Oxidized Lipoproteins Promote Resistance to Cancer Immunotherapy Independent of Patient Obesity

Abstract

Antitumor immunity is impaired in obese mice. Mechanistic insight into this observation remains sparse and whether it is recapitulated in patients with cancer is unclear because clinical studies have produced conflicting and controversial findings. We addressed this by analyzing data from patients with a diverse array of cancer types. We found that survival after immunotherapy was not accurately predicted by body mass index or serum leptin concentrations. However, oxidized low-density lipoprotein (ox-LDL) in serum was identified as a suppressor of T-cell function and a driver of tumor cytoprotection mediated by heme oxygenase-1 (HO-1). Analysis of a human melanoma gene expression database showed a clear association between higher HMOX1 (HO-1) expression and reduced progression-free survival. Our in vivo experiments using mouse models of both melanoma and breast cancer revealed HO-1 as a mechanism of resistance to anti-PD1 immunotherapy but also exposed HO-1 as a vulnerability that could be exploited therapeutically using a small-molecule inhibitor. In conclusion, our clinical data have implicated serum ox-LDL as a mediator of therapeutic resistance in patients with cancer, operating as a double-edged sword that both suppressed T-cell immunity and simultaneously induced HO-1-mediated tumor cell protection. Our studies also highlight the therapeutic potential of targeting HO-1 during immunotherapy, encouraging further translational development of this combination approach.See article by Kuehm et al., p. 227.

Figure 1.. Obesity and leptin levels do not accurately predict responses to checkpoint blockade immunotherapy.

Retrospective analysis of patient survival after treatment with anti-CTLA4 ipilimumab or antiPD-1/anti-PDL1. Kaplan-Meier curves show OS data for (A) 129 melanoma patients treated with anti CTLA4 and (B) 149 patients with various cancers treated with anti-PD1 or anti-PDL1. Patient-derived PBMC and TIL were analyzed by flow cytometry and gated on CD45⁺ live cells. (C) Representative FACS plots from lean (top) and obese (bottom) patients show PD1 surface expression on CD8⁺ PBMC (left) and CD8⁺ TIL (right). Linear regression analysis of PD1 MFI on CD8⁺ (D) PBMC and (E) TIL from 27 melanoma (black circles) and 11 breast cancer (pink circles) patients versus BMI. Correlative analysis of plasma leptin concentration versus (F) BMI or (G) PD1 MFI on CD8⁺ PBMC from these same patients is shown with inset numbers indicating the r values for each data set along with corresponding P values.

Figure 2.. Patient immune function correlates with lipid profile independent of BMI and leptin.

Serological analysis of plasma samples from melanoma (black) and breast cancer (pink) patients. Graphs display linear regression of (A) patient BMI, (B) leptin, (C) total cholesterol, (D) LDL, (E) HDL, and (F) CE vs immune score for 25 melanoma patients and 11 breast cancer patients. Linear regression of patient ox-LDL vs (G) immune score, (H) cholesterol, and (I) LDL for 17 melanoma patients and 11 breast cancer patients. Each circle represents data from an individual patient with inset numbers showing r values for each data set along with corresponding P values. (J) Kaplan-Meier curves display OS data for 30 cancer patients stratified by plasma LDL concentration treated with anti-PD1/L1.

Figure 3.. Oxidized LDL inhibits human CD8⁺ T-cell proliferation and effector responses.

Human PBMC were stimulated with anti-CD3 for 6 days. (A) Representative FACS plots show CD8⁺ and CD4⁺ T cells among all CD45⁺ lymphocytes (top) and Ki67⁺ CD8⁺ cells (bottom) gated on CD45 expression. Inset numbers indicate the percent of total CD45⁺ cells within the incribed regions. (B) Graphs show percent of Ki67⁺ CD8⁺ T cells from the indicated treatment groups (left) and as individual responses traced for each donor in a different color (right). (C) Representative FACS plots show the frequency of GzmB and perforin expression among all CD45⁺ CD8⁺ T cells (top) and IFNγ production by CD8⁺ T cells after restimulation (bottom). (D) Graphs show the cumulative effector score for indicated treatment groups (left) and individual responses for each donor (right). All graphs show data from the same 10 donors where each dot represents data from an individual donor and each color follows responses by a single donor over the four treatments. Exact P values are provided for the bracketed groups and error bars represent SEM.

Figure 4.. HO-1 is negatively associated with patient survival and promotes resistance to immunotherapy.

(A) Melanoma patient survival and disease stage relative to tumor HMOX1 gene expression data obtained from the Gene Expression Profiling Interactive Analysis (GEPIA). A total of 458 melanoma patients (229 high and 229 low) were analyzed independent of treatment and the indicated P value between these groups was determined by a Log-rank test. HO-1 protein expression in (B) mouse B16-F0 cells and (C) human A375 melanoma tumor cells incubated with the indicated concentration of LDL, ox-LDL or H-hemin, were analysed for HO-1 and β-actin expression by western blot. Densitometry results are shown below each band indicate the ratio of total HO-1 to β-actin and data shown are representative of 3 independent experiments. (D) HO-1 protein expression in various murine tumor lines measured by western blot. (E) Diagram of experimental setup where mice with established B16-F0 melanoma or E0771 breast tumors were treated with anti-PD1 on day 6, 10, 14 with or without treatment with an HO-1 inhibitor (iHO-1: OB24) on days 3, 6, 8, 12. (F) Subcutaneous bi-lateral B16-F0 tumor volumes (mm³) on day 16 are shown for 15 mice (30 tumors) per treatment group pooled from 3 independent experiments. (D) E0771 tumor volume (mm³) in mammary fat pads on day 16 are shown for 14 mice (28 tumors) per treatment cohort pooled from 3 different experiments. Exact P values are provided for the bracketed groups and error bars represent SEM.

Figure 5.. Inhibition of HO-1 overcomes resistance to immunotherapy independent of CD8⁺ TIL function.

B6 mice bearing B16-F0 melanoma tumors were treated as described in Figure 4E and TIL were assessed on day 16. (A) Frequency of CD8⁺ TIL from 15 mice per treatment cohort pooled from 3 independent experiments. (B) Representative FACS plots show granzyme B expression (top) and histograms show Ki67 expression (bottom) of CD8⁺ TIL for the different treatment groups in blue relative to PBS control in red. (C) Frequency of CD8⁺ TIL expressing GzmB, FNγ and Ki67 is shown for each treatment group and data are pooled from 3 independent experiments with each point representing data from an individual mouse. Exact P values are provided for the bracketed groups and error bars represent SEM.

Figure 6.. Oxidized-LDL and LDL induce HO-1-mediated cytoprotection in human melanoma tumors.

A375 human melanoma cells were pre-treated for 4 hours with ox-LDL, LDL or hemin in the absence or presence of iHO-1 (OB24; 30μg/mL) and then treated with doxorubicin (DXR) for 24h. (A) Representative FACS plots show expression of active caspase (CaspGlow) and annexin V in A375 tumor cells, and the (B) percent of double-positive cells is shown for the different treatment groups with data pooled from 3 independent experiments. (C) B16 melanoma cells were pretreated with 10μM Hemin, LDL or ox-LDL for 48h and then cocultured with activated Pmel T cells at a 10:1 effector to target ratio for 3 days. Tumor cell apoptosis was assessedby active caspase and annexin V staining and percent of double-positive cells is graphed for the different treatment groups with data pooled from 3 independent experiments. Each point represents data from an individual replicate well with exact P values provided for the bracketed groups and error bars represent SEM.