Inhibition of PCSK9 potentiates immune checkpoint therapy for cancer

Abstract

Despite its success in achieving the long-term survival of 10-30% of treated individuals, immune therapy is still ineffective for most patients with cancer1,2. Many efforts are therefore underway to identify new approaches that enhance such immune 'checkpoint' therapy3-5 (so called because its aim is to block proteins that inhibit checkpoint signalling pathways in T cells, thereby freeing those immune cells to target cancer cells). Here we show that inhibiting PCSK9-a key protein in the regulation of cholesterol metabolism6-8-can boost the response of tumours to immune checkpoint therapy, through a mechanism that is independent of PCSK9's cholesterol-regulating functions. Deleting the PCSK9 gene in mouse cancer cells substantially attenuates or prevents their growth in mice in a manner that depends on cytotoxic T cells. It also enhances the efficacy of immune therapy that is targeted at the checkpoint protein PD1. Furthermore, clinically approved PCSK9-neutralizing antibodies synergize with anti-PD1 therapy in suppressing tumour growth in mouse models of cancer. Inhibiting PCSK9-either through genetic deletion or using PCSK9 antibodies-increases the expression of major histocompatibility protein class I (MHC I) proteins on the tumour cell surface, promoting robust intratumoral infiltration of cytotoxic T cells. Mechanistically, we find that PCSK9 can disrupt the recycling of MHC I to the cell surface by associating with it physically and promoting its relocation and degradation in the lysosome. Together, these results suggest that inhibiting PCSK9 is a promising way to enhance immune checkpoint therapy for cancer.

Conflict of interest statement

Competing interest X.L. and C. L. are inventors on a patent application filed by Duke University that covers the use of anti-PCSK9 antibody in cancer immunotherapy. The other authors declare no competing interest.

Figures

Extended Fig.1.. CRISPR-Cas9 mediated knockout of PCSK9 gene and its effect on tumor cell growth in vitro and in vivo.

a. Western blot analysis of the expression of PCSK9 in murine tumor lines with PCSK9 knockout. GAPDH was used as protein loading control. Analysis was done twice with biologically independent samples. b. Cell growth of vector control or PCSK9 KO B16F10 tumor cells. Results from 5 biologically independent samples. Error bars, mean ± S.E.M. P value calculated by unpaired two-sided t test. c. Soft agar analysis of the colony formation ability of vector control or PCSK9 sgRNA-transduced B16F10 tumor cells. d. Quantitative representation of soft agar formation in c. n=4 biologically independent samples. Error bars, mean± S.E.M. P values calculated by unpaired two-sided t test. e. Diagram of in vivo competition assay. f. Change in ratios of mixed control-tdTomato and PCSK9 KO-EGFP B16F10 cells after 12 days grown in vivo (subcutaneously) in C57BL/6 mice, as determined by flow cytometry. g. Quantitative representation of the flow analysis in f. Error bar, mean± S.E.M. n=2 and 4 biologically independent tumor samples for the in vitro and in vivo groups, respectively. P value determined by unpaired two-sided t test.

Extended Data Fig. 2.. The effect of PCSK9 re-expression and the host immune system on tumor formation of PCSK9KO tumor cells.

a. Western blot analysis of the expression of exogenously transduced, HA-tagged PCSK9 in PCSK9KO B16F10 cells. The analysis was done once. b-c. Tumor formation (b) from B16F10-PCSK9KO cells transduced with either the vector control or the PSCK9 gene and Kaplan-Meier survival curve of host mice (c). About 2 × 105 tumor cells were injected subcutaneously into C57BL/6 mice and observed for tumor formation. n=5 tumors per group. Error bars, mean ± S.E.M. P values determined by two-way ANOVA in b and log-rank test in c. d-i. Growth rate, host survival and endpoint tumor weight of vector control and PCSK9KO 4T1 (d-f) and B16F10 (g-i) tumors. In each case, about 1 × 105 tumor cells were injected subcutaneously and observed for tumor formation in NCG mice. n=6 mice for d, e, g, and h and n=5 tumors for f and i. Error bars in d,f g, and i represent mean ± S.E.M. ns: not significant, as determined by two-way ANOVA (d,g), log-rank test (e,h), or unpaired two- sided t test (f, i). j-k. tumor growth from vector control and PCSK9KO B16F10 cells (j) and Kaplan-Meier survival curve of tumor-bearing host mice (k) in Rag1−/− C57BL/6 mice. About 1 × 105 vector 381 control or PCSK9KO B16F10 tumor cells were injected into Rag1−/− C57BL/6 mice and observed for tumor formation. ns: not significant. n=5 tumors per group. Error bars in j, mean ± S.E.M. P values calculated by two-way ANOVA in j and by log-rank in k.

Extended Data Fig. 3.. The influence of tumor or host cell LDLR and host cholesterol levels on tumor growth from control or PCSK9KO tumor cells in immunocompetent hosts.

a. Western blot analysis of CRISPR-Cas9 mediated knockdown (KD) of LDLR in B16F10 cells. The analysis was done once. b. Tumor growth from vector control and LDLR KD B16F10 cells in C57BL/6 mice. n=5 tumors per group. Error bars, mean ± S.E.M. P value calculated by two-way ANOVA. c. Kaplan Meier survival curve of mice (from b) bearing control and LDLR KD B16F10 tumors. n=5 mice per group. P value calculated by log-rank test. d. Tumor growth from vector control and PCSK9 knockout B16F10 cells in WT and LDLR−/− mice fed with high-fat diet. n=12, 12, 5, and 5 tumors in C57BL/6 mice inoculated with control and PCSK9KO tumor cells, and LDLR−/− mice inoculated with control and PCSK9 KO tumor cells. Error bars, mean ± S.E.M. P values calculated by two-way ANOVA with multiple comparisons. e. Kapan-Meier survival curve of wild type and LDLR−/− mice (from d) bearing vector control and PCSK9 knockout tumors. P value calculated by log-rank test.

Extended Data Fig. 4.. Additional data on anti-PD1 treatment in murine tumors.

a. Treatment schedule for PCSK9KO 4T1 tumors. Balb/c mice implanted with PCSK9KO 4T1 tumor cells but did not form visible tumors on day 9 after inoculation were excluded for treatment. b. Tumor growth delay in mice bearing PCSK9KO 4T1 tumors with or without anti- PD1 treatment. n=5 tumors per group. Error bars, mean ± S.E.M. P values calculated by two-404 way ANOVA test. c. Kaplan-Meier survival curve of tumor-bearing mice in b. P value calculated by log rank test. d. Treatment schedule for PCSK9KO CT26 tumors. Balb/c mice were implanted subcutaneously with PCSK9KO CT26 tumor cells and treated with an anti-PD1 antibody and observed for tumor formation. e. Tumor growth delay in mice bearing PCSK9KO CT26 tumors with or without anti-PD1 treatment. n=5 tumors per group. Error bars, mean ± SEM. P values were determined by two-way ANOVA test. f. Kaplan-Meier survival curve of tumor-bearing mice in e. Error bars, mean ± SEM. P value was determined by log-rank test. g. A scheme to develop anti-PD1 resistant MC38R tumor cells. h. Treatment scheme of anti-PD1 resistant MC38R tumors with evolocumab and an anti-PD1 antibody. i. Tumor growth kinetics from anti-PD1 resistant MC38R tumor treated with anti-PD1 and/or evolocumab. n=5 tumors per 414 group. Error bars: mean ± SEM. P values were determined by two-way ANOVA test. j. Kaplan-Meier survival curve for mice bearing MC38R tumors in i. P values determined by log-rank test. k. Treatment schedule for PCSK9KO MC38 tumors. l-m. Tumor growth delay (l) and host mice survival (m) among isotype- (iso) or evolocumab-treated mice bearing MC38-PCSK9KO tumors. n=5 tumors per group. P values were calculated by two-way ANOVA test in l and log-rank test in m.

Extended Data Fig.5.. Re-challenge of mice that were tumor free after initial tumor 4inoculation and gating strategy of intratumoral immune effector cells.

a-c. Treatment scheme (a), tumor growth (b), and survival of host mice (c) after re-challenge with wild type 4T1 tumor cells in Balb/c mice that remained tumor-free 43 days after initial challenge with PCSK9 deficient 4T1 cells. The control group consists of tumor-naive Balb/c mice challenged with wild type 4T1 cells. n=5 and 12 mice for naïve and re-challenged group, respectively. Error bars in b, mean ± S.E.M. P values in b and c calculated by two-way ANOVA test and log-rank test, respectively. d-f. Treatment scheme (d), tumor growth (e), and survival of host mice (f) after re-challenge with wild type B16F10 tumor cells in C57BL/6 mice that remained tumor-free 26 days after initial challenge with PCSK9 deficient B16F10 cells and the treatment with anti-PD1 antibody. The control group consisted of tumor-naïve C57BL/6 mice challenged with wild type B16F10 cells. n=5 and 13 mice for control and re-challenge groups, respectively. Error bars in e, mean ± S.E.M. P values in e and f calculated by two-way ANOVA test and log-rank test, respectively. g-i. Treatment scheme (g), tumor growth (h), and survival of host mice (i) after re-challenge with parental MC38 tumor cells in C57BL/6 mice that remained tumor-free 34 days after initial challenge with PCSK9-deficient MC38 cells and the treatment with anti-PD1 antibody. The control group consisted of tumor-naïve C57BL/6 mice challenged with wild type MC38 cells. n=5 mice per group. Error bars in h, mean ± S.E.M. P values calculated by two-way ANOVA test in h and log-rank test in i. j. Representative flow cytometry gating strategy to quantitate the numbers of various immune effector cell subsets in murine tumors.

Extended Data Fig. 6.. Additional data on the characterization of lymphocyte infiltration into murine tumors.

a. Immunofluorescence staining (left panel) and quantitative estimate (right panel) of CD45+ leukocytes in control and PCSK9KO tumors grown in syngeneic C57BL/6 mice. Scale bar = 50 μm. n=3 biologically independent samples. Four fluorescent fields for each of the three samples were counted. Error bars, mean ± S.E.M. P value calculated using unpaired two-sided t-test. b.Immunofluorescence staining (left panel) and quantitative estimate (right panel) of CD8a+ cells in control and PCSK9KO B16F10 tumors. Scale bar = 20 μm. n=3 biologically independent samples. Four fluorescent fields for each of the three samples were counted. Error bars, mean ± S.E.M. P value calculated using unpaired two-sided t test. c. Quantitative estimates of CD4+ and 452 CD8+ T cells in the spleens of mice bearing control and PCSK9KO B16F10 tumors as determined by flow cytometry. n=3 mice per group. Error bars, mean ± SEM. P values calculated using unpaired two-sided t-test. d. Flow cytometry determination of the percentage of intratumoral CD8+ T cells that were IFNγ+. n=6,5 tumors in the two groups. Error bars, mean ± S.E.M. P value calculated by unpaired two-sided t-test. e, f. Q-RT-PCR analysis of intratumoural IFNG (e) and GZMB (f) mRNA levels in control and PCSK9KO tumors. n=3 and 4 tumors for INFG and GZMB groups, respectively. Error bars, mean ± S.E.M. P values were determined by unpaired two-sided t test. g-i. Flow cytometry characterization of the cell surface expression levels of exhaustion markers for intratumoural CD8+ T cells in vector control and PCSK9KO tumors. n=6, 5 tumors. Error bars, mean ± S.E.M. P values were determined by unpaired two-sided t test. j. Evolocumab and anti-PD1 treatment schedule for syngeneic 4T1 tumor model. k. Growth of 4T1 tumors treated with anti-PD1 and/or evolocumab. n=5 mice per group. P values were determined by two-way ANOVA test. l. Kaplan-Meier survival curve for mice in k. P values were determined by log-rank test. m. Frequency of CD8+ T cells in 4T1 tumors treated with anti-PD1 and/or evolocumab. n=5 tumors per group. Error bars, mean ± SEM. P values were determined by unpaired two-sided t test. n. Frequency of IFNγ+CD8+ T cells in 4T1 tumors treated with anti-PD1 and/or evolocumab. n=5 tumors per group. Error bars, mean 469 ± SEM. P values were determined by unpaired two-sided t test.

Extended Data Fig. 7.. Additional data on the effect of PCSK9 inhibition on immune effector function and antigen presentation.

a. Injection schedule for antibody-mediated immune cell depletion of CD4+, CD8+, and NK cells. b-c. Growth rates (b) and host mice survival (c) of PCSK9KO tumors in mice administered with control or anti-CD4 antibody. n=5 tumors per group. Error bars, mean ± S.E.M. P values determined by two-way ANOVA test in b and log-rank test in c. d-e. Growth rates (d) and host mice survival (e) of PCSK9KO tumors in mice administered with control or anti-NK1.1 antibody. n=5 tumors per group. Error bars, mean ± S.E.M. P values in d and e determined by two-way ANOVA test and logrank test, respectively. f. Fluorescence images of the tdTomato-labeled tumor cells with or without the OVA antigen in the presence or absence of OVA-specific T cells. The experiments were completed twice with similar results. Scale bar = 200 μm for all image panels. g. Enhanced presentation of OVA antigen (SIINFEKL) by MHC-I in cultured B16F10 cells with PCSK9 deficiency. Control and PCSK9KO B16F10 cells transduced with the OVA gene were treated with IFNγ and assayed for the amount of cell surface H-2Kb-SIINFEKL complex using flow cytometry. Shown were representative results from analyses of 4 sets of biologically independent samples. h-i. Flow cytometry analysis of MHC II (h) and PD-L1(i) expression in control and PCSK9KO B16F10 cells. n=5 and 4 biologically independent samples, respectively. P values determined by unpaired two-sided t test. j. Western blot of PCSK9 expression in control or PCSK9KO MDA-MB-231 cells. The analyses were done twice. k. The effect of evolocumab and alirocumab on HLA-ABC expression on the surface of MDA-MB-231 human breast cancer cells. n=6, 6, and 5 biologically independent samples. Data represent mean ± S.E.M, P values were determined by unpaired two-sided t test. l. H2-Kd/Dd expression levels of 4T1 tumor cells that were exposed to anti-PD1and/or evolocumab in vivo. n=5 mice per group. Error bars, mean ± S.E.M. P values were determined by unpaired two-sided t test.

Extended Data Fig. 8.. Additional data on the analysis of PCSK9, H2-K1, and LDLR in murine tumor cells.

a. Lentivirus mediated over-expression of HA-tagged H2-K1 gene in B16F10 cells as determined by western blot analysis. The analysis was done once. b-c. Tumor growth delay (b) and Kaplan-Meier survival curve (c) of tumor-bearing C57BL/6 mice implanted with vector control or H2-K1 over-expression B16F10 cells. Error bars, mean ± S.E.M. n=5 tumors in each group. P values in b and c were determined by two-way ANOVA and log-rank test, respectively. d-e. Tumor growth delay (d) and Kaplan-Meier survival curves (c) in mice injected with vector control, H2-K1 KD, PCSK9 KO, or H2-K1KD/PCSK9KO B16F10 cells. n=5 mice per group. Error bars, mean ± S.E.M. P values were determined by two-way ANOVA and log-rank test, respectively. f. Western blot analysis of and LDLR knockdown in control and PCSK9KO B16F10 tumor cells. The analysis was done once. g-h. Tumor growth delay (g) and Kaplan-Meier survival curves (h) from LDLRKD and LDLRKD/PCSK9KO B16F10 tumors. n=5 mice per group. Error bars, mean ± S.E.M. P values were determined by two-way ANOVA and log-rank test, respectively. i. Flow cytometry analysis of MHC I expression in tumors formed from tdTomato-labeled control and LDLRKD B16F10 cells. n=6 biologically independent tumors. Error bars, mean ± S.E.M. P value calculated by unpaired two-sided t-test. j. Flow cytometry analysis of MHC I expression in tumors formed from tdTomato-labeled LDLRKD (n=6) and LDLRKD/PCSK9KO cells (n=4). Error bars, mean ± S.E.M. P value calculated by unpaired two-sided t-test.

Extended Data Fig. 9.. Additional data on mapping and functional characterization of interacting domains in PCSK9 and MHC I and the association of PCSK9 expression and the prognosis of TCGA cancer cohorts.

a. Domain structure of the mouse PCSK9 protein. SP, signal peptide; Pro, propeptide; Catalytic, catalytic domain; CRD; C-terminal domain. b. IP-Western blot analysis of the interaction between full-length Flag-labeled H2-K1 and full length or partially deleted mouse PCSK9-HA. Plasmids encoding the two genes were transfected into 293T cells in pairs and lysates from transduced cells were immunoprecipitated first with an anti-HA antibody and probed with an anti-Flag antibody by western blot analysis. The analyses were repeated twice with biologically independent samples with similar results. c. IP-Western blot analysis of the interaction between full-length HA-labeled mouse PCSK9-HA and full-length or partially deleted H2-K1-Flag (aa66–202) (α1-α2 domains). The analyses were repeated twice with biologically independent samples with similar results. d. IP-Western blot analysis of the interaction of HA-labeled mouse PCSK9-HA to full-length H2-K1 or H2-K1 with more limited deletions (aa 66–100, α1 domain; or aa68–70). The analyses were repeated twice with biologically independent samples with similar results. e-f. Tumor growth rates (e) and Kaplan-Meier survival curve (f) of mice inoculated with PCSK9KO B16F10 tumor cells re-expressed with wild type or partially (ΔM2) deleted PCSK9. n=5 tumors per group. Error bars, mean ± S.E.M. P values were determined by two-way ANOVA and log-rank test in e and f, respectively. g-h. Tumor growth rates (g) and Kaplan-Meier survival curve (h) of mice inoculated with H2-K1KO or H2-K1/PCSK9 DKO B16F10 tumor cells re-expressed with wild type or partially deleted (Δ68–70) H2-K1 gene. n=5 tumors per group. Error bars, mean ± S.E.M. P values were determined by two-way ANOVA and log-rank test, respectively. i. Higher levels of PCSK9 expression correlated to worse survival in 9 cancer patient cohorts including liver hepatocellular carcinoma (LIHC), pancreatic adenocarcinoma (PAAD), skin cutaneous melanoma (SKCM), uveal melanoma (UVM), bladder urothelial carcinoma (BLCA), lung adenocarcinoma (LUAD), kidney renal clear cell carcinoma (KIRC), kidney renal papillary cell carcinoma (KIRP), and ovarian carcinoma (OV). P values calculated by log-rank test. Data from TCGA datasets.

Extended Data Fig. 10.. A schematic diagram illustrating PCSK9-mediated degradation of MHC I in the lysosome.

In the presence of PCSK9, MHC I is transported into the lysosome and degraded (left panel). In the absence of PCSK9, either because of genetic deletion or antibody neutralization, MHC I levels on the surface remains high and is thus able to present tumor-specific peptic antigens more efficiently to T cells (right panel). Illustration by Stan Coffman.

Fig1. depletion attenuates tumor growth in syngeneic mice.

About 1 × 105 vector control and PCSK9 knockout murine tumor cells were inoculated subcutaneously into syngeneic mice and observed for tumor formation. Both tumor size and overall survival were monitored. a-b. 4T1 breast cancer line grown in Balb/c mice. n=9 and 20 mice for control and PCSK9KO tumor cells, respectively. c-d. B16F10 melanoma line grown in C57BL/6 mice. n=12 mice for both groups. e-f. CT26 colon cancer line grown in Balb/c mice. n=5 mice for both groups. g-h. MC38 colon cancer line grown in C57BL/6 mice. n=5 mice for both groups. Error bars: mean ± S.E.M. P values were calculated by two-way ANOVA in a, c, e, g and log-rank test in b, d, f, h, respectively.

Fig.2. inhibition overcomes tumor resistance to anti-PD1 therapy.

a-c. Treatment of vector control and PCSK9 knockout B16F10 melanoma with an anti-PD1 antibody in syngeneic mice. Experimental protocol (a), tumor growth curve (b), and overall survival (c) were shown. n=5, 5, 5, and 20 mice in the four groups, respectively. d-f. Treatment of vector control and PCSK9KO MC38 colon cancer with an anti-PD1 antibody in syngeneic mice. Experimental protocol (d), tumor growth curve (e), as well as overall survival (f) were shown. n=5 mice per group. g-i. Treatment of MC38 colon cancer with combined anti-PCSK9 (evolocumab, Evo; or alirocumab, Ali) and anti-PD1 antibodies in mice. Experimental protocol (g), tumor growth curve (h), as well as overall survival (i) were shown. n=10 mice per group. Shown were combined results from two separate experiments. Error bars represent mean ± S.E.M. P values were calculated by two-way ANOVA in b, e, h and logrank test in c, f, i, respectively.

Fig.3. depletion enhances intratumoral T-cell infiltration.

a-e. Quantitative estimate of various immune effector cells per mg of tumor tissue in vector control and PCSK9KO B16F10 tumors as determined by flow cytometry. n=7 tumors per group. f. Ratio of CD8+ T /CD4+Foxp3+ Treg cells in control and PCSK9KO B16F10 tumors. n=7 tumors per group. g-h. Average numbers of tumor-infiltrating IFNγ+ CD8+ T (g) and Gzmb+CD8+T (h) cells per mg of tumor tissue in control or PCSK9 KO tumors. n=7 tumors per group. i-j. Tumor growth (i) and host survival (j) from control and PCSK9KO B16F10 tumor cells in C57BL/6 mice depleted of CD8+ T cells. n=6, 5, 5 tumors in the three groups. k-o. Total TCR (k), unique TCR (l), productive clonality (m), and max productive frequency (n), and heatmap of top 5% TCR (o) in control and PCSK9 knockout tumors by use of TCRB CDR3 sequencing. n=3 tumors per group. p. Quantitative estimates of the fraction of live control or PCSK9KO B16F10-tdTomato tumor cells remaining after 24 hrs of incubation with activated OVA-specific T cells. n=3 biologically independent samples per group. Three representative fields from each sample were counted. q. Negative correlation of PCSK9 mRNA levels with that of CD8A mRNA levels in human esophageal carcinoma (ESCA, 47 samples), cervical squamous cell carcinoma and endocervical adenocarcinoma (CESC,112 samples), pancreatic adenocarcinoma (PAAD,173 samples), prostate adenocarcinoma (PRAD, 313 samples). Data from the GENT database. R represents Pearson correlation coefficient. Error bars represent mean ± S.E.M throughout the figure. P values in a-h, k-n, and p were calculated by unpaired two-sided t-test. P values in i and j determined by two-way ANOVA and log-rank test, respectively. P values in q were calculated by F test.

Fig. 4.. PCSK9 promotes lysosome-mediated degradation of MHC I in tumor cells.

a. Quantitative estimates of SIINFEKL-H-2Kb levels in B16F10-OVA cells with or without IFNγ treatment in control and PCSK9KO B16F10 cells. n=4,4,6,5 biologically independent samples, respectively. b-c. Flow cytometry estimates of H-2Kb/Db levels on the surface of subcutaneously grown control and PCSK9-deficient B16F10 cells. n=3 tumors per group. d. Quantitative estimate of MHC I levels on the surface of in tissue cultured control and PCSK9-deficient 4T1 cells treated with or without IFNγ. n=5 biologically independent samples. e. The effect of PCSK9 deficiency on HLA-A2 expression on the surface of control and PCSK9KO MDA-MB-231 cells. n=3 biologically independent samples. f. The effect of exogenous PCSK9 protein on HLA-A2 degradation on the surface of MDA-MB-231 PCSK9KO cells. n=4 biologically independent samples. g. Interaction of mouse PCSK9 and H2-K1 in 293T cells were transduced with FLAG-tag labeled H2-K1 gene in combination with HA-tag labeled full-length or various deleted PCSK9 genes. h. PCSK9-promoted H2-K1 migration into the lysosome. Representative fluorescence confocal images of MHC I (H2-K1Flag) distribution in PCSK9 (PCSK9-HA) over-expressing (top panels) or PCSK9 KO (lower panels) B16 F10 cells. Scale bar: 10 μm. Insets showed magnified areas with additional details of co-localization. i-j. Western blot quantification of H2-K1-Flag in the lysosome (i) and plasma membrane (j) fractions of PCSK9 overexpressing (PCSK9-OE) and PCSK9KO B16F10 cells. Bottom panels are the quantitative estimates of MHC I (H2-K1-Flag) expression levels based on data in the top panels. k-l. Western blot analysis of HLA-ABC expression in cycloheximide (CHX, k)- and bafilomycin A1 (BafA1, l)-treated vector control and PCSK9KO MDA-MB-231 cells. In k and l, the right panels showed the quantitative estimates of HLA-ABC levels based on WB analysis. Error bars in a-f, mean ± S.E.M. P values in a-f were determined by unpaired two-sided t test. Two independent experiments were done with similar results for g-l.