Lysophosphatidic acid modulates CD8 T cell immunosurveillance and metabolism to impair anti-tumor immunity

Abstract

Lysophosphatidic acid (LPA) is a bioactive lipid which increases in concentration locally and systemically across different cancer types. Yet, the exact mechanism(s) of how LPA affects CD8 T cell immunosurveillance during tumor progression remain unknown. We show LPA receptor (LPAR) signaling by CD8 T cells promotes tolerogenic states via metabolic reprogramming and potentiating exhaustive-like differentiation to modulate anti-tumor immunity. We found LPA levels predict response to immunotherapy and Lpar5 signaling promotes cellular states associated with exhausted phenotypes on CD8 T cells. Importantly, we show that Lpar5 regulates CD8 T cell respiration, proton leak, and reactive oxygen species. Together, our findings reveal that LPA serves as a lipid-regulated immune checkpoint by modulating metabolic efficiency through LPAR5 signaling on CD8 T cells. Our study offers key insights into the mechanisms governing adaptive anti-tumor immunity and demonstrates LPA could be exploited as a T cell directed therapy to improve dysfunctional anti-tumor immunity.

Fig. 1. Lysophosphatidic acid is a prognostic marker in solid tumor malignancies.

A Analysis of data from The Cancer Genome Atlas (TCGA) on progression free survival. Data was taken as pan-cancer data from all solid tumors in cBioPortal from the complete curated non-redundant studies and accessed on June 18, 2021. Cohorts were stratified based on genomic status of amplification of ENPP2, MYC, or wildtype for both genes. Amplification cohorts are tumors displaying amplification of either ENPP2 or MYC in the absence of a co-occurring alteration in the other gene. The ANOVA statistical test with post-hoc analysis was performed where ***p < 0.0001. B mRNA z-scores of exhaustion markers from TCGA data with samples stratified by high and low ENPP2 expression representing the top 25% and bottom 25% of ENPP2 expressing melanoma tumors. Descriptive statistics are as follows for low ENPP2: number of values = 91, minimum = −21.76, 25% percentile = −8.777, median = −2.278, 75% percentile = 3.966, maximum = 11.80, range = 33.56, mean = −2.863. Descriptive statistics are as follows for high ENPP2: number of values = 91, minimum = −24.17, 25% percentile = −6.980, median = 2.438, 75% percentile = 9.988, maximum = 17.75, range = 41.92, mean = 1.634. Statistics show the unpaired two-sided Student’s t-test analysis was performed where n = 363 samples and ***p < 0.0005 where p = 0.0004. C, D tSNE plots of (C) LPAR5 expression in melanoma and immune cells and (D) corresponding immune cell populations. tSNE plots were generated using the Single Cell Portal (https://singlecell.broadinstitute.org/single_cell). E Two-sided Spearman correlation analysis of LPAR5 expression and “exhaustion” signature from bulk RNA sequencing on TCGA melanoma tumors where p < 0.0001. F Relative abundance of LPA in stage IV melanoma responder patients (blue symbols; complete response and partial response where n = 3 patients) or non-responders (red symbols; stable disease and progressive disease where n = 6 patients) measured both pre- and post-treatment. The unpaired two-sided Student’s t-test analysis was performed where *p < 0.05 and p = 0.0313. Error bars represent standard error of the mean.

Fig. 2. Lpar5^−/− OT-I mice have improved tumor clearance and anti-tumor immunity.

A Schematic showing study design where B16.cOVA tumor cells and OT-I CD8 T cells are co-transferred in a 1:1 ratio into the mice on day 1. Mice were harvested on day 20 and evaluated for tumor burden and flow cytometric and histological analyses were performed on the lung bearing tumors. B Quantified tumor burden in the lung after intravascular injection of B16.cOVA cells. Tumor burden is presented as the number of tumor nodules in the lung and n = 8 mice per group and p = 0.0024. C Representative hematoxylin & eosin (H&E) histology images of the B16.cOVA tumor seeded in the lungs. Scale bars represent 100 µm. D Tumor area (µm²) quantified from H&E histology of B16.cOVA lung tumors (3 tumors per lung were analyzed and all measurements were averaged and p < 0.0001). E Circularity analysis of B16.cOVA lung tumors quantified from H&E histology (3 tumors per lung were analyzed and all measurements were averaged and p < 0.0001). F Representative images of immunohistochemistry for CD8 on lung sections. CD8 positive cells are shown as red to distinguish between melanin and positive stain. Scale bars represent 50 µm. G Quantification of intratumoral CD8 positive cells per high powered field (hpf, 400x, averaged values of 3 different tumors per mouse with technical error propagated, p = 0.0041). H Flow cytometric quantification of Tim3 expression by CD8⁺ CD44⁺ CD69⁺ PD1⁺ tumor-associated T cells expressed as geometric mean fluorescence intensity (gMFI) where p = 0.0417. Statistics for this entire figure were performed using the unpaired two-sided Student’s t-test analysis was performed where *p < 0.05, **p < 0.005, ***p < 0.0005, ****p < 0.0001. Error bars represent standard error of the mean.

Fig. 3. Lpar5 signaling on CD8 T cells promotes phenotypic tolerogenic states through exhaustive-like differentiation.

A Schematic of in vitro chronic stimulation. Effector CD8 T cells are persistently cultured with either anti-CD3 + LPA or LPA alone until Day 15. B–D Flow cytometric analysis and quantification of percent PD1⁺ Tim3⁺ from the CD8 T cell population. B, C Representative contour plots for (B) OT-I or (C) Lpar5^−/− OT-I CD8 T cells persistently cultured with LPA. D Quantification of percent of CD8⁺ T cells that are PD1⁺ Tim3⁺ (n = 3 mice per group). Exact p-values are as follows, OT-I CD3 + LPA vs OT-I LPA p = 0.0049; OT-I CD3 + LPA vs Lpar5^−/− OT-I LPA p < 0.0001; Lpar5^−/− OT-I CD3 + LPA vs Lpar5^−/− OT-I LPA p < 0.0001; OT-I LPA vs Lpar5^−/− OT-I LPA p = 0.0012^. (E–G) Flow cytometric analysis of Tim3 expression on OT-I or Lpar5^−/− OT-I CD8 T cells cultured in either anti-CD3 + LPA or LPA. E, F Representative flow cytometric histograms and (G) quantification the geometric mean fluorescence intensity (gMFI) of PD1 (n = 3 mice). Exact p-values are as follow, OT-I CD3 + LPA vs Lpar5^−/− OT-I CD3 + LPA p < 0.0001; OT-I CD3 + LPA vs OT-I LPA p < 0.0001; OT-I CD3 + LPA vs Lpar5^−/− OT-I LPA p < 0.0001; Lpar5^−/− OT-I CD3 + LPA vs OT-I LPA p = 0.0007; Lpar5^−/− OT-I CD3 + LPA vs Lpar5^−/− OT-I LPA p = 0.0003. H–J Flow cytometric analysis of Tim3 expression on OT-I or Lpar5^−/− OT-I CD8 T cells cultured in either anti-CD3 + LPA or LPA. H, I Representative flow cytometric histograms and (J) quantification the geometric mean fluorescence intensity (gMFI) of Tim3 (n = 3 mice). Exact p-values are as follows, OT-I CD3 + LPA vs Lpar5^−/− OT-I p = 0.0012; OT-I CD3 + LPA vs OT-I LPA p < 0.0001; OT-I CD3 + LPA vs Lpar5^−/− OT-I LPA p < 0.0001; Lpar5^−/− OT-I CD3 + LPA vs OT-I LPA p = 0.0044; Lpar5^−/− OT-I CD3 + LPA vs Lpar5^−/− OT-I LPA p = 0.0031. K Schematic of study design where B16.cOVA tumor cells and OT-I CD8 T cells are co-transferred in a 1:1 ratio into the mice on day 1. Mice were harvested on day 20 and evaluated for tumor burden and flow cytometric analysis for exhaustion markers. L Quantified tumor burden in the lung after intravascular injection of B16.cOVA cells. Tumor burden is presented as the number of tumor nodules in the lung where n = 5 mice per OT-I group and n = 7 mice per Lpar5^−/− OT-I group and p = 0.0124. (M,N) Flow cytometric quantification of Lag3 expression with a (M) representative histogram and (N) quantification of CD45.1⁺ CD8⁺ T cells represented as gMFI where n = 5 mice per OT-I group and n = 7 mice per Lpar5^−/− OT-I group and p = 0.0216. O, P Flow cytometric quantification of Tox expression with a (O) representative histogram and (P) quantification of CD45.1⁺ CD8⁺ T cells represented as gMFI where n = 5 mice per OT-I group and n = 7 mice per Lpar5^−/− OT-I group and p = 0.0151. Statistics for panels (D, G, J) were performed using a Two-way ANOVA with a Tukey’s post-hoc analysis where *p < 0.05, **p < 0.005, ***p < 0.0005, ****p < 0.0001. Statistics for panels (L, N, P) were performed using the unpaired two-sided Student’s t-test analysis where *p < 0.05. Error bars for panels (D, G, J, L, N, P) represent standard error of the mean.

Fig. 4. Signaling via Lpar5 modulates antigen-specific killing in vivo.

A Schematic of in vivo killing assay. B, C Representative flow cytometric histograms and dot plots of target cell input (left) and killing of target cells (right) pulsed with (B) N4 ovalbumin peptide or (C) HSV1 irrelevant peptide. D Frequency of ovalbumin-specific (tetramer⁺) CD8 T cells from the spleens of wildtype C57BL/6 mice and Lpar5^−/− mice immunized with N4 ovalbumin peptide 4 days earlier where n = 3 mice per group and p = 0.8065. E Quantitative analysis of percent specific in vivo killing 5 days after ovalbumin peptide immunization and 1 day after transfer of pulsed target cells from panels A and B where n = 3 mice per group and p = 0.0422. F Schematic of adoptive transfer for in vivo killing assay. G Representative flow cytometric dot plots target cell killing pulsed with N4 ovalbumin peptide. H Quantitative analysis of percent specific in vivo killing 4 days after ovalbumin peptide immunization and 2 h after transfer of pulsed target cells where n = 3 mice per group and p = 0.0123. Statistics for this entire figure were performed using the unpaired two-sided Student’s t-test analysis was performed where *p < 0.05. Error bars for panels (D, E, H) represent standard error of the mean.

Fig. 5. Lysophosphatidic acid rewires CD8 T cell metabolism and modulates reactive oxygen species.

A Mass spectrometry showing global metabolomic data on effector CD8 T cells given media without LPA (RPMI+Glutamine) or treated with 1 µM LPA for 30 min, 2 h, or 4 h prior to sample collection where n = 6 mice per group and shows Euclidean clustering analysis. B Metabolite set enrichment analysis (MSEA) performed on raw data with KEGG analysis to determine enriched metabolic pathways. C Metabolic pathway of γ-glutamyl cycle where blue represents the recycled atoms from γ-L-glutamyl-D-alanine to synthesize glutathione. D–G Relative intracellular abundancies of (D) γ-L-glutamyl-D-alanine and exact p-values are as follows, RPMI+Glutamine vs 30 min LPA p = 0.0313; 30 min LPA vs 2 h LPA p = 0.0347; 30 min LPA vs 4 h LPA p = 0.0290, (E) 5-oxoproline and exact p-values are as follows, 2 h LPA vs 4 h LPA p = 0.0143, (F) L-glutamate and exact p-values are as follows, 2 h LPA vs 4 h LPA p = 0.0262, and (G) glutathione with n = 6 mice per group. H, I Direct measurements of H₂O₂ in effector CD8 T cells after LPA treatment (H) at 1 µM for 30 min, 2 h, or 4 h or (I) at varying concentrations of LPA after 15 min of LPA treatment with n = 3 mice per group. Data measuring reactive oxygen species are normalized to cells cultured in the absence of LPA. J, K Measurements of lipid peroxidation in effector CD8 T cells after LPA treatment at 1 µM for 30 min, 2 h, or 4 h in (J) OT-I effector CD8 T cells and (K) Lpar5^−/− OT-I CD8 T cells with n = 3 mice per group. For (H–K) samples were measured in technical triplicates and error was propagated to biological replicate error where n = 3 mice per group performed in 3 independent experiments. Exact p-values for (H) are as follows, 0 vs 30 min LPA p = 0.0239; 30 min LPA vs 2 h LPA p = 0.0045. Exact p-values for (J) are as follows, RPMI+Glutamine vs 30 min LPA p = 0.0039; 30 min LPA vs 4 h LPA p = 0.0166. Statistics for this entire figure were performed using an ANOVA statistical test with Tukey’s post-hoc analysis was performed where *p < 0.05 and **p < 0.005.

Fig. 6. Lysophosphatidic acid shifts metabolism to consume fatty acids for oxidation.

A–D Oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) by both naïve and effector CD8 T cells given media without LPA (RPMI+Glutamine) or treated with 1 µM LPA for 30 min, 2 h, or 4 h prior to starting the Seahorse metabolic flux assay. Assay was performed with injections of oligomycin (oligo), (4-(trifluoromethoxy) phenyl) carbonohydrazonoyl dicyanide (FCCP), antimycin A (ant), and rotenone (rot) at 18-min intervals in media supplemented with 25 mM glucose. Data are representative and show n = 6 technical replicates. E–H Capacity calculations from Seahorse metabolic flux assay showing basal respiration, maximal respiration, ATP-linked production, and proton leak. Data show n = 5 independent experiments with technical replicate error propagated into biological replicate error. Exact p-values for (E) are as follows, RPMI+Glutamine vs 4 h LPA p = 0.0149. Exact p-values for (F) are as follows, RPMI+Glutamine vs 2 h LPA p = 0.0455; RPMI+Glutamine vs 4 h LPA p = 0.0489. Exact p-values for (G) are as follows, RPMI+Glutamine vs 30 min LPA p = 0.0491. Exact p-values for panel (H) are as follows, 30 min LPA vs 4 h LPA p = 0.0129; 2 h LPA vs 4 h LPA p = 0.0347. I, J Flow cytometric analysis of BODIPY in effector CD8 T cells given media without LPA (RPMI+Glutamine) or treated with 1 µM LPA for 30 min, 2 h, or 4 h. I Shows representative histogram and (J) shows quantitative analysis of normalized geometric mean fluorescence intensity (gMFI) across n = 3 independent experiments with 3 mice per group. Exact p-values are as follows, RPMI+Glutamine vs 2 h LPA p = 0.0002; RPMI+Glutamine vs 4 h LPA p = 0.0002. K Seahorse metabolic flux analysis performed with acute injection of etomoxir to a final concentration of 1 µM. Effector CD8 T cells were cultured in normal media (RPMI+Glutamine) or 1 µM LPA for 4 h prior to starting the assay. n = 6 technical replicates. Exact p-values are as follows, at t = 60 min RPMI+Glutamine vs 4 h LPA p = 0.0049; 4 h LPA vs 4 h LPA + Etomoxir p = 0.0044, at t = 66 min RPMI+Glutamine vs 4 h LPA p = 0.0045; 4 h LPA vs 4 h LPA + Etomoxir p = 0.0042, at t = 72 min RPMI+Glutamine vs 4 h LPA p = 0.0030; 4 h LPA vs 4 h LPA + Etomoxir p = 0.0030. Statistics for (E–K) were performed using an ANOVA statistical test with a Tukey’s post-hoc analysis was performed where *p < 0.05, **p < 0.005, and ***p < 0.0005. Error bars for panels (A–H, J, K) represent standard error of the mean.

Fig. 7. Lysophosphatidic acid receptor 5 modulates metabolic adaptability and efficiency in effector CD8 T cells.

A, B Seahorse metabolic flux assay performed on effector CD8 T cells from (A) an OT-I mouse or Lpar5^−/− OT-I mouse in the absence of LPA treatment and (B) given normal media without LPA (RPMI+Glutamine), treated with 1 µM LPA for 4 h prior to starting the assay, or co-treatment of 1 µM LPA and the LPA receptor antagonist (TC LPA5 4 at 1 µM) for 4 h prior to starting the assay. Data are representative and show n = 6 technical replicates. C–G Capacity calculations from Seahorse metabolic flux assay showing (C) ratio of maximal respiratory capacity / basal respiratory capacity where exact p-values are as follows, OT-I vs Lpar5^−/− OT-I p = 0.0026; Lpar5^−/− OT-I vs OT-I 4 h LPA p = 0.0015; Lpar5^−/− OT-I vs OT-I 4 h LPA + TC LPA5 4 p = 0.0002, D basal respiration where exact p-values are as follows, OT-I vs Lpar5^−/− OT-I + 4 h LPA p = 0.048; OT-I vs OT-I 4 h LPA p = 0.0052; Lpar5^−/− OT-I vs Lpar5^−/− OT-I + 4 h LPA p = 0.0010; Lpar5^−/− OT-I vs OT-I 4 h LPA p = 0.0010, (E) maximal respiration where exact p-values are as follows, OT-I vs Lpar5^−/− OT-I p = 0.0052; OT-I vs Lpar5^−/− OT-I + 4 h LPA p = 0.07; OT-I vs OT-I 4 h LPA p = 0.0055; Lpar5^−/− OT-I + 4 h LPA vs OT-I 4 h LPA p = 0.0049; Lpar5^−/− OT-I + 4 h LPA vs OT-I + 4 h LPA + TC LPA5 4 p = 0.0042; OT-I 4 h LPA vs OT-I 4 h LPA + TC LPA5 4 p = 0.0049, (F) ATP-linked production where exact p-values are as follows, OT-I vs Lpar5^−/− OT-I 4 h LPA p = 0.0001; Lpar5^−/− OT-I vs Lpar5^−/− OT-I + 4 h LPA p = 0.0041; Lpar5^−/− OT-I + 4 h LPA vs OT-I + 4 h LPA p = 0.0050; Lpar5^−/− OT-I + 4 h LPA vs OT-I + 4 h LPA + TC LPA5 4 p = 0.0050, and (G) proton leak where exact p-values are as follows, OT-I vs Lpar5^−/− OT-I p = 0.0080; OT-I vs OT-I + 4 h LPA p = 0.0050; Lpar5^−/− OT-I vs OT-I + 4 h LPA p < 0.0001; Lpar5^−/− OT-I + 4 h LPA vs OT-I + 4 h LPA p = 0.0001; OT-I + 4 h LPA vs OT-I + 4 h LPA + TC LPA5 4 p < 0.0001. Data show n = 3 independent experiments with technical replicate error propagated into biological replicate error. H, I Seahorse metabolic flux assay on effector CD8 T cells from an Lpar5^−/− OT-I mouse cells given media without LPA (RPMI+Glutamine) or treated with 1 µM LPA for 30 min, 2 h, or 4 h prior to starting the assay. Data are representative and show n = 6 technical replicates. J–M Capacity calculations from Seahorse metabolic flux assay showing basal respiration, maximal respiration, ATP-linked production, and proton leak. Data show n = 3 independent experiments with technical replicate error propagated into biological replicate error. Exact p-values for (J) are as follows, RPMI+Glutamine vs 4 h LPA p = 0.0328. Exact p-values for (L) are as follows, RPMI+Glutamine vs 30 min LPA p = 0.0495; RPMI+Glutamine vs 2 h p = 0.0485; RMPI+Glutamine vs 4 h LPA p = 0.0215. Statistics for this entire figure were performed using an ANOVA statistical test with a Tukey’s post-hoc analysis was performed where *p < 0.05, **p < 0.005, ***p < 0.0005, and ****p < 0.0001. Error bars for panels (A-M) represent standard error of the mean.