Autotaxin suppresses cytotoxic T cells via LPAR5 to promote anti–PD-1 resistance in non–small cell lung cancer

Non–small cell lung cancers that harbor concurrent KRAS and TP53 (KP) mutations are immunologically warm tumors with partial responsiveness to anti–PD-(L)1 blockade; however, most patients observe little or no durable clinical benefit. To identify novel tumor-driven resistance mechanisms, we developed a panel of KP murine lung cancer models with intrinsic resistance to anti–PD-1 and queried differential gene expression between these tumors and anti–PD-1–sensitive tumors. We found that the enzyme autotaxin (ATX), and the metabolite it produces, lysophosphatidic acid (LPA), were significantly upregulated in resistant tumors and that ATX directly modulated antitumor immunity, with its expression negatively correlating with total and effector tumor-infiltrating CD8+ T cells. Pharmacological inhibition of ATX, or the downstream receptor LPAR5, in combination with anti–PD-1 was sufficient to restore the antitumor immune response and efficaciously control lung tumor growth in multiple KP tumor models. Additionally, ATX was significantly correlated with inflammatory gene signatures, including a CD8+ cytolytic score in multiple lung adenocarcinoma patient data sets, suggesting that an activated tumor-immune microenvironment upregulates ATX and thus provides an opportunity for cotargeting to prevent acquired resistance to anti–PD-1 treatment. These data reveal the ATX/LPA axis as an immunosuppressive pathway that diminishes the immune checkpoint blockade response in lung cancer.


9
ordered from Sigma, targeting murine Enpp2. Viral transduction of Enpp2 targeting and non-10 targeting control shRNAs was performed as described above with the 344SQ PD1R2 cell line 11 (puromycin 10 µg/ml).

28
Splenocytes were plated onto tumor cells in 24 well plates in replicates and cultured for 4-5 days 29 at 37 o C at 1:10 ratio. After culturing, splenocytes were collected from the media and stained for 30 flow cytometry, described below.

32
For CD8+ T cell isolation, splenocytes were thawed from cryopreservation. CD8+ T cells were 33 purified using a CD8+ T cell negative isolation kit (Miltenyi) following the recommended protocol.

34
After magnetic sorting, purified CD8+ T cells were then stimulated with 3 µg/mL of anti-CD3 and 35 anti-CD28 alone or in the presence of 1 μM LPA. Samples were incubated for various times over 36 a 60-minute time course of stimulation, at which point they were fixed using 1% 37 paraformaldehyde. These samples were then stained for flow cytometry analysis, described 38 below.

41
Immune cells collected from co-culture assays were plated into round bottom 96 well plates and 42 stained with cell membrane T cell markers as described previously (25). After membrane staining, 43 samples were washed, fixed, permeabilized, and then stained with intracellular markers. The 44 purified CD8+ T cells stimulated with anti-CD3 and anti-CD28 with and without LPA were fixed 45 and permeabilized before staining with a phospho-ERK antibody. An IgG control antibody was 46 used for one sample as a negative control. Cells were then washed and stained with AlexaFluor-47 488 secondary. An unstained control was also included. These samples were acquired using the 48 BD LSR Fortessa (BD Biosciences) and analyzed using FlowJo software (version 10.6.1).

50
For LPAR staining of CD8+ T cells, 344SQ and 344SQ PD1R1 tumors and spleens were collected 51 and processed as described in methods section. Single cell suspensions were then stained with 52 a live/dead marker, CD45, CD3, CD8 and CD4 antibodies. Samples were fixed and 53 permeabilized, then separated into 3 different tubes. Each one was stained with a different LPAR 54 antibody (LPAR2, LPAR5, or LPAR6). An isotype control was used to account for non-specific 55 binding. After washes, the experimental samples and the IgG controls were stained with a 56 fluorophore conjugated secondary antibody (AlexaFluor-488). Two additional controls included 57 were an unstained control and an LPAR fluorescence minus one (FMO) sample, which included 58 all T cell markers without the LPAR antibody, IgG control antibody, or the secondary antibody.

59
Samples were acquired using BD LSR Fortessa (BD Biosciences) and analyzed using FlowJo 60 software.

62
Semi-quantitative real-time PCR

63
RNA was extracted from cells in culture using Trizol Reagent (Thermo Fisher) following the 64 recommended protocol. All RNA samples were quantified, and reverse transcription was 65 performed with 2 μg of RNA using qSCRIPT cDNA SuperMix (Quantabio). Real-time PCR was 66 performed using primer sets specific for each gene (sequences listed in Table S2) and the SYBR®

71
For the conditioned media samples collection, cells were plated in a 6 well plate at equal cell 72 numbers, grown overnight to ~95% confluency, then the media was changed from complete 73 media to serum free media and incubated for 24 hours. The media was collected, and 1 mL was 74 processed using acetone precipitation. Samples were stored at -80 o C overnight. The samples  Table S2). Secondary antibodies were added the next day after washes, and signal was 83 observed using ECL reagents and a film processor.

85
Immunohistochemistry 86 Paraffin-embedded tumor sections were processed following a standard protocol. Following heat-87 mediated antigen retrieval (Dako Technologies) and blocking with 5% goat serum, the primary 88 antibodies were added to tissue based upon manufacturer recommendations. Images were 89 acquired using brightfield microscopy on the Olympus IX73. For the CD8 and granzyme B stains,

90
ImageJ software was used to quantify the number of positive cells per field of view. Image 91 deconvolution was used to view DAB staining separately from the nuclear stain, and a threshold 92 was applied to all images. The "analyze particles" feature was used to quantify the number of     Cryopreserved splenocytes, collected as described above, were thawed and CD8+ T cells were 133 specifically extracted using the CD8+ T cell Isolation Kit from Miltenyi. Purified CD8+ T cells were 134 then plated at 100,000 cells per well in a 24 well plate containing glass coverslips. The plate was 135 spun for 10 min at 2,000 rpm to adhere T cells to the coverslips. Cells were then fixed in 1% 136 paraformaldehyde for 15 min at room temperature, washed with glycine, then permeabilized in 137 0.1% triton-X. Blocking was performed using 5% normal goat serum for 1 hour, then individual 138 LPAR antibodies were added overnight. Secondary was added the next day and Alexa-Fluor 488 139 goat anti-rabbit was used for all. A control well with secondary only was used to control for non-140 specific secondary binding. Coverslips were mounted onto glass slides using Prolong-Gold with 141 DAPI to stain the cell nuclei. Images were obtained using an Olympus IX71 widefield microscope.

142
Quantification of LPAR+ cells was done using ImageJ software. A threshold was applied to 143 images and the "analyze particles" feature was used to identify LPAR+ cells on the GFP channel 144 and the total number of cells on the DAPI channel.

145
Fig S1. 344SQ PD1R cells display no overt differences in morphology or EMT status and are resistant to anti-PD-L1 re-challenge. A. 344SQ lung cancer cells were treated in vivo with either IgG control or anti-PD-1 antibody until the development of resistance, at which point tumors were excised and cell lines developed from them (see Figure 1A). Phase contrast microscopy was performed to demonstrate the morphology of the 6 different IgG treated cell lines and the 5 different anti-PD-1 treated cell lines. B. The 6 IgG treated 344SQ cell lines (termed 344SQ PD1S1-6 or S1-6) and the 5 anti-PD-1 treated 344SQ cells lines (termed 344SQ PD1R1-5 or R1-5) were collected for RNA. QPCR analysis was performed to analyze the expression of the epithelial marker Ecadherin (Cdh1) and the mesenchymal markers N-cadherin (Cdh2), Vimentin, and Zeb1. L32 was used as an internal normalization control, and values were then normalized the parental 344SQ expression levels. C. The anti-PD-1 sensitive and resistant models (cell lines and tumors) were collected for protein and analyzed via western blotting for epithelial and mesenchymal markers described in B. Actin was used as a loading control. Red dashed line used to demarcate the sensitive versus resistant samples. D-E. Two representative 344SQ anti-PD-1 sensitive and resistant cell lines (S1 and S2; R1 and R2) were implanted into mice and treated with either IgG control or anti-PD-L1 antibody. n = 4-5 mice/ group. D. Tumor growth was measured weekly via caliper measurements. The 2 sensitive models are shown on top, and the 2 resistant models are shown below. *p<0.05, **p<0.01 by multiple t tests (one per timepoint). E. Total macroscopic lung metastases were counted at necropsy. *p<0.05 by t test.

Fig S2. 344SQ PD1R cells maintain key inflammatory molecules and pathways, including PD-L1 expression, IFNγ response and antigen presentation machinery. A.
The cell surface expression of PD-L1 was measured on 344SQ PD1S1-6 and 344SQ PD1R1-5 cells via flow cytometry. The value is graphed as the percentage of total live single cells. B. 344SQ PD1S1-6 and 344SQ PD1R1-5 cells were collected for protein and analyzed via western blotting for phosphorylated and total levels of JAK2, STAT1, and STAT3. Actin was used as a loading control. C. The parental 344SQ, 2 of the anti-PD-1 sensitive (S1 and S2) and 3 of the anti-PD-1 resistant (R1-3) cell lines were grown in culture with the addition of murine IFNγ to the culture media or in control media. After 48 hours, cells were collected for qPCR analysis of Irf1. L32 was used as an internal control, and all samples are normalized to the 344SQ control (no IFNγ) sample. D. The 344SQ parental cells or representative anti-PD-1 sensitive and resistant lines were collected and analyzed via flow cytometry for surface expression of MHCI. The values are graphed as a percentage of total live single cells. E. The Tap1 and Tap2 gene expression was analyzed using qPCR in the 344SQ parental, PD-1 sensitive (S1 and S2) and the PD-1 resistant (R1-3) cells. F. The 344SQ anti-PD-1 sensitive (S1-6) and anti-PD-1 resistant (R1-5) cells were analyzed via qPCR for the expression of CD38. L32 was used as an internal normalization control, and values were then normalized the parental 344SQ expression levels. G. The protein expression of CD38 was analyzed across the 344SQ anti-PD-1 sensitive and resistant panels (cell lines and tumors) via western blotting. Actin was used as a loading control. The cell line actin blot was also shown in Figure S1C. The tumors generated from the 344SQ, 344SQ PD1S , and 344SQ PD1R models were analyzed via flow cytometry for tumor-infiltrating CD4+ subsets (Figure 2A). CD4+ T cells are graphed as a percentage of total T cells. CD4+ effectors (CD278+) and CD4+ Tregs (FoxP3+/CD25+) are graphed as a percentage of CD4+ T cells. *p<0.05 by t test. C. The myeloid populations from the experiment described in Figure 2A were also analyzed. Dendritic cells are depicted as a percentage of CD11c+ (from CD45+). Myeloidderived suppressor cells (MDSCs) are graphed as a percentage of the CD45+ subset. Macrophages are gated underneath CD45+/CD11b+ cells. Total macrophages (not shown) are then divided into M1 (iNOS+/MHCII+/CD86+) or M2 (Arg1+/MHCII-). *p<0.05, ****p<0.0001 by t test. D-E. Additional flow cytometry analysis from the experiment depicted in Figure 2D. D. CD4+ T cell subsets are shown as described in panel B. E. Myeloid cell populations were analyzed as described in panel C. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001 by 1-way ANOVA. Fig S4. Increased Enpp2/ATX is associated with activated immune signature and anti-PD-1 resistance in mouse and human tumors. A. The KP IgG and KP PDL1 cell lines were collected for RNA and protein to analyze Enpp2/ATX expression via qPCR (top) and western blotting (bottom). ***p<0.001 by t test. B. 344SQ parental, anti-PD-1 sensitive, and anti-PD-1 resistant conditioned media was collected for western blot analysis of secreted ATX expression. Ponceau S stain shows total protein levels and acts as a loading control. Densitometry was performed by measuring the ATX band compared to the Ponceau S stain and all values were normalized to the 344SQ parental line. This densitometry was quantified in the graph to the right. *p<0.05 by t test. C. 344SQ PD1S and 344SQ PD1R tumors were collected for RNA and the expression of Plpp1-3 were analyzed via qPCR. L32 was used as an internal control, and expression levels in all samples were normalized to the S4 sample. *p<0.05, **p<0.01, ***p<0.001 by t tests with grouped samples (PD1S versus PD1R). D. MD Anderson ICON patients were analyzed for expression of ENPP2 correlated to numerous clinical features, including overall survival (OS), T cell cytolytic score (CYT), histology, treatment, tumor mutational burden (TMB). E-F. Analysis of the published dataset from melanoma patients on nivolumab therapy (GSE91061)(34). Individual paired tumors (pre-treatment and on-treatment) were analyzed for ENPP2 expression levels. F. The change in ENPP2 expression (delta) between pre-treatment and on-treatment samples was calculated. Patients were separated into progressive/stable disease (PDSD) or partial/complete responders (PRCR), and the delta in ENPP2 expression was graphed in each of these patient categories.

Fig S5. ATX expression promotes invasion of 344SQ lung cancer cells and decreases CD8+ T cell phenotypes in tumors. A. 344SQ vector (ctrl) or ATX-overexpressing cells were analyzed via qPCR to confirm
Enpp2 levels. L32 was used as an internal control, and values are normalized to control cells. B. 344SQ ctrl and ATX-overexpressing cells were plated for a WST-1 proliferation assay over 3 days. Raw absorbance values are graphed. C. 344SQ ctrl and ATX cells were plated for a WST-1 assay as described in B. At the time of plating, DMSO control or lysophosphatidylcholine (LPC) was added to cells. Absorbance values are graphed relative to the 24hr no LPC control **** p<0.0001 by 2-way ANOVA with Tukey's correction. D. The migratory and invasive capabilities of the 344SQ ctrl and ATX cells were analyzed using Boyden Transwell assays without Matrigel (migration; left) or with Matrigel (invasion; right). n= 3 chambers/cell line. E. 344SQ ctrl and ATX overexpressing cells were plated in 3-D assays using either 100% Matrigel (left) or a 50:50 mixed Matrigel/Collagen type I matrix (right). Representative brightfield images are shown in each condition. Invasive structures were calculated as a percentage of total structures. n = 2-3 wells/condition, 5 images/well. *p<0.05 by t test. F. 344SQ ctrl and ATXoverexpressing cells were implanted into mice. After 3 weeks, tumors were excised, and immune populations were characterized by flow cytometry. Total CD8+ T cells were calculated as a fraction of live cells, and Granzyme B+ cells were calculated underneath the CD8+ population. *p<0.05 by t test. G. The tumors from F were analyzed for the CD4+ subpopulations. H. The 344SQ PD1R2 -shATX#4 and scrambled control cells (described in Figure 4) were implanted into mice and after 3 weeks, tumors were excised and processed for flow cytometry as described in Figure 2. *p<0.05, **p<0.01 by t test. I. The tumors from H were analyzed for CD4+ subpopulations by flow cytometry .   Fig S6. ATX inhibition with anti-PD-1 increased CD4+ proliferation and efficaciously controlled tumor growth in KP anti-PD-1 sensitive and resistant tumor models. A. 344SQ and 344SQ PD1R1 cells were plated for an LPA ELISA. To the 344SQ PD1R1 cells, the ATX inhibitor PF-8380 was added at 1μM for 24 hours prior to media collection for ELISA. *p<0.05, **p<0.01 by one-way ANOVA corrected for multiple comparisons. B. 344SQ tumors collected from mice treated with the PF-8380 ATX inhibitor or vehicle control (from the experiment in Figure 5C) were processed for LPA ELISA. C. Tumor growth data for flow experiment described in Figure 5A. **p<0.01, ****p<0.0001 by 2-way ANOVA. D. Additional flow analysis from the experiment shown in Figure 5A. Total CD4+ T cells were gated under total T cells (CD45+/CD3+) (left). Proliferating CD4+ T cells were gated under total CD4+ T cells and were Ki67+ (right). **p<0.01 by one-way ANOVA corrected for multiple comparisons. E. The KP IgG cell line (anti-PD-L1 sensitive) was implanted into wildtype mice and treated with vehicle control, the ATX inhibitor PF-8380, anti-PD-L1, or the combination. Tumor growth over time was measured weekly via calipers (left) and at endpoint (right). n = 4-5 mice/treatment. *p<0.05 by 1-way (right graph) or 2-way ANOVA (left graph). F. The KP PDL1 cell line (anti-PD-L1 resistant) was implanted into wildtype mice. Mice were treated and tumor growth measured as described in D. Open shapes are from one experiment that lasted 3 weeks, and closed shapes are from another experiment that went for 4 weeks. n = 4-5mice/treatment. ***p<0.001 by 2-way ANOVA (left graph), *p<0.05 by 1-way ANOVA (right graph). Fig S8. LPA diminishes CD8+ T cell functions. A. Naïve CD8+ T cells were purified from murine splenocytes and activated using anti-CD3 and anti-CD28 antibodies, alone or with treatment of exogenous lysophosphatidic acid (LPA). T cells were then collected at various timepoints of stimulation and stained for flow cytometry analysis of phospho-ERK levels. n = 2 replicates/condition. B. Naïve immune cells from murine splenocytes (left) or human PBMCs from healthy donors (right) were stimulated with anti-CD3 and anti-CD28 in the presence of LPA (20 µM) or solvent control for 24 hours. Conditioned media was then collected for IFN-γ ELISA analysis. *p<0.05, **p<0.01, ***p<0.001 by multiple t tests. C. Naïve immune cells isolated from mouse splenocytes were stimulated with anti-CD3 and anti-CD28 in the presence of LPA (20 µM) and/or the LPAR5 inhibitor AS2717638 (5 µM). After 72 hours, immune cells were collected and stained for multicolor flow cytometry analysis. *p<0.05, **p<0.01 by 1-way ANOVA. D. 344SQ anti-PD-1 sensitive (S1 and S2) and anti-PD-1 resistant (R1 and R2) were analyzed via qPCR for expression of Lpar1, Lpar2, Lpar3 (no amplification; not shown), Lpar4, and Lpar6. L32 was used as an internal normalization control, and samples were all normalized to the S1 line. E. The cells from D were stimulated with increasing concentrations of LPA (1µM and 10µM) or control and collected for western blot analysis of downstream signaling pathways, including phospho-PKA, phospho-PKC, phospho-AKT, and phospho-ERK. Actin was used as a loading control. F. Mice bearing 344SQ tumors were treated with the pan-LPAR inhibitor BrP-LPA or the LPAR5 inhibitor AS2717638 alone and in combination with anti-PD-1 as described in Figure 6G. *p<0.05, ****p<0.0001 by 2-way ANOVA with multiple corrections.