cancer (NSCLC, n = 146), colorectal cancer (n = 65), breast cancers (n = 36), pancreatic ductal adenocarcinoma (n = 58), biliary tract cancer (n = 35), and pancreatic neuroendocrine tumor (n = 36). We then evaluated the expression of cholesterol regulatory enzymes in the liver and small intestine, as these organs are the main regulators of cholesterol uptake, synthesis, and excretion (29). mRNA expression of Hmgcr, which encodes a rate-limiting enzyme in cholesterol biosynthesis, and Ldlr, which encodes the receptor of LDL (24), showed no consistent changes in livers from MN/MCA1, B16-F10, LLC, and MC38 mice (Supplementary Fig. S1A; Supplementary Table S1). In contrast, the expression levels of genes encoding key regulators of blood cholesterol clearance (i.e., Vldlr, Apoe, and Apob; ref. 24) were downmodulated in tumor-bearing (TB) mice. Likewise, the expression of several key enzymes mediating the early steps of oxysterol and bile acid formation (i.e., Cyp7a1, Cyp27a1, and Cyp7b1; ref. 30) and their efflux and biliary excretion (i.e., Abcg5, Abcg8, and Abcb11; Supplementary Fig. S1A; Supplementary Table S1; refs. 24, 31) was decreased. In the small intestine, we observed decreased expression of cholesterol efflux transporter genes Abcg5/Abcg8, whereas the expression of Npc1l1, involved in the absorption of cholesterol, remained unchanged (Supplementary Fig. S1B). Consistently, MN/MCA1 mice showed reduced total bile acid (TBA) levels in their feces (Supplementary Fig. S1C). In agreement with the increased LDL cholesterol levels, hepatic PCSK9 mRNA (Supplementary Fig. S1A) and protein levels (Supplementary Fig. S1D) were upregulated in livers from AD TB mice, which correlated with downregulation of hepatic LDLR protein (Supplementary Fig. S1D). Modulation of hepatic cholesterol metabolism gene expression was also confirmed in healthy liver parenchyma from patients with colorectal cancer (n = 23) compared with distant healthy liver parenchyma of benign adenomas or angiomas (control, n = 6) from non-dyslipidemic patients (Fig. 1F). Moreover, treatment of primary hepatocytes with MN/MCA1 tumor-conditioned medium (TCM) increased Pcsk9 gene expression, whereas it reduced both Vldlr and Abcg8 gene expression (Supplementary Fig. S1E; Supplementary Table S1).

Figure 1.

Tumor progression alters cholesterol metabolism. A and B, Total HDL and LDL blood cholesterol levels in mice bearing MN/MCA1 fibrosarcomas (n = 6; A), B16-F10 melanomas (n = 11), or LLC lung (n = 5) or MC38 colon (n = 5) cancers (B), at time points corresponding to AD stage. C and D, Total cholesterol, HDL, and LDL blood levels in MN/MCA1 (n = 4; C), B16-F10, LLC, and MC38 mice (D) at time points corresponding to ED and AD stages compared with healthy, age- and sex-matched TF mice (n = 5). E, Total cholesterol (top), HDL (middle), and LDL (bottom) blood levels in patients with early (I–II) vs. advanced (III–IV) NSCLC (n = 76 vs. n = 70), CRC (n = 30 vs. n = 35 for total cholesterol; n = 15 vs. n = 16 for HDL and LDL), BRC (n = 18 vs. n = 18), PDAC (n = 26 vs. n = 32), BTC (n = 20 vs. n = 15), or PNET (n = 20 vs. n = 16) compared with healthy donors (HD, n = 35). F, mRNA expression levels of cholesterol metabolism genes in healthy liver parenchyma from patients with CRC (n = 23) compared with healthy liver parenchyma from patients with benign lesions (n = 6). Data are representative of at least four independent (A–C) or one (D) experiments. Data are presented as the mean ± SEM (A–D) or Box-and-whisker min-to-max plots (E and F). Statistic by t test [A, B, and F (PCSK9, CYP27A1, and ABCG8)], Mann–Whitney (F), two-way ANOVA (C and D), and one-way ANOVA or Kruskal–Wallis (E). BRC, breast cancer; BTC, biliary tract cancer; Ch, cholesterol; CRC, colorectal cancer; CTRL, control; PDAC, pancreatic ductal adenocarcinoma; PNET, pancreatic neuroendocrine tumor.

Hypercholesterolemia Exacerbates Cancer-Related Cholesterol Dysmetabolism

We next assessed the effects of diet on plasma cholesterol levels by feeding mice bearing transplantable fibrosarcoma MN/MCA1 (Fig. 2A and B; Supplementary Fig. S1F) or conditional Kras^LSL-G12D/+;Trp53^fl/fl (KP) mice developing NSCLC (Supplementary Fig. S1G and S1H; ref. 32) on either normal chow diet (NCD) or hypercholesterolemic diet (HCD). We observed that cancer development further enhanced HCD-induced hypercholesterolemia. Furthermore, whereas Vldlr, Cyp7a1, Cyp27a1, Abcg5, and Abcg8 gene expression was upregulated in HCD-fed tumor-free (TF) mice, MN/MCA1 tumor development prevented this upregulation, as well as that of the Apoe and Abcb11 genes (Fig. 2C; Supplementary Table S1). Of note, AD marked the most pronounced changes in gene expression, including Pcsk9, whose mRNA expression (Fig. 2C; Supplementary Table S1) and blood protein levels (Fig. 2D) increased regardless of diet type, correlating with reduced levels of LDLR protein expression (Fig. 2E). Fittingly, tumor development in KP mice increased the circulating levels of PCSK9 (Supplementary Fig. S1I). We next evaluated whether the increase in blood cholesterol was also associated with changes in the hepatic levels of mevalonate and squalene, crucial intermediates of the cholesterol biosynthesis chain (23). MN/MCA1 progression resulted in reduced levels of both metabolites (Supplementary Fig. S1J), suggesting that hepatic cholesterol biosynthesis does not contribute to the observed increase in blood cholesterol in cancer bearers. Furthermore, regardless of the type of diet, oral administration of fluorescence-labeled cholesterol [BODIPY-cholesterol (BODIPY-Ch)] clearly showed a decrease in hepatic cholesterol during tumor progression (Fig. 2F), indicating its accumulation in extrahepatic tissues, such as tumor cells (CD45⁻CD31⁻), TAMs, myeloid progenitors [both common myeloid progenitors (CMP) and granulocyte–macrophage progenitors (GMP)], and circulating CD11b⁺Ly6C^hi monocytic cells (Fig. 2G and H). This was further supported by reduced excretion of TBAs in both diet regimens (Supplementary Fig. S1K). Finally, serum from TB mice increased PCSK9 expression in primary hepatocytes (Supplementary Fig. S1L), indicating that tumors secrete factors that modulate hepatic cholesterol metabolism, particularly upregulating PCSK9. In line with these findings, circulating PCSK9 was significantly elevated in patients with advanced cancers (Fig. 2I).

Figure 2.

Hypercholesterolemia exacerbates alterations in cholesterol metabolism. A, The experimental design involved adult mice being fed with HCD for 8 weeks before being injected with MN/MCA1 tumor cells. These mice were maintained on HCD for an additional 23–24 days, after which they were sacrificed for sample collection and analysis. B, Blood levels of total cholesterol, HDL, and LDL in NCD- or HCD-fed MN/MCA1 mice vs. TF mice. Starting from time 0 (T0), mice were maintained on their respective dietary regimens for 8 weeks, injected with MN/MCA1 tumor cells (T1), and then sacrificed at AD stage (T2; n = 9). C and D, Heatmap showing differential mRNA expression of genes involved in hepatic metabolism (n = 5; C) and circulating PCSK9 levels (n = 5; D) in ED or AD MN/MCA1 vs. TF mice fed with either NCD or HCD. E, Immunoblot analysis of LDLR protein in the livers from NCD- or HCD-fed MN/MCA1 mice at AD vs. TF similarly fed (n = 2). F, FACS quantification of BODIPY-Ch ΔMFI of hepatic CD45⁻CD31⁻ cells from MN/MCA1 mice at ED (n = 3) or AD (n = 5) on NCD or HCD compared with TF mice (n = 3) similarly fed. G and H, FACS quantification of BODIPY-Ch ΔMFI of intratumoral CD45^–CD31⁻ cells and CD11b⁺Ly6C^lo/−F4/80⁺ TAMs (G,) and in BM myeloid progenitors (both CMPs and GMPs) or blood CD11b⁺Ly6C^hi monocytic cells (H) from TF (n = 3), ED (n = 3), or AD (n = 5) MN/MCA1 mice on NCD or HCD. I, Circulating PCSK9 levels in patients with early (I–II) vs. advanced (III–IV) NSCLC (n = 15 vs. n = 21), CRC (n = 16 vs. n = 20), BRC (n = 14 vs. n = 14), PDAC (n = 9 vs. n = 10), BTC (n = 9 vs. n = 10), or PNET (n = 10 vs. n = 10) in comparison with healthy donors (HD, n = 9). J–L, FACS analysis of PCSK9 and LDLR protein expression (J and K) or BODIPY-Ch (L) in Hepa1-6 cells treated with MN/MCA1 TCM, IL6, or IL1β alone or in combination with anti-IL6 (mAb) or anti-IL1β (IL1 receptor antagonist, IL1Ra) agent (n = 3). M, Circulating PCSK9 levels in TF (n = 3) or MN/MCA1 bearing mice in AD (n = 5) on NCD or HCD in the absence (control) or presence of anti-IL6 or anti-IL1β. N, Immunoblot (top) and relative densitometry quantification (bottom, n = 2) of hepatic LDLR protein using protein extracts from mice treated as described above. Data are representative of three (B) or two (E) independent experiments. C, D, and F–N, One experiment. Data are presented as the mean ± SEM (A–H and J–M) or Box-and-whisker min-to-max plots (I). Statistic by two-way ANOVA (B), one-way ANOVA (D–N), or Kruskall–Wallis (I, BRC). BRC, breast cancer; BTC, biliary tract cancer; CRC, colorectal cancer; PDAC; pancreatic ductal adenocarcinoma; PNET, pancreatic neuroendocrine tumor.

Emerging evidence has linked inflammatory conditions (e.g., infections, rheumatoid arthritis, and acute coronary syndrome) and their mediators to altered lipid metabolism (33–35) and PCSK9 upregulation (36). In particular, IL6 and IL1β, which play critical roles in both cancer development (37) and metabolic alterations (11), have been found to influence the activity of the PCSK9/LDLR axis as well as cholesterol levels (38–40). As shown in Supplementary Fig. S1M, TCM contained significant levels of IL6 and IL1β. Fittingly, stimulation of hepatocytic cells (Hepa1-6) with TCM, IL6, or IL1β enhanced PCSK9 expression and decreased LDLR protein expression, whereas treatment with an anti–IL6 or anti–IL1β agent reversed this effect (Fig. 2J and K), restoring the ability of the hepatocytes to uptake extracellular cholesterol (Fig. 2L). In line with these findings, both liver (Supplementary Figs. S1N–S1Q) and adipose tissues (Supplementary Fig. S1R) from TB mice and the healthy liver parenchyma from patients with colorectal cancer (Supplementary Fig. S1S) showed enriched levels of both IL6 and IL1β. Additionally, the administration of anti–IL6 or anti–IL1β to MN/MCA1 mice led to the suppression of PCSK9 upregulation (Fig. 2M; Supplementary Fig. S1T) and restoration of LDLR protein levels (Fig. 2N). Interestingly, these effects were consistent with the inhibition of tumor growth and reduction in blood cholesterol levels observed in NCD- or HCD-fed mice (Supplementary Fig. S1U and S1V).

Hypercholesterolemia Boosts Protumoral Myelopoiesis

Interestingly, whereas HCD did not affect primary MN/MCA1 tumor growth (Supplementary Fig. S2A), it did lead to an increased number of circulating (mCherry⁺) tumor cells (Supplementary Fig. S2B) and a higher frequency of CD31⁺ endothelial cells in both primary and metastatic lungs (Supplementary Fig. S2C), ultimately resulting in a pronounced exacerbation of lung metastasis formation (Fig. 3A). Similarly, hypercholesterolemia also promoted the formation of lung metastasis in the K1735-M2 melanoma model (Supplementary Fig. S2D) and contributed to the growth of tumor lesions in the genetic KP lung cancer model (Supplementary Fig. S2E).

Figure 3.

Hypercholesterolemia boosts protumoral myelopoiesis. A, Lung metastatic area in NCD- or HCD-fed MN/MCA1 mice (n = 3). Representative images are shown. Scale bar, 1 mm. B, Representative FACS plot (left) and frequencies (right) of intratumoral CD11b⁺Ly6G⁻Ly6C^hi M-MDSCs and CD11b⁺Ly6G⁻Ly6C^loF4/80⁺ TAMs from NCD- or HCD-fed MN/MCA1 mice (n = 6). C, FACS analysis of M1 (top: TNFα, MHC-II, iNOS, and CCR7) and M2 (bottom: CD206, PD-L1, CD204, and IDO1) polarization markers of TAMs from NCD- or HCD-fed MN/MCA1 mice (n = 5). D, Frequencies of CD4⁺ (top) and CD8⁺ (bottom) T cells and relative PD-1⁺ subsets and expression levels of CTLA4 and IFNγ in tumors (n = 5). E, Effects of TAMs (top right) or M-MDSCs (bottom right), FACS-sorted from NCD- or HCD-fed MN/MCA1 mice, on the proliferation of CFSE-labeled CD8⁺ T cells activated with anti-CD3/anti-CD28 (n = 3). Left, Representative histogram plot of CD8⁺ T-cell proliferation in coculture with TAMs. Data are representative of at least five (A and B) or two (C–E) independent experiments. Data are presented as the mean ± SEM. Statistic by t test [A, B, C (TNFα, MHC-II, iNOS, CCR7, CD206, PD-L1, CD204), D], Mann–Whitney (C, IDO1), or one-way ANOVA (E).

Because obesity (9, 11, 41) and high cholesterol (26, 27) are associated with preferential differentiation of hematopoietic stem cells into the monocyte–macrophage cell lineage, and given that myeloid cells exhibit tumor-promoting activities (2–4), we further investigated the myeloid compartment of hypercholesterolemic MN/MCA1 mice. HCD increased the frequency of blood and splenic CD11b⁺Ly6C^hi monocytic cells under both TF and TB conditions (Supplementary Fig. S2F), whereas, in contrast with previous observations (12), the frequency of CD11b⁺Ly6G⁺ granulocytic cells remained unaffected (Supplementary Fig. S2F). Flow cytometry analysis of tumor-infiltrating myeloid cells (see “Methods”) confirmed that HCD specifically increased the presence of TAMs and monocytic MDSCs (M-MDSC; Fig. 3B), which showed higher lipid-load levels as evidenced by LipidTOX neutral lipid (i.e., cholesterol, cholesterol esters, triglycerides) staining (Supplementary Fig. S2G) and by BODIPY-Ch uptake (Fig. 2G). In-depth lipidomic analysis of TAMs revealed significant imbalances in numerous lipid species induced by HCD (Supplementary Fig. S2H–S2J). Remarkably, the lipidomic analysis revealed 52 significantly enriched lipids and 148 downregulated lipids (Supplementary Fig. S2I). Importantly, this lipidomic profiling showed an overall enrichment of triglyceride and sterol lipid classes, with a notable increase in cholesterol levels at the expense of its esterified forms (Supplementary Fig. S2K). Notably, the lipid changes observed in TAMs from HCD-fed mice were accompanied by a prominent shift toward M2 polarization (42), as evidenced by the downregulation of M1 markers (i.e., TNFα, MHC-II, iNOS, and CCR7; Fig. 3C, top) and simultaneous increase in M2 indicators (i.e., CD206, PD-L1, CD204, and IDO1; Fig. 3C, bottom).

As macrophages and MDSCs are known to contribute to the formation of premetastatic niches (3, 4, 43), and there is evidence that cholesterol metabolites and, more general obesity may promote cancer metastasis (12, 23, 44–46), we analyzed the frequency of lung myeloid cells in mice fed NCD or HCD (Supplementary Fig. S2L). HCD-fed TB mice showed an increased frequency of CD11b⁺Ly6C^hi monocytic cells (Supplementary Fig. S2M), as well as of bone marrow (BM)-derived interstitial macrophages (IM) and resident alveolar macrophages (AM; Supplementary Fig. S2N; see “Methods”; refs. 47, 48). In contrast, conventional CD103⁺ (cDC1) and CD11b⁺ (cDC2) dendritic cell (DC) subsets (see “Methods”; ref. 47) remained unaffected (Supplementary Fig. S2O). Furthermore, HCD induced a shift toward M2 polarization (TNFα^loMHC-II^loCD206^hiPD-L1^hi) of IM and AM cells (Supplementary Fig. S2P). These HCD-induced alterations in myeloid cell frequency and phenotype were similarly observed in the K1735-M2 metastatic melanoma model (Supplementary Fig. S3A–S3D). Additionally, whereas HCD improved the frequency of CD4⁺ and CD8⁺ T cells in both primary tumors (Fig. 3D) and metastatic lungs (Supplementary Fig. S3E), their phenotype displayed signs of dysfunction and exhaustion (49), with higher expression of PD-1 and cytotoxic T-lymphocyte antigen 4 (CTLA4) and lower expression of IFNγ. We further assessed the proliferation rate of T cells cocultured with TAMs or M-MDSCs (Fig. 3E), FACS-sorted from HCD- or NCD-fed mice, and found that myeloid cells from HCD-fed mice significantly reduced CD8⁺ T-cell proliferation compared with their NCD-fed counterparts, indicating that hypercholesterolemia induces the expansion of myeloid cells endowed with suppressive activity.

Lowering Cholesterol Mitigates Immunosuppressive Myelopoiesis in Cancer Bearers

Given the observed hepatic elevation of PCSK9 promoted by tumor growth (Fig. 2C, D, and I; Supplementary Fig. S1I) and its association with immunosuppressive myeloid cells (Fig. 3B and C; Supplementary Fig. S2F and S2L–P), we explored the potential impact of administration of the cholesterol-lowering anti-PCSK9 mAb, mAb1 (50), on protumoral myelopoiesis and tumor development in both MN/MCA1 (Supplementary Fig. S3F) and KP cancer models (Fig. 4A). Strikingly, anti-PCSK9 antibody treatment lowered blood cholesterol levels in MN/MCA1 (Supplementary Fig. S3G) and KP mice (Fig. 4B) and decreased primary MN/MCA1 (Supplementary Fig. S3H) and KP tumors (Fig. 4C), as well as MN/MCA1 lung metastasis (Supplementary Fig. S3I), with a prominent effect in the context of HCD. These beneficial effects were accompanied by a reduced frequency of M-MDSCs, TAMs, and lung metastasis–associated IMs and AMs in MN/MCA1 (Supplementary Fig. S3J–S3L), as well as decreased M-MDSCs, IMs, and AMs in primary KP lung tumors (Fig. 4D–F).

Figure 4.

Lowering cholesterol in cancer bearers mitigates immunosuppressive myelopoiesis. A–G, NCD- or HCD-fed mice bearing the genetically induced Kras^LSL-G12D/+; Trp53^fl/fl (KP) lung cancer and treated with anti-PCSK9 or isotype control (n = 5). Experimental design (A): KP mice at 6 weeks of age were conditioned with HCD or NCD for 8 weeks before intranasal inoculation of Ad5-CMV-Cre. After further 25 days, an anti-PCSK9 blocking antibody was administered once a week. After 12 weeks, mice were sacrificed for sample collection and analysis: total cholesterol blood levels (B); lung metastatic area. Representative images are shown. Scale bar, 1 mm (C); blood M-MDSC frequency (D). E and F, FACS quantification of IMs (E) and AMs (F) and their relative MFI of TNFα, MHC-II, CD206, PD-L1, and RORγ. G, Frequency of pulmonary CD8⁺ T cells and relative PD-1 and IFNγ expression levels (MFI). H and I, Experimental design (H): wt or PCSK9-deficient (Pcsk9^−/−) mice were lethally irradiated (9 Gy) and respectively transplanted with either wt (wt > wt; wt > Pcsk9^−/−) or Pcsk9^−/− (Pcsk9^−/− > wt; Pcsk9^−/− > Pcsk9^−/−) BM cells. After 4 weeks for complete hematopoietic reconstitution, mice were conditioned for an additional 8 weeks with HCD and then engrafted with MN/MCA1 cells. After 25 days of tumor growth, mice were sacrificed for sample collection and analysis, including the evaluation of lung metastatic areas (n = 5). Representative images are shown. Scale bar, 1 mm (I). A–G, Data are representative of two independent experiments. H and I, One experiment was performed. Data are presented as the mean ± SEM. Statistic by one-way ANOVA [B, D, E (MHC-II and CD206), F (TNFα, CD206, and PD-L1), and G], Welch ANOVA [C, E (IM, TNFα, and PD-L1), and F (AM and MHC-II)], or two-way ANOVA (I).

Consistent with the reduction in circulating cholesterol levels (Supplementary Fig. S3G), anti-PCSK9 treatment in MN/MCA1-bearing mice decreased neutral lipid content in TAMs (Supplementary Fig. S3K). This restoration of lipid levels corresponded with a shift in TAMs toward an M1 phenotype (MHC-II^hiCD206^lo; Supplementary Fig. S3K), as well as in IMs and AMs from the metastatic lungs (Supplementary Fig. S3L). Likewise, anti-PCSK9 blockade in KP lung cancer bearers reduced blood cholesterol (Fig. 4B), whereas it increased expression of the M1 markers TNFα, to a lesser extent, and MHC-II and decreased the M2 markers CD206 and PD-L1 in IMs and AMs (Fig. 4E and F). These anti-PCSK9–mediated events were associated with reduced expression of the exhaustion markers PD-1 and CTLA4, accompanied by increased production of IFNγ by CD8⁺ T cells in both primary MN/MCA1 tumors (Supplementary Fig. S3M) and metastatic lungs (Supplementary Fig. S3N), as well as in the KP lung cancer model (Fig. 4G).

Given that we did not detect any significant difference in PCSK9 expression by tumor cells between NCD and HCD feeding and that anti-PCSK9 treatment had no effect on tumor cell viability and proliferation in vitro (Supplementary Fig. S3O), to rule out possible tumor- or immune cell–intrinsic roles of PCSK9 and confirm the centrality of tumor inflammatory signals in the induction of PCSK9 expression in the liver, we transplanted wild-type (wt) or Pcsk9^−/− BM cells into lethally irradiated wt or Pcsk9^−/− recipient mice, and vice versa, and evaluated the development of MN/MCA1 under HCD conditions (Fig. 4H). Extramedullary deficiency of PCSK9 markedly lowered circulating cholesterol levels (Supplementary Fig. S3P) and inhibited tumor growth (Fig. 4I; Supplementary Fig. S3Q), along with a decrease in the frequency and M2 polarization of TAMs (Supplementary Fig. S3R). As expected, Pcks9 deletion in the hematopoietic compartment did not produce these effects, whereas anti-PCSK9 treatment elicited cholesterol-lowering and antitumoral activities only in extramedullary PCSK9-proficient recipient mice (Fig. 4I; Supplementary Fig. S3P ans S3Q). Altogether, these results indicate hepatic PCSK9 as the main driver of cholesterol metabolism alterations in cancer bearers.

Hypercholesterolemia Affects Differentiation and Mobilization of Myeloid Cells

We then proceeded to characterize the roles of the two main pathways of macrophage accumulation, CSF1/CSF1R (51) and CCL2/CCR2 (3), in hypercholesterolemic MN/MCA1 mice. Even though HCD did not affect CSF1R (CD115) expression in TAMs, IMs, and AMs (Supplementary Fig. S4A), in agreement with previous reports (52, 53), it did increase the frequency of CCR2⁺ myeloid cells by augmenting their CCR2 surface expression levels (Supplementary Fig. S4B). Furthermore, we detected higher CSF1 (M-CSF) and CCL2 mRNA expression levels in the TME of HCD-fed mice (Supplementary Fig. S4C). Accordingly, treatment with anti-CSF1R or anti-CCR2 agents reduced primary MN/MCA1 growth and lung metastatic burden (Supplementary Fig. S4D and S4E), paralleled by a marked reduction in M-MDSCs (Supplementary Fig. S4F), as well as TAMs (Supplementary Fig. S4G), IMs, and AMs (Supplementary Fig. S4H), which, however, displayed higher levels of TNFα (Supplementary Fig. S4G and S4H). Anti-CSF1R or anti-CCR2 treatment increased the frequency of CD8⁺ T cells and their IFNγ expression levels (Supplementary Fig. S4I). Given the significance of the CSF1/CSF1R and CCL2/CCR2 axes in the differentiation and mobilization of myeloid progenitors (1), we next evaluated the lineage commitment of Lin⁻c-Kit⁺Sca-1⁻ BM hematopoietic progenitors in HCD-fed TB mice. HCD increased the lipid and cholesterol load of myeloid progenitors (Fig. 2H; Supplementary Fig. S4J), leading to enhanced commitment of CMPs to GMPs in both MN/MCA1 (Supplementary Fig. S4J and S4K) and KP mice (Supplementary Fig. S4L). Both anti-PCSK9 treatment (Supplementary Fig. S4J and S4L) and extramedullary Pcks9 deletion (Supplementary Fig. S4M) prevented these effects as result of a drastic reduction of systemic cholesterol levels (Fig. 4B; Supplementary Fig. S3G). Moreover, anti-CSF1R and anti-CCR2 treatments also hampered the HCD-induced commitment of GMPs (Supplementary Fig. S4N).

Taken together, these results suggest that PCSK9 plays a pivotal role in the cholesterol-induced modulation of myeloid progenitor differentiation, ultimately affecting TAMs and myeloid cell populations within the TME.

RORγ Tunes Myelopoiesis to Cholesterol Metabolism

As we have previously reported RORC1/RORγ to be a marker of advanced cancer inflammation and a promoter of both MDSC and TAM differentiation (5), and given that cholesterol precursors and oxysterol derivatives not just enhance RORγ intrinsic activity (28) but are also aberrantly expressed in both cancer (54) and obesity (55, 56), we asked whether increased cholesterol levels would influence cancer myelopoiesis in a RORγ-dependent manner. Analysis of tumors, lungs, BM, and spleens from HCD-fed MN/MCA1 mice showed increased RORγ expression in BM-derived CCR2⁺ subsets of TAMs (Fig. 5A), M-MDSCs, IMs, and AMs (Supplementary Fig. S5A–S5D) but not in their CCR2⁻ counterparts (Supplementary Fig. S5E). Congruently, HCD led to the upregulation of RORγ in both CMPs and GMPs (Supplementary Fig. S5F).

Figure 5.

RORγ bridges hypercholesterolemia and protumoral myelopoiesis. A, RORγ expression and RORγ⁺ cells in CCR2⁺ TAMs from NCD- or HCD-fed MN/MCA1-bearing mice (n = 5). B, Linear Pearson correlation between total cholesterol levels and M-MDSC (HLA-DR^lo/−CD14⁺; left) or M-MDSC RORγ⁺ cells (right) in PBMCs from healthy donors (HD) or patients with NSCLC (CA; stages III–IV). Pearson correlation coefficient (r) and corresponding P value are indicated. C and D, NCD- or HCD-fed MN/MCA1 wt mice transplanted with either wt (wt > wt) or Rorc^−/− (Rorc^−/− > wt) BM cells (n = 7). C, Primary tumor growth (left) and lung metastatic areas (right). Representative images are shown. Scale bar, 1 mm. D, Intratumoral M-MDSC and TAM frequency (top) and MHC-II and CD206 expression in TAMs (bottom). E, Primary tumor growth (left) and lung metastatic area (right) in Rorc^fl/fl and Rorc^fl/flLyz2-Cre MN/MCA1 mice on NCD or HCD (n = 5). Representative images are shown. Scale bar, 1 mm. F, Top, Experimental design: wt mice on NCD or HCD were coinjected with MN/MCA1 and FACS-sorted wt or Rorc^−/− TAMs (dose 1). A second adoptive transfer of FACS-sorted TAMs was intratumorally administered after 13 days of tumor growth (dose 2; n = 5). Bottom, lung metastatic areas. Representative images are shown. Scale bar, 1 mm. G, Principal component analysis (PCA) based on transcriptional profiles from RNA-seq analysis on wt or Rorc^−/− BMDMs differentiated with M-CSF, with or without cholesterol (Ch) supplementation, and then treated or not [control (ctrl)] with IL4 (n = 3). Data are representative of five (A) or two (E) independent experiments. B–D, F, and G, One experiment was performed. A and C–F, Data are presented as the mean ± SEM. Statistic by t test (A), Pearson’s correlation with two-tailed P value (B), two-way ANOVA [C (left) and E (left)], Kruskal–Wallis [C (right), E (right), and D (TAM)], Welch ANOVA [D (M-MDSC) and F], and one-way ANOVA [D (MHC-II, CD206]. G, Detailed description in the “Methods” section.

To validate our findings in patients with cancer, we assessed monocytic cell subsets in peripheral blood mononuclear cells (PBMC; see “Methods”) from healthy donors (n = 18) and patients with NSCLC (stage III-IV, n = 23). In line with our observation in the mouse model, the frequency of M-MDSCs and their RORγ⁺ subset correlated with total cholesterol or LDL levels in patients with lung cancer (Fig. 5B; Supplementary Table S2). To test whether circulating cholesterol would modulate RORγ transcriptional activity, we cotransfected HEK293 cells with a vector constitutively expressing the Rorc gene and a plasmid expressing the luciferase reporter gene under the control of the minimal promoter of the Il17a gene (minIL17Aprom-Luc; ref. 57), a transcriptional target of RORγ (Supplementary Fig. S5G). Transfected cells were treated with free cholesterol, LDL, the RORγ agonist SR0987, or desmosterol, an immediate cholesterol precursor acting as endogenous RORγ agonist (28). All four conditions significantly induced luciferase activity (Supplementary Fig. S5H), indicating that circulating cholesterol is crucial for the activation of RORγ-dependent cellular programs. Next, we treated transfected HEK293 cells with serum from NCD- or HCD-fed TB mice or with serum from TF mice and measured luciferase activity. Whereas exposure to serum from NCD-fed TB mice led to increased RORγ transcriptional activity, this response was consistently higher when cells were stimulated with serum from HCD-fed TB mice (Supplementary Fig. S5I). Notably, TCM alone poorly affected RORγ activity, and its increase by cholesterol supplementation was inhibited by RORγ inhibitor SR2211 (Supplementary Fig. S5J; refs. 5, 58).

To confirm in vivo the key role of PCSK9 as a key interface between inflammation and cholesterol in inducing the expansion of RORγ⁺ myeloid cells, we chronically treated PCSK9-proficient (wt) or PCSK9-deficient TF mice with IL6 and IL1β cytokines. Coherently with the downregulating effect by anti–IL6 and anti–IL1β on PCSK9 and cholesterol levels (Fig. 2M; Supplementary Fig. S1T and S1V) and in line with previous observation (40, 59), both IL6 and IL1β chronic treatment induced PCSK9 upregulation (Supplementary Fig. S5K) as well as blood cholesterol levels in wt mice (Supplementary Fig. S5L). Consistently with previous reports (60), chronic administration of both IL6 and IL1β increased the myeloid cell output as displayed by the enhanced commitment of CMP into GMP cells (Supplementary Fig. S5M) in association with a higher frequency of RORγ⁺ GMP cells and monocytic cells in the BM, as well as in the peripheral blood (Supplementary Fig. S5N). Noteworthy, these IL6/IL1β–mediated effects on cholesterol increase and induction of myelomonocytic cells were lost in PCSK9-null mice.

Accordingly, reduction of PCSK9 and cholesterol levels in tumor settings, via blockade of IL6 and IL1β (Fig. 2M; Supplementary Fig. S1T and S1V), was associated with reduced CMP engagement in GMP RORγ⁺ cells in the BM (Supplementary Fig. S5O), which was reflected in a lower frequency of circulating RORγ⁺ monocytic cells and intratumoral RORγ⁺ TAMs (Supplementary Fig. S5P), to a greater extent under HCD conditions (Supplementary Fig. S5Q). Taken together, our data clearly indicate that the cancer-related inflammation/PCSK9/cholesterol axis plays a central role in RORγ⁺ monocytic cell expansion.

Next, considering that RORγ transcriptional activation by cholesterol-rich serum correlated with the expansion of monocytic cells seen in hypercholesterolemic tumor bearers (Fig. 3B; Supplementary Fig. S2F), we aimed to evaluate the extent of the interplay between RORγ and hypercholesterolemia in NCD- or HCD-fed MN/MCA1 RORγ-deficient mice (Rorc^−/−). Whereas RORγ deficiency did not affect cholesterol levels (Supplementary Fig. S6A and S6B), it reduced primary tumor growth (Supplementary Fig. S6C) and inhibited the formation of metastasis and circulating mCherry⁺ MN/MCA1 tumor cells (Supplementary Fig. S6D and S6E), regardless of the diet type. Consistent with our previous findings, a predominant M2-polarized immune profile was observed in the TME of HCD- versus NCD-fed mice, as judged by increased mRNA expression of Il4, Tgfb, Ccl22, Mrc1 (CD206), Ccl2, and Csf1 (M-CSF; Supplementary Fig. S6F). This M2 polarization was further supported by an increased frequency of TAMs, lung IMs and AMs, and tumor-infiltrating M-MDSCs (Supplementary Fig. S6G). Supporting a key role of RORγ in bridging hypercholesterolemia and protumoral myelopoiesis, all the differences were lost in both NCD- or HCD-fed RORγ-null mice (Supplementary Fig. S6C–S6G), concomitant with a predominant expression of M1 cytokines (i.e., Tnfa, Cxcl9, Cxcl10, and Cxcl11; Supplementary Fig. S6F). In contrast, no variation was observed in the granulocytic MDSC subset (PMN-MDSC; Supplementary Fig. S6G). Regardless of RORγ expression, TAMs (Supplementary Fig. S6H), IMs, and AMs (Supplementary Fig. S6I) from HCD-fed mice exhibited an accumulation of neutral lipids. Cholesterol-specific Filipin-III staining confirmed the cholesterol overload in TAMs from hypercholesterolemic mice (Supplementary Fig. 6J). Consistent with the role of RORγ as a mediator of cholesterol-induced M2 polarization, the M2-like activation state of TAMs, IMs, and AMs induced by hypercholesterolemia was lost in RORγ null mice, as evidenced by increased expression of TNFα and MHC-II and downmodulation of CD206 and PD-L1 (Supplementary Fig. S6K and S6L), as well as by a shift of TAMs toward an IL12p40^high/IL10^low phenotype (Supplementary Fig. S6M).

Myeloid-Specific RORγ Regulates Tumor-Promoting Immunosuppressive Myelopoiesis

To rule out the possibility of any antitumoral effect resulting from extramedullary deficiency of RORγ, we transplanted lethally irradiated wt recipient mice with either Rorc^−/− or wt BM cells. These mice were then subjected to HCD or NCD conditioning before engrafting them with MN/MCA1 cells (Supplementary Fig. S7A). Hematopoietic RORγ deficiency (Rorc^−/− > wt) curtailed HCD-induced tumor progression (Fig. 5C), which was accompanied by a reduced frequency of M-MDSCs, TAMs (Fig. 5D), IMs, and AMs (Supplementary Fig. S7B), all of which displayed a polarization shift toward an M1 phenotype (Fig. 5D; Supplementary Fig. S7B).

In light of cholesterol-induced expression of RORγ in both CMPs and GMPs (Supplementary Fig. S5F), which suggested a direct link between lipid metabolism and RORγ-dependent emergency myelopoiesis (5), we assessed the frequency of BM myeloid progenitors in hypercholesterolemic TB and lethally irradiated wt mice transplanted with Rorc^−/− BM cells. In agreement with a previous report (5), we observed a reduction in Lin⁻c-kit⁺Sca-1⁺ hematopoietic cells in the BM of TB mice reconstituted with wt BM cells (wt > wt) under HCD conditions. By contrast, RORγ deficiency in hematopoietic cells (Rorc^−/− > wt) increased the frequency of Lin⁻c-kit⁺Sca-1⁺ cells in a diet-independent manner (Supplementary Fig. S7C). Similar results were observed in total RORγ-deficient mice (Supplementary Fig. S7D). In addition, whereas CMPs and GMPs displayed elevated intracellular neutral lipids and cholesterol content under HCD conditions, regardless of RORγ expression (Supplementary Fig. S7E and S7F), the increased transition of CMPs to GMPs observed in hypercholesterolemic mice (Supplementary Fig. S4K) was significantly inhibited by RORγ deficiency (Supplementary Fig. S7C), indicating that RORγ acts as a key transcriptional effector in monocyte/macrophage differentiation induced by hypercholesterolemia.

During our experiments, we also noticed that both whole RORγ-deficient mice (Rorc^−/−; Supplementary Fig. S7G) and Rorc^−/− BM cell–transplanted mice (Rorc^−/− > wt; Supplementary Fig. S7H) displayed an increased number of tumor-infiltrating CD4⁺ and CD8⁺ T cells expressing high IFNγ levels compared with their wt counterparts. To ascertain whether RORγ directly influenced T-cell differentiation, we used myeloid-specific RORγ−deficient mice (Rorc^fl/flLyz2-Cre). These mice displayed reduced MN/MCA1 primary tumor growth and metastasis formation (Fig. 5E), as well as B16-F10 (Supplementary Fig. S8A) and MC38 (Supplementary Fig. S8B) metastasis development, particularly under HCD conditions. Correspondingly, specific ablation of RORγ in myeloid cells (Rorc^fl/flLyz2-Cre mice) reprogrammed TAMs toward an M1 phenotype, as evidenced by increased MHC-II and decreased CD206 and PD-L1 expression levels (Supplementary Fig. S8C–S8E). This phenotypic change was associated with enhanced effector functions of infiltrating CD8⁺ T cells, as judged by the higher expression levels of IFNγ and granzyme-B (Supplementary Fig. S8F–S8H). Moreover, in contrast with wt TAMs, adoptive transfer of Rorc^–/– TAMs into either NCD- or HCD-fed MN/MCA1 wt mice (Fig. 5F) led to increased IFNγ and granzyme-B expression in infiltrating CD8⁺ T cells (Supplementary Fig. S8I), effectively hindering tumor progression (Fig. 5F).

RORγ Acts as a Cholesterol Sensor and Modulator of Myeloid-Mediated Immunosuppression

To investigate the role of cholesterol-induced RORγ activity in monocyte-to-macrophage differentiation driven by M-CSF and alternative polarization, we performed RNA sequencing (RNA-seq) on RNA purified from wt or Rorc^−/− BM-derived macrophages (BMDM) treated with IL4 versus untreated control to induce M2 polarization in the presence or absence of cholesterol supplementation. Principal component analysis and unsupervised hierarchical clustering revealed distinct responses to cholesterol between wt and Rorc^−/− cells (Fig. 5G; Supplementary Fig. S9A). RORγ-deficient cells had a similar transcriptional profile regardless of cholesterol supplementation. In contrast, cholesterol addition profoundly changed the transcriptional profile in wt macrophages, with a more pronounced effect in IL4-treated cells (Fig. 5G). Specifically, IL4-treated wt macrophages supplemented with cholesterol displayed the highest number of differentially expressed genes (1539 upregulated and 1351 downregulated; Supplementary Fig. S9B and S9C). Consistent with our findings in vivo (Fig. 5D; Supplementary Figs. S6K–M and S8C–E), the intersection analysis of gene sets identified in wt and Rorc^−/− macrophages revealed opposite modulation of gene expression upon cholesterol supplementation (Supplementary Fig. S9B and S9C). Specifically, there was an inverse regulation of genes involved in the “IFNα/γ response” pathway in wt versus Rorc^−/− macrophages (Supplementary Fig. S9B and S9C; Combo 2 and 8). Furthermore, whereas the “inflammatory response” and “TNF signaling” pathways were only associated with downregulated genes in wt cells (Supplementary Fig. S9C; Combo 7 and/or 8), genes of the “TGFβ signaling” pathway were upregulated exclusively in wt macrophages in response to cholesterol treatment (Supplementary Fig. S9B; Combo 3 and 6). In addition, we cultured peritoneal macrophages (PEC) from wt or Rorc^−/− mice with TCM, supplemented or not with cholesterol. As shown in Supplementary Fig. S9D, RORγ-null PECs displayed decreased expression of M2 genes (i.e., Il4, Tgfb, Ccl22, Mrc1, Ccl2, and Csf1) while upregulating M1 genes (i.e., Tnfa, Cxcl9, Cxcl10, and Cxcl11).

In agreement with Supplementary Fig. S4A, the expression of the M-CSF receptor CD115 (Csf1r) was not affected by either cholesterol or RORγ deficiency (Supplementary Fig. S9E), whereas cholesterol-dependent CCR2 upregulation was reduced in the absence of RORγ (Supplementary Fig. S9F). TAM analysis corroborated RORγ-dependent regulation of CCR2 (Supplementary Fig. S9G). Because CCR2 and CSF1R/CD115 are chemotactic receptors, we assessed the migration of wt versus Rorc^−/− PECs, preconditioned with cholesterol for 24 hours, in response to M-CSF, CCL2, or TCM. In line with the unchanged expression of CSF1R/CD115 (Supplementary Fig. S9E and S9G), the chemotaxis ability of cholesterol-conditioned Rorc^−/− PECs was severely impaired in response to CCL2 and TCM but not M-CSF (Supplementary Fig. S9H). Consistent with the M1 skewing observed in Rorc^−/− PECs (Supplementary Fig. S9D), RORγ deficiency in TAMs or M-MDSCs from hypercholesterolemic mice enhanced the proliferation rate of cocultured CD4⁺ and CD8⁺ wt T cells (Supplementary Fig. S9I). Likewise, BMDMs from Rorc^−/− mice, conditioned with TCM plus cholesterol, displayed increased CD4⁺ and CD8⁺ T-cell proliferation (Supplementary Fig. S9J). Collectively, these findings couple the cholesterol sensing function of RORγ to its role in driving the differentiation and tumor infiltration of M2-like immunosuppressive myeloid cells.

Given that cholesterol reduction by anti-PCSK9 treatment (Fig. 4B; Supplementary Fig. S3G) or extramedullary PCSK9 deficiency (Supplementary Fig. S3P) prevented the CMP-to-GMP commitment (Supplementary Fig. S4K–S4M) and M1-to-M2 macrophage polarization associated with higher levels of RORγ induced by HCD (Fig. 4E and F; Supplementary Fig. S3K and S3R), to dissect the relative contribution of RORγ and cholesterol to tumor development, we treated hypercholesterolemic wt or Rorc^−/− mice with an anti-PCSK9 antibody. Tumor progression was reduced to a similar extent in both anti-PCSK9–treated wt mice and untreated Rorc^−/− mice (Supplementary Fig. S10A and S10B). Moreover, the antitumor activity driven by RORγ deficiency was not further improved by anti-PCSK9 treatment, indicating that cholesterol-lowering (Supplementary Fig. S10C) did not enhance the impact of RORγ deficiency on tumor inhibition (Supplementary Fig. S10A and S10B) and the frequency of tumor-infiltrating M-MDSCs, TAMs, IMs, and AMs (Supplementary Fig. S10D–S10F). In this regard, the unchanged transition of CMPs to GMPs observed in the context of RORγ deficiency (Supplementary Figs. S7C, S7D, and S10G) was consistent with the unaltered effects of cholesterol-lowering under these conditions. Additionally, anti-PCSK9 treatment reduced the lipid content in TAMs without affecting CCR2, TNFα, MHC-II, and CD206 expression levels in RORγ-null animals (Supplementary Fig. S10H). Similar results were observed for IM and AM cell subsets (Supplementary Fig. S10F). In keeping with these findings, whereas anti-PCSK9 treatment of HCD-fed mice enhanced IFNγ and reduced both CTLA4 and PD-1 expression by CD8⁺ T cells, these effects were hindered by RORγ deficiency (Supplementary Fig. S10I). Furthermore, we confirmed the importance of myeloid cell–dependent suppression of specific antitumor immunity by depleting CD4⁺ and CD8⁺ T lymphocytes in HCD-fed TB mice, which curbed the antitumor activity of RORγ deficiency (Supplementary Fig. S10J and S10K).

Overall, these results demonstrate that the cholesterol–RORγ axis plays a critical role in myeloid cell–mediated suppression of specific antitumor immunity.

RORγ Inhibition Improves the Response to Immunotherapy

To validate the potential of targeting RORγ as a pharmacologic intervention in hypercholesterolemia-induced protumoral myelopoiesis, we treated NCD- or HCD-fed mice with the RORγ inhibitor SR2211 (5, 58) and evaluated MN/MCA1 tumor development. SR2211 treatment led to a substantial reduction in lung metastasis formation (Supplementary Fig. S10L) and only had a mild effect on primary tumor growth (Supplementary Fig. S10M), without altering cholesterol levels (Supplementary Fig. S10N). Furthermore, whereas SR2211 treatment of HCD-fed TB mice reduced both TAM accumulation and their CCR2 expression levels (Supplementary Fig. S10O), it increased the expression of TNFα and MHC-II, leaving CD206 expression unchanged (Supplementary Fig. S10P). Similar results were obtained in both IMs and AMs (Supplementary Fig. S10Q). In good agreement with the observed decrease in TAMs in HCD-fed mice (Supplementary Fig. S10O), SR2211 also effectively blocked the transition of CMPs to GMPs, particularly under hypercholesterolemic conditions (Supplementary Fig. S10R). SR2211-induced reprogramming of TAMs, IMs, and AMs was paralleled by increased IFNγ and reduced PD-1 and CTLA4 expression levels in tumor-infiltrating CD8⁺ T cells (Supplementary Fig. S10S). The antitumor activity of the RORγ inhibitor was further demonstrated in K1735-M2 metastatic melanoma–bearing mice (Fig. 6A–C), in which SR2211 treatment led to a reduction in the frequency of TAMs, M-MDSCs, IMs, and AMs (Fig. 6D–F), restored the TNFα^high M1 phenotype of pulmonary macrophages (Fig. 6F), and reactivated the immune functions of CD8⁺ T cells (CD8⁺IFNγ^highPD-1^low; Fig. 6G). In line with the more pronounced protumoral phenotype of RORγ-proficient TAMs from hypercholesterolemic mice (Fig. 5D and F; Supplementary Figs. S6K and S8C–E), SR2211 treatment significantly improved the antitumor efficacy of anti–PD-1 immunotherapy in MN/MCA1 mice, particularly under HCD conditions (Fig. 6H and I). Thus, these findings underscore the significant therapeutic potential of targeting the cholesterol–RORγ axis as an effective strategy to counteract protumoral myelopoiesis.

Figure 6.

RORγ inhibition improves the response to immunotherapy. A, Experimental design: adult wt C3H/HeJ mice were fed with HCD for 8 weeks before being injected with K1735-M2 tumor cells. When the tumor volume reached ∼0.80–1.00 cm³, typically 20–24 days after tumor cell injection, primary tumors were surgically removed. Twenty-eight days after tumor removal, mice were sacrificed for sample collection and analysis. SR2211 was administered to a first group of mice 10 days before tumor removal (SR2211_PRE). A second group of mice received SR2211 treatment 5 days after tumor removal (SR2211_POST). B–G, HCD-fed K1735-M2-bearing wt mice treated with SR2211 or vehicle control: tumor weight (n = 6; B); lung metastatic areas (n = 5). Representative images are shown. Scale bar, 1 mm (C). D–G, FACS quantification of TAMs (n = 4; D), tumor-infiltrating M-MDSCs (n = 4; E), IM and AM frequencies and relative TNFα expression (MFI; n = 5; F), IFNγ, PD-1, and CTLA4 expression (MFI) by lung CD8⁺ T cells (n = 5; G). H and I, tumor growth (volume and weight; H) and lung metastatic areas (I) in NCD- or HCD-fed wt mice treated with vehicle, SR2211, anti–PD-1, or SR2211 plus anti–PD-1 (n = 5). Representative images are shown. Scale bar, 1 mm. One experiment was performed. Data are presented as the mean ± SEM. Statistic by t test (B, D, and E), Kruskal–Wallis (C), Welch ANOVA [F (AM and IM) and I], one-way ANOVA [F (TNFα of IM and AM), G, and H (right)], or two-way ANOVA (H, left).

Discussion

Our work identifies RORC1/RORγ as a key immunometabolic sensor, linking tumor-induced dyslipidemia to myelopoietic alterations that fuel disease progression. Tumor-dependent IL6/IL1β-mediated induction of PCSK9 leads to hepatic LDLR downregulation, elevating circulating cholesterol levels and activating suppressive RORγ-dependent myeloid populations. Cholesterol-induced RORγ activation triggers the transition of CMPs to GMPs, resulting in increased intratumoral accumulation of TAMs and M-MDSCs, as well as of lung M-MDSCs, IMs, and AMs, through a CSF1R- and CCR2-dependent mechanism. Inhibition of these pathways restores specific antitumor immunity. Thus, whereas cholesterol is a fundamental component of phospholipid bilayers, which enables membrane biogenesis and proliferation of cancer cells (61), the activation of RORγ, consequent to lipid dysmetabolism, triggers a protumoral myelopoiesis that establishes immunosuppressive conditions in the host, facilitating both metastatic spread and seeding. Indeed, hypercholesterolemia induces the expression of PD-L1 in TAMs, IMs, and AMs (Figs. 3C, 4E, and F), as well as of PD-1 in CD4⁺ and CD8⁺ lymphocytes (Fig. 3D; Supplementary Fig. S3E), and these effects are driven by RORγ (Fig. 6G; Supplementary Figs. S6K–M and S10S).

The scientific literature on the connection between cholesterol and cancer is full of contradictory observations (62) and the lipid alterations that we observed during tumor development can in principle be further affected by more advanced stages of the disease, as it is known that cachexia promotes an increase in lipolysis (63). Thus, because different types of tumors and stages of progression present important metabolic differences, further studies will need to characterize this connection in other neoplasms.

In this scenario, the observation that statins do not reduce cancer incidence or mortality (64) may be due, at least in part, to the inhibition of cholesterol biosynthesis observed in cancer bearers, as assessed by the reduction of the cholesterol precursors, mevalonate and squalene (Supplementary Fig. S1J), which would suggest PCSK9 inhibition as a preferential cholesterol-lowering treatment in patients with cancer. Furthermore, because PCSK9 can disrupt the recycling of MHC-I by cancer cells (25), its inhibition would offer a double mechanism of immune reactivation, adding to the increased recognition of tumor antigen mediated by MHC-I and the reduced expansion of immunosuppressive myeloid cells.

Furthermore, because we observed that cholesterol downregulation by anti-PCSK9 antibody itself promotes antitumor activity, future evaluation of circulating PCSK9 levels as a predictive indicator of response to therapy, alone or in combination with immune checkpoint inhibitors (ICI), could provide benefits for personalized therapeutic treatments. The latter investigation could also help to clarify the relative contribution offered by inhibition of RORC1/RORγ-mediated protumor myelopoiesis versus increased MHC expression to the antitumor activity of anti-PCSK9 treatment.

The proatherogenic lipid profile that we observe in patients with cancer (total cholesterol^high, LDL^high and HDL^low) is consistent with the observation that more patients with cancer die from cardiovascular disease than from other causes (65), whereas tumor growth associates with subclinical development of atherosclerosis (66). In this scenario, the role of IL1β in PCSK9-mediated cholesterol dysregulation seems in agreement with the dose-dependent reductions in incident and fatal lung cancer (∼75%) observed in patients with previous myocardial infarction (67).

Our work establishes that, whereas tumor progression is fueled by cholesterol-driven protumor myelopoiesis, hypercholesterolemia may favor the efficacy of ICI due to increased levels of PD-L1. Whereas our observations should be reasonably extended and validated in other types of tumors as well, our conclusions are in agreement with recent evidence indicating a positive correlation between overweight and the efficacy of ICIs (49, 68).

In this study, we show that the myeloid-specific deletion of RORγ gene (Rorc^fl/flLyz2-Cre; Fig. 5E; Supplementary Fig. S8A and S8B), as well as adoptive transfer of RORγ-deficient macrophages (Fig. 5F), hinders tumor progression. In addition, we report that pharmacologic inhibition of RORγ reduces tumor growth while improving the antitumor efficacy of anti–PD-1 treatment. Because conflicting results have been reported on the role of the Th17 response in cancer (69), as well as on the contribution of RORγt/RORC2 agonists to ICI efficacy (70, 71), future studies will need to clarify possible differences in the relative influence that cholesterol exerts on myeloid cells versus lymphoid cells. Further investigations should also unravel the precise mechanisms by which cholesterol promotes the RORγ-oriented M2 phenotype of myeloid cells.

Along with the intense efforts to develop strategies targeting suppressor myeloid cells and to recognize the significance of PD-L1 as a biomarker of responsiveness to ICB treatment (72), especially in obese/hypercholesterolemic patients with cancer (49, 73), our findings provide crucial insights into the potential adjuvant role of RORC1/RORγ inhibitors and cholesterol-lowering drugs in anticancer immunotherapy.

Methods

Animals

All mice used in this study were 7 to 18 weeks of age, with both males and females included. C57BL/6J wt (RRID: MGI:3028467) and C3H/HeJ (RRID: IMSR_JAX:000659) mice were obtained from Charles River Laboratories. Pcsk9-deficient mice (B6;129S6-Pcsk9^tm1Jdh/J; RRID: IMSR_JAX:005993) and B6.129P2-Kras^tm4Tyj/J (Kras^LSL-G12D); Trp53^tm1Brn/J (p53^flox/flox; KP; RRID: IMSR_JAX:032435) mice were purchased from The Jackson Laboratory. Rorc mutant mice [B6.129P2(Cg)-Rorc^tm1Litt/J] were kindly provided by Dr. Dan Littman (New York University). Rorc^flox/flox [Rorc^fl/fl, B6(Cg)-Rorc^tm3Litt/J; The Jackson Laboratory; RRID: IMSR_JAX:008771] mice were crossed with Lyz2-Cre [B6.129P2-Lyz2^tm1(cre)Ifo/J; The Jackson Laboratory; RRID: IMSR_JAX:004781] mice to generate Rorc^fl/flLyz2-Cre mice. All colonies were housed and bred in a specific pathogen-free animal facility at the Humanitas Research Hospital (Rozzano, Milan, Italy) in individually ventilated cages. Mice were housed under a 12/12-hour light/dark cycle at an ambient room temperature (RT) of 22°C ± 1°C with 52% to 55% humidity. Mice were randomized based on sex, age, and weight. All procedures involving handling and care of C57BL/6J mice followed approved protocols by the Humanitas Research Hospital and were in compliance with national (Legislative Decree No. 116, G.U., supplement 40, 02-18-1992 and No. 26, G.U., 03-04-2014) and international law and policies (European Economic Community, EEC, Council Directive Nos. 2010/63/EU, OJ L 276/33, and 09-22-2010; NIH Guide for the Care and Use of Laboratory Animals, U.S. National Research Council, 2011). The study received approval from the Italian Ministry of Health (Nos. 97/2014-PR and 25/2018-PR). Every effort was made to minimize the number of animals used and their suffering.

Cell Lines

The following cell lines were used: 3-MCA–derived sarcoma MN/MCA1; B16-F10 (RRID: CVCL_0159) and K1735-M2 (RRID: CVCL_F684) melanoma; LLC (RRID :CVCL_4358); and MC38 colon adenocarcinoma (RRID: CVCL_B288). B16-F10, LLC, Hepa1-6 (RRID: CVCL_0327), and HEK293T (RRID: CVCL_0063) cells were all purchased from the ATCC. K1735-M2 cells were kindly provided by L. Carminati (Laboratory of Tumor Microenvironment, Istituto di Ricerche Farmacologiche Mario Negri IRCCS, Bergamo, Italy). The 3-MCA–derived sarcoma cell line MN/MCA1 was derived from stock periodically renewed through primary cells isolated from tumors implanted in wt mice, as previously described (7). All cell lines were verified to be free from Mycoplasma contamination.

Cancer Models

Mice were injected intramuscularly in their left hind limb with MN/MCA1 cells (1 × 10⁵ per mouse in 100 μL of PBS) or subcutaneously with 100 μL of PBS containing 5 × 10⁵ murine melanoma B16-F10 cells, 1 × 10⁶ MC38 cells, or 1 × 10⁶ LLC cells. When tumors became palpable (∼9–12 days after tumor cell injection), tumor growth was monitored three times per week with a caliper. ED and AD were established based on the metastatic rate in MN/MCA1 fibrosarcoma model, as described in previous experiments (see also Supplementary Fig. S11A; ref. 5). The primary tumor growth of the MN/MCA1 model was then used as reference to define ED or AD in the B16-F10, MC38, and LLC models. For the spontaneous K1735-M2 lung metastasis assay, C3H/HeJ (RRID: IMSR_JAX:000659) mice were subcutaneously injected in the flank with 100 μL of PBS containing 1 × 10⁵ murine melanoma K1735-M2 cells. The growth of the primary tumors was monitored with a caliper, and when the tumors reached 800 to 1,000 mm³, mice were anesthetized, and the tumors surgically removed. For the analysis of spontaneous lung metastasis, all mice were autopsied 28 days after surgery (43). The Kras/p53-driven lung cancer model was generated as previously described (33). Briefly, KP (RRID: IMSR_JAX:032435) mice were intranasally inoculated with 2.5 × 10⁷ infectious particles of an adenoviral vector constitutively expressing the Cre recombinase (Ad5-CMV-Cre) to induce sporadic mutations and lung tumor development. MN/MCA1 mCherry⁺ (1 × 10⁵ cells per mouse in 100 μL of PBS) were injected intramuscularly in the left hind limb of wt or Rorc^−/− (RRID: IMSR_JAX:007571) mice. Blood was collected 20 days after tumor cell injection and analyzed by RT-PCR for mCherry expression and by FACS to quantify circulating mCherry⁺ MN/MCA1 tumor cells. For the B16-F10 lung metastatic model, 3 × 10⁵ tumor cells in 200 μL of PBS were intravenously injected into Rorc^fl/fl/Lyz2-Cre mice (RRID: IMSR_JAX:008771; RRID: IMSR_JAX:004781), and the mice were sacrificed for metastasis evaluation after 14 days. An MC38 liver metastasis model was established on ketamine/xylazine-anesthetized mice whose spleens were exposed to allow direct intrasplenic injections of 3 × 10⁵ luciferase-transduced MC38 cells (MC38-Luc) in 50 μL of PBS. Ten minutes after the injection of tumor cells, spleens were removed using a cauterizer. After 30 days of tumor cell injection, mice were administered by intraperitoneal injection with d-luciferin (150 mg/kg; PerkinElmer) for ex vivo imaging. Eight minutes later, mice were sacrificed, and the livers were collected and acquired using counted using the ex vivo IVIS Lumina III system (PerkinElmer, RRID: SCR_025239). Collected images were analyzed using Living Image 4.3.1 software (PerkinElmer, RRID: SCR_014247): ROIs were drawn around each liver, and radiant efficiency values were calculated.

Patients

To measure cholesterol and PCSK9, human peripheral blood sera were isolated from patients with NSCLC, colorectal cancer, breast cancer, pancreatic ductal adenocarcinoma, biliary tract cancer, and pancreatic neuroendocrine tumor at both early (I-II) and advanced (III-IV) stages according to the AJCC cancer staging systems, eighth edition, at the Humanitas Research Hospital. For the colorectal cancer patient cohort (n = 65), HDL and LDL measurements were performed according to volume sample availability (n = 31). The patients for PCSK9 measurement were randomly selected based on the residual sample volume available. Frozen biopsies of nonneoplastic liver tissues from patients with metastatic colorectal cancer and healthy liver tissues from patients with benign lesions (control; i.e., adenomas, angiomas, and hyperplasia) were provided by Biobanca Humanitas. The samples were selected from non-obese (body mass index < 30) and non-dyslipidemic subjects (Supplementary Table S3). Human peripheral blood leukocytes were isolated from patients with NSCLC at stages III-IV, healthy normocholesterolemic subjects, or hypercholesterolemic patients with dyslipidemia and stable coronary artery disease. Samples from the cohorts described in this study were obtained with written informed consent from the patients. The study was conducted according to the Declaration of Helsinki Principles and following approval by the Ethics Committee of Humanitas Research Hospital, Milan, Italy. More detailed information about the patients, including tumor staging and therapy, are in Supplementary Table S3.

Blood Cholesterol Measurement

Mouse blood samples were collected at the described time points by puncturing the facial vein into microcentrifuge tubes and kept on ice for 20 minutes. They were then centrifuged for 20 minutes at 13,000 rpm at 4°C to separate the serum. Human blood samples were obtained from patients and healthy donors and collected in BD Vacutainer SST tubes. The samples were allowed to clot for 20 minutes at RT and then centrifuged for 15 minutes at 2,000 rpm at RT to separate the serum from the cells. Serum samples were measured for total cholesterol, LDL, and HDL content at the Clinical Analysis Laboratory of Humanitas Hospital. The test assay kits used were purchased from Abbott (Cholesterol 7D62; Direct LDL 1E31-20; and Ultra HDL 3K33-21).

ELISA Measurement

Murine and human IL6 and IL1β were quantified using R&D DuoSet ELISA Development Systems. Human and murine PCSK9 levels were measured using Abcam kits (ab209884 and ab215538). Murine Very Low-Density Lipoprotein (VLDL) was quantified using a Biorbyt kit (orb692924).

Lung Histology

Lungs were collected following intracardiac perfusion with cold PBS, formalin-fixed for 24 hours, dehydrated, and paraffin-embedded for histologic analysis. Histology was performed on two or three longitudinal serial sections spaced 100 to 150 μm apart, with a width of 7 μm, from each lung. The sections were stained with hematoxylin and eosin and scanned using a VS120 Dot-Slide BX61 virtual slide microscope (Olympus Optical; RRID: SCR_018411). The areas of lung lesions were identified by manually tracing the perimeter of lesions using Image Pro-Premiere software 9.2 (Media Cybernetics; RRID: SCR_016497). In graphs showing percentages of lung metastatic areas, each dot value represents the mean area of two contiguous sections obtained from each longitudinal level of a single lung.

BM Transplantation

BM transplantation was performed by intravenously injecting 5 × 10⁶ BM cells into lethally irradiated (two doses of 4.50 Gy each) 8-week-old mice. For Rorc-deficient BM cell transfer, CD45.1 recipient wt mice were transplanted with BM cells from either wt or Rorc^−/− CD45.2 donors. BM engraftment was verified 4 weeks later through FACS analysis (Supplementary Fig. S11B) of blood cells stained with CD45.1 (RRID: AB_492864) and CD45.2 antibodies (RRID: AB_893352; Supplementary Table S4).

TAM Adoptive Transfer

FACS-sorted TAMs (2 × 10⁵) from wt or Rorc^−/− MN/MCA1-bearing mice on NCD were intramuscularly injected together with 1 × 10⁵ MN/MCA1 tumor cells into the hind limb of wt mice fed NCD or HCD (dose 1). After 13 days of growth, when tumors were palpable, a second injection of FACS-sorted wt or Rorc^−/− TAMs (1 × 10⁵) was administered intratumorally. At day 23, mice were sacrificed and further analyses were performed.

In vivo Treatments

To model diet-induced obesity and hypercholesterolemia, 7- to 8-week-old mice were assigned to either a high-fat, high-cholesterol (HCD; Ssniff EF D12079: 21% w/w butterfat, 0.2% w/w cholesterol) or a standard low-fat, low-cholesterol (NCD; Special Diet Service VRF1(P): 5% w/w fat, 0.006% w/w cholesterol) irradiated diet. After 8 weeks, the animals were injected with tumor cells and maintained on their respective dietary regimen throughout the experiments. BODIPY-Ch (TopFluor cholesterol, Avanti Polar Lipids) was dissolved in a mixture of DMSO and corn oil (1:1). Mice were fasted 24 hours before sacrifice and concomitantly administered BODIPY-Ch (250 μg/mouse) using an intragastric syringe. Starting from day 10 after tumor cell injection, MN/MCA1-bearing mice received intraperitoneal treatment with the following reagents: anti–IL6 mAB (Bio X Cell, clone MP5-20F3; RRID: AB_1107709) at a dose of 200 μg twice a week; anti–IL1 (IL1 receptor antagonist, IL1Ra, anakinra, Sobi; RRID: AB_2125559) at a dose of 200 μg twice a week; anti-PCSK9 mAb mAb1 (kindly provided by Amgen Inc; RRID: AB_2921539) at a dose of 200 μg per mouse twice a week (MN/MCA1 tumor) or once a week (KP mice); anti-CSF1R mAb (Bio X Cell, clone AFS98; RRID: AB_2687699) at an initial dose of 400 μg per mouse, followed by twice-weekly doses of 200 μg for the duration of the experiment; anti-CCR2 inhibitor (Tocris) at a dose of 75 μg per mouse twice a week; and the RORγ inverse agonist SR2211 (Tocris) at a dose of 50 μg per mouse twice a week; recombinant IL6 and IL1β cytokines (R&D Systems) were chronically administered at a dose of 500 ng per mouse, once per day, for 20 days. When specified, starting from the day before tumor injection, mice received intraperitoneal injection of 300 μg of anti-mouse CD4 (Bio X Cell, clone GK1.5; RRID: AB_1107636) and 300 μg of anti-mouse CD8 (Bio X Cell, clone 2.43; RRID: AB_1125541) once a week. FACS analysis of peripheral blood samples confirmed the depletion of CD4⁺ and CD8⁺ T cells over a period of 7 days. When specified, MN/MCA1-bearing mice were intraperitoneally injected with 100 μg of anti–PD-1 antibody (Bio X Cell, clone RPM1-14; RRID: AB_10949053), twice a week, starting 10 days after tumor cell injection, either alone or in combination with SR2211 treatment. For the spontaneous metastasis assay, SR2211 (50 μg, intraperitoneally, twice a week) was administered either before or after surgical removal of K1735-M2 primary melanomas or only after surgical removal of primary melanoma (43).

qRT-PCR

Liver tissues from mice and patients, small intestine and adipose tissues from mice, and MN/MCA1 tumor tissues were disrupted by TissueLyser II (QIAGEN; RRID: SCR_018623) with stainless steel beads following the manufacturer’s instructions. Total RNA from tissues, primary hepatocytes, and PECs was extracted using TRIzol reagent (Invitrogen) following the manufacturer’s instructions. Total RNA from FACS-sorted TAMs was extracted using an RNA/DNA isolation kit (Zymo). Complementary DNA was synthesized by reverse transcription using High-Capacity cDNA Archive Kit (Applied Biosystems), and qRT-PCR was performed using SybrGreen PCR Master Mix (Applied Biosystems; RRID: SCR_023358) through ViiA-7 Fast Real-Time System (Applied Biosystems). Data were processed using ViiA 7 Software (Applied Biosystems; RRID: SCR_023358) and analyzed with the 2^(−ΔCT) method. Data were normalized to β-actin or 18S expression and represented as fold change over control. Primer sequences used in the manuscript are available upon request.

Immunoblotting

Liver tissues were disrupted by TissueLyser II (QIAGEN) and lysed in RIPA buffer supplemented with protease and phosphatase inhibitors (Sigma) for 2 minutes at 25 Hz at 4°C, repeated twice. The lysates were centrifuged at 13,000 rpm at 4°C for 15 minutes, and the supernatants were run on SDS–PAGE (30 μg of total proteins per lane) and transferred to polyvinyldifluoride membranes. Immunoblotting was performed using the following antibodies: (primary) rabbit anti-mouse LDLR (Invitrogen; RRID: AB_2809369), mouse anti-mouse PCSK9 (Invitrogen; RRID: AB_2802490), mouse anti-mouse vinculin (Santa Cruz), and goat anti-mouse actin (Santa Cruz) and (secondary) goat anti-mouse and anti-rabbit IgG horseradish peroxidase–conjugated (Amersham). ChemiDoc Touch Imaging System (RRID: SCR_021693) and ImageJ software (NIH, RRID: SCR_003070) were used for image acquisition and densitometric analysis of blots.

Lipid Extraction from Feces and TBA Assay

Fecal lipids were extracted using a modified Folch method, as previously described (74). Briefly, feces were collected from MN/MCA1-bearing mice housed for 4 days (from day 18 to day 22 of tumor growth) or from age-matched TF mice. Three replicates of 1 g of feces per experimental group were homogenized in 5 mL of saline in a 15-mL tube. To this, 5 mL of a chloroform/methanol mixture (2:1) was added and thoroughly mixed. After centrifugation at 400 × g for 10 minutes at RT, the separation into three layers occurred, and the bottom layer, containing the organic phase, was recovered and dried. Lipid extracts were resuspended in 300 μL of ethanol. The solution was analyzed by the commercially available TBA assay kit purchased from Abcam (ab239702) following the manufacturer’s instructions.

Murine Primary Hepatocyte Isolation

Primary hepatocytes were isolated from 12-week-old mice anesthetized and sacrificed by bleeding as previously described (75). Briefly, liver was first perfused with Hank’s balanced salt solution (HBSS) without magnesium or calcium supplemented with 0.5 mmol/L Ethylene Glycol Tetraacetic Acid (EGTA) and then with digestion medium (low-glucose DMEM with 100 U/mL penicillin/streptomycin, 15 mmol/L HEPES, and 0.8 mg/mL of collagenase-IV). The digestion was performed at 37°C for about 7 to 8 minutes, after which the liver was excised, finely minced using forceps into a Petri dish containing digestion medium under sterile conditions, and then filtered through a 70-μm cell strainer. The cells were centrifuged at 1,000 rpm for 5 minutes and washed twice with DMEM:HAM’S F-12 (1:1) with 100 U/mL penicillin/streptomycin and 2 mmol/L glutamine without FBS. Cell viability was assessed by trypan blue exclusion. Viable cells were resuspended in DMEM:HAM’S F-12 (1:1) supplemented with 10% FBS and then seeded in six-well plates (5 × 10⁵ cells/well) that had been precoated the day before with type I collagen (Sigma) dissolved in 0.02 N acetic acid. Cells were incubated for 2 hours at 37°C. After the cells had attached, they were washed and incubated in fresh medium for 1 hour before treating them with the following compounds: TCM (30%), cholesterol (50 μg/mL), or mouse serum (20%). Cells were then incubated for 24 hours.

Metabolomic Analyses of Liver Tissues

Mevalonate quantification was assessed as previously described (76). Briefly, liver samples (15 mg) were homogenized in water, and HCl 6 mol/L was added. The samples were then incubated for 30 minutes in the dark at RT to enable complete lactonization of mevalonate, eluted with methanol, and centrifuged at 4°C at 14,500 × g for 15 minutes. The supernatants were collected and dried using a speed vacuum concentrator. Dried samples were then resuspended in methanol and quantified. For squalene quantification, livers were homogenized in water, and hexane was added. After incubating for 30 minutes at 15 rpm, the supernatants were collected and then dried, concentrated with a speed vacuum concentrator, and finally resuspended in hexane and analyzed. Each sample was spiked with internal standards for data normalization and instrument stability monitoring. Quality control samples containing spiked standards were also acquired. Mevalonate and squalene were quantified using a GCXGC/TOFMS (Leco Corporation). The first-dimension column was a 30 m Rxi-5Sil MS capillary column (Restek Corporation) with an internal diameter of 0.25 mm and a stationary phase film thickness of 0.25 mm. Thiol-gold-dimension of chromatographic column was a 2 m Rxi-17Sil MS (Restek Corporation) with a diameter of 0.25 mm and a film thickness of 0.25 mm. High-purity helium (99.9999%) was used as the carrier gas, with a flow rate of 1.4 mL/minutes. For both molecules, 1 mL of sample was injected in splitless mode at 250°C. Sample analysis was performed using two different temperature programs. For mevalonate, the initial temperature program started at 70°C and increased gradually at a rate of 4°C/minutes until reaching 280°C. For squalene, the initial temperature program began at 150°C and was raised to 250°C at a faster rate of 40°C/minutes, which was maintained for 2 minutes. Then, it was further ramped to 285°C at 5°C/minutes and finally heated to 300°C at a rate of 15°C/minutes. Throughout both analyses, the secondary column was consistently maintained at a temperature 5°C higher than the GC oven temperature of the first column to ensure proper functioning and accurate results. Electron impact ionization (70 eV) was applied, with the ion source temperature set at 250°C. The mass range was 25 to 550 m/z, with an extraction frequency of 32 kHz and acquisition rates of 200 spectra/seconds. The modulation period for the entire run was 4 seconds. The modulator temperature offset was set at +15°C relative to the secondary oven temperature, whereas the transfer line was set at 280°C. The chromatograms were acquired in total ion current mode, and m/z 69 and m/z 58 at 616 and 358 s were selected to identify squalene and mevalonate, respectively. The raw data were processed with ChromaTOF v5.31 (RRID: SCR_023077). Mass spectral assignment was performed by matching data with NIST MS Search 2.3 libraries (RRID: SCR_014668) and FiehnLib. M/z and retention times were also confirmed with standards. External calibration curves and internal standards were used for quantification of mevalonate and squalene.

Lipidomic Analysis

Lipids were extracted from 5 × 10⁵ FACS-sorted TAMs using a 1 mL solution of 75:25 IPA/H₂O after the addition of deuterated lipid standard (Splash Lipidomix). The samples were vortexed and sonicated for 2 minutes and then incubated for 30 minutes at 4°C under gentle and constant shaking. To remove debris and other impurities, the samples were centrifuged for 10 minutes at 3,500 × g at 4°C. Subsequently, 1 mL of the supernatant was dried using a speed vacuum concentrator and then reconstituted in 100 μL of MeOH containing the internal standard CUDA (12.5 ng/mL). The reconstituted lipids were analyzed by UHPLC Vanquish system (RRID: SCR_025713) coupled with Orbitrap Q-Exactive Plus (Thermo Fisher Scientific; RRID: SCR_020556). A reverse phase column was used for lipid separation (Hypersil Gold 150 × 2.1 mm, particle size 1.9 μm), with the column maintained at 45°C at a flow rate of 0.260 mL/minutes. Mobile phases and mass spectrometry parameters were set as previously reported (77). The acquired raw data from the untargeted analysis were processed using MSDIAL software version 4.24 (Yokohama City; RRID: SCR_023076). This involved the detection of peaks, MS2 data deconvolution, compound identification, and the alignment of peaks across all samples. To obtain an estimated concentration expressed in μg/mL, the normalized areas were multiplied by the concentration of the internal standard. An in-house library of standards was also used for lipid identification. Statistical analysis was performed using MetaboAnalyst 5.0 software (www.metaboanalyst.org; RRID: SCR_015539) and GraphPad Prism v.8 (RRID: SCR_002798).

FACS Analysis of Murine Samples

Primary tumors and explanted lungs were cut into small pieces, disaggregated with 0.5 mg/mL collagenase-IV and 150 IU/mL DNase-I in RPMI 1640 for 30 minutes at 37°C, and filtered through a 70-μm strainer. Splenocytes were collected from spleens after disaggregation and filtration through a 70-μm strainer. BM cells were isolated from the tibias and femurs of both TF and TB mice. Whole blood samples were collected from the facial vein or heart in EDTA-coated collection tubes and directly processed using a standardized red blood cell lysis protocol. The resulting cells were resuspended in HBSS supplemented with 0.5% FBS. Staining was performed at 4°C for 20 minutes using the antibodies listed in Supplementary Table S4. In addition, LipidTOX Green Neutral Lipid Stain (1:200, Invitrogen) was used after fixing the cells with paraformaldehyde (PFA) 1% for 30 minutes. LipidTOX staining was performed in PBS for 30 minutes up to 2 hours at RT. Cell viability was determined using either Zombie Aqua Fixable Viability Kit (1:800, BioLegend) or LIVE/DEAD Fixable Violet Dead Cell Stain Kit (1:1,000); negative cells were considered viable. For intracellular staining of TNFα, iNOS, CD206, IDO1, IFNγ, RORγ, PCSK9, and LDLR, a Foxp3/Transcription Factor Staining Buffer Set (eBioscience) was used. Expression of TNFα and IFNγ was analyzed by flow cytometry following 4 hours of treatment with brefeldin A (5 μg/mL), phorbol 12-myristate 13-acetate (50 ng/mL), and ionomycin (1 μg/mL). PCSK9 mouse and LDLR rabbit mAbs were followed by incubation with secondary goat anti-mouse Alexa Fluor 647–conjugated (RRID: AB_2535804) and goat anti-rabbit Alexa Fluor 488–conjugated antibodies (RRID: AB_143165; Thermo Fisher Scientific), respectively. Cell detection was performed using either BD FACSCanto II (RRID: SCR_018056), BD LSRFortessa (RRID: SCR_018655), or BD FACSymphony (RRID: SCR_022538), and data were analyzed with FlowJo 9.9.6 software (RRID: SCR_008520). Gating strategies for myeloid cells are shown in Supplementary Fig. S11C and S11D. The results are reported as the mean fluorescence intensity (MFI) or MFI normalized to isotype control or fluorescence minus one (ΔMFI). To assess BODIPY-Ch fluorescence in tissues or cells, flow cytometry (AF488 channel) was performed, and the relative MFI was normalized to the biological vehicle control (ΔMFI). For t-distributed stochastic neighbor embedding analysis (78), a unique computational barcode was assigned to single samples. Events gated on live CD45⁺ cells were subsequently concatenated and visualized with t-distributed stochastic neighbor embedding visualization (Barnes–Hut implementation; iterations, 1,000; perplexity, 20; initialization, deterministic; theta, 0.5; and eta: 200) was carried out. The expression of the following markers was analyzed: CD11b, CD11c, CD103, CD64, Ly6C, Ly6G, and F4/80.

FACS Analysis of Human PBMC Samples

Human PBMCs were obtained through Ficoll density gradient centrifugation (Ficoll-Paque PLUS, GE Healthcare). Multicolor staining for cell populations was performed on cryopreserved samples. Cells were thawed in RPMI medium supplemented with 10% FBS, incubated at 37°C for 2 hours, and then washed and resuspended in staining buffer (HBSS with 0.5% FBS). Staining was performed at 4°C for 20 minutes using the antibodies listed in Supplementary Table S5. Cell viability was determined using LIVE/DEAD Fixable Near-IR Dead Cell Stain Kit (1:1,000, Thermo Fisher Scientific), in which negative cells were considered viable. A Foxp3/Transcription Factor Staining Buffer Set (eBioscience) was used for intracellular staining of RORγ. Cells were detected using BD LSR Fortessa and analyzed with FlowJo 9.9.6 software. The gating strategy is shown in Supplementary Fig. S11E.

Purification of Mouse Leukocytes

PECs were harvested by peritoneal lavage from mice injected with 1 mL of 3% (weight/vol) hioglycolate medium (Difco) 5 days before isolation as previously described (40). PECs were cultured in RPMI 1640 medium containing 10% FBS, 2 mmol/L glutamine, and 100 U/mL penicillin/streptomycin, with the addition of TCM (30%) with or without cholesterol (50 μg/mL) for 48 hours. Isolation of TAMs was performed as previously described (40). Briefly, 18 to 21 days after tumor cell injection, primary tumors were cut into small pieces, disaggregated with 0.5 mg/mL collagenase-IV and 150 U/mL DNase I in RPMI 1640 for 30 minutes at 37°C, and filtered through a strainer. The cell suspension was enriched in CD11b⁺ cells through positive selection using CD11b microbeads (MACS, Miltenyi Biotec). The purity of CD11b⁺ cells was ∼90% as determined by FACS. To obtain high-purity TAM populations, CD11b⁺ cells were further stained (Live/Dead-Pacific Blue, CD45-FITC, CD11b-PerCP-Cy5.5, Ly6C-PE-Cy7, Ly6G-APC, and F4/80-PE) and sorted using a FACSAria cell sorter (BD Bioscience). Myeloid progenitor cells (CMPs and GMPs) were isolated from the BM of MN/MCA1-bearing mice after 23 days of tumor growth. Briefly, BM cells were flushed from femurs and tibias of different mice and then pooled for each experimental group. To obtain high-purity CMP and GMP populations, after red blood cell lysis, the cells were stained (Live/Dead-APC-Cy7, cKit-PE-Cy7, Lin-eFluor450, Sca1-BV711, CD34-FITC, and CD16/32-APC) and sorted on a FACSAria cell sorter. The purity of each sorted population was ≥95%. The resulting sorted cells were processed for adoptive transfer, lipid immunofluorescence staining, or mRNA extraction. BM cells were collected from femurs and tibias of healthy TF C57BL/6J mice (RRID: MGI:3028467) and differentiated for 6 days into BMDMs in RPMI 1640 medium (10% FBS, 2 mmol/L glutamine, and 100 U/mL penicillin/streptomycin) supplemented with 30 ng/mL of M-CSF (plus cholesterol at 50 μg/mL, when described). After 3 days of culturing in the M-CSF–supplemented medium, BMDMs underwent medium refreshment. Following a total culturing period of 6 days, cells were stimulated with IL4 (20 ng/mL) for 20 hours, or with TCM (30%) for 48 hours after which they were analyzed as described above.

Neutral Lipid and cholesterol Immunofluorescence

FACS-sorted TAMs, CMPs, or GMPs were seeded on poly-L-lysine (Sigma-Aldrich)–coated sterile rounded glass at a density of 2 × 10⁵ cells/mL in RPMI medium and incubated for 2 hours at 37°C. Cells were then fixed with 4% PFA for 10 minutes at RT. Cells were washed twice with PBS and stained with fluorescent lipid staining reagents: LipidTOX Green (1:500, Invitrogen), for total neutral lipids, or Filipin-III solution (1:20, Sigma-Aldrich), for cellular cholesterol. The staining was performed for 1 hour at RT. For LipidTOX-stained cells, nuclei were counterstained with DAPI (Invitrogen), whereas for Filipin-III–stained cells, nuclei were visualized with SYTO Green (Invitrogen). Coverslips were mounted using the antifade medium FluorPreserve Reagent (EMD Millipore) and analyzed using an SP8 Laser Confocal Microscope (Leica; RRID: SCR_024563) equipped with fine-focus oil immersion lens (×4/1.3 numerical aperture) and operated with lasers with 405 and 488 excitations. The intensity of fluorescence was analyzed using LAS X software (Leica; RRID: SCR_013673).

Cell Treatments

To obtain MN/MCA1 TCM, digested MN/MCA1 tumors from wt mice were seeded at a density of 4 × 10⁶ cell/mL in RMPI 1640 supplemented with 10% FBS, 2 mmol/L glutamine, and 100 U/mL penicillin/streptomycin. After 24 hours of incubation, the supernatant was collected and filtered through a 0.2-μm filter. For the preparation of mouse serum for cell treatment, mice were anesthetized and sacrificed by bleeding. Blood was collected with a syringe via heart puncture to maintain sterility. The collected blood was centrifuged for 20 minutes at 13,000 rpm, 4°C, to separate the serum, which was then heat-inactivated for 15 minutes at 55°C and filtered as above. For cell stimulation, the following compounds and murine recombinant cytokines were used: cholesterol, desmosterol, LDL cholesterol, and the RORγ agonist SR0987 (all purchased from Sigma-Aldrich), as well as IL6, IL1β, IL4, M-CSF, and CCL2 (all purchased from Peprotech). Hepa1-6 cells were treated for 24 hours with TCM (30%), IL6 (10 ng/mL), or IL1β (10 pg/mL), along with anti–IL6 mAb (300 ng/mL) and anti–IL1 (300 ng/mL) where applicable. In vitro administration of BODIPY-Ch (0.5 μg/mL) was performed 3 hours before cell collection. Subsequently, the cells were fixed with PFA 1% and subjected to FACS analysis. MN/MCA1 cells were treated in vitro with cholesterol (50 μg/mL) and/or anti-PCSK9 mAb (300 ng/mL). The cell cycle was then assessed by DAPI staining. Briefly, cells were washed with PBS, fixed with 80% ethanol, and permeabilized for 10 minutes with 0.1% Triton X-100 (PBS). Nuclei were stained with 1 μg/mL DAPI (PBS) for 3 minutes. Alternatively, cell viability was measured by Annexin V-FITC/PI-staining. Samples were analyzed by flow cytometry.

RNA-seq and Analysis

Total RNA was purified from BMDMs treated as described above using the Zymo Research kit (No. R2050) according to the manufacturer’s instruction. The quality of the RNA was assessed using a High Sensitivity RNA ScreenTape Assay with a 4200 TapeStation System (Agilent). mRNA-seq library preparation was performed using SMART-Seq v4 PLUS Kit (R4000752, Takara Bio). Single-end multiplexed libraries were sequenced using the NextSeq 2000 instrument (Illumina), resulting in approximately 76 ± 14 million reads/sample. The 75-bp single-end reads were aligned to the GENCODE Mus musculus reference genome (build GRCm38/mm10) using STAR v2.7.2b (RRID: SCR_004463; ref. 79). Raw read counts were normalized with TMM implemented in edgeR (80), and low-expressed genes were filtered out with the filterByExpr function (minute count = 19). Differential expression analysis of read counts was performed using voom, lmFit, and eBayes (robust = T) function of limma v.3.46 (RRID: SCR_010943; ref. 81) package in R. Significant differential genes were chosen based on an FDR < 0.05 and |log₂FC| > 0.5. Hierarchical clustering of significantly modulated genes was performed using the hclust (RRID: SCR_009154) and dist R functions on log₂ CPM. Clustering was performed with the ward.D method and Pearson’s correlation distance to generate a heatmap using pheatmap (82), with row scaling applied. Principal components analysis was performed using principal component analysis function of FactoMineR v2.7 (RRID: SCR_014602; ref. 83) package in R. The UpSet plots were obtained using the R package UpSetR (84). Overrepresentation analysis was performed using enrichR (85) with the “MsigDB Hallmark 2020” gene set library, and the results were visualized using the “ggplot2” package.

T-Cell Suppression Assay

Splenocytes from wt mice were labeled with 3 μg/mL of CellTrace CFSE (Life Technology) for 10 minutes at 37°C protected from light. Subsequently, cells were washed, resuspended in RPMI 1640 complete medium, and seeded on wells coated with anti-CD3 (2 μg/mL; RRID: AB_468851) and anti-CD28 antibodies (3 μg/mL). Myeloid cells (i.e., TAMs and M-MDSCs FACS-sorted, and BMDMs) were cocultured with CFSE-labeled splenocytes at different ratios of splenocyte:myeloid cells. After 72 hours of coculture, cells were collected, stained with anti-CD4 (RRID: AB_312697) and anti-CD8 (RRID: AB_312760) antibodies, and analyzed by flow cytometry. CellTrace signal from anti-CD3/CD28–activated lymphocytes, which were not cocultured with myeloid cells, was used to evaluate cell proliferation.

Plasmid Transfection and Luciferase Assay

A plasmid vector constitutively expressing the Rorc gene under the control of the CMV promoter (pCMV6-Rorc; OriGene, No. MR222309) was cotransfected with a plasmid expressing the luciferase gene (Luc) under the control of the minimal promoter of the Il17a gene (minIL17prom-Luc; Addgene, No. 20124; RRID: Addgene_20124). Transfections were performed using calcium phosphate method. After 18 hours of transfection, HEK293T cells were washed, fresh medium was added, and the cells were allowed to rest for 1 hour before being treated with SR0987 (20 μmol/L), desmosterol (20 μmol/L), cholesterol (20 μmol/L), LDL (50 μg/mL), mouse serum (20%), TCM (30%), or SR2211 (20 μmol/L). After 24 hours, the cells were collected and subjected to luciferase assay following the manufacturer’s instructions (Promega, E1500). Luminescence was measured through Synergy 2 (BioTek).

PEC Migration

PECs were seeded for 18 hours in low-adherence plates in RPMI 1640 with 10% FBS, 2 mmol/L glutamine, and 100 U/mL penicillin/streptomycin, with or without the supplementation of cholesterol (50 μg/mL). Cells were then collected, washed, resuspended in RPMI 1640 medium with 1% FBS, and seeded (2 × 10⁵) on the upper chamber of a 5-μm-pore transwell insert in 24-well plates, whereas the bottom chamber was filled with RPMI 1640 with 10% FBS in the presence or absence of M-CSF (50 ng/mL), CCL2 (50 ng/mL), and TCM (30%), with or without cholesterol (50 μg/mL). After 5 hours of incubation, the chambers were fixed and stained with Diff-Quik (Baxter).

Statistical Information

Statistical analysis was performed using GraphPad Prism v.8.2.1 software. P values below 0.05 were considered statistically significant: * P ≤ 0.05, ** P ≤ 0.01, *** P ≤ 0.001, **** P ≤ 0.0001. The results are represented as the mean ± SEM, unless otherwise stated. Statistical significance is indicated in each graph except for the heatmap graphs, for which the statistical analysis is provided in Supplementary Table S1. For the experiments comparing two independent grouping variables, two-way ANOVA was applied, and Šidák or Tukey multiple comparisons tests were performed based on the nature of the comparison. Data with more than two independent groups were analyzed using one-way ANOVA, Kruskal–Wallis, or Brown–Foresythe and Welch (refereed as Welch) ANOVA test with Holm–Šidák, Tukey, or Dunn multiple comparisons, depending on the specific experimental design, data distribution (Shapiro–Wilk test) or variance (Bartlett test or F-test). When comparing only two groups, an unpaired t test or Mann–Whitney test was applied, depending on the nature of data distribution. RNA-seq analysis is described in the corresponding method section above. The number of biological replicates and the statistical methods used are specified in all figure legends.

Data Availability

RNA-seq data supporting the findings of this study are available from the NCBI BioProject database (https://www.ncbi.nlm.nih.gov/bioproject/; RRID: SCR_004801) under accession code PRJNA1000461.

Supplementary Material

Table S1

Statistical analysis relative to the heatmap graphs of Figures 2C, S1A, S1E and S1L

Table S2

Statistical analysis of flow cytometry data of blood PBMC from NSCLC patients. Related to Figure 5B

Table S3

Detailed information about the patients used in the study.

Table S4

Murine antibodies for flow cytometry analysis

Table S5

Human antibodies for flow cytometry analysis

Figure S1

Hypercholesterolemia exacerbates cancer-induced alterations in cholesterol metabolism

Figure S2

Hypercholesterolemia exacerbates cancer progression and shapes the immunosuppressive myeloid compartment

Figure S3

Cholesterol levels modulate cancer-driven myeloid immunosuppression

Figure S4

Hypercholesterolemia induces bone marrow myelopoiesis, CMPs to GMPs transition, and CSF-1/CSF-1R and CCL2/CCR2 axes

Figure S5

IL-6/IL-1β-PCSK9-cholesterol axis regulates RORγ expression in myeloid cells

Figure S6

RORγ couples hypercholesterolemia to protumoral myelopoiesis

Figure S7

RORγ deficiency limits the effects of cholesterol on both immunosuppressive activity and myeloid cell maturation

Figure S8

Myeloid-specific genetic deletion of RORγ regulates tumor-promoting immunosuppressive myelopoiesis

Figure S9

RORγ acts as a cholesterol sensor and modulator of myeloid cell immunosuppressive functions

Figure S10

Blockade of cholesterol-RORγ axis restrains tumor progression and myeloid cell immunosuppressive functions

Figure S11

Flow cytometry gating strategis. Related to Methods

Authors’ Disclosures

No disclosures were reported.

Authors’ Contributions

A. Bleve: Conceptualization, data curation, formal analysis, validation, investigation, methodology, writing–original draft, writing–review and editing. M. Incerti: Data curation, formal analysis, investigation, methodology, writing–review and editing. F.M. Consonni: Investigation, methodology. V. Garlatti: Investigation, methodology. G. Ballerini: Investigation, methodology. C. Pandolfo: Investigation, methodology. M.N. Monari: Formal analysis, investigation, methodology. S. Serio: Data curation, software, formal analysis, bioinformatic analysis. D. Pistillo: Resources. M. Sironi: Methodology. C. Alì: Methodology. M. Manfredi: Methodology. E. Barberis: Methodology. G. Finocchiaro: Resources, data curation, formal analysis, investigation. M.A. Cassatella: Writing–review and editing. C. Panico: Resources, visualization. G. Condorelli: Resources, visualization. A. Sica: Conceptualization, resources, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing.