PKM2-TMEM33 axis regulates lipid homeostasis in cancer cells by controlling SCAP stability

Abstract

The pyruvate kinase M2 isoform (PKM2) is preferentially expressed in cancer cells to regulate anabolic metabolism. Although PKM2 was recently reported to regulate lipid homeostasis, the molecular mechanism remains unclear. Herein, we discovered an ER transmembrane protein 33 (TMEM33) as a downstream effector of PKM2 that regulates activation of SREBPs and lipid metabolism. Loss of PKM2 leads to up-regulation of TMEM33, which recruits RNF5, an E3 ligase, to promote SREBP-cleavage activating protein (SCAP) degradation. TMEM33 is transcriptionally regulated by nuclear factor erythroid 2-like 1 (NRF1), whose cleavage and activation are controlled by PKM2 levels. Total plasma cholesterol levels are elevated by either treatment with PKM2 tetramer-promoting agent TEPP-46 or by global PKM2 knockout in mice, highlighting the essential function of PKM2 in lipid metabolism. Although depletion of PKM2 decreases cancer cell growth, global PKM2 knockout accelerates allografted tumor growth. Together, our findings reveal the cell-autonomous and systemic effects of PKM2 in lipid homeostasis and carcinogenesis, as well as TMEM33 as a bona fide regulator of lipid metabolism.

Keywords: PKM2; SCAP degradation; TMEM33; total cholesterol levels; tumor growth.

Figure 1. Lipid metabolism pathways are impaired in PKM2 deficient cells

1. Heatmap showing the altered metabolic pathways in PKM2 KO MCF7 and MDA‐MB‐231 cells relative to the parental cells (n = 4). Fatty acid metabolism (Acyl Carnitine) and long‐chain fatty acid pathways were significantly changed in both cell lines.

2. Heatmap showing the differential fatty acid β‐oxidation pathway metabolite levels between PKM2 KO MCF7 and MDA‐MB‐ 231 cells, and their respective parental counterparts (n = 4).

3. Decreased levels of free carnitine and modified carnitine species in PKM2 KO cell lines as compared with their parental counterparts (n = 4).

4. Heatmap showing the decreased levels of long‐chain fatty acids in PKM2 KO cell lines relative to their parental counterparts (n = 4).

5. Increased acetyl‐CoA in PKM2 KO MCF7 and MDA‐MB 231 cells as compared with their parental counterparts (n = 4).

6. Decreased total cholesterol in PKM2 KO MCF7 and MDA‐MB 231 cells as compared with their parental counterparts (n = 3).

7. Confocal imaging of neutral lipid droplets stained with BODIPY493/503 or Nile red dye in PKM2 KO cells relative to the parental cells. Data represent one of three independent experiments with similar results. Scale bar, 25 μm.

Data information: Data are displayed (C, E and F) as means ± SD. Statistical significance was assessed using two‐tailed Student’s t‐test, *P < 0.05, **P < 0.01 and ***P < 0.001. n.s means no significance. Source data are available online for this figure.

Figure 2. SREBP activation is impaired in PKM2 KO breast cancer cells

1. Heatmap showing the differential gene expression patterns between MDA‐MB‐231 PKM2 KO and parental cells (n = 3).

2. Gene set enrichment analysis of the top enriched gene signatures associated with PKM2 knockout in (A). Hallmark 50 gene sets and KEGG 186 gene sets were used for this analysis.

3. Decreased cholesterol homeostasis signature in MDA‐MB‐231 PKM2 KO cells. Cholesterol homeostasis signature was generated by GSEA (v4.1.0) with adjusted P‐value and normalized enrichment score (NES).

4. Heatmap showing the differentially expressed genes related to cholesterol homeostasis in PKM2 KO versus parental MDA‐MB‐231 cells (n = 3).

5. Volcano plot showing proteomic changes in PKM2 KO versus parental MDA‐MB‐231 cells. Blue triangle symbol indicated PKM2 level; TMEM33, PKM1, and RNF5 were labeled in red dots. Horizontal dashed line indicated P‐value 0.05 and vertical dashed lines indicated 2‐fold cutoff (n = 3).

6. Correlation of PKM2 deficiency‐induced protein changes between MCF7 and MDA‐MB‐231 cells. Green highlights down‐regulated lipid‐related proteins. PKM1, TMEM33, OCIAD2, and TIMM8A were up‐regulated and highlighted in red dots.

7. Western blot analysis of TMEM33, ABCA1, SREBP1, HMGCR, SCAP, FASN, LDLR, FDFT1, I DI1, RNF5, INSIG 1 and 2, and calnexin in PKM2 KO MCF7 and MDA‐MB‐231 cells, and their respective parental cells. Arrow indicates nuclear SREBP1. P: precursor, N: nuclear.

Source data are available online for this figure.

Figure EV1. PKM2 deficiency results in decreased lipid synthesis in breast cancer cells

1. Heatmap showing differential gene expression profiling in PKM2 KO versus wild‐type MCF7 cells (n = 3).

2. Gene set enrichment analysis of the top enriched signatures affected by PKM2 KO in MCF7. Hallmark 50 gene sets and KEGG 186 gene sets were used for this analysis.

3. The cholesterol homeostasis signature was decreased by knocking out PKM2 in MCF7 cells. Adjusted P‐value and normalized enrichment score (NES) were generated by GSEA.

4. Decreased expression of lipid synthetic genes in PKM2 KO MCF7 cells analyzed by qRT‐PCR (n = 3).

5. Decreased expression of lipid synthetic genes in PKM2 KO MDA‐MB‐231 cells analyzed by qRT‐PCR (n = 3).

6. Ingenuity pathway analysis (IPA) of proteins affected by PKM2 KO in MDA‐MB‐231 cells in Fig 2F. Cholesterol synthesis pathways were highlighted in red.

Data information: Data are displayed (D and E) as means ± SD. Statistical significance was assessed using two‐tailed Student’s t‐test, *P < 0.05, **P < 0.01 and ***P < 0.001.Source data are available online for this figure.

Figure 3. TMEM33 is the downstream effector of PKM2 controlling lipid metabolism gene expression

1. A

   Immunofluorescence of SREBP1 in MCF7 parental and PKM2 KO cells by confocal imaging. Scale bar, 5 μm.

2. B

   Quantitation of nuclear SREBP1 intensity in A. n ≥ 25 cells as in (A) from three different fields.

3. C

   Western blot analyses of nuclear and cytoplasmic SREBP1 levels in parental and two PKM2 KO clones. PCNA serves as a nuclear protein loading control.

4. D

   Western blot analyses of TMEM33, RNF5, SCAP, SREBP1, and IDI1 in PKM2 KO MCF7 and MDA‐MB‐231 cells stably transfected with shControl (shCtrl) or shTMEM33 constructs. Arrow indicates nuclear SREBP1, N: nucleus and P: precursor.

5. E

   Cell proliferation assays of the PKM2 KO MCF7 cells stably transfected with shCtrl or shTMEM33 constructs (n = 3). shCtrl: shControl.

6. F

   Cell migration assays of PKM2 KO MDA‐MB‐231 cells stably transfected with shCtrl or shTMEM33 constructs (n = 3). Scale bars, 50 µm.

7. G

   Western blot analysis of TMEM33, RNF5, SCAP, SREBP1, HMGCR, LDLR, FDFT1, IDI1, and PKM2 in wild type, PKM2 KO, TMEM33 OE (pool), and two TMEM33OE clones (clones #1 and #8) MDA‐MB‐231 cells. Arrow indicates nuclear SREBP1. N: nucleus and P: precursor.

8. H, I

   Gene set enrichment analysis of the top enriched gene signatures (H), including (I) Cholesterol homeostasis signature in MDA‐MB‐231 TMEM33 OE (#1) cells using KEGG and hallmark gene sets in GSEA databases. Adjusted P‐value and normalized enrichment score (NES) were generated by GSEA.

9. J

   Volcano plot showing the altered transcripts induced by TMEM33 OE (clone #1) (P < 0.05) and PKM2 KO (P < 0.05) in MDA‐MB‐231 cells. Green depicts shared down‐regulated genes by both TMEM33 OE and PKM2 KO; red depicts shared up‐regulated genes by both TMEM33 OE and PKM2 KO. Variance analysis (ANOVA) was used to calculate the P‐value.

10. K

    Volcano plot showing protein changes induced by TMEM33 OE (P < 0.05) and PKM2 KO (P < 0.05) in MDA‐MB‐231 cells. TMEM33 labeled in red dot was up‐regulated in both TMEM33 OE and PKM2 KO cells and PKM2 labeled in purple had no change in TMEM33 OE cells. ANOVA was used to calculate the P‐value.

Data information: Data are displayed (B and E) as means ± SD. Statistical significance was assessed using two‐tailed Student’s t‐test, *P < 0.05 and ***P < 0.001. Source data are available online for this figure.

Figure EV2. The PKM2‐TMEM33 axis regulates cholesterol and fatty acid synthesis‐related gene expression

1. Co‐immunoprecipitation of endogenous PKM2 and sec61β in MCF7 cells treated with DMSO or 20 µM TEPP‐46.

2. Reciprocal co‐immunoprecipitation of TMEM33 and sec61β in MCF7 cells. TMEM33 and sec61β were detected by western blotting.

3. TMEM33 expression is regulated by PKM2, not PKM1, in MCF7 cells. MCF7 cells were stably transfected with a shRNA targeting PKM 3′ untranslated region (UTR) to eliminate endogenous PKM. Either PKM1 or PKM2 was restored using a retroviral expression system. A scramble shRNA vector was used as a control (shCtrl).

4. Knocking down TMEM33 led to increased total cholesterol in PKM2 KO MDA‐MB 231 cells (n = 3).

5. Knocking down TMEM33 led to increased lipid droplets in PKM2 KO MDA‐MB 231 cells. A representative confocal imaging data from three independent experiments is shown. Scale bar, 25 µm.

6. Western blotting of TMEM33 in MCF7 cells stably expressing control (Ctrl) or FLAG‐TMEM33 vectors. Exo. denotes exogenous and endo. denotes endogenous.

7. Cell growth curves of MCF7 cells stably transfected with Ctrl or FLAG‐TMEM33 vectors (n = 4).

8. Volcano plot showing the proteome changes induced by TMEM33 OE (#1) relative to control in MDA‐MB‐231 cells (n = 3). Two‐tailed Student’s t‐test was used to calculate the P‐value. TMEM33 level was up‐regulated and marked in red.

9. Down‐regulation of cholesterol homeostasis (upper panel) and fatty acid metabolism (lower panel) signatures in TMEM33 OE (#1) MDA‐MB‐231 cells relative to the control cells. Proteomics data in panel (I) were used to extract gene signature. Adjusted P‐value and normalized enrichment score (NES) were generated by GSEA.

Data information: Data are displayed (D and G) as means ± SD. Statistical significance was assessed using two‐tailed Student’s t‐test, *P < 0.05 and **P < 0.01.Source data are available online for this figure.

Figure 4. TMEM33 recruits RNF5 to promote SCAP polyubiquitination

1. Increased SCAP polyubiquitination in TMEM33‐transfected HEK293T cells. 293T cells were transiently transfected with Halo‐tagged TMEM33 for 18 h followed by treatment with MG‐132 for 3 h.

2. Western blot analysis of PKM2, PKM1, NRF1, TMEM33, RNF5, and SCAP levels in MEF (PKM2^fl/fl, Cre‐ER^T2) cells treated with 4‐OHT in a time course experiment.

3. Co‐immunoprecipitation of endogenous SCAP and RNF5 in Flag‐tagged TMEM33 stably expressed MCF7 cells. SEC61β is an ER protein, which is found to interact with TMEM33.

4. Reciprocal co‐immunoprecipitation of TMEM33 and RNF5 in MDA‐MB‐231 cells. TMEM33, RNF5, and SCAP were detected by Western blotting.

5. Proximity ligation assays of TMEM33 and RNF5 in MDA‐MB‐231 cells using confocal imaging. NC means negative control. Scale bar, 10 μm.

6. Interaction between in vitro translated TMEM33 and RNF5. FLAG‐tagged vector or TMEM33 and HA‐tagged RNF5 were expressed using an in vitro transcriptional and translational kit, followed by FLAG antibody pull‐down.

7. Increased polyubiquitination of SCAP in RNF5‐transfected HEK293 cells. 293T cells were transiently transfected HA‐tagged RNF5 for 18 h followed by MG‐132 treatment for 3 h. Arrow indicates HA‐tagged RNF5.

8. RNF5 is involved in TMEM33‐mediated SCAP polyubiquitination in 293T cells. 293T cells were stably transfected with shCtrl or shRNF5 constructs; then, the indicated plasmids were co‐transfected into shCtrl or shRNF5 expressing 293T cells for 18 h followed by MG‐132 treatment for 8 h. FLAG M2 beads were used for pull‐down.

9. The relative binding affinity of FLAG‐tagged full‐length and truncated SCAP proteins with endogenous TMEM33 and RNF5. The band intensity of TMEM33 and RNF5 in Appendix Fig S2D was quantified and normalized with the FLAG control. SCAP protein encompasses eight transmembrane (TM) helices in N terminus (TM 2‐6: sterol‐sensing domain) and a WD40 domain in C terminus with total of 1,279 aa.

10. The relative binding affinity of HA‐tagged full‐length and truncated TMEM33 with endogenous SCAP and RNF5. The band intensity of SCAP and RNF5 in Appendix Fig S2E was quantified and normalized to the HA control. TMEM33 protein encompasses three transmembrane (TM) helices in N terminus and a C‐terminal cytosolic domain with a total of 247 aa.

Source data are available online for this figure.

Figure 5. RNF5 promotes SCAP degradation in breast cancer cells

1. Lysine 48 linked polyubiquitinated SCAP levels in parental and PKM2 KO MDA‐MB‐231 cells.

2. SCAP‐GFP stability in parental and PKM2 KO MDA‐MB‐231 cells (n = 5). Cells stably expressing SCAP‐GFP were treated with cycloheximide (CHX). GFP signal was measured at the indicated time.

3. Western blot analyses of RNF5 and SCAP in shCtrl and shRNF5 PKM2 KO MDA‐MB‐231 cells.

4. qRT‐PCR analyses of selected SREBPs downstream genes in shCtrl and shRNF5 PKM2 KO MDA‐MB‐231 cells (n = 3).

5. Western blot analyses of RNF5, SCAP, and IDI1 in Ctrl and HA‐RNF5 OE MCF7 cells.

6. qRT‐PCR analyses of SREBPs downstream genes in Ctrl and HA‐RNF5 OE MCF7 cells (n = 3).

7. Cell proliferation of Ctrl and HA‐RNF5 OE MCF7 cells measured by MTT assays (n = 8).

Data information: Data are displayed (B, D, F, and G) as means ± SD. Statistical significance was assessed using two‐tailed Student’s t‐test, *P < 0.05, **P < 0.01, and ***P < 0.001. Source data are available online for this figure.

Figure EV3. RNF5 promotes K48 linked SCAP ubiquitination and reduces SCAP stability

1. Increased K48 linked polyubiquitination of SCAP in RNF5‐transfected HEK293 cells. 293T cells were transiently transfected FLAG‐SCAP, V5‐tagged RNF5, and HA‐tagged ubiquitin (WT or K48) for 18 h followed by incubation with MG‐132 for 8 h. UbK48 depicts a mutant Ub with all the Ks mutated to R, except for K48.

2. qRT‐PCR analyses of several SREBPs downstream genes in MDA‐MB‐231 cells expressing shCtrl or shRNF5 (n = 3). Data are displayed as means ± SD. Statistical significance was assessed using two‐tailed Student’s t‐test, **P < 0.01 and ***P < 0.001.

3. SCAP stability measured by Western blotting in PKM2 KO MDA‐MB‐231 cells expressing shCtrl and shRNF5 treated with cycloheximide for the indicated time.

4. SCAP stability measured by Western blotting in MCF7 cells expressing Ctrl and HA‐RNF5 vectors treated with cycloheximide for the indicated time.

Source data are available online for this figure.

Figure 6. PKM2/VCP/RNF1 axis regulated TMEM33 expression

1. Western blot analyses of NRF1, TMEM33, and PKM2. The two bands of NRF1 denote precursor and cleaved NRF1.

2. The relative mRNA expression levels of TMEM33, RNF5, MT1A, and MT2A in parental and PKM2 KO MDA‐MB‐231 cells (n = 3).

3. Regulation of TMEM33 promoter with or without an ARE motif by NRF1 in luciferase assays in 293T cells (n = 3).

4. Recruitment of NRF1 to TMEM33 promoter analyzed by ChIP‐qPCR in parental and PKM2 KO MCF7 cells. IgG was used as a negative control. NRF1 enrichment on the TMEM33 promoter region was normalized to IgG control (n = 3).

5. Western blot analyses of TMEM33 and RNF5 in NRF1 overexpressed MCF7 clones.

6. Western blot analyses of TMEM33 and RNF5 in shCtrl and shNRF1 PKM2 KO MCF7 cells.

7. Co‐immunoprecipitation of VCP with FLAG‐tagged PKM1/2 in PKM2 KO 293T cells. FLAG‐tagged PKM1 or PKM2 construct was transiently transfected into PKM2 KO or parental 293T cells for 24 h, followed by pull‐down using FLAG‐M2 beads. PKM1, PKM2, and VCP antibodies were used for immunoblotting.

8. Decreased VCP and PKM2 in vitro interaction by treatment with TEPP‐46.

9. TEPP‐46 treatment attenuates accumulation of polyubiquitinated proteins by NMS‐873 detected with anti‐Ub Western blot. MCF7 cells were treated with DMSO, NMS873 (5 μM), TEPP‐46 (10 μM), or NMS873 (5 μM) plus TEPP‐46 (10 μM) for 7 h followed by co‐treatment with cycloheximide (CHX, 50 μg/ ml) for 1 h.

10. Western blotting of VCP, NRF1, TMEM33, and PKM2 in MCF7 treated with TEPP‐46 for 24 h.

11. A schematic diagram depicting the regulation of lipogenesis by PKM2. ER located PKM2 inhibits VCP activity, which is essential for NRF1 precursor cleavage. Cleaved NRF1 acts as a transcriptional factor for TMEM33. ER‐localized TMEM33 inhibits SREBPs activity through promoting SCAP degradation in RNF5‐dependent manner. Knocking out PKM2 leads to increased NRF1 cleavage and TMEM33 expression. Consequentially, SCAP is destabilized by RNF5‐mediated ubiquitination and expression levels of lipid synthetic genes are decreased.

Data information: Data are displayed (B‐D) as means ± SD. Statistical significance was assessed using two‐tailed Student’s t‐test, **P < 0.01 and ***P < 0.001. Source data are available online for this figure.

Figure EV4. TMEM33 is transcriptionally regulated by NRF1, whose cleavage and activation were mediated by VCP on ER membrane in breast cancer cell lines

1. A

   NRF1 target genes were up‐ or down‐regulated in PKM2 KO MCF7 cells. Data were from MCF7 PKM2 KO RNA‐seq dataset (n = 1).

2. B

   The relative mRNA levels of TMEM33 and MT2 in parental and PKM2 KO MCF7 cells (n = 3).

3. C

   Mapping the interaction domains in VCP that binds to PKM2. Halo‐tagged full‐length or truncated VCP‐HA were transiently transfected into 293T cells for 24 h. Halo linkTM resins were used to pull down full‐length or truncated VCP along with associated proteins. TEV enzyme was used to release the VCP‐HA and its truncations from the resin. Then, full‐length or truncated VCP and PKM2 were analyzed by Western blotting using HA and PKM2 antibodies, respectively.

4. D, E

   Western blotting of VCP and PKM2 in parental or PKM2 KO breast cancer cell lines after NMS‐873 treatment for 3 h.

5. F

   Western blotting of NRF1 and TMEM33 after 3‐h NMS‐873 treatment in parental or PKM2 KO MCF7 cells.

6. G

   Proximity ligation assays detecting interaction between PKM2 and VCP in MDA‐MB‐231 cells. Cells were treated with TEPP‐46 (10 µM) for 24 h, and representative confocal imaging data from one of the three independent experiments are shown. Scale bar 10 µm.

7. H

   Measurement of the VCP activity by Ub‐GFP reporter assays. GFP intensities in MCF7 cells stably expressing UbG76V‐GFP fusion protein were measured after treatment with DMSO, NMS873 (5 μM), TEPP‐46 (10 μM), or NMS873 (5 μM) plus TEPP‐46 (10 μM) for 8 h (n = 8).

8. I

   Western blotting of NRF1 in MCF7 cells treated with TEPP‐46 or TEPP‐46 plus NMS‐873.

Data information: Data are displayed (B and H) as means ± SD. Statistical significance was assessed using two‐tailed Student’s t‐test, *P < 0.05 and ***P < 0.001.Source data are available online for this figure.

Figure 7. TMEM33 expression is positively correlated with plasma cholesterol levels in mice

1. A

   Representative images from the Adeno vector control and TMEM33‐expressing groups. Scale bar, 1.5 cm.

2. B

   Time course of bioluminescent intensity of the mice injected with a control adenovirus and TMEM33‐expressing adenovirus (n = 7).

3. C

   The total serum cholesterol levels in mice before and after control adenovirus and TMEM33‐expressing adenovirus injection on days −1 and 5 (n = 7).

4. D

   Genotyping of CRISPR/Cas9 generated TMEM33 knockout mice using fluorescent probes for PCR. FAM⁺, Cy5⁻ indicates the TMEM33 ^+/+ genotype; FAM⁺, Cy5⁺ indicates the TMEM33 ^+/− genotype; and FAM⁻, Cy5⁺ indicates the TMEM33^−/− genotype (n ≥ 5).

5. E

   Representative H&E and IHC staining images of TMEM33 and SCAP in WT and TMEM33 KO mouse livers. Scale bars, 100 µm.

6. F, G

   Significant decrease of total plasma cholesterol (F) and LDL‐C (G) levels in TMEM33 ^−/− and +/− mice relative to WT counterparts (n ≥ 5).

Data information: Data are displayed (B, C, F, and G) as means ± SD. Statistical significance was assessed using two‐tailed Student’s t‐test, *P < 0.05 and ***P < 0.001. n.s denotes no significance. Source data are available online for this figure.

Figure EV5. PKM2 regulates cholesterol homeostasis in vitro and in vivo

1. Genotyping detects the deletion of PKM2 in PKM2fl/fl; Cre‐ERT2 mice by treatment of tamoxifen on days 1, 3, 5, 7, and 9.

2. qRT‐PCR analyses of several lipid synthetic genes in parental py8119 cells and two PKM2 KO clones (n = 3).

3. Western blotting of PKM2, TMEM33, HMGCR, and FASN in two different PKM2 KO clones relative to parental py8119 cells.

4. Decreased total cholesterol in two PKM2 KO clones relative to parental py8119 cells (n = 3).

5. Tumor growth curves of py8119 allografts in systemic PKM2 KO or WT mice treated with DMSO or pitavastatin (n ≥ 5).

6. Quantification of the averaged tumor weights in (E) (n ≥ 5).

7. Decreased LDL‐C levels in PKM2 KO or WT mice by treating with pitavastatin (n = 5).

8. Western blotting of SCAP and TMEM33 in allografted tumors from vehicle or TEPP‐46 treated mice groups.

Data information: Data are displayed (B, D, and E‐G) as means ± SD. Statistical significance was assessed using two‐tailed Student’s t‐test, *P < 0.05, **P < 0.01 and ***P < 0.001.Source data are available online for this figure.

Figure 8. Systemic inhibition of PKM2 in vivo increases serum cholesterol levels and promotes allografted tumor growth

1. Systemic PKM2KO results in increased total plasma cholesterol and LDL‐C levels in mice (n = 8).

2. Tumor growth curves of PKM2 WT or KO py8119 cell line allografts in systemic PKM2 KO or WT mice (n = 5).

3. Quantification of the averaged tumor weights in (B) (n = 5).

4. TEPP‐46 treatment increased total plasma cholesterol and LDL‐C levels in mice (n = 8).

5. The growth curves of py8119 allografts treated with vehicle and TEPP‐46 (50 mg/kg) for 17 days in mice (n = 7).

6. Tumor weights of PKM2 WT py8119 allografts treated with vehicle or TEPP‐46 (n = 7).

Data information: Data are displayed (A‐F) as means ± SD. Statistical significance was assessed using a two‐tailed Student’s t‐test, *P < 0.05, **P < 0.01 and ***P < 0.001. Source data are available online for this figure.