Treatment of colorectal peritoneal metastases with oxaliplatin induces biomarkers predicting response to immune checkpoint blockade

Abstract

Background: Colorectal cancer (CRC) patients with inoperable peritoneal metastases (PM) have a dismal prognosis with limited treatment options. Local treatment of CRC-PM with oxaliplatin is commonly applied, but biomarkers steering patient selection, or informing potentially effective combination therapies are lacking. A novel potentially effective treatment strategy is Pressurized IntraPeritoneal Aerosol Chemotherapy (PIPAC) in which CRC-PM are exposed to cyclic treatment with high concentrations of locally applied oxaliplatin. However, it is unclear whether and how CRC-PM respond to PIPAC.

Methods: Here, we generated a biobank from 20 patients receiving PIPAC with oxaliplatin for CRC-PM. The biobank contains biopsies from 3 PM per patient, repeatedly sampled prior to each treatment cycle, and ascites. Anti-tumor effects were analyzed by shallow single-cell karyotype sequencing (sc-karyoSeq). RNA-sequencing and proteomics were performed to assess changes in gene and protein expression. Immunohistochemistry was performed to assess treatment-induced changes in tissue histology. Ascites was used to assess immunoglobulin content and reactivity.

Results: PIPAC reduced genomic heterogeneity and aneuploidy scores among PIPAC-surviving tumor cells. Furthermore, PIPAC reduced immunosuppressive signals (hypoxia, interleukin-10, transforming growth factor β), and induced an influx of B and T lymphocytes, which organized into metastasis-associated Tertiary Lymphoid Structures (TLS). TLS are biomarkers predicting response to Immune-Checkpoint Inhibitors (ICIs). The T cells residing in PIPAC-induced TLS expressed high levels of the checkpoints PD-1, TIGIT and EBI3. PIPAC also caused the generation of plasma cells producing tumor-reactive antibodies.

Conclusion: PIPAC shows modest anti-tumor activity and induces immune parameters predicting response to ICIs. Patients with inoperable CRC-PM may therefore benefit from PIPAC in combination with ICIs.

Keywords: CRC; Colorectal; Heterogeneity; Karyotype; Oxaliplatin; PIPAC; Peritoneal metastasis; Reverse translation; Tertiary lymphoid structure.

Fig. 1

PIPAC reduces intra-tumor genetic heterogeneity and selects for cells with simpler karyotypes. Single cell shallow sequencing data (Figure S1) were analyzed for genetic heterogeneity scores (A) and aneuploidy scores (B). To visualize the longitudinal effect of PIPAC all scores of all available samples (Figure S2) were plotted over time. Cycle 1 contains 13 samples (regions/metastases) from 7 patients. Cycle 2 contains 5 samples from 3 patients. Cycle 3 contains 8 samples from 3 patients. Cycle 4 contains 2 samples from 1 patient. Each dot represents a single chromosome and displays the average heterogeneity and aneuploidy scores per chromosome (22 dots).

Fig. 2

PIPAC reduces hypoxia-related gene expression and alters the immune microenvironment of peritoneal metastases. Dotplots showing expression of (A) HIF-1 alpha target genes [12] (B) CA9, (C) B cell signature [14], (D) plasma cell signature [16], (E) CD8 T cells [14], (F) CTLs [15] in treatment-naïve (red dots) versus PIPAC-treated (blue dots) peritoneal metastases. (G) XY plot showing the correlation between expression signatures reflecting B cell and T cell infiltration into peritoneal metastases. Treatment status is color coded: ePIPAC-naïve (red dots) and PIPAC-treated (blue dots. H) Heatmap showing the inverse relationship between expression of immune cell signatures and hypoxia-related gene expression. Pearson r values are color-coded from −1 (blue) to +1 (red). (I) CLUE-GO analysis of the 46 PIPAC-induced proteins identified by proteomics in both patient 3 and patient 6. See Table S4 for the complete protein list and Table S5 for statistical significance of the overrepresentation of GO terms and associated proteins in PIPAC-treated metastases.

Fig. 3

Proteomic and phosphoproteomic analysis of PIPAC-treated peritoneal metastases. (A) Unsupervised hierarchical clustering of 739 ANOVA-significant proteins based on z-scored log2 abundance values for patient 03 and 06 over the time-course of the treatment (t1, pretreatment, t2 and t3, both posttreatment). Patient 03 timepoint 1 (t1) is represented with n = 2. (B) Statistical overrepresentation test for proteins assigned to cluster 2 (patient 3) using GO-terms (biological process). Overrepresented terms are filtered for a Benjamini-Hochberg FDR < 0.05. (C) Time course of immunoglobuline protein expression. Log2-normalized protein abundances for 32 immunoglobulines grouped in cluster2 are shown. A mean immunoglobuline expression value for each timepoint was calculated and the trendline is indicated in red. (D) Selected time courses of STAT1 and STAT3 phosphopeptide isoform expression. Individual log2-normalized phosphopeptide isoform abundances per timepoint are shown, median values are indicated with horizontal dash (-). Patient 03 timepoint 1 (t1) is represented with n = 2.

Fig. 4

PIPAC causes reduced expression of immunosuppressive cytokines and increased expression of immune checkpoint genes. Dotplots showing RNA expression levels of (A) TGFB2, (B) IL10, (C) TGFB3, (D) TIGIT, (E) EBI3, and (F) CXCL9, in treatment naïve peritoneal metastases versus all PIPAC-treated samples. See Table S7 for the complete gene list examined. (G) Heatmap showing the inverse relationship between expression of immunosuppressive cytokines (IL10, TGFB2, TGFB3) and immune checkpoint genes (TIGIT, CXCL9, EBI3). Pearson r values are color-coded from −1 (blue) to +1 (red). (H) Heatmap showing the positive correlation of expression of signatures reflecting immune cell infiltration and checkpoint activation. Pearson r values are color-coded from 0 (white; no correlation) to +1 (red; perfect correlation). (I) XY plots showing the correlation between expression of signatures reflecting B cell (left) and T cell (right) infiltration with expression of the immune checkpoint gene TIGIT.

Fig. 5

PIPAC causes formation of metastasis-associated Tertiary Lymphoid Structures (TLS). Immunohistochemistry analysis of (A) CD3 (T cells), (B) CD20 (B cells), (C) CD138 (plasma cells) in biopsies from PIPAC-naïve and -treated samples over time. All stained sections were quantified using QuPath, and the resulting data are represented in (D) CD3, (E) CD20, and (F) CD138. (G) Immunofluoresence analysis of the expression of CD3 (T cells), CD20 (B cells) and HLA (MHC class II) in three peritoneal metastases over time. Low expression of all three markers in PIPAC-naïve metastases (cycle 1). Clusters of B and T cells in ePIPAC-treated metastases start to form after one (B) and two (C) treatment cycles. (J) Overview of a metastasis (dotted white line) surrounded by various TLS following 2 cycles of PIPAC. Bar=250μm. Significance of the differential staining between treatment cycles was assessed using student’s T test. **p < 0.001. ***p < 0.0001. ****p < 0.00001. Bar=250 mm.

Fig. 5

PIPAC causes formation of metastasis-associated Tertiary Lymphoid Structures (TLS). Immunohistochemistry analysis of (A) CD3 (T cells), (B) CD20 (B cells), (C) CD138 (plasma cells) in biopsies from PIPAC-naïve and -treated samples over time. All stained sections were quantified using QuPath, and the resulting data are represented in (D) CD3, (E) CD20, and (F) CD138. (G) Immunofluoresence analysis of the expression of CD3 (T cells), CD20 (B cells) and HLA (MHC class II) in three peritoneal metastases over time. Low expression of all three markers in PIPAC-naïve metastases (cycle 1). Clusters of B and T cells in ePIPAC-treated metastases start to form after one (B) and two (C) treatment cycles. (J) Overview of a metastasis (dotted white line) surrounded by various TLS following 2 cycles of PIPAC. Bar=250μm. Significance of the differential staining between treatment cycles was assessed using student’s T test. **p < 0.001. ***p < 0.0001. ****p < 0.00001. Bar=250 mm.