Extracellular matrix educates an immunoregulatory tumor macrophage phenotype found in ovarian cancer metastasis

Recent studies have shown that the tumor extracellular matrix (ECM) associates with immunosuppression, and that targeting the ECM can improve immune infiltration and responsiveness to immunotherapy. A question that remains unresolved is whether the ECM directly educates the immune phenotypes seen in tumors. Here, we identify a tumor-associated macrophage (TAM) population associated with poor prognosis, interruption of the cancer immunity cycle, and tumor ECM composition. To investigate whether the ECM was capable of generating this TAM phenotype, we developed a decellularized tissue model that retains the native ECM architecture and composition. Macrophages cultured on decellularized ovarian metastasis shared transcriptional profiles with the TAMs found in human tissue. ECM-educated macrophages have a tissue-remodeling and immunoregulatory phenotype, inducing altered T cell marker expression and proliferation. We conclude that the tumor ECM directly educates this macrophage population found in cancer tissues. Therefore, current and emerging cancer therapies that target the tumor ECM may be tailored to improve macrophage phenotype and their downstream regulation of immunity.

Steps of the cancer immunity cycle Score (AU)

High Expression of COL11A1 Low Expression of COL11A1
Step 4: Immune cell trafficking to tumours Step 1: Release of cancer cell antigens Step 2: Cancer antigen presentation Step 3: Priming and activation Step 5: Infiltration of immune cells to tumours Step 6: T cell recognition of cancer cells Step 7 Steps of the cancer immunity cycle Score (AU)

High Expression of VCAN Low Expression of VCAN
Step 4: Immune cell trafficking to tumours Step 1: Release of cancer cell antigens Step 2: Cancer antigen presentation Step 3: Priming and activation Step 5: Infiltration of immune cells to tumours Step 6: T cell recognition of cancer cells Step 7: Killing of cancer cells Step 4 Step 1: Release of cancer cell antigens Step 2: Cancer antigen presentation Step 3: Priming and activation Step 5: Infiltration of immune cells to tumours Step 6: T cell recognition of cancer cells Step 7 Steps of the cancer immunity cycle Score (AU)

High Expression of MXRA5 Low Expression of MXRA5
Step 4: Immune cell trafficking to tumours Step 1: Release of cancer cell antigens Step 2: Cancer antigen presentation Step 3: Priming and activation Step 5: Infiltration of immune cells to tumours Step 6: T cell recognition of cancer cells Step 7 Step 3:

Priming and activation
Step 4: Immune cell trafficking to tumours Step 5: Infiltration of immune cells to tumours Step 6: T cell recognition of cancer cells Step 7: Killing of cancer cells ty cycle ** : of ells rs Step 6: T cell recognition of cancer cells Step 7 Steps of the cancer immunity cycle Score (AU) Step 1: Release of cancer cell antigens Step 2: Cancer antigen presentation Step 3:

Priming and activation
Step 4: Immune cell trafficking to tumours Step 5: Infiltration of immune cells to tumours Step 6: T cell recognition of cancer cells Step 7 Steps of the cancer immunity cycle Score (AU) ** * Step 1:

Release of cancer cell antigens
Step 2: Cancer antigen presentation Step 3:

Priming and activation
Step 4: Immune cell trafficking to tumours Step 5: Infiltration of immune cells to tumours Step 6: T cell recognition of cancer cells Step 7: Killing of cancer cells

Steps of the cancer immunity cycle
High Expression of M0 immunomatrisome signature Low Expression of M0 immunomatrisome signature ** * Step 2: Cancer antigen presentation Step 3:

Priming and activation
Step 4: Immune cell trafficking to tumours Step 5:

Infiltration of immune cells to tumours
Step 6: T cell recognition of cancer cells Step 7: Killing of cancer cells unity cycle of M0 immunomatrisome signature f M0 immunomatrisome signature ** tep 5: ration of une cells umours Step 6: T cell recognition of cancer cells Step 7 Steps of the cancer immunity cycle

Release of cancer cell antigens
Step 2: Cancer antigen presentation Step 3:

Priming and activation
Step 4: Immune cell trafficking to tumours Step 5:

Infiltration of immune cells to tumours
Step 6: T cell recognition of cancer cells Step 7 Steps of the cancer immunity cycle

Score (AU)
High Expression of M0 immunomatrisome signature Low Expression of M0 immunomatrisome signature **

Release of cancer cell antigens
Step 2: Cancer antigen presentation Step 3:

Priming and activation
Step 4: Immune cell trafficking to tumours Step 5:

Infiltration of immune cells to tumours
Step 6: T cell recognition of cancer cells Step 7  Steps of the cancer immunity cycle Score (AU) High Expression of SFRP2 Low Expression of SFRP2 Step 4: Immune cell trafficking to tumours Step 1: Release of cancer cell antigens Step 2: Cancer antigen presentation Step 3: Priming and activation Step 5: Infiltration of immune cells to tumours Step 6: T cell recognition of cancer cells Step 7: Killing of cancer cells Steps of the cancer immunity cycle 1: e of cell ns Step 2: Cancer antigen presentation Step 3:

Priming and activation
Step 4: Immune cell trafficking to tumours Step 5: Infiltration of immune cells to tumours Step 6: T cell recognition of cancer cells Step 7: Killing of cancer cells munity cycle ** Step 5: Infiltration of immune cells to tumours Step 6: T cell recognition of cancer cells D C B A Supplementary Figure 6. Laser capture demonstrates enrichment for matrisome molecules associated with M0 macrophages in the stroma. Representative images of tissue with areas marked for A) tumor and B) stroma laser capture. C) Validation of tumor and stroma area laser capture using known malignant cell and fibroblast markers, e.g. malignant cell markers: PAX8, EPCAM, and fibroblast marker: ACTA2. D) Tumor/Stroma ratio for matrisome proteins associated with M0 macrophages. N=2 (G33, G75).  Figure 7. M0 macrophage ECM signature associates with poorer prognosis in HGSOC and across multiple cancer types. A) Kaplan-Meier survival curves with overall survival, divided by low and high gene expression levels of the averaged ECM signature associated with M0 macrophages. Serous ovarian cancer patients N = 523. Median survival time is calculated at 50% survival probability. B) Multivariate hazard ratio (HR) with 95% CI, derived using multiple datasets across a range of cancer types, with patients divided by low and high gene expression levels of the M0 macrophage-associated ECM signature averaged. HR > 1 indicates that the M0 macrophage-associated ECM signature gene expression is inversely correlated with OS, while HR < 1 shows positively correlated OS. Log ranked pvalues significances are presented by asterisks; **** p < 0.0001, ***p < 0.001, **p < 0.01 and *p < 0.            More stringent analysis used fold change greater than 2 (logFC ≥ 1 or logFC ≤ -1). E) Volcano plot of p-values and Log 2 FC of high versus low disease tissue cultured macrophages. T tests for differential expression. G276_D4_HighDisease G150_D4_HighDiseas e G302_D4_HighDis ease G164_D4_HighDisease G276_D3_HighDisease    Figure 1D and E). A limitation of our evaluation of the bulk deconvolution methods is that immune cell type proportions were assessed using IHC for only six markers: CD3 + , CD4 + , CD8 + , CD68 + , CD45RO + and FOXP3 + . As 22 immune cell types are computed by CIBERSORT and CIBERSORTx, and 34 immune cell types are computed by xCell, computed immune cell types were combined to assess their correlation against the IHC cell counts (e.g. xCell CD8 + naïve T cells, CD8 + Tcm and CD8 + Tem computed values were combined to correlate against CD8 + IHC immune cell counts). The gold standard to evaluate bulk deconvolution would be to compare against cell type proportions measured using single cell RNA seq or high resolution multiparameter flow cytometry which has not been possible to perform here but has been performed comprehensively in the literature 19,20,22 .
Clinical outcome analysis. Delogged HGSOC RNA-seq data was input to estimate the tumor progression within the cancer immunity cycle, using the tracking tumor immunophenotype (TIP) meta-server 31 .
Samples were separated into the top 12 high and bottom 12 low expressing samples. Kaplan-Meier plots, hazard ratios and log ranked p significance values were generated using gene expression omnibus (GEO), European genome archive (EGA) and the cancer genome atlas (TCGA) RNA-seq databases, via the KM plotter online meta-analysis tool 71 .
RNA isolation and sequencing. Total RNA was extracted from macrophage cultures on decellularized tissue using RLT buffer (Qiagen) and rigorously vortexed. Samples were processed using the RNeasy Micro Kit (Qiagen) following manufacturer's instructions. RNA quality and integrity assessments were performed at Oxford Genomics. RNA sequencing was performed at Oxford Genomics. Material was quantified using RiboGreen (Invitrogen) on the FLUOstar OPTIMA plate reader (BMG Labtech) and the size profile and integrity analysed on the 2200 or 4200 TapeStation (Agilent, RNA ScreenTape). RIN estimates (where available) were between 7 and 9.7. Input material was normalised to 10 ng (or maximum mass available) prior to library preparation. Polyadenylated transcript enrichment and strand specific library preparation was completed using NEBNext Ultra II mRNA kit (NEB) following manufacturer's instructions. Libraries were amplified (18 cycles) on a Tetrad (Bio-Rad) using in-house unique dual indexing primers (based on DOI: 10.1186DOI: 10. /1472. Individual libraries were normalised using Qubit, and the size profile was analysed on the 2200 or 4200 TapeStation. Individual libraries were normalised and pooled together accordingly. The pooled library was diluted to ~10 nM for storage. The 10 nM library was denatured and further diluted prior to loading on the sequencer. Paired end sequencing was performed using a NovaSeq6000 platform (Illumina, NovaSeq 6000 S2/S4 reagent kit v1.5, 300 cycles), generating a raw read count of >23 million reads per sample.
Histochemical analysis. Frozen tissues were fixed in in 4 % paraformaldehyde (PFA) and cryosectioned to 8-10 µm slices. All tissue sections were scanned using a 3DHISTECH Panoramic 250 digital slide scanner (3DHISTECH, Hungary) and the resulting scans were analysed using Definiens software (Definiens AG, Germany). Disease scores were determined firstly by manually defining regions of interest in the tissue that represented tumor (PAX8), stroma, and fat (adipocytes) and then training the software to recognize these regions of interest. Disease score was expressed as a percentage of the whole tissue area that contained tumor and/or stroma ( Figure 2B-D). Paraffin embedded tissues were submerged in xylene and then a series of ethanol washes of decreasing concentration for 2 x 2 min each (100 %, 90 %, 70 %, and 50 %). Antigen retrieval was performed for 10 min using vector antigen unmasking buffer and a pressure cooker. Tissue sections were then washed with DAKO wash buffer followed by application of H2O2 for 5 min. Blocking was performed using 5 % BSA for 20 min at RT followed by incubation with primary antibody in biogenex antibody diluent for 30 min. After 3 x washes, biogenex super enhancer was added for 20 min and then washed off before addition of biogenex ss label poly-HRP for 30 min. Tissues were washed three times before addition of DAB chromagen for 3 min followed by washing to stop further DAB development. Tissues were counterstained with haematoxylin followed by washing with H2O and ethanol solutions of increasing concentration for 2 min each (50 %, 70 %, 90 %, 100 %) and then 2 x xylene. Samples were then mounted and scanned using the 3DHISTECH Panoramic digital slide scanner.
Immune cells were counted using QuPath.
Matrix staining. Immunohistochemical staining for ECM proteins was performed on 4 µm slides of FFPE human omentum tissue as described above.
Antibodies. The following antibodies were used for immunohistochemical analyses: anti-PAX8 (clone Disease score quantification. The level of disease was calculated via IHC analysis of whole tissue sections as a disease score (Supplementary Figure 9B-D). The disease score was devised by a scoring system that quantified the sum of the area of tumor cells (in this case we used PAX8 as marker of ovarian cancer cells) and tissue stroma (Supplementary Figure 9B-C). The tissue stroma was the area of desmoplasia (or tumor ECM) resulting from the invasive tumor within the tissue. Whole tissue sections were analyzed using Definiens® digital image software, which calculated the percentage of the tumor, stroma and adipose content within the tissues (Supplementary Figure 9C-D). The 39 samples were clinically graded as stage III-IV and were collected predominantly from patients with HGSOC, but also patients with high grade clear cell carcinoma, malignant Brenner tumor, Borderline serous cystadenoma, benign fibroma, benign mucinous cystadenoma, sero-mucinous borderline tumor and mucinous adenocarcinoma (Supplementary Table 5 and Supplementary Figure 13). QuPath was used to quantify PAX8 + cells (Supplementary Figure 9E) and PAX8 + cell counts positively correlated with the disease score (Supplementary Figure 9F), indicating the quantity of tumor cells associated with the level of ECM remodeling. The clinical staging of samples did not correlate with the tissue disease score. We used the disease score to account for differences between tissues that may result from the level of disease present.