Hypoxia activates SREBP2 through Golgi disassembly in bone marrow‐derived monocytes for enhanced tumor growth

Abstract

Bone marrow‐derived cells (BMDCs) infiltrate hypoxic tumors at a pre‐angiogenic state and differentiate into mature macrophages, thereby inducing pro‐tumorigenic immunity. A critical factor regulating this differentiation is activation of SREBP2—a well‐known transcription factor participating in tumorigenesis progression—through unknown cellular mechanisms. Here, we show that hypoxia‐induced Golgi disassembly and Golgi‐ER fusion in monocytic myeloid cells result in nuclear translocation and activation of SREBP2 in a SCAP‐independent manner. Notably, hypoxia‐induced SREBP2 activation was only observed in an immature lineage of bone marrow‐derived cells. Single‐cell RNA‐seq analysis revealed that SREBP2‐mediated cholesterol biosynthesis was upregulated in HSCs and monocytes but not in macrophages in the hypoxic bone marrow niche. Moreover, inhibition of cholesterol biosynthesis impaired tumor growth through suppression of pro‐tumorigenic immunity and angiogenesis. Thus, our findings indicate that Golgi‐ER fusion regulates SREBP2‐mediated metabolic alteration in lineage‐specific BMDCs under hypoxia for tumor progression.

Low‐oxygen conditions induce lineage‐specific activation of SREBP2 via Golgi‐ER fusion, promoting lipid biosynthesis, myeloid cell infiltration, and tumorigenesis.

Introduction

Bone marrow‐derived cells (BMDCs) have been shown to infiltrate into and differentiate within hypoxic tumor regions, thereby stimulating angiogenesis, matrix organization, and pro‐tumorigenic immunity (Coussens et al, 2000; Ferrara & Kerbel, 2005; Shibuya & Claesson‐Welsh, 2006; Shojaei et al, 2007; Du et al, 2008; Ostrand‐Rosenberg, 2008). The hypoxia‐inducible factor 1α (HIF1α) is a master regulator of hypoxia‐induced transcriptional regulation (Semenza, 2000; Ke & Costa, 2006) and function by regulating glycolysis and angiogenetic immunity (Semenza, 2003) during cellular adaptation to an anaerobic microenvironment (Gatenby & Gillies, 2004; Trédan et al, 2007). Although the role of HIF1α in metabolic/immune reprogramming for tumor progression has been extensively investigated (Semenza, 2010; Palazon et al, 2017), the role of HIF1α‐dependent as well as HIF1α‐independent organelle‐mediated regulation in hypoxic maturation of BMDCs during tumor progression remains unclear.

SREBP2 is an essential regulator of cholesterol biosynthesis (Brown & Goldstein, 1997) and is activated by intracellular sterol deficiency. SREBP2 translocates from the endoplasmic reticulum to Golgi apparatus (Yang et al, 2002) and is cleaved for nuclear translocation. SREBP2 translocation preferentially leads to cholesterol biosynthesis, while SREBP1 promotes unsaturated fatty acid and triglyceride biosynthesis (Horton et al, 1998). SREBP1 and SREBP2 regulate tumor growth through membrane biosynthesis (Menendez & Lupu, 2007), fatty acid metabolism (Kim et al, 1998), cholesterol biosynthesis, and acetate switching (Kondo et al, 2017). Although cholesterol biosynthesis is essential for the regulation of differentiation process in multiple cell types such as osteoclasts and enterocytes (Herold et al, 1995; Parhami et al, 2002), it is still unknown how SREBP2 is activated for the differentiation of BMDCs in tumor progression. In this study, we elucidate the lineage‐specific activation of SREBP2 in BMDCs via Golgi disassembly in a hypoxic tumor microenvironment.

Results

Cholesterol biosynthesis pathway is upregulated in monocyte/macrophage lineage cells under hypoxia

We observed infiltration of bone marrow‐derived myeloid cells such as monocytes, macrophages, and neutrophils (CD11b⁺, F4/80⁺, Gr1⁺, respectively) into hypoxic tumor regions in a pre‐angiogenic state (Fig 1A), wherein pro‐tumorigenic maturation and immune activation occurs (Coussens et al, 2000; Ferrara & Kerbel, 2005; Shibuya & Claesson‐Welsh, 2006; Shojaei et al, 2007; Du et al, 2008; Ostrand‐Rosenberg, 2008; Muramatsu et al, 2010; Osawa et al, 2011, 2013). Hypoxia triggered different transcriptional gene expressions in monocyte/macrophage lineage cells, cancer cells, and fibroblasts, which are three major types of cells in tumor (Figs 1B and EV1A, Dataset EV1). In consistency with previous reports, expression levels of HIF1α target genes involved in angiogenesis and glycolysis, including LDHA, PDK1, and VEGF, were upregulated in all monocyte/macrophage lineage cells, cancer cells, and fibroblasts under hypoxia, suggesting that HIF1α is one of the key transcriptional regulators in all three cell types activated in a hypoxic environment (Fig EV1B) (Semenza, 2000; Ke & Costa, 2006). However, hypoxia only showed a minor effect on cellular growth of all monocyte/macrophage lineage cells, cancer cells, and fibroblast cells (Fig EV1C). We also examined whether hypoxia differentially affects cell cycle in cancer cells, fibroblasts, and monocytes/macrophages. Consistently, we found that hypoxia did not differentially influence the cell cycle in cancer cells, fibroblasts, and monocytes/macrophages (Fig EV1D).

Figure 1. Hypoxia triggers cholesterol biosynthesis in bone marrow‐derived monocyte/macrophage lineage cells.

• A
  Left, immunostaining of pre‐/post‐angiogenic allograft tumor of HSML cells. Endothelial cells indicated by CD31, myeloid cells by CD11b, macrophages by F4/80, and granulocytes by Gr‐1. Nucleus indicated by DAPI. Scale bars, 100 μm. Right, quantification of cell number or CD31⁺ area. Data represent mean ± SEM of n = 3 biological replicates.
• B
  Venn diagram of hypoxia‐regulated genes in monocyte/macrophage lineage, cancer and fibroblast cells, up‐ or downregulated over twofold as measured by DNA microarray.
• C
  Pathway analysis of twofold upregulated genes under hypoxic conditions measured by DNA microarray.
• D
  Schematics of cholesterol biosynthesis pathway. Enzymes upregulated in hypoxic THP‐1 cells shown in red.
• E
  mRNA expression level of cholesterol biosynthetic enzymes in THP‐1 cells. Data represent mean ± SEM of n = 3 biological replicates.
• F, G
  Relative percentage inputs of histone, (F) methylation (H3K4me3), and (G) acetylation (H3K27Ac), on promoter and enhancer regions of HMGCS1, IDI1, MSMO1, and LSS in THP‐1 cells. Data represent mean ± SEM of n = 3 biological replicates.
• H
  Total cellular cholesterol content in THP‐1 cells. Data represent mean ± SEM of n = 3 biological replicates.

Data information: Indicated P‐values were obtained using (A) one‐way analysis of variance (ANOVA) followed by Student–Newman–Keuls multiple comparisons for post hoc test or (E–H) paired Student's t‐tests. *P < 0.05, **P < 0.01, ***P < 0.001.

See also Fig EV1A–K.

Source data are available online for this figure.

Figure EV1. Hypoxia triggers cholesterol biosynthesis in bone marrow‐derived monocyte/macrophage lineage cells but not cancer or fibroblast cells.

1. Heatmap representation of differentially expressed genes in monocyte/macrophage lineage cells, cancer cells, and fibroblast cells under hypoxic conditions. Gene expression was measured by DNA microarray. Fold change of gene expression is indicated by color bar. Representative genes from the commonly and differentially regulated sections were listed.
2. mRNA expression levels of PDK1, LDHA, and VEGF in response to 24 h of control or hypoxic condition in monocyte/macrophage lineage (THP‐1), cancer (HeLa), and fibroblast (NHDF) cells. Data represent mean ± SEM of n = 3 biological replicates.
3. Relative cell growth of THP‐1, HeLa, and NHDF during 72 h of culture under control or hypoxic condition. Cell growth is assessed every 24 h by sulforhodamine (SRB) assay. Data represent mean ± SEM of n = 3 biological replicates.
4. Cell cycle analysis of THP‐1, HeLa, and NHDF under 24 h of control or hypoxic condition.
5. Pathway analysis of twofold downregulated genes in monocyte/macrophage lineage, cancer, and fibroblast cells under hypoxic conditions as measured by DNA microarray.
6. mRNA expression levels of HMGCS1 and LSS in response to 24 h of control or hypoxic condition in HeLa and NHDF cells. Data represent mean ± SEM of n = 3 biological replicates.
7. mRNA expression levels of HMGCS1, MSMO1, and LSS in response to 24 h of control or hypoxic condition in human acute lymphoblastoid cell line (BALL‐1). Data represent mean ± SEM of n = 3 biological replicates.
8. Relative % input of histone methylation (H3K4me3) on promoter regions of HMGCS1, MSMO1 and LSS measured by ChIP‐seq in human acute lymphoblastoid cell line (BALL‐1) with 24 h of exposure to hypoxic conditions. Data represent mean ± SEM of n = 3 biological replicates.
9. Relative % input of histone acetylation (H3K27Ac) on enhancer regions of HMGCS1, MSMO1, and LSS measured by ChIP‐seq in human acute lymphoblastoid cell line (BALL‐1) with 24 h of exposure to hypoxic conditions. Data represent mean ± SEM of n = 3 biological replicates.
10. Validation of HIF1α‐siRNA‐mediated effect in THP‐1 cells under control and hypoxic conditions. Data represent mean ± SEM of n = 3 biological replicates.
11. Effect of HIF1α‐siRNA on hypoxia‐induced expression of HIF1α target genes (PDK1, LDHA) and cholesterol biosynthetic genes (HMGCS1, HMGCR, IDI1, and LSS) in response to 24 h of control or hypoxic condition in THP‐1 cells. Data represent mean ± SEM of n = 3 biological replicates.

Data information: Indicated P‐values were obtained using paired Student's t‐tests. *P < 0.05, **P < 0.01, ***P < 0.001.

Source data are available online for this figure.

Although HIF1α and glycolytic pathways were universally upregulated in all three types of cells, cholesterol biosynthesis‐related pathways showed upregulation only in monocyte/macrophage lineage cells under hypoxia (Figs 1C and EV1E). To confirm the results of comprehensive differential gene expression analysis, we examined the mRNA expression level of each enzyme in the cholesterol biosynthesis pathway (Fig 1D) by quantitative PCR. The expression levels of HMGCS1, IDI1, LSS, MSMO1, and most other cholesterol biosynthetic enzymes were highly upregulated under hypoxia in monocyte/macrophage lineage cells (Fig 1E). However, mRNA expression of cholesterol biosynthetic enzymes such as HMGCS1 and LSS was not increased in cancer or fibroblast cells under hypoxia (Fig EV1F), indicating that hypoxia‐induced upregulation of cholesterol biosynthesis pathway was a unique response of monocyte/macrophage lineage cells.

To elucidate epigenetic modification of hypoxia‐responsive genes, we performed chromatin immunoprecipitation (ChIP) PCR analysis to investigate histone methylation (H3K4me3) and acetylation (H3K27Ac) on active promoters and enhancers, respectively. We observed a significantly increased enrichment of H3K4me3 (Fig 1F) and H3K27Ac (Fig 1G) on HMGCS1, IDI1, MSMO1, and LSS in monocyte/macrophage lineage cells, as well as in immune cells originating from another lineage (Fig EV1G–I), suggesting that cholesterol biosynthetic genes are activated under hypoxia via epigenetic regulation.

To investigate whether HIF1ɑ is involved in hypoxia‐induced cholesterol biosynthesis, we silenced HIF1ɑ in monocyte/macrophage lineage cells using siRNA (Fig EV1J). Silencing of HIF1ɑ caused significant downregulation of HIF1ɑ target gene LDHA and PDK1 but did not cancel hypoxia‐induced upregulation of genes in the cholesterol synthesis pathway, indicating that HIF1ɑ is not involved in the transcriptional activation of hypoxia‐induced cholesterol biosynthesis (Fig EV1K). In addition, cellular cholesterol level in monocyte/macrophage lineage cells increased under hypoxic treatment, which was then recovered after 24 h of normoxia (Fig 1H). Together, these results indicate that in response to hypoxia, cholesterol biosynthetic genes in monocyte/macrophage lineage cells, but not in cancer cells and fibroblasts, are transcriptionally activated via histone modifications of promoter/enhancer regions.

Hypoxia‐induced Golgi disassembly activates SREBP2 for cholesterol biosynthesis in monocytic myeloid cells

Cholesterol biosynthetic genes are well known to be regulated by either SREBP1 or SREBP2 (Brown & Goldstein, 1997; Madison, 2016). To examine whether hypoxia‐induced upregulation of cholesterol biosynthesis was mediated by SREBP1 or SREBP2, we measured the mRNA expression levels of cholesterol biosynthetic enzymes in monocyte/macrophage lineage cells where SREBP1 or SREBP2 was silenced using two different siRNAs. We found that silencing of SREBP2, but not SREBP1, offset the hypoxia‐induced upregulation of HMGCR, HMGCS1, and LSS in monocytic myeloid cells (Fig 2A and B), suggesting that SREBP2 is involved in the regulation of hypoxia‐induced cholesterol biosynthesis.

Figure 2. Hypoxia‐induced Golgi disassembly activates SREBP2 for cholesterol biosynthesis in monocytic myeloid cells.

• A
  Validation of SREBP2 and SREBP1 siRNA‐mediated effect in THP‐1 cells. siRNA knockdown was performed in THP‐1 cells under both normoxia and hypoxia. Data represent mean ± SEM of n = 3 biological replicates.
• B
  Effect of siRNA‐SREBP2 on hypoxia‐induced expression of HMGCR, HMGCS1, and LSS mRNA in THP‐1 cells. siRNA knockdown was performed in THP‐1 cells under both normoxia and hypoxia. Data represent mean ± SEM of n = 3 biological replicates.
• C
  Venn diagram of genes up‐ or downregulated over twofold and pathway analysis on genes upregulated under hypoxia and downregulated after SREBP2 silencing in THP‐1 cells.
• D, E
  Active form (N) and inactive precursor form (P) of SREBP2 in nuclear extract (Nuc. Ext.) and membrane (Membrane) fractions of THP‐1 cells under (D) control (C), hypoxic (H) conditions, and (E) cholesterol‐deprived (Induced: Ind.) and cholesterol‐rich (Suppressed: Supp.) conditions.
• F
  Hypoxia‐induced binding of SREBP2 on the promoter regions of HMGCS1, IDI1, LSS, and MSMO1 in THP‐1 cells. Data represent mean ± SEM of n = 3 biological replicates.
• G
  Effect of cholesterol (10 μg/ml)/25OH‐cholesterol (1 μg/ml) treatment and hypoxia on SREBP2 target genes HMGCR and HMGCS1 in THP‐1 cells. Data represent mean ± SEM of n = 3 biological replicates.
• H
  Immunofluorescent staining of TGN46—a Golgi marker (green), KDEL—an ER marker (red), and DAPI (blue) for nucleus in THP‐1 cells under control and hypoxic conditions. All scale bars, 5 μm. Quantification data represent mean ± SEM of n = 24 biological replicates.
• I
  Proximity ligation assay (PLA) staining of TGN46 (Golgi), KDEL (ER), and DAPI (nucleus) in THP‐1 cells under control and hypoxic conditions. All scale bars, 20 μm. Quantification data represent mean ± SEM of n = 30 biological replicates.
• J
  Immunofluorescent staining of TGN46 (Golgi), SREBP2, and DAPI (nucleus) in THP‐1 cells under control and hypoxic conditions. All scale bars, 5 μm. Quantification data represent mean ± SEM of n = 32 biological replicates.

Data information: Indicated P‐values were obtained using (A, B, F and G) paired Student's t‐tests or (H–J) one‐way analysis of variance (ANOVA) followed by Student–Newman–Keuls multiple comparisons for post hoc test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

See also Figs EV2A–I and EV3A–H.

Source data are available online for this figure.

To further investigate the roles of SREBP2 in monocytic myeloid cells under hypoxia, we performed transcriptomic analysis after gene silencing of SREBP2 (si‐SREBP2) (Fig EV2A). By comparing the differential gene expression patterns of monocyte/macrophage lineage cells exposed to hypoxia and si‐SREBP2 by RNA‐seq, we identified 138 SREBP2 target genes that were upregulated under hypoxia but downregulated by treatment with two different siRNAs against SREBP2 (Fig 2C, Table EV1). Pathway analysis of these genes showed an enrichment in cholesterol and sterol biosynthetic pathways genes (Fig 2C), suggesting that silencing of SREBP2 downregulated cholesterol biosynthesis pathways in monocyte/macrophage lineage cells under hypoxic conditions.

Figure EV2. Hypoxia‐induced Golgi disassembly activates SREBP2 in monocytic myeloid cells in a SCAP‐independent manner.

1. Heatmap representation of differentially expressed genes in monocyte/macrophage cells (THP‐1) when they undergo 24 h of hypoxia or knockdown of SREBP2 using two different siRNAs. Fold change of gene expression is indicated by color bar. Representative genes from differentially regulated section were listed.
2. Total protein staining (Ponceau S) confirming equal loading in the Western blot analysis of SREBP2 expression in THP‐1 cells, related to Fig 2D.
3. Total protein staining (Ponceau S) confirming equal loading in the Western blot analysis of SREBP2 expression in THP‐1 cells, related to Fig 2E.
4. Immunofluorescent staining of TGN46 (green), KDEL (red), and nucleus by DAPI (blue) in THP‐1 treated with vehicle or 1 μg/ml 25‐hydroxycholesterol (25‐OH) under control and hypoxic condition. TGN46⁺ area/cell indicating the extent of Golgi disassembly is quantified. All scale bars, 5 μm. Quantification data represent mean ± SEM of n = 24 biological replicates.
5. Immunofluorescent staining of TGN46 (green), KDEL (red), and nucleus by DAPI (blue) in THP‐1 treated with SCAP‐siRNA under control and hypoxic condition. TGN46⁺ area/cell indicating the extent of Golgi disassembly is quantified. All scale bars, 5 μm. Quantification data represent mean ± SEM of n = 24 biological replicates.
6. Immunofluorescent staining of TGN46 (green), SREBP2 (red), and nucleus by DAPI (blue) in THP‐1 treated with SCAP‐siRNA under control and hypoxic condition. SREBP2 intensity in nucleus is quantified. All scale bars, 5 μm. Quantification data represent mean ± SEM of n = 28 biological replicates.
7. Validation of SCAP‐siRNA‐mediated effect in THP‐1 cells under control or hypoxic conditions. Data represent mean ± SEM of n = 3 biological replicates.
8. Effect of SCAP‐siRNA on cholesterol biosynthetic genes (HMGCS1, HMGCR, IDI1, LSS, and MSMO1) in response to 24 h of control or hypoxic condition in THP‐1 cells. Data represent mean ± SEM of n = 3 biological replicates.
9. Validation of SCAP‐siRNA‐mediated effect in THP‐1 cells under vehicle (0.2% methanol) or 2 μg/ml Brefeldin A (BFA) treatment. Data represent mean ± SEM of n = 3 biological replicates.
10. Effect of SCAP‐siRNA on cholesterol biosynthetic gene (HMGCS1, HMGCR, IDI1, LSS, and MSMO1) in THP‐1 cells under vehicle (0.2% methanol) or 2 μg/ml BFA treatment. Data represent mean ± SEM of n = 3 biological replicates.

Data information: Indicated P‐values were obtained using one‐way analysis of variance (ANOVA) followed by Student–Newman–Keuls multiple comparisons for post hoc test. *P < 0.05, ***P < 0.001, ****P < 0.0001; ns: not significant.

Source data are available online for this figure.

Nuclear translocation is a necessary precursor event for SREBP2 activation (Sakai et al, 1996). Thus, we investigated whether hypoxia induces translocation of SREBP2 into the nucleus for activation in monocyte/macrophage lineage cells. We found that under hypoxia, levels of the transcriptionally active form of SREBP2 (65 kDa) increased in the nuclear fraction, with a concomitant decrease in levels of the inactive precursor form of SREBP2 (120 kDa) in the endoplasmic reticulum (ER) membrane fraction of the monocytic myeloid cells (Figs 2D and EV2B). Similar results were observed under cholesterol deprivation too, which resulted in SREBP2 processing within the Golgi apparatus (Figs 2E and EV2C) (Sakai et al, 1996). Thus, the nuclear translocation‐mediated activation of SREBP2 is responsible for hypoxia‐induced upregulation in cholesterol biosynthesis.

To examine whether SREBP2 binds to promoter and enhancer regions to activate cholesterol biosynthetic genes under hypoxic conditions, we conducted quantitative ChIP‐PCR analysis. The binding percentage of SREBP2 to promoter regions of HMGCS1, IDI1, and MSMO1 was significantly enhanced in monocytic myeloid cells under hypoxia compared to normoxic conditions (Fig 2F), indicating that SREBP2 transcriptionally activates cholesterol biosynthetic genes through promoter binding. Cholesterol and its hydroxylated derivative—25‐hydroxycholesterol—are known to suppress the activity of SREBPs (Adams et al, 2004). Treatment with both cholesterol and 25‐hydroxycholesterol inhibited the upregulation of HMGCR and HMGCS1 genes under hypoxic conditions (Fig 2G), suggesting that the cholesterol biosynthetic pathway is regulated by canonical SREBP2 activation in monocytic myeloid cells under hypoxia.

Because the activation process of SREBP2 is associated with ER‐to‐Golgi transport (DeBose‐Boyd et al, 1999; Nohturfft et al, 2000; Osborne, 2001), we investigated the effects of hypoxia on the dynamics of Golgi apparatus in monocytic myeloid cells. Interestingly, hypoxia markedly stimulated Golgi disassembly (Fig 2H) and fusion of Golgi apparatus and endoplasmic reticulum (ER) (Fig 2I) in monocytic myeloid cells, concomitant with SREBP2 nuclear translocation (Fig 2J). Thus, Golgi disassembly‐induced Golgi‐ER fusion may trigger SREBP2 nuclear translocation and SREBP2‐mediated cholesterol biosynthesis in hypoxic monocytic myeloid cells.

25‐hydroxycholesterol inhibited the upregulation of SREBP2 target genes under hypoxic conditions (Fig 2G). To assess whether 25‐hydroxycholesterol reverses hypoxia‐induced Golgi disassembly, we compared the morphology of Golgi in monocytic myeloid cells under hypoxia with or without treatment with 25‐hydroxycholesterol. We found that the addition of 25‐hydroxycholesterol did not reverse hypoxia‐induced Golgi disassembly (Fig EV2D), suggesting that 25‐hydroxycholesterol likely prevents the maturation of SREBP during hypoxia by limiting SREBP movement from the ER to the Golgi apparatus, which is still required even if Golgi disassembly is induced under hypoxia.

In sterol‐depleted cells where canonical SREBP2 processing occurs, SCAP escorts SREBP2 from the ER to the Golgi, where SREBP2 is cleaved by Site‐1 protease (S1P) and activated for nuclear translocation. However, SREBP2 processing can occur independently of SCAP and cellular cholesterol levels as long as active S1P are in the same subcellular compartment as SREBP2 (Debose‐Boyd et al, 1999). We propose that hypoxia induces Golgi‐ER fusion and results in the colocalization of SREBP2 and S1P, which is sufficient for activating SREBP2 processing in monocytic myeloid cells.

To prove that SCAP is not involved in hypoxia‐induced SREBP2 activation, we knocked down the gene expression of SCAP in monocytic myeloid cells using siRNAs and examined whether SREBP2 was still activated under hypoxia. We observed significant Golgi disassembly (Fig EV2E), nuclear translocation of SREBP2 (Fig EV2F), and upregulation of downstream target genes (Fig EV2G and H) under hypoxia treatment despite SCAP knockdown, indicating that SCAP is not necessary for hypoxia‐induced SREBP2 activation in monocytic myeloid cells.

In addition, as a positive control, we treated monocytic myeloid cells with Brefeldin A (BFA), which is known to induce Golgi disassembly and Golgi‐ER fusion (Debose‐Boyd et al, 1999). We found significant upregulation of SREBP2 downstream target genes (Fig EV2I and J) after BFA treatment despite SCAP knockdown, indicating that SCAP is not necessary for BFA‐induced SREBP2 activation. Together, these results suggest that in monocytic myeloid cells, hypoxia induces SCAP‐independent SREBP2 activation in a similar way to BFA.

To investigate whether Golgi disassembly and Golgi‐ER fusion are required for SREBP2 activation, we treated monocytic myeloid cells with mepanipyrim. Mepanipyrim has been reported to inhibit Golgi disassembly and BFA‐induced retrograde Golgi‐to‐ER trafficking through a mechanism that is not fully understood (Nakamura et al, 2003). We found that mepanipyrim prevented hypoxia‐induced Golgi disassembly (Fig EV3A), SREBP2 nuclear translocation (Fig EV3B), and upregulation of SREBP2 downstream targets (Fig EV3C), suggesting that Golgi disassembly is required for hypoxia‐induced activation of SREBP2 processing.

Figure EV3. Hypoxia‐induced Golgi disassembly and Golgi‐ER fusion are required for SREBP2 activation.

1. Immunofluorescent staining of TGN46 (green), KDEL (red), and nucleus by DAPI (blue) in THP‐1 treated with vehicle (0.1% DMSO) or 40 μg/ml Mepanipyrim under control or hypoxic conditions. All scale bars, 5 μm. TGN46⁺ area/cell, indicating the extent of Golgi disassembly is quantified. Quantification data represent mean ± SEM of n = 24 biological replicates.
2. Immunofluorescent staining of TGN46 (green), SREBP2 (red), and nucleus by DAPI (blue) in THP‐1 treated with vehicle or 40 μg/ml Mepanipyrim under control or hypoxic conditions. All scale bars, 5 μm. SREBP2 intensity in nucleus is quantified. Quantification data represent mean ± SEM of n = 32 biological replicates.
3. mRNA expression level of cholesterol biosynthetic genes (HMGCS1, IDI1, LSS, and MSMO1) in THP‐1 cells treated with vehicle or 40 μg/ml Mepanipyrim under control or hypoxic conditions. Data represent mean ± SEM of n = 3 biological replicates.
4. Immunofluorescent staining of TGN46 (green), KDEL (red), and nucleus by DAPI (blue) in THP‐1 treated with vehicle or 10 μM Nocodazole. All scale bars, 5 μm. Quantification data represent mean ± SEM of n = 24 biological replicates.
5. mRNA expression level of cholesterol biosynthetic genes (HMGCS1, IDI1, LSS, and MSMO1) in THP‐1 cells treated with vehicle or 10 μM Nocodazole. Data represent mean ± SEM of n = 3 biological replicates.
6. Immunofluorescent staining of TGN46 (green), KDEL (red), and nucleus by DAPI (blue) in THP‐1 treated with vehicle or 2 μg/ml Brefeldin A (BFA). All scale bars, 5 μm. Quantification data represent mean ± SEM of n = 24 biological replicates.
7. Immunofluorescent staining of TGN46 (green), SREBP2 (red), and nucleus by DAPI (blue) in THP‐1 treated with vehicle or 2 μg/ml BFA. All scale bars, 5 μm. Quantification data represent mean ± SEM of n = 24 biological replicates.
8. mRNA expression level of cholesterol biosynthetic genes (HMGCS1, IDI1, LSS, and MSMO1) in THP‐1 cells treated with vehicle or 2 μg/ml BFA. Data represent mean ± SEM of n = 3 biological replicates.

Data information: Indicated P‐values were obtained using (A, B, D, F and G) one‐way analysis of variance (ANOVA) followed by Student–Newman–Keuls multiple comparisons for post hoc test or (C and H) paired Student's t‐tests. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; ns: not significant.

Source data are available online for this figure.

To further support our hypothesis that hypoxia‐induced SREBP2 translocation occurs due to Golgi‐ER fusion and consequent Golgi‐to‐ER retrograde transport, we treated monocytic myeloid cells with nocodazole and Brefeldin A (BFA), which are two commonly used chemicals to induce Golgi disassembly through different mechanisms (Debose‐Boyd et al, 1999). Nocodazole depolymerizes microtubules and plays a crucial role in Golgi‐to‐ER retrograde trafficking (Thyberg & Moskalewski, 1985). Therefore, nocodazole causes Golgi disassembly without inducing fusion of the Golgi apparatus and the ER. BFA induces both Golgi disassembly and Golgi‐ER fusion by preventing the association of COP‐I coat to the Golgi membrane (Helms & Rothman, 1992). We expected that treatment of monocytic myeloid cells with either nocodazole or BFA would induce Golgi disassembly, but only BFA treatment and hypoxia would cause SREBP2 activation and upregulation of downstream target genes. As expected, BFA induced Golgi disassembly (Fig EV3F), Golgi‐ER fusion, SREBP2 nuclear translocation (Fig EV3G), and upregulation of genes in the cholesterol biosynthetic pathway (Fig EV3H) in similar way to hypoxia. In contrast, nocodazole treatment only resulted in Golgi disassembly (Fig EV3D) and did not upregulate SREBP2 target genes (Fig EV3E). In summary, hypoxia‐induced SREBP2 activation requires both Golgi disassembly and Golgi‐ER fusion and is independent of SREBP cleavage‐activating protein (SCAP).

Hypoxia‐induced Golgi‐ER fusion and SREBP2 activation is lineage‐dependent

Cholesterol biosynthesis is reported to be associated with the regulation of differentiation in bone marrow‐derived cells (Parhami et al, 2002). In the present study, we used THP‐1, A‐THP‐1^early and A‐THP‐1^late cells as models for monocytes, early macrophages, and late macrophages, respectively, to investigate the relationship between differentiation and activation of cholesterol biosynthesis under hypoxia (Fig 3A) (Tominaga et al, 1998). First, we confirmed the degree of differentiation using monocyte and macrophage markers (CD11b, CD35, CD32) (Fig 3B) (Forrester et al, 2018). To determine whether hypoxia‐induced upregulation of cholesterol biosynthesis was associated with the extent of differentiation in the monocyte/macrophage lineage cells, we measured mRNA expression levels of cholesterol biosynthetic genes in the macrophage lineage (A‐THP‐1^early/late) cells. Although proliferation rate under hypoxia (Figs EV4A and EV1C) and HIF1α‐mediated hypoxic responses of macrophages (A‐THP‐1^late) were similar to those of monocytic myeloid cells (Figs EV4B and EV1B), the expression level of cholesterol biosynthetic genes was much less upregulated in early macrophages (A‐THP‐1^early) compared to monocytic myeloid cells (Figs 3C and 1E). This decline in upregulation was even more significant in late macrophages (A‐THP‐1^late), correlating with the degree of maturation (Fig 3D). Thus, differentiation cancels hypoxia‐induced upregulation of cholesterol biosynthetic genes in monocyte/macrophage lineage cells.

Figure 3. Hypoxia‐induced Golgi‐ER fusion and SREBP2 activation is lineage‐dependent.

• A
  Schematics of the cellular model used to study monocyte/macrophage differentiation.
• B
  mRNA expression of monocyte/macrophage markers in THP‐1, A‐THP‐1^early, and A‐THP‐1^late cells. Data represent mean ± SEM of n = 3 biological replicates.
• C, D
  mRNA expression of cholesterol biosynthetic enzymes in (C) A‐THP‐1^early and (D) A‐THP‐1^late cells. Data represent mean ± SEM of n = 3 biological replicates.
• E
  Active form (N) and inactive precursor form (P) of SREBP2 in nuclear extract (Nuc. Ext.) and membrane (Membrane) fractions of THP‐1, A‐THP‐1^early, and A‐THP‐1^late cells under control, hypoxic conditions.
• F, G
  Hypoxia‐induced binding of SREBP2 on the promoter regions of HMGCS1, IDI1, and MSMO1 in A‐THP‐1^early and A‐THP‐1^late cells. Data represent mean ± SEM of n = 3 biological replicates.
• H, I
  Immunofluorescent staining of TGN46 (green), DAPI (blue) for nucleus, and (H) KDEL (red) or (I) SREBP2 (red) in A‐THP‐1^early and A‐THP‐1^late cells under hypoxic conditions. All scale bars, 5 μm. Quantification data represent mean ± SEM of n = 24 (I) and 32 (H) biological replicates.

Data information: Indicated P‐values were obtained using (B–D, F and G) paired Student's t‐tests or (H and I) one‐way analysis of variance (ANOVA) followed by Student–Newman–Keuls multiple comparisons for post hoc test. *P < 0.05, **P < 0.01, ***P < 0.001, ns: not significant.

See also Fig EV4A–H.

Source data are available online for this figure.

Figure EV4. Hypoxia‐induced Golgi‐ER fusion and SREBP2 activation are lineage‐dependent.

1. Relative cell growth of A‐THP‐1^late cells during 72 h of culture under control or hypoxic condition. Cell growth is assessed every 24 h by sulforhodamine (SRB) assay. Data represent mean ± SEM of n = 3 biological replicates.
2. mRNA expression levels of VEGF, PDK1, and LDHA in response to 24 h of control or hypoxic condition in A‐THP‐1^late cells. Data represent mean ± SEM of n = 3 biological replicates.
3. Total protein staining (Ponceau S) confirming equal loading in the Western blot analyses of SREBP2 expression in THP‐1, A‐THP‐1^early and A‐THP‐1^late cells, related to Fig 3E.
4. (Left) Active form (N) and inactive precursor form (P) of SREBP2 in nuclear extract (Nuc. Ext.) and membrane (Membrane) fractions of A‐THP‐1^late cells under cholesterol‐deprived (Induced: Ind.) and cholesterol‐rich (Suppressed: Supp.) conditions. (Right) Total protein staining (Ponceau S) confirming equal loading.
5. Relative % input of histone methylation (H3K4me3) and acetylation (H3K27Ac) on promoter regions and enhancer regions, respectively, of HMGCS1, IDI1, and MSMO1 in A‐THP‐1^early cells with 24 h of exposure to hypoxia. Data represent mean ± SEM of n = 3 biological replicates.
6. Relative % input of histone methylation (H3K4me3) and acetylation (H3K27Ac) on promoter regions and enhancer regions, respectively, of HMGCS1, IDI1, and MSMO1 in A‐THP‐1^late cells with 24 h of exposure to hypoxia. Data represent mean ± SEM of n = 3 biological replicates.
7. Relative % input of histone methylation (H3K4me3) and acetylation (H3K27Ac) on promoter regions and enhancer regions, respectively, of HMGCR, MVD, FDPS, and FDFT1 in THP‐1, A‐THP‐1^early, or A‐THP‐1^late cells with 24 h of exposure to hypoxia. Data represent mean ± SEM of n = 3 biological replicates.
8. Immunofluorescent staining of TGN46 (green), KDEL (red), and nucleus by DAPI (blue) in A‐THP‐1^late cells treated with 10 μM Nocodazole. TGN46⁺ area/cell indicating the extent of Golgi disassembly is quantified. All scale bars, 5 μm. Quantification data represent mean ± SEM of n = 24 biological replicates.
9. Total protein staining (Ponceau S) confirming equal loading in the Western blot analyses of SREBP2 expression in PMA‐stimulated macrophages, related to Fig 4D, respectively.

Data information: Indicated P‐values were obtained using (B, E and F) paired Student's t‐tests or (H) one‐way analysis of variance (ANOVA) followed by Student–Newman–Keuls multiple comparisons for post hoc test. **P < 0.01, ***P < 0.001, ****P < 0.0001, ns: not significant.

Source data are available online for this figure.

Next, we questioned whether the decrease in cholesterol biosynthetic gene expression in macrophages was due to declined SREBP2 nuclear translocation. Indeed, the nuclear translocation of SREBP2 was not observed in early or late macrophages (A‐THP‐1^early/A‐THP‐1^late) in response to hypoxia (Figs 3E and EV4C), although cholesterol‐deprived conditions activated SREBP2 processing (Fig EV4D). ChIP‐PCR revealed that promoter binding of SREBP2 to its target genes, concomitant with active histone modifications on the promoter (H3K4me3) and enhancer regions (H3K27Ac), also decreased with lineage progression in the macrophage lineage (A‐THP‐1^early and A‐THP‐1^late) cells (Figs 3F and G, and EV4E and F). In addition, hypoxia‐induced active histone modifications on the promoters and enhancers of other cholesterol biosynthesis genes such as HMGCR, MVD, FDPS, and FDFT1, also decreased with the lineage progression (Fig EV4G). Furthermore, we observed no significant Golgi disassembly nor SREBP2 nuclear translocation in macrophages (A‐THP‐1^early and A‐THP‐1^late) in response to hypoxia (Fig 3H and I), while nocodazole treatment stimulated Golgi disassembly of late macrophages (A‐THP‐1^late) (Fig EV4H). Altogether, these observations suggest that Golgi disassembly and Golgi‐ER fusion stimulate SREBP2‐mediated cholesterol biosynthesis under hypoxia in a lineage‐dependent manner.

Hypoxia‐induced Golgi‐ER fusion and SREBP2 activation are impaired by phorbol 12‐myristate‐13‐acetate (PMA)‐induced differentiation

To test whether PMA‐induced differentiation downregulates cholesterol biosynthesis in myeloid cells, we treated monocytic myeloid cells with different combinations of PMA‐treatment and resting time (Fig 4A) (Daigneault et al, 2010). First, we confirmed the degree of differentiation using monocyte and macrophage markers (CD11b, CD14) (Fig 4B) (Chen et al, 2022). Next, we investigated whether PMA‐induced differentiation alters expression of cholesterol biosynthetic genes under hypoxia, and found their expressions to be downregulated in differentiated macrophages (Fig 4C). Unlike monocytic myeloid cells, SREBP2 did not translocate to nucleus in the hypoxic macrophages (Figs 4D, 2D and EV4I), which lead to a significant decrease in SREBP2 binding to promoter and enhancer regions of downstream target genes (Fig 4E). Thus, SREBP2‐mediated activation of cholesterol biosynthesis is impaired in hypoxic macrophages. Furthermore, Golgi disassembly and Golgi‐ER fusion were suppressed in hypoxic PMA‐treated macrophages compared to monocytic myeloid cells (Fig 4F and G), thereby resulting in inactivation of SREBP2 processing for nuclear translocation (Fig 4H).

Figure 4. Hypoxia‐induced Golgi‐ER fusion and SREBP2 activation are impaired by phorbol 12‐myristate‐13‐acetate (PMA)‐induced differentiation.

1. Schematics of THP‐1 differentiation upon PMA and resting treatments.
2. mRNA expression of monocyte/macrophage markers in THP‐1 cells with/without PMA and resting treatments. Data represent mean ± SEM of n = 3 biological replicates.
3. mRNA expression of cholesterol biosynthetic enzymes in THP‐1 cells with/without PMA and resting treatments under control/hypoxic conditions. Data represent mean ± SEM of n = 3 biological replicates.
4. Active form (N) and inactive precursor form (P) of SREBP2 in nuclear extract (Nuc. Ext.) and membrane (Membrane) fractions of PMA‐stimulated macrophages under control/hypoxic conditions.
5. Hypoxia‐induced binding of SREBP2 on the promoter regions of HMGCS1, IDI1, LSS and MSMO1 in PMA‐stimulated macrophages. Data represent mean ± SEM of n = 3 biological replicates.
6. Immunofluorescent staining of TGN46—a Golgi marker (green), KDEL—an ER marker (red), and DAPI (blue) for nucleus in PMA‐stimulated macrophages under control/hypoxic conditions. All scale bars, 10 μm. Quantification data represent mean ± SEM of n = 24 biological replicates.
7. Proximity ligation assay (PLA) staining of TGN46 (Golgi), KDEL (ER), and DAPI (nucleus) in PMA‐stimulated macrophages under control/hypoxic conditions. All scale bars, 20 μm. Quantification data represent mean ± SEM of n = 30 biological replicates.
8. Immunofluorescent staining of TGN46 (Golgi), SREBP2, and DAPI (nucleus) in PMA‐stimulated macrophages under control/hypoxic conditions. All scale bars, 10 μm. Quantification data represent mean ± SEM of n = 32 biological replicates.

Data information: Indicated P‐values were obtained using (B, C and E) paired Student's t‐tests or (F–H) one‐way analysis of variance (ANOVA) followed by Student–Newman–Keuls multiple comparisons for post hoc test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, ns: not significant.

See also Fig EV4I.

Source data are available online for this figure.

Cholesterol biosynthesis is enriched in hypoxic bone marrow and tumor cells

We next asked whether bone marrow cells, such as hematopoietic stem cells (HSCs), endogenously residing in a hypoxic bone marrow environment (Parmar et al, 2007), are enriched in cholesterol biosynthesis. Immunohistochemical analyses of mouse bone marrow revealed a significant overlap between the hypoxic and SREBP2⁺/HMGCS1⁺ regions (Fig 5A), suggesting that cholesterol biosynthesis is stimulated in bone marrow cells in hypoxic regions.

Figure 5. Cholesterol biosynthesis is enriched in hypoxic bone marrow and tumor cells.

1. Immunohistochemical staining of SREBP2, HMGCS1, and hypoxic regions in murine bone marrow. SREBP2⁺/HMGCS1⁺ cells (red); hypoxia probe (green); Topro3 for nucleus (blue). Scale bars, 100 μm. Quantification data represent mean ± SEM of n = 5 biological replicates.
2. Uniform Manifold Approximation and Projection (UMAP) of human adult bone marrow cells, classified by single‐cell RNA‐sequencing analysis.
3. Pathway analysis of cholesterol biosynthetic genes in each of human bone marrow cell subcluster.
4. Expression profile of macrophage marker (CD35) and monocyte marker (CD32) on human bone marrow cells.
5. Immunohistochemical staining of SREBP2, HMGCS1, and hypoxic regions in tumors formed by subcutaneous injection of HSML cells. SREBP2⁺/HMGCS1⁺ cells (red); hypoxia probe (green); DAPI for nucleus (blue). Scale bars, 100 μm. Quantification data represent mean ± SEM of n = 5 biological replicates.

Data information: Indicated P‐values were obtained using one‐way analysis of variance (ANOVA) followed by Student–Newman–Keuls multiple comparisons for post hoc test. *P < 0.05, **P < 0.01, ***P < 0.001.

See also Fig EV5A–E.

Source data are available online for this figure.

To investigate which fractions of the bone marrow cells exhibit hypoxia‐induced upregulation of cholesterol biosynthesis, we analyzed single‐cell RNA‐sequencing data of human bone marrow cells from a public database (Data ref: Human Cell Atlas Data Portal, 2020) and identified 19 sub‐clusters based on differential gene expression patterns (Fig 5B). Although bone marrow cells are universally responsive to hypoxia (Fig EV5A), we found that single‐cell expression level of HMGCS1, IDI1, HMGCR, and LSS was higher in cholesterol biosynthesis high clusters (7, 0, 17, 4, 12) than in low clusters (15, 19, 14, 5, 16) (Fig EV5B). Pathway analysis and signature score prediction revealed that bone marrow HSCs (sub‐clusters 0, 17, 11) and monocytes (sub‐clusters 0, 12, 4) were enriched in the cholesterol biosynthesis pathway, while macrophages (subcluster 13, 18, 19) were not (Figs 5C and D, and EV5C and D), suggesting that SREBP2‐mediated cholesterol biosynthesis is a lineage‐specific response specifically activated in immature bone marrow cells residing in hypoxic regions.

Figure EV5. Cholesterol biosynthesis is enriched in hypoxic bone marrow cells.

1. Uniform Manifold Approximation and Projection (UMAP) of human adult bone marrow cells with corresponding expression levels of hypoxic signature genes.
2. Single‐cell expression level of cholesterol biosynthetic genes (HMGCS1, HMGCR, IDI1, LSS) in human adult bone marrow cells, compared between cholesterol biosynthesis high clusters (7, 0, 17, 4, 12) and low clusters (15, 19, 14, 5, 16), related to Fig 5C.
3. UMAP of human bone marrow cells and their single‐cell expression profile of signature genes for hematopoietic stem cells, monocytes, and macrophages.
4. Signature scores of hematopoietic stem cells, monocytes, and macrophages for each subcluster of human bone marrow cells. Data represent mean ± SEM of at least three independent data sets.
5. Cellular content of cholesterol biosynthesis pathway metabolites in mouse bone marrow‐derived monocyte/macrophage cells treated under control or hypoxic conditions for 24 h. Data represent mean ± SEM of n = 3 biological replicates.

Data information: Indicated P‐values were obtained using paired Student's t‐tests. *P < 0.05, **P < 0.01. (B and D) Central bands indicate medium, boxes indicate interquartile range, whiskers indicate minimum and maximum for all data points, respectively.

Source data are available online for this figure.

For direct metabolic analysis of cholesterol biosynthesis pathway in bone marrow‐derived monocyte/macrophage lineage cells, we measured the cellular levels of HMG‐CoA, mevalonate, mevalonate‐5P, 7‐dehydrocholesterol, desmosterol, and cholesterol in mouse monocyte/macrophage lineage cells under normoxia and hypoxia using CE‐MS and LC‐MS‐based metabolome analyses depending on the hydrophobicity of metabolites. We found that desmosterol and cholesterol levels were increased in bone marrow‐derived monocyte/macrophage under hypoxia (Fig EV5E).

Immunohistochemical analyses of tumor tissues revealed a similar significant overlap between the hypoxic and SREBP2⁺/HMGCS1⁺ regions (Fig 5E), suggesting that cholesterol biosynthesis is also stimulated in tumor cells residing in the hypoxic regions of the tumor microenvironment.

Cholesterol biosynthesis is associated with tumor growth through infiltration of immune cells and enhanced angiogenesis

We hypothesized that the hypoxia‐induced cholesterol biosynthesis pathway may influence infiltration of monocyte/macrophage lineage cells during tumor progression. Indeed, inhibiting cholesterol biosynthesis with atorvastatin pre‐treatments (Fig 6A) significantly reduced infiltration of monocyte/macrophage lineage cells into tumor during angiogenesis (Fig 6B). Moreover, atorvastatin post‐treatments (Fig 6C) also significantly reduced tumor growth, angiogenesis, infiltration of monocyte/macrophage lineage cells, and expression of Ccl2 (Figs 6D and E, and EV6A and B), suggesting that disruption of cholesterol biosynthesis suppresses tumor progression. In addition, CCL2 inhibitor (bindarit) significantly reduced infiltration of myeloid cells (CD11b⁺) and macrophages (F4/80⁺) into the tumor and the extent of angiogenesis (% CD31⁺ area) (Fig EV6C), resulting in decreased tumor volume compared to vehicle treatment (Fig EV6D). Interestingly, treatment with atorvastatin did not alter the composition of immune cells including neutrophils, monocytes, and macrophages in peripheral blood or bone marrow of mice (Fig EV6E), suggesting that statins may influence migration of immune cells. We then hypothesized that inhibition of cholesterol biosynthesis through inhibition of SREBP2 may influence migration of monocytic cells under hypoxia. To assess the same, we performed trans‐well migration assay using THP‐1 cells under control or hypoxic conditions. Disruption of cholesterol synthesis through SREBP2 silencing significantly inhibited hypoxia‐induced migration of immune cells (Fig EV6F).

Figure 6. Cholesterol biosynthesis is associated with tumor growth through infiltration of immune cells and enhanced angiogenesis.

1. Schematics of subcutaneous tumor model with atorvastatin pre‐treatment.
2. Immunostaining of HSML allograft tumor at pre‐angiogenic (Day 2) and post‐angiogenic state (Day 4) with pre‐treatment of atorvastatin. Endothelial cells indicated by CD31, myeloid cells by CD11b, macrophages by F4/80, granulocytes by Gr‐1, and nucleus by DAPI, respectively. Scale bars, 100 μm. Quantification data represent mean ± SEM of n = 3 biological replicates.
3. Schematics of subcutaneous tumor model with atorvastatin post‐treatment.
4. Tumor growth of HSML and B16 tumor allografts with continuous atorvastatin administration. Data represent mean ± SEM of n = 5 biological replicates.
5. Immunostaining of HSML allograft tumor with post‐treatment of atorvastatin. Endothelial cells indicated by CD31, myeloid cells by CD11b, macrophage cells by F4/80, granulocytes by Gr‐1, and nucleus by DAPI, respectively. Scale bars, 50 μm. Quantification data represent mean ± SEM of n = 3 biological replicates.
6. Kaplan–Meier analysis of hypoxia‐related genes in multiple cancer types. Overall survival rates of patients with breast, brain, colorectal, liver, ovarian, lung, sarcoma, and pancreatic cancer.

Data information: Indicated P‐values were obtained using (B and E) one‐way analysis of variance (ANOVA) followed by Student–Newman–Keuls multiple comparisons for post hoc test or (D) paired Student's t‐tests. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

See also Fig EV6A–G and Dataset EV2.

Source data are available online for this figure.

Figure EV6. Cholesterol biosynthesis is associated with tumor growth through infiltration of immune cells and enhanced angiogenesis.

1. Immunostaining of B16 allograft tumor with post‐treatment of atorvastatin. Endothelial cells are indicated by CD31, myeloid cells by CD11b, macrophage cells by F4/80, granulocytes by Gr‐1, and nucleus by DAPI, respectively. Scale bars, 50 μm. Quantification data represent mean ± SEM of n = 3 biological replicates.
2. Effect of atorvastatin on Ccl2 expression in B16 tumor allografts. Data represent mean ± SEM of n = 7 biological replicates.
3. Immunostaining of B16 allograft tumor with treatment of CCL2 inhibitor (bindarit) at 10 mg/kg every 5 days. Endothelial cells indicated by CD31 (red), myeloid cells by CD11b (green), macrophage cells by F4/80 (green), and nucleus by DAPI, respectively. Scale bars, 100 μm. Quantification data represent mean ± SEM of n = 3 biological replicates.
4. Tumor growth of B16 tumor allografts with continuous CCL2 inhibitor administration. Data represent mean ± SEM of n = 5 biological replicates.
5. Composition of immune cells (% in CD45⁺ cells) in mouse blood and bone marrow after treatment of vehicle or atorvastatin (10 mg/kg/day) for 7 days. The population of immune cells was distinguished by flow cytometry analysis: Neutrophils (CD45⁺Ly‐6C⁺Ly‐6G⁺); monocyte (CD45⁺Ly‐6C^HighLy‐6G⁻); dendritic cells (CD45⁺CD11c⁺); macrophages (CD45⁺F4/80⁺CD11b⁺); CD4⁺ T cells (CD45⁺CD4⁺); CD8⁺ T cells (CD45⁺CD8⁺); B cells (CD45⁺B220⁺); NK cells (CD45⁺NK1.1⁺). Data represent mean ± SEM of n = 3 biological replicates.
6. Trans‐well migration assay of THP‐1 under control or hypoxic condition with silencing of SREBP2. Data represent mean ± SEM of n = 3 biological replicates.
7. Survival rate of six types of cancer patients with low or high expression of nine cholesterol biosynthesis signature genes (HMGCS1, HMGCR, MVD, IDI1, FDPS, FDFT1, SQLE, LSS, MSMO1), normalized by the expression of monocyte marker gene CD11b.

Data information: Indicated P‐values were obtained using (A and C) one‐way analysis of variance (ANOVA) followed by Student–Newman–Keuls multiple comparisons for post hoc test, (B, D and F) paired Student's t‐tests, or (G) Mann–Whitney U‐test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Source data are available online for this figure.

To assess whether hypoxia‐regulated genes in monocytic myeloid cells were associated with reduced survival in cancer patients, we performed Kaplan–Meier survival analysis on data from the Broad TCGA GDAC web site (http://gdac.broadinstitute.org/) and the Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/). Expression level of hypoxia‐induced genes predicted poor prognosis in multiple cancer types, including breast, brain, colorectal, liver, ovarian, lung, sarcoma, and pancreatic cancer (Fig 6F, Dataset EV2). Furthermore, we analyzed whether expression of cholesterol synthesis genes in monocytes correlates with survival rate of patients with cancer. We found that higher expression of cholesterol synthesis‐related signature genes normalized by the expression of the monocyte marker gene (CD11b) significantly correlated with lower patient survival in six types of cancer (ACC: adenoid cystic carcinoma, CESC: cervical squamous cell carcinoma, HNSC: head and neck squamous cell carcinoma, MESO: mesothelioma, SARC: sarcoma, and SKCM: skin cutaneous melanoma) (Fig EV6G), suggesting that hypoxia‐responsive cholesterol biosynthesis in monocytic myeloid cells correlates with cancer progression and poor clinical outcomes.

Discussion

Hypoxia induces metabolic rewiring, such as elevated glycolysis, in cancer cells and fibroblasts within tumor microenvironment (Eales et al, 2016; Becker et al, 2020); however, the cellular response of immune cells to hypoxia remains unknown. Our investigations reveal that SREBP2 activation via Golgi disassembly occurs specifically in hypoxic monocytes. SREBPs are transcription factors that modulate lipid homeostasis through a complex feedback mechanism involving intracellular cholesterol levels (Brown & Goldstein, 1997; Horton et al, 2002; Yang et al, 2002). Our group recently reported SREBP2 activation in response to an acidic extracellular pH to provide a growth advantage to cancer cells (Kondo et al, 2017). In addition, a previous study showed that SREBP may act as an oxygen sensor in yeast (Hughes et al, 2005). In the present study, we report SREBP2 activation as a novel oxygen sensory mechanism in bone marrow‐derived cells (BMDCs) in response to hypoxia through the fusion of Golgi apparatus and endoplasmic reticulum (ER).

Tumor‐associated macrophages (TAMs) are a major component of the tumor microenvironment that promote cancer progression (Mantovani & Sica, 2010). Studies have linked hypoxia with the enhanced pro‐tumor effect of TAMs via stimulation of immune surveillance and angiogenesis (Casazza et al, 2013). TAMs are recruited to and restrained in a hypoxic tumor environment by hypoxia‐induced regulation of chemo‐attractants and chemokine receptors (Murdoch & Lewis, 2005). In the present study, we observed that bone marrow‐derived cells infiltrate into hypoxic tumor centers and differentiate to TAMs in the pre‐angiogenic phase of tumor formation, consistent with literature (Coussens et al, 2000; Ferrara & Kerbel, 2005; Shibuya & Claesson‐Welsh, 2006; Shojaei et al, 2007; Du et al, 2008; Ostrand‐Rosenberg, 2008). During the differentiation of BMDCs to TAMs, we demonstrated cholesterol synthesis may be essential in the regulation of tumor‐promoting effect. Macrophages also function as anti‐tumor M1 macrophages (Liu et al, 2021). However, the mechanism of M1 and M2 macrophage conversion in the context of cholesterol biosynthesis remains unclear.

HSCs face a tightly orchestrated regulation of cell fate between quiescence, self‐renewal, apoptosis, and differentiation (Pollard & Kranc, 2010). Although a detailed mechanism of cell fate regulation has not been elucidated, environmental cues, such as hypoxia, are proposed to play a pivotal role in regulating the dynamics of HSC activity (Simsek et al, 2010; Takubo et al, 2010). The results of the present study showed that hypoxia‐induced SREBP2 activation only occurs in HSCs and early lineage of myeloid cells. This observation sheds new light on how a hypoxic environment may regulate bone marrow cell differentiation.

Finally, we demonstrated that Golgi‐ER fusion‐mediated SREBP2 transcriptional activation is a key regulator in the oxygen sensory mechanism of HSCs and monocytic myeloid cells. Inhibition of cholesterol biosynthesis abolished infiltrations of pro‐tumor BMDCs, angiogenesis, and thus suppressed tumor growth. Notably, high expression of hypoxia‐responsive signature genes in BMDCs correlated with poor patient prognosis, further supporting the significance of hypoxia‐induced SREBP2 pathway in clinical prospective. In conclusion, we have shown strong evidence that SREBP2 activation through lineage‐specific organelle contacts in BMDCs promotes tumor progression. Thus, inhibitors targeting cholesterol biosynthesis can be used in combination with conventional chemotherapy for immune modulation.

Materials and Methods

Reagents and Tools table

Reagent/Resource	Reference or source	Identifier or catalog number
Experimental models
List cell lines, model organism strains, patient samples, isolated cell types etc. Indicate the species when appropriate.
HeLa	ATCC	Cat# ATCC CCL‐2
		RRID: CVCL_0030
HSML	Kudoh (Hirosaki University)	N/A
THP‐1	RIKEN	Cat# RCB3686
		RRID: CVCL_0006
A‐THP‐1	Cell Resource Center for Biomedical Research, Institute of Development, Aging and Cancer Tohoku University	TKG 0296
		RRID: CVCL_5115
B16	ATCC	Cat# ATCC CRL‐6322
		RRID: CVCL_0604
NHDF	Lonza	Cat# CC‐2511
BALL‐1	RIKEN	Cat# RCB1882
		RRID: CVCL_1075
C57BL/6 (M. musculus)	CLEA Japan Inc	C57BL/6
Antibodies
Include the name of the antibody, the company (or lab) who supplied the antibody, the catalogue or clone number, the host species in which the antibody was raised and mention whether the antibody is monoclonal or polyclonal. Please indicate the concentrations used for different experimental procedures.
Anti‐mouse PECAM‐1 Hamster mAb	Sigma‐Aldrich	Cat# MAB1398Z
		RRID: AB_94207
Anti‐mouse CD11b Rat mAb	BD Pharmingen	Cat# 550282
		RRID: AB_393577
Anti‐mouse F4/80 Rat mAb	Bio‐Rad Laboratories	Cat# MCA497GA
		RRID: AB_323806
Anti‐mouse Ly6G Rat mAb	BioLegend	Cat# 127602
		RRID: AB_1089180
Anti‐H3K4me3 Rabbit pAb	Active Motif	Cat# 39159
		RRID: AB_2615077
Anti‐H3K27ac Mouse mAb	FUJIFILM Wako	Cat# 308‐34843
		RRID: AB_2819244
Anti‐SREBP2 Mouse mAb	Takao Hamakubo (University of Tokyo)	N/A
Anti‐Histone H3 Rabbit pAb	Abcam	Cat# ab1791
		RRID: AB_302613
Anti‐GRP78 BiP Rabbit pAb	Abcam	Cat# ab21685
		RRID: AB_2119834
Anti‐SREBP2 Rabbit pAb	Cayman Chemical	Cat# 10007663
		RRID: AB_2615896
Anti‐TGN46 Mouse mAb	Sigma‐Aldrich	Cat# SAB4200355
		RRID: AB_10762671
Anti‐KDEL Rabbit mAb	Abcam	Cat# ab176333
		RRID: AB_2819147
Anti‐SREBP2 Rabbit pAb	Abcam	Cat# ab30682
		RRID: AB_779079
Anti‐HMGCS1 Rabbit pAb	GeneTex	Cat# GTX112346
		RRID: AB_11179696
Alexa Fluor 488–conjugated goat anti‐Rat IgG	Thermo Fisher Scientific	Cat# A‐11006
		RRID: AB_2534074
Alexa Fluor 568–conjugated goat anti‐Hamster IgG	Thermo Fisher Scientific	Cat# A‐21112
		RRID: AB_2535761
Alexa Fluor 568–conjugated goat anti‐Rabbit IgG	Thermo Fisher Scientific	Cat# A‐11036
		RRID: AB_10563566
Peroxidase‐conjugated affinity‐purified goat anti‐mouse IgG	Sigma‐Aldrich	Cat# A4416
		RRID: AB_258167
Peroxidase‐conjugated affinity‐purified goat anti‐rabbit IgG	Cell Signaling Technology	Cat# 7074
		RRID: AB_2099233
PerCP‐Cy5.5 anti‐CD11b mAb (M1/70)	Biolegend	Cat# 101228
PE anti‐Ly6C mAb (HK1.4)	Biolegend	Cat# 128008
APC anti‐CD11c mAb (N418)	Biolegend	Cat# 117310
FITC anti‐NK1.1 mAb	Biolegend	Cat# 156508
FITC anti‐F4/80 mAb (BM8)	Biolegend	Cat# 123108
Pacific Blue anti‐Ly6G mAb (1A8)	Biolegend	Cat# 127612
APC‐Cy7 anti‐CD45 mAb (IM7)	Biolegend	Cat# 103115
PE anti‐CD4 mAb (GK1.5)	Biolegend	Cat# 100408
PerCP‐Cy5.5 anti‐CD8 mAb (53‐6.7)	Biolegend	Cat# 100734
APC anti‐B220 mAb (RA3‐6B2)	TOMBO	20‐0452‐U025
Oligonucleotides and sequence‐based reagents
For long lists of oligos or other sequences please refer to the relevant Table(s) or EV Table(s)
qPCR primers	This study	Table EV2A
ChIP‐PCR primers	This study	Table EV2B
Sequences for siRNA	This study	Table EV2C
Chemicals, enzymes and other reagents
e.g., drugs, peptides, recombinant proteins, dyes etc.
Fibroblast Basal Medium	Lonza	Cat# CC‐3131
DMEM (High Glucose)	Nacalai Tesque	Cat# 08459‐64
RPMI Medium 1640	Nacalai Tesque	Cat# 30264‐56
Fetal Bovine Serum	Biosera	Cat# FB‐1285/25
Newborn Calf Lipoprotein‐deficient Serum	Sigma‐Aldrich	Cat# S5394
Sulforhodamine B sodium salt	Sigma‐Aldrich	Cat# S1402‐5G
Ribonuclease	Nippon Gene	Cat# 313‐01461
Propidium Iodide	FUJIFILM Wako	Cat# 25535‐16‐4
Mevastatin	Sigma‐Aldrich	Cat# 474700
Mevalonolactone	Sigma‐Aldrich	Cat# M4667
Cholesterol	Sigma‐Aldrich	Cat# C8667
25‐hydroxycholesterol	Sigma‐Aldrich	Cat# H1015
Calpain Inhibitor I	Nacalai Tesque	Cat# 07036‐24
DMSO	Nacalai Tesque	Cat# 09659‐14
Methanol	FUJIFILM Wako	Cat# 134‐11821
Ethanol	FUJIFILM Wako	Cat# 057‐00456
Mepanipyrim	FUJIFILM Wako	Cat# 137‐12651
Nocodazole	FUJIFILM Wako	Cat# 140‐08531
Brefeldin A	LKT Laboratories	Cat# B6816
Phorbol 12‐myristate 13‐acetate (PMA)	AdipoGen	Cat# AG‐CN2‐0010‐M010
Pimonidazole Hydrochloride	Shinsuke Sando (The University of Tokyo)	N/A
Atorvastatin calcium salt trihydrate	Tokyo Chemical Industry	Cat# A2476
Bindarit	Cayman	Cat# 11479
Isogen reagent	Nippon Gene	Cat# 311‐02504
5X PrimeScript RT Master Mix	Takara	Cat# RR036A
Lipofectamine® RNAiMAX Reagent	Thermo Fisher Scientific	Cat# 13778‐100
Opti‐MEM Reduced Serum Medium	Thermo Fisher Scientific	Cat# 31985‐070
DAPI‐Fluoromount‐G	Southern Biotech	Cat# 0100‐20
cOmplete™, Mini, EDTA‐free Protease Inhibitor Cocktail	Sigma‐Aldrich	Cat# 11836170001
Precision Plus Protein™ Dual Color Standards	Bio‐Rad	Cat# 1610374
SuperSignal™ West Dura Extended Duration Substrate	Thermo Fisher Scientific	Cat# 34075
Ponceau S	Nacalai Tesque	Cat# 28322‐72
Formaldehyde solution	Wako	Cat# 064‐00406
4% paraformaldehyde phosphate‐buffered solution	FUJIFILM Wako	Cat# 163‐20145
Software
Include version where applicable
IMPaLa	Kamburov et al (2011)	http://impala.molgen.mpg.de/
GeneChip Operating Software	Thermo Fisher Scientific	https://www.thermofisher.com/jp/en/home/life‐science/microarray‐analysis/
LAS X (v. 4.2.1)	Leica	https://www.leica‐microsystems.com/
Seurat 4.0.1	Butler et al (2018)	https://satijalab.org/seurat/
Scanpy 1.7.2	Wolf et al (2018)	https://scanpy.readthedocs.io/en/stable/
GSVA package 1.38.2	Hänzelmann et al (2013)	https://www.bioconductor.org/packages/release/bioc/html/GSVA.html
fgsea package 1.16.0	Korotkevich et al (preprint: Korotkevich et al, 2016)	http://bioconductor.org/packages/release/bioc/html/fgsea.html
glmnet package 4.0.2	Hurvich & Tsai (1995)	https://cran.r‐project.org/web/packages/glmnet/index.html
GEPIA2	Tang et al (2019)	http://gepia2.cancer‐pku.cn
Kaplan‐Meier Plotter	Lánczky et al (2016)	http://kmplot.com/analysis/
Other
Kits, instrumentation, laboratory equipment, lab ware etc. that are critical for the experimental procedure and do not fit in any of the above categories can be listed here.
RNeasy microKit	Qiagen	Cat# 74134
TruSeq RNA Library Prep	Illumina	Cat# RS‐122‐2001 or RS‐122‐2002
Genechip Human Genome U133 plus 2.0 oligonucleotide arrays	Thermo Fisher Scientific	Cat# 900466
THUNDERBIRD® Probe qPCR Mix	TOYOBO	Cat# QPS‐101
Total Cholesterol Assay Kit (Fluorometric)	Cell Biolabs	Cat# STA‐390
QIA quick PCR purification kit	QIAGEN	Cat# 28181
Duolink® Proximity Ligation Assay Kit	Sigma‐Aldrich	Cat# DUO92008
Hypoxyprobe™‐1 Plus Kit	Hypoxyprobe, Inc.	Cat# HP2‐100
6.5 mm Transwell® with 8.0 μm Pore Polycarbonate Membrane Insert	Corning	Cat# 3464

Methods and Protocols

Cell culture

Human cervical cancer cell line HeLa and mouse melanoma cell line B16 was purchased from the American Type Culture Collection. Normal human dermal fibroblast (NHDF) was purchased from Lonza. Murine uterine cancer cell line HSML was kindly provided by Dr. Kudoh (Hirosaki University, Japan). Human acute monocytic leukemia cell line THP‐1, A‐THP‐1, and human B‐cell acute lymphoblastoid leukemia cell line BALL‐1 were purchased from RIKEN. HeLa and B16 were maintained in Dulbecco's modified Eagle's medium (DMEM) (Nacalai Tesque), supplemented with 10% fetal bovine serum (FBS) (Biosera). NHDF was maintained in Fibroblast Basal Medium (Lonza). HSML, THP‐1, A‐THP‐1, and BALL‐1 were maintained in Roswell Park Memorial Institute 1640 (RPMI 1640) medium (Nacalai Tesque) supplemented with 10% FBS (Gibco). A‐THP‐1^early refers to A‐THP‐1 cells that are within five passages upon purchase. A‐THP‐1^late cells were prepared by passaging A‐THP‐1^early for more than 20 times under normal culture condition. All cells were cultured at 37°C in a 5% CO₂ atmosphere in a humidified incubator unless specified otherwise.

Mice

Male C57BL/6 mice aged 8 weeks were purchased from CLEA Japan, Inc. The animals were housed in individual cages in a temperature‐ and light‐controlled environment and had ad libitum access to chow and water. All mouse experiments were approved by the University of Tokyo Animal Care and Use Committee.

Hypoxic culture condition

Cells were seeded in the corresponding medium described above at 2 × 10⁵ cells per dish for adherent cell lines (NHDF, HeLa) or 2 × 10⁵ cells/ml per flask for suspension cell lines (THP‐1, A‐THP‐1, BALL‐1). After 24 h of culture in normoxic (pO₂ = 21%) incubator, cells were transferred to Hypoxia Workstation Invivo2 400 (Ruskinn Technology Ltd) and cultured for another 24 h in hypoxia (pO₂ = 1%). In recovery condition, cells were cultured first in hypoxia (pO₂ = 1%) for 24 h, then in normoxia (pO₂ = 21%) for another 24 h.

Inducing/suppressing condition under cholesterol deprivation

The medium for inducing conditions was RPMI 1640 supplemented with 5% (v/v) newborn calf lipoprotein‐deficient serum (Sigma‐Aldrich), 50 μM mevastatin (Sigma‐Aldrich), 50 μM mevalonolactone (Sigma‐Aldrich), and 0.2% (v/v) ethanol. The medium for suppressing conditions was supplemented with a mixture of 1 μg/ml 25‐hydroxycholesterol (Sigma‐Aldrich) and 10 μg/ml cholesterol (Sigma‐Aldrich), in addition to the inducing medium, as previously described (Hua et al, 1995). Cells were cultured in inducing or suppressing condition for 24 h before harvested for cell fractionation and protein extraction. 25 μg/ml of calpain inhibitor I (Nacalai Tesque) was also supplemented to cell culture 3 h before harvest. Cell fractionation was carried out as previously described (Sakai et al, 1996) with minor modification.

Cell proliferation assay

Cells were seeded on 96‐well plate in control medium at 10³ cells per well. After cells were attached, plates for hypoxic condition were treated with hypoxia (pO₂ = 1%) in Hypoxia Workstation Invivo2 400 (Ruskinn technology Ltd). The viability of attached cells was measured at 24, 48, and 72 h after treatment to each condition using the sulforhodamine B (SRB, Sigma‐Aldrich) cell proliferation assay as previously described (De Silva et al, 2000).

Cell cycle assay

Cells were harvested and washed in PBS then fixed in cold 70% ethanol with gentle vortexing. After incubation at 4°C for 30 min, cells were washed twice in PBS and resuspended in 500 μl of 100 μg/ml ribonuclease (Nippon Gene) and 50 μg/ml propidium iodide (FUJIFILM Wako). After incubation at room temperature for 30 min, samples were analyzed using FACSVerse (BD BioSciences) with propidium iodide fluorescence being measured at 694 nm.

Tumor allograft models

Eight‐week‐old C57BL/6 mice were anesthetized and injected subcutaneously with murine uterine cancer cell line HSML or murine melanoma cell line B16 at 1 × 10⁷ cells in 100 μl phosphate‐buffered saline (PBS). Tumors were isolated either on Day 2/Day 4 after injection for pre/post‐angiogenic studies, or extracted after 20 days of atorvastatin/CCL2 inhibitor treatment. In pre‐atorvastatin treatment model, mice were treated with 10 mg/kg/day of atorvastatin (Tokyo Chemical Industry) or vehicle (PBS) intraperitoneally for 7 days before tumor injection (Fig 6A). In post‐atorvastatin treatment model, mice were treated with 10 mg/kg/day of atorvastatin (Tokyo Chemical Industry) or vehicle (PBS) intraperitoneally for 10 days after Day 10 of tumor injection (Fig 6C). In CCL2 inhibitor model, mice were treated with 10 mg/kg every 5 days of bindarit (Cayman) or vehicle (5% DMSO in PBS) intraperitoneally after tumor injection. To estimate the tumor volume, the long (L) and short (S) axis of tumor were measured, and the tumor volume was calculated as L × S × S/2 (Sápi et al, 2015).

Immunohistochemical analysis of tumor tissue, bone marrow tissue, and ells

Tumors were isolated, directly embedded in optimal cutting temperature (OCT) compound, and stored in −80°C until analysis. Frozen tumor tissues were cut 15 μm thick by using Cryostat CM1950 (Leica), fixed with 4% paraformaldehyde, and stained with hamster anti‐mouse CD31 (Sigma‐Aldrich, clone 2H8, 1:100), rat anti‐mouse CD11b (BD Biosciences, clone M1/70, 1:100), rat anti‐mouse F4/80 (Bio‐Rad Laboratories, clone CI:A3‐1, 1:100), rat anti‐mouse Ly6G (BioLegend, clone 1A8, 1:100), rabbit anti‐mouse HMGCS1 (GeneTex, 1:100), or rabbit anti‐mouse SREBP2 (Abcam, 1:100) primary antibodies overnight at 4°C. The sections were then incubated for 60 min at room temperature with Alexa Fluor 488‐conjugated goat anti‐rat (Thermo Fisher Scientific), Alexa Fluor 568‐conjugated goat anti‐hamster (Thermo Fisher Scientific), or Alexa Fluor 568‐conjugated goat anti‐rabbit (Thermo Fisher Scientific) secondary antibodies, which were diluted at 1:1,000 in PBS. Tumor sections were mounted on Dapi‐Fluoromount‐G (SouthernBiotech) and then analyzed using a confocal microscope (STELLARIS 5, Leica).

Bone tissues were dissected from 8‐week‐old C57BL/6 mice, fixed with 4% paraformaldehyde, embedded in OCT compound, and stored in −80°C until analysis. Frozen bone marrow tissues were cut 15 μm thick by using Cryostat CM1950 (Leica) and stained with rabbit anti‐mouse HMGCS1 (GeneTex, 1:100) or rabbit anti‐mouse SREBP2 (Abcam, 1:100) primary antibodies overnight at 4°C. The sections were then incubated for 60 min at room temperature with Alexa Fluor 568‐conjugated goat anti‐rabbit (Thermo Fisher Scientific, 1:1,000) secondary antibody. Bone marrow sections were mounted on Dapi‐Fluoromount‐G (SouthernBiotech) and then analyzed using a confocal microscope (STELLARIS 5, Leica). To label hypoxic regions in bone marrow and tumor tissues, 60 mg/kg pimonidazole hydrochloride was intravenously injected into mice 2 h before sacrifice and collection of tissues. Pimonidazole was detected on bone marrow and tumor sections using the Hypoxyprobe™‐1 Plus Kit (Hypoxyprobe, Inc.) according to the manufacturer's instruction.

THP‐1 and A‐THP‐1 cells were fixed in pre‐warmed 4% PFA 10 min, washed with PBS, and then permeabilized in 0.1% Triton X‐100 in PBS for 15 min. Cells were stained with rabbit polyclonal anti‐TGN46 antibody (Sigma‐Aldrich, clone TGN46‐8, 1:100), rabbit polyclonal anti‐SREBP2 antibody (Abcam, 1:100) or rabbit polyclonal anti‐KDEL antibody (Abcam, 1:100). The cells were incubated with appropriate secondary antibodies, mounted on Dapi‐Fluoromount‐G (SouthernBiotech), and then analyzed using a confocal microscope (STELLARIS5, Leica). Excitation for DAPI, Alexa 488, and Alexa 568 chromophores was provided by a 405‐, 488‐, 561‐, and 637‐nm laser, respectively, and a Plan‐Apochromat oil objective (×63, NA 1.4). The Lightning mode (Leica) was used to generate deconvolved images. Microscope acquisitions were controlled by LAS X (v. 4.2.1) software from Leica. FL‐IHC images, TGN46‐, KDEL‐, SREBP2‐, HMGCS1‐, CD31‐, CD11b‐, F4/80‐, and Gr‐1‐positive area were determined with Fiji software (Schneider et al, 2012). Each positive area was binarized and quantified using > the Analyze particle tool.

Expression array analysis

GeneChip Human Genome U133 plus 2.0 oligonucleotide arrays (Thermo Fisher Scientific) were used to measure genomic expression level of HeLa, NHDF, and THP‐1 cells. Data were collected by GeneChip Scanner 3000 (Thermo Fisher Scientific) and analyzed by GeneChip Operating Software v1.3 by MAS5 algorithms, where the average signal in each array was normalized to 100. Genes expressed more than 2‐fold under hypoxia compared to normal (530 genes in THP cells, 561 genes in HeLa cells, 537 genes in fibroblast) and genes expressed less than 0.5‐fold (433 genes in THP cells, 334 genes in HeLa cells, 432 genes in fibroblast) were performed for pathway overexpression analysis using 4,588 publicly available gene sets obtained from the IMPaLA (http://impala.molgen.mpg.de/) (Kamburov et al, 2011) database. The obtained P‐values were corrected using the FDR method (Benjamini & Hochberg, 1995). The detailed results of the pathway overexpression analysis are summarized in Dataset EV1.

Gene expression analysis using real‐time PCR

Total RNA was extracted from cells using the Isogen reagent (Nippon Gene), converted to cDNA by using the Prime Script reverse transcriptase (Takara) as per the manufacturer's instructions, and used for quantitative real‐time PCR amplification with SYBR Green (Toyobo) and indicated primers (Table EV2A). All results of mRNA expression level were normalized by the expression level of ACTB.

Chromatin immunoprecipitation

THP‐1 and A‐THP‐1 (1 × 10⁷ cells) were crosslinked with 1% formaldehyde for 10 min at room temperature. Fixation was stopped by adding 0.125 M glycine. Nuclear pellets were prepared for ChIP as described previously. Pre‐washed magnetic Dynabeads (Invitrogen) were incubated with anti‐H3K4me3 antibody (Active Motif, 1:300), anti‐H3K27ac antibody (FUJIFILM Wako, clone MABI0309, 1:300), or anti‐SREBP2 antibody (Cayman, 1:100) in ChIP Dilution buffer (16.7 mM Tris–HCl (pH 8.0), 167 mM NaCl, 1.2 mM EDTA, 1.1% Triton X‐100, Protease Inhibitor Cocktail (Sigma‐Aldrich)) for 6 h with wheel rotating at 4°C. Subsequently, sonicated crosslinked nuclear lysates were added and incubated overnight at 4°C by wheel rotating. The beads were washed several times and eluted with elution buffer (25 mM Tris–HCl (pH 7.5), 5 mM EDTA, 0.5% SDS, 150 mM NaCl), and the eluent was incubated with pronase (1 mg/ml) at 42°C for 2 h and then incubated at 65°C overnight. DNA was purified using QIA quick PCR purification kit (QIAGEN). ChIP quantitative PCR was performed with indicated primers (Table EV2B).

Silencing of SREBP1/2, HIF1ɑ, and SCAP

siRNAs designed against human SREBP1, SREBP2, HIF1ɑ, and SCAP were obtained commercially (Thermo Fisher Scientific). The sequences of siRNAs are summarized in Table EV2C. Cells were transfected with siRNAs against target genes or negative (scramble) control siRNA using Lipofectamine RNAiMAX transfection reagent (Thermo Fisher Scientific) according to the instructions of the manufacturer. The ability of the siRNA to inhibit target gene expression was assessed 48 h post‐transfection. After the transfection, cells were treated under normoxia or hypoxia for 24 h for subsequent analyses.

Western blotting

Whole‐cell extracts were separated into nuclear extracts and the membrane fractions from THP‐1 and A‐THP‐1 cells as previously described (Sakai et al, 1996) with minor modification. For immunoblot analysis, aliquots of proteins were separated by SDS–PAGE and transferred to nitrocellulose membrane (Bio‐Rad Laboratories). Immunodetection was carried out with mouse anti‐human SREBP2 antibody (generated by Prof. Hamakubo in the University of Tokyo, 1:100), rabbit anti‐histone H3 antibody (Abcam, 1:2,000), rabbit anti‐GRP78 BiP antibody (Abcam, 1:1,000) in combination with peroxidase‐conjugated affinity‐purified donkey anti‐mouse (Sigma‐Aldrich, 1:1,000), or anti‐rabbit (Cell Signaling Technology, 1:1,000) IgG, and then visualized using SuperSignalTM West Dura Extended Duration Substrate (Thermo Fisher Scientific). Luminescence images were analyzed by luminescent image analyzer (Fusion FX, Vilber).

Proximity ligation assay (PLA) staining

THP‐1 cells were fixed in pre‐warmed 4% paraformaldehyde in PBS for 10 min and permeabilized in 0.2% Triton X‐100 in PBS for 15 min when necessary. After the cells were incubated with rabbit polyclonal anti‐TGN46 antibody (Sigma‐Aldrich, clone TGN46‐8, 1:100), or rabbit polyclonal anti‐KDEL antibody (Abcam, 1:100) overnight at 4°C, in situ protein interactions were detected using the Duolink proximity ligation assay kit according to the manufacturer's instructions (Sigma‐Aldrich) (Aki et al, 2015).

Cellular cholesterol measurement assay

Cellular total cholesterol was measured as per the manufacture's protocols of Total Cholesterol Assay Kit (Fluorometric) (Cell Biolabs). Cells were washed three times with cold PBS prior to lysis. 10⁶ cells were extracted with 200 μl lysing reagent (a mixture of chloroform, isopropanol, and NP‐40 at the ratio of 7:11:0.1) in a micro‐homogenizer. The extract was centrifuged 10 min at 15,000 g and the liquid (organic phase) were transferred to a new tube, air dried at 50°C, and vacuumed for 30 min to remove organic solvent. The dried lipids were dissolved in 200 μl of 1× Assay Diluent, and the solution is vortexed until homogenous. 50 μl of extracted solution were mixed with 50 μl of the prepared Cholesterol Reaction Reagent and incubated for 45 min at 37°C. After incubation, fluorescent intensity of excitation at the 544 nm range and emission at the 590 nm range was measured by ARVO X3 (PerkinElmer).

PMA‐induced differentiation of THP‐1

Differentiation of THP‐1 monocytes to macrophages was induced by treatment of THP‐1 cell line (2 × 10⁵ cells/ml) with 200 nM Phorbol 12‐myristate 13‐acetate (PMA) (AdipoGen) for 24 or 48 h. After PMA stimulation, differentiation was enhanced by removing PMA‐containing medium then incubating cells in fresh RPMI 1640 medium (supplemented with 10% FBS) for another 24 or 48 h (Daigneault et al, 2010).

RNA‐sequencing data analysis

RNA‐seq reads were aligned to human transcriptome (UCSC gene) and genome (GRCh37/hg19) references, respectively, using Burrows‐Wheeler Aligner. After transcript coordinate was converted to genomic positions, an optimal mapping result was selected either from transcript or genome mapping by comparing the minimal edit distance to the reference. Local realignment was performed within in‐house short reads aligner with smaller k‐mer size (k = 11). Finally, fragments per kilobase of exon per million fragments mapped (FPKM) values were calculated for each UCSC gene while considering strand‐specific information.

Single‐cell RNA‐seq data processing

Single‐cell RNA‐seq data of hematopoietic stem cells from human adult bone marrow (preprint: Mende et al, 2020) were downloaded from the Human Cell Atlas Data Portal (ERP120138/PRJEB36885) and were analyzed with the Seurat R package (Butler et al, 2018) (version 4.0.1). We used reciprocal principal component analysis (PCA) and reference‐based integration for integrating 58,535 cells from 12 different datasets. This workflow was available from the Seurat web site (https://satijalab.org/seurat/articles/integration_large_datasets.html). The parameters for the all Seurat functions for reference‐based integration (Normalize Data, Find Variable Features, Select Integration Features, Scale Data, Run PCA, Find Integration Anchors, and Integrate Data) were used with default settings. Dimensionality reduction for the integrated dataset was then performed using PCA, and UMAP plots were generated by the Seurat function Run UMAP with the first 30 principal components (PCs) as input. We calculated signature scores for hematopoietic stem cells, monocytes, and macrophages using scanpy.tl.score_genes implemented in the scanpy Python package (Wolf et al, 2018) (version 1.7.2). The marker genes of each cell type derived from Panglao DB (Franzén et al, 2019) for gene_list arguments in scanpy.tl.score_genes. When we visualize the expression patterns in UMAP representation, we set the maximum values to 99% quantiles of the original expression values.

Single‐sample enrichment analysis, cell clustering, and correlation analysis

To compute an enrichment score for gene set related to THP‐1 hypoxia‐related genes in individual cell, single‐sample GSEA (ssGSEA) was performed with the R package GSVA (Hänzelmann et al, 2013) (version 1.38.2). Next, in order to determine cell clusters, unsupervised clustering was performed using the Seurat functions Find Neighbors with default parameters and Find Clusters with resolution 0.5 to generate 20 clusters. For 28,287 genes and each cluster, we identified statistically significant differences between two states (cells with high and low THP‐1 hypoxia enrichment) by performing two sample t‐test. Here, cells with high THP‐1 hypoxia enrichment were defined as those with THP‐1 hypoxia enrichment scores above the 90th percentile, and cells with low THP‐1 hypoxia enrichment were defined as those with THP‐1 hypoxia enrichment scores below the 10th percentile. In order to analyze the correlation between hypoxia and cholesterol biosynthesis for each cluster, gene set enrichment analysis by using estimated t‐test statistics of 28,287 genes was performed for investigating an enrichment of gene set related to cholesterol biosynthesis with the R package fgsea (preprint: Korotkevich et al, 2016) (version 1.16.0).

Metabolite extraction and metabolomic analysis using CE‐MS/LC‐MS

2 × 10⁷ monocyte/macrophage lineage cells were isolated from bone marrow of C57BL/6 mice, washed twice with PBS, and extracted with a mixture of methanol and chloroform at 1,000:400. Hydrophilic metabolites of cholesterol pathway (HMG‐CoA, Mevalonate, Mevalonate‐5‐P) in mouse‐derived monocyte/macrophage lineage cells were measured using capillary electrophoresis‐mass spectrometry (CE‐MS) (Agilent Technologies, Santa Clara, CA). Hydrophobic metabolites of cholesterol pathway (7‐dehydrocholesterol, desmosterol, cholesterol) in same cells were measured using liquid chromatography mass spectrometry (LC‐MS) (Agilent Technologies, Santa Clara, CA). For CE‐MS, commercially available COSMO(+) (chemically coated with cationic polymer) capillary (50 mm i.d. × 105 cm) was used with a 50 mM ammonium acetate solution (pH 8.5) as the electrolyte (Nacalai Tesque, Kyoto, Japan). Methanol/5 mM ammonium acetate (50% v/v) containing 0.1 mM hexakis (2,2‐difluoroethoxy) phosphazene was run as the sheath liquid at 10 ml/min. ESI‐TOFMS was performed in negative ion mode, and the capillary voltage was set to 3.5 kV. For the anion analysis, trimesate and CAS were used for the reference of the internal standards, respectively. CE‐TOFMS raw data were analyzed using Master Hands software (ver, 2.17.0.10). For each experiment, data conversion, binning data into 0.02 m/z slices, baseline elimination, peak picking, integration, and elimination of redundant features were conducted to yield the possible peaks. Data matrices were aligned based on corrected migration times, and metabolites were assigned to the aligned peaks by matching m/z and the corrected migration times using standards metabolite library. Relative peak areas were calculated by the ratio of peak area divided by the internal standards. Metabolite concentrations were calculated based on the relative peak area between the sample and the standard. LC‐MS measurement and data analysis were performed as previously reported (Yamashita et al, 2010).

Immune cell profiling in blood or bone marrow by flow cytometry analysis

Eight‐week‐old C57BL/6 mice were treated with vehicle or atorvastatin (10 mg/kg/day) for 7 days before sacrifice and isolation of cells from peripheral blood or bone marrow. The resultant cell suspension was passed through cell strainer (BD Falcon) and treated with RBC lysis buffer (Invitrogen). After washing by PBS, cells were first incubated with anti‐CD16/32 antibody (BioLegend) for 5 min on ice. The following antibodies were purchased from Biolegend and used: APC or PerCP‐Cy5.5 anti‐CD11b mAb (M1/70); PE anti‐Ly6C mAb (HK1.4); APC anti‐CD11c mAb (N418); FITC anti‐NK1.1 mAb (PK136); FITC anti‐F4/80 mAb (BM8); Pacific Blue anti‐Ly6G mAb (1A8); Pacific Blue or APC‐Cy7 anti‐CD45 mAb (IM7); PE anti‐CD4 mAb(GK1.5); APC anti‐B220 mAb (RA3‐6B2); and PerCP‐Cy5.5 anti‐CD8 mAb(53‐6.7). The cells were then stained with antibodies in PFE (PBS with 2% FBS and 1 mM of EDTA) for 20 min on ice and analyzed by BD LSR Fortessa (BD Biosciences). Flow cytometry data were collected by FACSDiva (BD Biosciences). Collected data were analyzed with FlowJo software (v10.2, BD BioSciences).

Migration assay of THP‐1

10⁵ THP‐1 cells transfected with siRNA negative control or siRNA targeting SREBP2 for 48 h were seeded on the upper chamber of 8‐μm pore trans‐well plate (Corning) in 200 μl of RPMI 1640 supplemented with 1% FBS. In the lower chamber, 500 μl RPMI 1640 supplemented with 10% FBS was added. Cells were cultured for 8 h under either control or hypoxic condition, and the number of migrated cells in lower chamber was counted by TC20 Automatic Cell Counter (Bio‐Rad).

Survival analysis

Survival analysis was performed using the expression of genes related to hypoxic conditions in THP‐1 cells. We defined hypoxia‐related genes as the 530 genes whose expression levels in the case were up‐regularized more than two times under hypoxic conditions compared to that in the control, according to the results of our previous microarray analysis of THP‐1 cells (Table EV3). To construct the prognostic model of cancer patients based on the expression profiles of hypoxia‐related genes, we downloaded 31 normalized RNA‐seq expression datasets (adrenocortical carcinoma, bladder urothelial carcinoma, breast invasive carcinoma, cervical squamous cell carcinoma and endocervical adenocarcinoma, cholangiocarcinoma, colon adenocarcinoma, lymphoid neoplasm diffuse large B‐cell lymphoma, esophageal carcinoma, glioblastoma multiforme, head and neck squamous cell carcinoma, kidney chromophobe, kidney renal clear, cell carcinoma, kidney renal papillary cell carcinoma, acute myeloid leukemia, brain lower grade glioma, liver hepatocellular carcinoma, lung adenocarcinoma, lung squamous cell carcinoma, mesothelioma, ovarian serous cystadenocarcinoma, pancreatic adenocarcinoma, prostate adenocarcinoma, rectum adenocarcinoma, sarcoma, skin cutaneous melanoma, stomach adenocarcinoma, thyroid carcinoma, thymoma, uterine corpus endometrial carcinoma, uterine carcinosarcoma, and uveal melanoma) and 226 RNA microarray datasets (two datasets of adrenocortical carcinoma, five datasets of bladder cancer, one dataset of osteosarcoma, 12 datasets of brain tumor, 81 datasets of breast cancer, one dataset of cervical cancer, 21 datasets of colorectal cancer, one dataset of uveal melanoma, four datasets of gastric cancer, 16 datasets of lymphoma, five datasets of head and neck squamous cell carcinoma, three datasets of acute myeloid leukemia, six datasets of hepatocellular carcinoma, 21 datasets of lung cancer, two datasets of neuroblastoma, 29 datasets of ovarian cancer, three datasets of pancreatic adenocarcinoma, two datasets of prostate cancer, four datasets of renal cell carcinoma, four datasets of sarcoma, and three datasets of melanoma) of cancer cohort studies from the Broad TCGA GDAC web site (http://gdac.broadinstitute.org/) and the Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/). Then, we used a Lasso‐regularized Cox proportional hazard model with the glmnet package (version 4.0.2) in the R statistical environment (version 4.0.0) to build a prognostic classifier based on hypoxia‐related genes. The tuning parameter λ in the Lasso regularization was chosen by corrected Akaike's information criterion (Sugiura, 1978; Hurvich & Tsai, 1995). The results of the Cox model in each cohort are summarized in Dataset EV2. Patients were classified into two groups based on whether the risk score in the Cox model was more than 0 (high risk and worse prognosis) or less than 0 (low risk and better prognosis). To evaluate the prognostic significance of the Cox model, we used the Kaplan–Meier method and the P‐value was calculated using the log‐rank test. For adjustment for multiple hypothesis testing, the P‐values were corrected by the Benjamini–Hochberg procedure (Hochberg & Benjamini, 1990).

Survival analysis in all tumor types of TCGA was also performed using the Gene Expression Profiling Interactive Analysis (Tang et al, 2019) (GEPIA2, http://gepia2.cancer‐pku.cn) tool. Overall survival (OS) based on the expression of cholesterol synthesis‐related genes normalized by the expression of monocyte marker genes (CD11b) was observed, and Kaplan–Meier curves were drawn for tumor types with log‐rank P‐values < 0.05. The cutoff for the groups was 50% of the expression of cholesterol synthesis‐related genes.

Statistics

Subjective bias was minimized when allocating animals/samples to treatment using randomization procedure. No data points were excluded or omitted from analysis. Plotted values are shown as means ± S.E.M. throughout this study. Indicated P‐values were obtained using one‐way analysis of variance (ANOVA) followed by Student–Newman–Keuls multiple comparisons for post hoc test, paired Student's t‐tests, or Mann–Whitney U‐test. P‐values are denoted as *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns: not significant.

Author contributions

Ryuichi Nakahara: Data curation; formal analysis; validation; investigation; methodology; writing – original draft; writing – review and editing. Sho Aki: Data curation; formal analysis; investigation; visualization; methodology. Maki Sugaya: Data curation; formal analysis; validation; investigation; methodology. Haruka Hirose: Software; formal analysis; validation; investigation; methodology. Miki Kato: Data curation; formal analysis. Keisuke Maeda: Data curation; validation. Daichi M Sakamoto: Visualization; methodology. Yasuhiro Kojima: Data curation; software; formal analysis; investigation. Miyuki Nishida: Data curation; formal analysis; validation. Ritsuko Ando: Software; visualization. Masashi Muramatsu: Data curation; formal analysis. Melvin Pan: Data curation. Rika Tsuchida: Formal analysis; validation. Yoshihiro Matsumura: Data curation; validation. Hideyuki Yanai: Data curation; formal analysis; validation; investigation; methodology. Hiroshi Takano: Investigation; methodology. Ryoji Yao: Investigation; methodology. Shinsuke Sando: Investigation; methodology. Masabumi Shibuya: Investigation; methodology. Juro Sakai: Formal analysis; validation; investigation; methodology. Tatsuhiko Kodama: Investigation; methodology. Hiroyasu Kidoya: Data curation; formal analysis; investigation; visualization. Teppei Shimamura: Data curation; software; formal analysis; visualization; methodology. Tsuyoshi Osawa: Conceptualization; supervision; funding acquisition; validation; investigation; writing – original draft; project administration; writing – review and editing.

Disclosure and competing interests statement

The authors declare that they have no conflict of interest.

Supporting information

Expanded View Figures PDF

Table EV1

Table EV2

Table EV3

Dataset EV1

Dataset EV2

Source Data for Expanded View

PDF+

Source Data for Figure 1

Source Data for Figure 2

Source Data for Figure 3

Source Data for Figure 4

Source Data for Figure 5

Source Data for Figure 6

The EMBO Journal (2023) 42: e114032

Data availability

DNA microarray data related to Fig 1B and C are available in Gene Expression Omnibus (GEO) under the accession number GSE239711 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE239711). RNA‐sequencing data related to Fig 2C are available in Gene Expression Omnibus (GEO) under the accession number GSE239740 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE239740).