Introduction

Metastatic colorectal cancer (CRC) is a major cause of cancer-related deaths2,3,4. Although less frequent than hematogenous metastasis, peritoneal metastasis (PM) occurs in up to 20% of CRC patients5. Systemic therapy remains the treatment of choice for patients with any metastatic CRC, however there is evidence that hematogenous metastasis is better controlled than PM3,4. The reason for this is unknown but might be attributable to different molecular subtypes among metastatic sites6 or the specific microenvironment in the peritoneal cavity7. In patients with CRC PM only a highly selected subset of patients qualify for radical resection8. In those highly selected patients, so-called cytoreductive surgery (CRS) where all metastatic lesions are resected, followed by local application of hyperthermic intraperitoneal chemotherapy (HIPEC), can be performed. Although clinical trials remain unclear about the role, duration, and composition of HIPEC9,10, the benefit of surgery as a part of multimodal treatment of PM is highly evident11,12. In many cohorts, median survival rates significantly increased up to 50 months for a disease which was considered terminal until not so long ago4,9.

The current concept of CRS/HIPEC follows a two-step process, where in the first step the macroscopic tumor lesions are removed through CRS and in the second step, heated chemotherapy is applied locally to ensure destruction of remnant microscopic cancer lesions. Drugs for HIPEC are selected based on their cytotoxic ability to kill tumor cells, usually in a combination with mild hyperthermia (41–43°C) for 30–90 min to increase the cytotoxic effect13. In patients with CRC PM, a variety of protocols evolved historically and include drugs such as Mitomycin C, Doxorubicin or Oxaliplatin14,15,16. Since some of these drugs can induce immunogenic effects17,18,19, we assumed, long-term survival after CRS/HIPEC for CRC PM may result from induction of anti-tumor immunity. In the present study, we first analyzed the accumulation of CD8 + T cells in human primary CRC samples and PM lesions and their impact on disease free (DFS) and overall survival (OS). In a next step, we turned our focus on in-vitro and in-vivo assays to investigate the influence of heated chemotherapy (mimicking HIPEC) on CRC cell lines and patient-derived tumor organoids. Using in-vitro assays, we discerned that heated chemotherapy induced immunogenic changes on cancer cells that activated dendritic cells (DCs) and subsequently primed CD8 + T cells. Using a PM mouse model, we finally assessed the accumulation of functional CD8 + T cells within PM lesions after HIPEC and could show that CD8 + T cells are essential to control PM lesions.

Materials and methods

Patient samples

This study includes patients with visible peritoneal metastasis (PCI > 0) from colorectal origin. All patients underwent the CRS/HIPEC procedure at the University Hospital of Zurich. Patients with synchronous metastatic disease were operated for the primary tumor and the PM lesion at the same time at the University Hospital Zurich. Patients with a metachronous disease were operated before the CRS/HIPEC procedure at the University Hospital Zurich or at another Hospital in Switzerland. Patients with a MSI or BRAF mutated primary tumor were not included. All patients gave an informed consent for the further analysis of their samples. The study was approved by the ethical committee Zurich, Stampfenbachstrasse 121, 8090 Zurich (cantonal ethics number: 2019 − 01031). From 19 patient, paired samples were selected from the primary tumor and from PM lesion and from additional 18 patients only PM samples were included. The most important criteria was the size of vital microscopic tumor area on an H&E stain. These samples were than analyzed using digital pathology.

Tissue samples or ascites were collected during CRS to generate patient-derived tumor organoids (cantonal ethics number 2019 − 01031. Organoids were prepared at the laboratory of Prof. Chantal Pauli at the Department of Pathology and Molecular Pathology (University Hospital Zurich, Switzerland). The organoids were expanded by splitting every 3–4 weeks. Organoids were cultured in Matrigel in suspension plates (6-well TC plates from Sarstedt, Nümbrecht, Germany) with WRN media (provided by Chantal Pauli`s Laboratory, exact details are listed in Supplementary Table 1). The cell-cell connection and cell-Matrigel connection was detached with Triple-LE (Gibco, Life Technologies, Zug, Switzerland). After a few washing steps, the cells were dissolved in a WRN/Matrigel ratio of 1/1 and distributed on a new dish. The Matrigel was ordered at Corning (Lot number: 9238003). The patient-derived tumor organoids were included in experiments after 3–4 passage cycles.

Cancer cell-lines

Human CRC cells HCT-8 and HT-29, a gift from Prof. M. Scharl’s Laboratory (University Hospital Zurich, Switzerland), were used for in-vitro studies. HCT-8 cells were cultured in RPMI 1640 medium and HT-29 cells in Dulbecco`s modified Eagles medium (DMEM) (both from Gibco, Life Technologies, Zug, Switzerland), respectively. The medium was supplemented with 10% fetal bovine serum (FBS; Gibco, Life Technologies, Zug, Switzerland) and penicillin/streptomycin (100U/ml). In-vivo studies were performed with syngeneic mycoplasma negative (tested with PCR Mycoplasma Test Kit, PromoCell, Heidelberg, Germany) murine colorectal cancer MC-38-OVA cells obtained from Prof. M. van den Broek (University of Zurich, Switzerland). MC-38-OVA cells were also cultured in complete DMEM media.

Mice

C57BL/6 mice (8–10 weeks old) were purchased from Envigo (Horst, Netherlands). All mouse experiments and treatments were performed in accordance with the Swiss Federal Animal Regulations and approved by the Veterinary Office of Zurich (no. 165/2017 and 022/2021). OT-I transgenic mice were purchased from Jackson laboratories.

Heated chemotherapy treatment in vitro

Human cancer cells (0.5 × 106) were seeded into 6-well culture plates (TPP, Sigma-Aldrich, Schaffhausen, Switzerland) containing 1 ml of the corresponding media. After 24 h cells were treated either with control (a carrier solution used for the chemotherapy) or with the chemotherapy at 37 °C or at 43 °C. Chemotherapies - either Oxaliplatin 300 mg/l or the combination of MitomycinC/Doxorubicin 10 mg/l - were used for respective experiments. After 30 min of the treatment, the medium was removed, the cells were once washed with phosphate-buffered saline (PBS; Gibco, Life Technologies, Zug, Switzerland) then fresh corresponding medium was added to the wells. The cells were then incubated for additional 72 h.

For RT-PCR, cells were lysed with TRIzol (15596026, Thermo Fisher Scientific) and consequently lysate was stored at -80 °C until the RNA extraction was performed. For western blotting, cells were lysed by adding 400ul of 5 ml RIPA buffer + 1 Roche Protease Inhibitor tablet + 50ul PMSF (200nM). For flow cytometry analysis, cells were detached with 0.05% Trypsin-EDTA and transferred as a cell-suspension to flow cytometry tubes for further processing.

Co-culture experiments

Peripheral blood from a healthy donor was collected in an EDTA-containing vial. Only HLA matching samples were included in these experiments. Peripheral blood mononuclear cells (PBMC) were isolated with a Ficoll gradient (Ficoll Histopaque-1077: Sigma-Aldrich, Schaffhausen Switzerland). Monocytes were isolated with the magnetic cell separation (MACS) technology as per manufacturer`s instructions. The purity > 95% of the monocyte fraction was determined with flow cytometry. 3 × 105 monocytes were added to each well of a 12- well culture plate. To generate monocyte-derived dendritic cells (Mo-DC`s), monocytes were cultured for 7 days with DC medium supplemented with Cytokine A (Dendritic Cell Generation Medium, PromoCell, Schaffhausen, Switzerland). Every second day, the medium was exchanged. HCT-8 cells were seeded in 6-well plates on day 6 after monocyte purification and treated as described above. Mo-DC`s were added to treated tumor cells (ratio Mo-DC`s/tumor cells 1:5). 24 h after co-culture, Mo-CD`s were collected for flow cytometry analysis. The subsequent effect of Mo-DC maturation was further assessed on CD8 + T-cells. This experimental set-up was identical. At day 8 after monocyte purification, CD8 + T-cells were purified from the same healthy volunteer with MACS technology (CD8 + MicroBeads: Miltenyi Biotec, Adliswil, Switzerland). 1 × 105 CD8 + T-cells were cultured in 96 round bottom tissue culture plates (TPP, Sigma-Aldrich, Schaffhausen Switzerland) and Mo-DC`s exposed to different treated HCT-8 tumor cells were added to the CD8 + T-cells. The positive control condition for cytokine induction was PMA and Ionomycin treated CD8 + T-cells. In last 3 h of the culture, Brefeldin A (BioLegend, United Kingdom) was added to block the vesicular transport to measure intracellular IFN-γ production by CD8 + T-cells.

For the murine co-culture experiments, splenocytes from wt (wildtype) C57BL/6 mice and from OT-I transgenic on C57BL/6 background were used to set-up the co-cultures. Spleens were harvested from the mice and meshed through a 70 μm filter (Corning cell strainer, Sigma-Aldrich, Schaffhausen Switzerland) to create a single cell suspension. Red blood cells were lysed with 1 ml RBC Lysis Buffer (RBC Lysis Buffer, BioLegend, United Kingdom). 1 day before the harvest, 1 × 105 murine tumor cells (MC-38, MC-38-Ova) were seeded into 24 well tissue culture plates and treated 24 h after in different conditions. Directly after the treatment of the tumor cells, 2.5 × 105 splenocytes suspended in DMEM supplemented with IL-2 100U/ml were added to the tumor cells. 6 h before the collection of the splenocytes, Brefeldin A was added to the cultures. The splenocytes were collected 48 h after co-culture set-up and processed for flow cytometry analysis.

For the co-culture in vitro killing essay, spleens were harvested from OT-I transgenic mice. CD8 + T-cells were purified using MACS technology (CD8 + MicroBeads: Miltenyi Biotec, Adliswil, Switzerland). MC-38-Ova cells were seeded into 96 well tissue culture plates and treated in different conditions 24 h before adding the purified CD8 + T-cells. Specific CD8 + T-cells were added in a1:1 ratio to the cancer cells. 6 h after co-culturing, cancer cell-viability was determined using CellTiter-Glo 2.0 (Cat. No. G9241 from Promega, Dübendorf, Switzerland).

RT-PCR

RNA was extracted from treated and untreated cancer cells using TRIzol Reagent (Invitrogen, Basel, Switzerland). RNA from tumor organoid was extracted with RNA columns (Qiagen, Hombrechtikon, Switzerland) 1 µg RNA was reverse-transcribed to cDNA (ThermoScript reverse transcription PCR system; Invitrogen, Basel, Switzerland). PCR amplification was performed with the ABI Prism Sequence Detector System using TaqMan gene expression assays. Results are illustrated as fold induction relative to the 18s ribosomal RNA transcription.

Western blotting

After protein isolation from different treated cancer cell suspensions, the protein concentration was measured using a DC Protein Assay Reagent Package (Bio-Rad, Hercules, CA, USA). Protein aliquots were separated by SDS-PAGE electrophoresis and blotted using a V3 Western Workflow system by BioRad (Hercules, CA, USA). PVDF membranes were blocked with TBST (containing 5% BSA) and incubated with the primary Cyclin A1 antibody (Abcam, clone ab53699) overnight at 4 °C. Protein expression was measured by densitometry and illustrated relative to α-Tubulin as a reference protein.

Flow cytometry

Cells were detached from culture plates and transferred to flow cytometry tubes. Cells washed with PBS. The single cell-suspension was stained with surface antibody cocktail for 30 min at 4 °C. After staining, the samples were washed with PBS and then fixed with 4% formaldehyde and stored at 4 °C. For intracellular cytokine staining (ICS), Brefeldin A was added 5 h prior to block vesicular transport. For ICS first cells were stained with surface antibodies, later cells were washed with PBS, fixed with 4% PFA for 5 min and then permeabilized with 1% Saponin-PBS solution for 5 min. Cells were subsequently stained with antibodies against cytokines for 2 h at 4 °C. The samples were analyzed on the same day either on the BD FACS Canto II or BD Fortessa (BD Biosciences, LSR II Fortessa 4 L). Data analysis was carried out in FlowJo (V10.7.1, BD, Ashland, OR, USA).

In vivo experiments

Mice were intraperitoneally injected with 0.5 × 106 MC-38-Ovalbumin+ (MC-38-Ova) murine colon carcinoma cells. The time-line of the experiment is shown in Supplementary Fig. 5a. Macroscopic peritoneal tumor formation occurred mostly by day 7 or 8. The anesthetized mice underwent a median laparotomy to assess PM lesions. PM-lesions bearing mice were randomly assigned to different treatment groups (heated M/D, M/D, heated PBS, PBS). The treatments were performed in an open abdomen coliseum technique. Temperature during treatment was constantly measured with a thermometer. The abdomen was rinsed with saline solution after the treatment. The abdomen was closed with a two layered continuous suture technique with ethilon 4.0 (Ethicon, Zug, Switzerland). Six days after the surgery, the mice were sacrificed and the tumor load was assessed with the peritoneal cancer index (PCI)19,20 by an independent researcher, who was blinded to the treatment group. Peritoneal tumors were harvested for flow cytometry analysis and histology.

Immune cells depletion

CD-4 + and CD8+-T-cell depletion in mice was achieved by the intraperitoneal injection of 100ug CD4 or CD8a depletion antibody (BioXCell, USA; clone GK 1.5 for CD4 + T-cells and clone YTS 169.4 was used for CD8 + T-cells) 1 day prior to tumor cells injection and 1 day prior to the treatment. The macrophages were depleted using anti-CSFR1 antibody (BioXCell, USA; clone: 5A1; 150 µg/mice) injected 1 day prior to tumor cells injection and every 3rd day until the end of the experiment.

Immunohistochemistry and multispectral imaging

Tissue samples were collected in 4% buffered formaldehyde and paraffin-embedded. Mouse tumor tissue blocks were sliced into 4 μm and Haematoxylin and eosin (H&E) at our department according to a standard protocol. Additionally, murine tumor samples were immunohistochemically stained for CD8a+ (ab 209775, Dilution 1:500 abcam, USA), Granzyme B (ab 2555598, Dilution 1:1000 abcam, USA) and Macrophages F4/80 (ab 100790, Dilution 1:100 abcam, USA) with the autostainer Link 48 from Dako. Tumor blocks were subsequently sectioned at 4 μm and stained at the Department of Pathology and Molecular Pathology, University Hospital Zurich, Switzerland. Haematoxylin and eosin (H&E) stains were performed according to standard protocol. Immunohistochemistry was performed with double-stain for CD8 (Dako/Agilent M710301, Dilution 1:40, pre-treatment with EDTA buffer (pH8.4), at 100 °C for 32 min, OptiView Kit Ventana) and with pan cytokeratin antibody (panCK, Dako/Agilent M351501, Dilution 1:100, processed with no further pre-treatment, UltraView Red Kit Ventana) using a Ventana Benchmark Ultra platform with Haematoxylin counterstaining. Primary tumor and the corresponding PM lesion were stained accordingly and scanned using a 3D Histech Pannoramic 250 Flash III Scanner (3DHISTECH Ltd., Budapest, Hungary) at 40x and a resolution of 0.24 μm/pixel. Selected patient samples were stained with anti CD68 (Dilution1:50, Dako/Agilent, Basel, Switzerland).

Paraffin embedded mouse tumor samples were used for the multiplex immunostaining. The staining was performed using the Opal 7-color Manual IHC Kit (Cat. No. NEL811001KT, Akoya Bioscience, USA) according to the manufacturer’s protocol. In brief, slides were pressure cooked for 17 min at high heat in AR Buffer pH6 (Cat. No. AR6001KT, Akoya Bioscience, USA) for antigen retrieval. Afterwards, to avoid unspecific antibody binding, slides were blocked with 3% H2O2 for 15 min at RT. After washing with TBS-T three times for two minutes, slides were incubated with the primary antibody CD8a (ab217344, Dilution 1:2000, abcam, USA), Granzyme B (ab255598, Dilution 1:2000, abcam, USA) overnight at 4 °C. Subsequently, an Opal HRP polymer (Cat. No. ARH1001EA, Akoya Bioscience, USA) was added for 10 min at RT. After washing, following Opal Fluorophores were used to bind the polymer (10 min at RT): for CD8a Opal 620 (FP1495001KT, Akoya Bioscience, USA), for Granzyme B Opal 480 (FP1500001KT, Akoya Bioscience, USA). Finally, for nuclear staining, slides were incubated with DAPI (two drops in 1 ml TBS-T, Cat. No.FP1490, Akoya Bioscience, USA) for 10 min at RT, washed again and mounted with Vectashield Vibrance Antifade medium (Cat. No. H-1800-10, Vector Laboratories, Zürich, Switzerland). Slides were scanned using the automated imaging system Vectra Polaris (Akoya Bioscience, USA). Afterwards images were processed using the software QuPath (V0.3.2).

Digital pathology

CD8-panCK double stained slides were scanned and the tumor area was annotated. Artificial intelligence (AI)-based histomorphological tissue and CD8 + T-cell classification was performed using deep neural net algorithm (DNN) to quantify tissue area and to count CD8 + T-cells within the corresponding area in HALO (Indica Labs, Albuquerque, NM, USA). DNN classification was used to.

segmental annotated tumor areas into the following compartments and to quantify the tissue area in mm221: Background (white space and tissue folds, excluded from subsequent analysis), Necrosis, Epithelium (intraepithelial area), Stroma. Cell nuclei in each compartment were segmented, and CD8 + T-cells were identified based on Ultra View Red Chromogen signal. Density of CD8 + T-cells cells in the stromal and intraepithelial compartment was calculated as cells / mm2 of tissue and analyzed with clinicopathological variables and outcome.

Statistics

The CD8 + T-cell counts were normalized to their respective area as shown in the figure legends. Due to limited availability of paired samples, non-parametric Wilcoxon-test was used for analysis with categorical values. To define high versus low CD8 + T-cells count, a linear regression of the normalized CD8 + T-cell count to the intraepithelial area and the overall survival was performed. Due to its strong correlation, the median of the normalized CD8 + T-cell count was used to define the groups. A CD8 + T-cell count > the median was defined as the high CD8 + T-cell group and ≤ the median as the low CD8 + T-cell group. Based on these two groups, disease-free survival (DFS) and overall survival (OS) were compared and the log-rank test was performed to determine significance between the groups. The stromal content was calculated by the division of stromal area to the annotated area. The cut-off calculation of poor versus rich stroma was performed by a ROC-curve analysis including the CD8 + T-cell high and low group. The cut-off value of 0.67 with the highest likelihood ratio was taken and applied for the further statistical analysis. Disease-free and overall survival of poor versus rich stroma were also compared using log-rank statistical analysis. The data of normalized CD8 + T-cell counts and stromal content were used to distinguish between stromal rich and CD8 + T-cell high or low and stromal loose and CD8 + T-cell high or low groups. The disease-free and overall survival data of four groups were compared and a log-rank test was performed.

Descriptive statistics were illustrated as mean +-SD and the individual values as dots. Groups were compared by parametric unpaired Students t test or one-way analysis of variance (ANOVA) with multiple comparisons. Non-parametric numeric data was analyzed using the Wilcoxon test and categorical data was analyzed applying the Fisher`s exact test. GraphPad prism (version 9.3.1) and SPSS (IBM, version 29) were used to calculate statistical differences. P values were **** = p ≤ 0.0001, *** = p ≤ 0.001, ** = p ≤ 0.01, * = p ≤ 0.05, ns = p > 0.05 indicates no significant difference.

Results

Patient characteristics

Due to limited availability of paired samples, we included 19 samples from patients with PM originating from CRC (Table 1) and 18 samples from PM lesions only for the analysis of CD8 + quantities and compartment distribution. 37 PM lesions were collected during CRS before the HIPEC treatment. In case of synchronous disease, tumor tissue from the primary tumor was sampled during CRS. If the PM occurred metachronous, the primary tumor was resected before CRS/HIPEC. The majority of patients had a T4 stage colorectal cancer with nodal metastasis (Table 2).

Table 1 Patient characteristics of the paired sample cohort. NOS: not otherwise specified. The data is shown as median and interquartile range or numbers with percentage of the paired cohort (n = 19 patients).
Full size table
Table 2 Patient characteristics of the whole cohort. NOS: not otherwise specified, CC-Score: completeness of Cytoreduction score (0 stands for completed tumor resection). The data is shown as median and interquartile range or numbers with percentage of the whole cohort (n = 37 patients).
Full size table

All patients had a MSS type of the colon cancer. The extent of the disease had a median PCI of seven point five, matching to preferred criteria to qualify for the CRS/HIPEC treatment. The majority of patients was treated with Mitomycin C/Doxorubicin HIPEC. All patients were radically resected with complete cytoreduction-score (CC-Score) of zero. The two groups based on the CD8 + T-cell number normalized to intraepithelial area of the PM lesion into high versus low had no significant differences in PCI or driver mutations (K-Ras or N-Ras) (Table 3). Further, the adjuvant systemic chemotherapy regimen was similar between the groups in terms of number of cycles and drug combinations.

Table 3 Comparison of patient characteristics between CD8 + T-cell high and CD8 + T-cell low. The data is.
Full size table

CD8 + T-cells within the Pan CK + intraepithelial area of PM lesions are associated with prolonged patient survival

We noticed that the analyzed annotated tumor area was significantly larger for the primary tumor than for the PM lesions (p = 0.0018) (Supplementary Fig. 1a). However, intraepithelial and stromal area within annotated tumor area were similar between the primary and the PM lesion (Supplementary Fig. 1b and 1c). Interestingly, the number of CD8 + T-cells normalized to the corresponding area (within annotated tumor area, intraepithelial and stroma) were also similar between primary tumor and PM lesions (Supplementary Fig. 1d, 1e, and 1f).

We noticed that compared to intraepithelial area, the stromal area harbored significantly higher CD8 + T cells in both primary tumors and PM lesions (Fig. 1a and b). We then first classified primary tumors based on the presence of CD8 + T cells within the intraepithelial area allowing creation of CD8 + T-cell high and low groups (Supplementary Fig. 2a, dotted line shows the median). We noticed that intraepithelial CD8 + T-cells infiltration in primary tumors has no impact on DFS and OS between CD8 high and CD8 low groups (p = 0.18, respectively p = 0.59) (Supplementary Fig. 2b and 2c). However, high CD8 + T cells numbers in the stroma in primary tumors seems to be associated with significantly longer DFS than those with low stromal CD8 + T cell numbers, but the OS was similar in between CD8 high and CD8 low groups (Supplementary Fig. 2d – 2f).

Fig. 1
figure 1

Assessment of CD8 + T cells in patient samples. (a and b) analysis of CD8 + T cells in stroma and epithelium of the primary tumor of 19 patients (a) and the PM lesions of 37 patients. The graphs illustrate the number of CD8 + T-cells normalized to the corresponding area of stroma or epithelium. (c).The bar graph shows the distribution of intraepithelial CD8 + T-cells normalized by area in PM-lesions among the 37 patients. The dotted line indicates the median and divide the cohort into CD8 + T-cell high and low. (d) The scanned histological slide of PM lesions. The upper left picture presents an example with high CD8 + T-cell infiltration and the corresponding HALO classified picture below. The left upper example is CD8 + T-cell low. The classified pictures show the different areas of the tumor (green: stroma, violet: necrosis, red: intraepithelial, yellow: white space). DFS (e) and OS (f) based on intraepithelial CD8 + T-cell counts of PM lesions. 18 patients belong to the CD8 + T-cell high group and 19 patients to the CD8 + T-cell low group. DFS (g) and OS (h) based on the stroma content and CD8 + T-cell distribution. 15 patients with a CD8 + T-cell high PM lesion were associated with low stromal content (continuous line), 9 patients with a CD8 + T-cell low PM-lesion had a rich stroma (fine dotted line),10 patients with a CD8 + T-cell low PMlesion had a poor stroma (bold dotted line) and 3 patients with a CD8 + T-cell high PM-lesion were stroma rich (dotted line). Error bars represent the median and the lines the interquartile range. Each dot represents a patient. **** = p ≤ 0.0001, *** = p ≤ 0.001, ** = = p ≤ 0.01, * = p ≤ 0.05, ns = p > 0.05.

Full size image

Assessment of PM lesions based on CD8 high and CD8 low groups revealed that the number of stromal CD8 + T-cells in PM lesions is not associated with increase in DFS or OS (Supplementary Fig. 4a – 4c). Conversely, comparing CD8 high and CD8 low groups (Fig. 1c) in the intraepithelial area of PM lesions seems to associate with longer DFS (log rank p = 0.003) as well as OS (log rank p = 0.046) (Fig. 1e and f). This is supported by the zoomed pictures showing the different numbers of intraepithelial CD8 + T-cells and the HALO classified areas of the tumor (green: stromal area, violet: necrotic area, red: intraepithelial area, yellow: white space area) (Fig. 1d). These results suggest that the presence of intraepithelial CD8 + T cells in PM lesions are associated with the positive outcome of these patients. Assessment of macrophages in selected patients’ samples did not show any difference individually or when categorized in CD8 + high and CD8 low groups (Supplementary Fig. 4d – 4e). Since stromal density is also known to influence outcome for patients with colorectal cancer22 we added this additional prognostic factor in our assessment.

The cut-off value for the stroma poor and stroma rich group was determined according to the ROC curve illustrated in Supplementary Fig. 3a and 3b. Thus, dual assessment of stromal density with CD8 + T-cells allowed the creation of four groups namely1 CD8 + T-cell high/stroma poor group (n = 15)2, CD8 + T-cell low/stroma poor group (n = 10)3, CD8 + T-cell low/stroma rich group (n = 9) and4 CD8 + T-cell high/stroma rich group (n = 4) (Supplementary Fig. 3c). The comparison of these groups showed significantly longer DFS for CD8 + T-cell high/stroma rich and CD8 + T-cell high/stroma poor groups (Fig. 1g and h). The median DFS for groups 1–4 was 21, 9, 7 and 24 months, respectively, whereas the OS for groups 1–4 was 47, 38, 32 and 48 months and didn`t differ significantly.

Heated chemotherapy prevented growth of PM lesions in a CD8 + T-cell dependent manner in a PM mouse model

Our patient data revealed the importance of CD8 + T-cells in human peritoneal tumor tissues and their influence on patient survival. As it is not possible to study the impact of HIPEC treatment on CD8 + T cell-mediated immunity in patients, we decided to discern this aspect using a PM mouse model. We intraperitoneally injected MC-38-Ova (murine colon cancer cells) cells in C57BL/6 mice to establish microscopic PM lesions. Eight days after the cell injection, mice were treated in four different conditions PBS, heated PBS, M/D, heated M/D at day 8 (Supplementary Fig. 5a).

Compared to the other treatment groups, the peritoneal tumor load, measured using murine PCI (peritoneal cancer index), was significantly reduced after heated M/D treatment to a mean PCI of 6.90, (Fig. 2a and Supplementary Fig. 5b). The mean PCI after PBS, heated PBS or M/D treatment was 20.33, 21.86 and 13.00. Mice that received heated M/D treatment showed significantly more CD8 + T-cells and Granzyme B + cells in PM lesions (Fig. 2b and d), while other immune cells such as macrophages did not change (Fig. 2b and Supplementary Fig. 5c). The phenotype of Granzyme B + cells depended on the treatment and we observed a trend towards CD8 + positivity after heated M/D therapy (Supplementary Fig. 5g and h). Furthermore, depletion of CD8 + T-cells (Supplementary Fig. 5f) before heated M/D treatment abrogated its anti-tumor effect (Fig. 2e and Supplementary Fig. 5d and 5e). This data suggests that CD8 + T cells are crucial to control growth of PM lesions after heated M/D treatment. Moreover, the depletion of CD4 + T cells and macrophages did not therapeutic effects of heated M/D, suggesting these cells are not important for therapeutic effects.

Fig. 2
figure 2

Impact of HIPEC treatment in PM mouse model. (a) measurement of peritoneal tumor load as PCI. mice were treated with PBS ( n = 9) or heated PBS ( n = 7) or with M/D (n = 10) or with heated M/D (n = 11). (b) Staining of tumor tissues for the presence of CD8 + T cells, Granzyme B + cells and macrophages. (c and d) Quantification of CD8 + T cells and GZMB + cells. (e) PCI of the treated mice with and without CD8 + T-cells. Each dot represents one mouse. Error bars show the mean +/-SD. **** = p ≤ 0.0001, *** = p ≤ 0.001, ** = = p ≤ 0.01,* = p ≤ 0.05, ns = p > 0.05.

Full size image

Heated chemotherapy treatment induces immunogenic changes in human cancer cells and patient-derived tumor organoids

While we could discern the role of CD8 + T cells in controlling PM lesions, the direct impact of heated chemotherapies on tumor cells leading to induction of immunogenic changes was not clear. Therefore, to understand effects of HIPEC treatment on cancer cells, we carried out in-vitro experiments, where human cancer cells (HCT-8, micro satellite instable cells) were exposed to short-term treatment with heated chemotherapy mimicking HIPEC in patients. We treated colorectal cancer cell-lines with PBS, heated PBS, chemotherapy (Oxaliplatin 300 mg/l or M/D 15 mg/l) and short-term (30 min) heated chemotherapy, respectively. After treatment, cells were washed to remove dead cells due to direct cytotoxicity by chemotherapy (not shown). Cells were examined for immunogenic changes after 48–72 h. We noticed that surviving cancer cells showed enhanced expression of MHC-class I molecules (Fig. 3a Supplementary Fig. 6a). Which is likely induced by a higher amount of nucleotide-binding domain and leucine-rich repeat (NLR) protein C5 (NLRC5) (Supplementary Fig. 6k)23,24. In addition, a panel of nine Cancer Testis Antigens CTAs was screened with RT-PCR (Fig. 3b and Supplementary Fig. 6b). Heated chemotherapy with Oxaliplatin or M/D enhanced the expression of the CTA Cyclin A1 and SSX-4 at mRNA and Cyclin A1 at protein levels (Fig. 3c − 3f and Supplementary Fig. 6c and 6d). We also noticed similar changes in Cyclin A1 and SSX-4 expression in another colorectal cancer cell line (HT-29, micro satellite stable cells Supplementary Fig. 6e).

Fig. 3
figure 3

The effects of hyperthermic chemotherapy on colorectal cancer cell-lines and on patient derived tumor organoids. (a) MHC-I expression on treated HCT-8 cells. (b)The heat-map shows the expression profile of nine different CTAs of untreated and treated HCT-8 cell-line samples. (c and d). The fold induction of Cyclin A1 and SSX-4 expression. (e and f) Western Blot for Cyclin A1 and the quantification of the protein expression. The original blots are presented in Supplementary Fig. 8. (g) CTA expression of four different patient derived tumor organoids after Oxaliplatin treatment. The experiments were performed in triplicates. One representative experiment out of three is shown. Error bars show the mean +/-SD. **** = p ≤ 0.0001, *** = p ≤ 0.001, ** = = p ≤ 0.01, * = p ≤ 0.05, ns = p > 0.05.

Full size image

Furthermore, to understand the effects of heated chemotherapy directly on primary patient material, we utilized patient-derived tumor organoids (for patient information please see Supplementary Fig. 6i) and treated tumor organoids as we treated cells in the experiments above. Similar to cell lines data, tumor organoids from colorectal cancer patients (2 to 4) depicted higher expression of CyclinA1 upon treatment with heated chemotherapy while patient 1 with a gastric tumor did not show any change in CTA expression (Fig. 3g).

Heated chemotherapy enhances anti-cancer immunity

To confirm data obtained from our PM mouse model and to understand functional consequences of immunogenic changes on human cancer cells and patient-derived tumor organoids via heated chemotherapies, we created a multistep in-vitro setup shown in Fig. 4a, c and e. This co-culture setting allows to study how immune cells can recognize treatment-induced immunogenic changes on cancer cells.

Fig. 4
figure 4

Induction of antigen-specific CD8 + T cells via heated chemotherapy. (a)Time-line of the experiment. (b) flow cytometry data on the maturation state of Mo-DC`s depending on the cancer-cell treatment. (c)Time-line of the experiment with additional co-culture of CD8 + T-cells. (d) Shows flow cytometry data of CD8 + T-cells and their IFN-γ production depending on the cancer-cell treatment. (e) Shows the time-line of a similar experiment using splenocytes from OT-I mice, which have a specific TCR for the ovalbumin. (f) Presents the ratio of IFN-γ positive CD8 + T-cells after co-culturing with PBS, or M/D or heated M/D treated MC-38-Ova cancer cells with splenocytes from a OT-I mouse. Cancer cell killing by specific CD8 + T-cells. (g) time-line of the co-culture experiment. The % of cancer cell killing by the treatment with or without specific CD8 + T-cells is shown in (h).The experiments were performed in triplicates. Error bars show the mean +/-SD. **** = p ≤ 0.0001, *** = p ≤ 0.001, ** == p ≤ 0.01, * = p ≤ 0.05, ns = p > 0.05.

Full size image

After co-culturing different treated cancer cells with MACS purified CD14 + monocytes, we assessed activation/maturation of monocytic DCs (Mo-DC`s) by analyzing expression of CD83 and HLA-DR molecules using flow cytometry (see gating strategy in Supplementary Fig. 7a). As shown in Fig. 4b, we noticed marked expression of HLA-DR and CD83 on Mo-DCs with heated chemotherapy when compared to chemotherapy without heat, while treatment with PBS± heat failed to enhance expression of these molecules. These results suggest that immunogenic changes following chemotherapy treatment are able to mature dendritic cells. To assess, whether matured DCs are able to activate purified autologous CD8 + T cells, we carefully collected Mo-DCs from cancer cells co-cultures and then added them to MACS-purified autologous CD8 + T cells (Fig. 4c). After 48 h, we collected cells from co-cultures for flow cytometry and stained them for CD8 and for intracellular IFNγ. Gated CD8 + T cells (please see gating strategy in Supplementary Fig. 7b) showed high levels of IFNγ when co-cultured with Mo-DCs that were primed with cancer cells treated with heated chemotherapy (Fig. 4d). Interestingly, chemotherapy without heat induced significant Mo-DCs maturation (Fig. 4b) but did not activate CD8 + T cells to produce IFNγ.

Furthermore, to elaborate whether CD8 + T cells responded in antigen-specific manner, we utilized an artificial antigen-specific in-vitro model as we lacked human T cells clones that would recognize an antigen (such as Cyclin A1) on treated cancer cells. In this setup, murine MC-38 OVA cancer cells were co-cultured with Ova-specific OT-1 CD8 + T cells. We used single cell suspension from whole spleen and did not purify CD8 + T cells, so other cells within spleen could act as antigen presenting cells. In flow cytometry assessments, we noticed that gated CD8 + T-cells from spleen of OT-1 mice responded best when treated with heated M/D (Fig. 4f), while wild-type CD8 + T cells completely failed to respond to MC-38-Ova cells in all treatment conditions (Supplementary Fig. 7c). This is expected as CD8 + T cells from WT spleens were not exposed to Ova antigen before. Overall, such an in-vitro setup allowed us to show that compared to unheated chemotherapies, heated chemotherapy is able to induce more potent tumor-specific immunity. To support further and to show CD8 + T cell mediated killing of CRC cells, we performed in-vitro experiment using MC-38-Ova cells and MACS purified CD8 + T-cells from OT-I mice. The timeline of this experiment is illustrated in the Fig. 4g. We determined the killing of the MC-38-Ova cells using the cell-titer glo assay. The addition of antigen-specific CD8 + T-cells showed enhanced cancer cell killing in chemotherapy groups with no significant difference between co-cultured M/D and co-cultured heated M/D, probably due to lack of inhibitory cells that stop T cells functions in-vivo often requiring combination treatment for longer T cells activation.

Increased intraepithelial CD8 + T-cells in the PM lesions after HIPEC treatment

PM samples of 2 patients with a limited recurrent disease on the abdominal wall could be double stained with CD8 + and panCK. Both patients had a recurrent disease 13 months or 17 months after the first CRS/HIPEC. Interestingly, both patients show a massive increase in intraepithelial CD8 + T-cells after the first procedure (Fig. 5a), which could be associated with the HIPEC treatment. While patient 1 had a long-term survival of 105 months is patient 2 still alive without any recurrent disease (Fig. 5b).

Fig. 5
figure 5

Intraepithelial CD8 + T-cell number after HIPEC in 2 patients. (a) Numbers of intraepithelial CD8 + T-cells before first HIPEC and after the first HIPEC treatment of 2 different patient and the DFS and OS of these two patients (b).

Full size image

Discussion

A subset of patients with CRC PM after CRS/HIPEC presents with long-term survival. The mechanistic role of HIPEC is largely unknown. Chemotherapeutics that are used in HIPEC procedure and also systemically can have immunogenic impact. The tumor microenvironment (TME) of PM lesions in the context of immune cells composition is partially explored. In case of CRC it was noted that the presence of intraepithelial CD8 + T cells is associated with prolonged survival of the patients25, while for metastatic CRC, at advanced stages the data of the role of CD8 + T cells within primary tumors and associated metastasis is not completely clear. In this regard, our data showed that presence of26,27,28. CD8 + T-cells into the epithelium of PM lesions was associated with prolonged disease free and the overall survival of patients with PM originating from colorectal cancer. Interestingly, compared to previous studies, in our cohort of advanced metastatic CRC the presence of CD8 + T cells in primary tumors were not of prognostic value. In addition, similar to other tumor studies the density of stroma can impact the infiltration of immune cells within the tumors. Our data is in line with as we noted higher intraepithelial CD8 + T-cell infiltration in tumors with low stromal content. Macrophages within assessed PM lesions did not significantly differ in between CD8 high and CD8 low groups suggested that role of unknown factor allowing accumulation of the immune cells within PM lesions. On one hand assessment of these selected cells is interesting to show a critical role of CD8 + T cells but is of limited value without assessment of many other immune cells such as DCs, NK cells, B cells, Tregs, NKT cells and their associated subsets including functional state of the immune cells. The complete immune cells assessment was not in the scope of this study as the examination of CD8 + T cells and macrophages was only to have an initial idea about these cells in PM lesions and to test immunological benefits post HIPEC.

Using heated (HIPEC-like conditions) chemotherapy, we noticed immunological changes (CTA upregulation and MHC-I expression) on both MSS and MSI cancer cells and CTA upregulation on patient derived tumor organoids. These results suggest that HIPEC like treatment is immunogenic independent of MSS and MSI status of the cells. While most of the CRC and PM lesions show similar distribution in their MSS and MSI status29, it seems that CRS/HIPEC is more beneficial for PM patients with MSI30. Since, molecular classification of MSS and MSI status of CRC and associated PM is important for selecting patients to treat with immunotherapy but not for chemotherapies, thus, we assessed whether HIPEC is able to enhance tumor-specific CD8 + T cells responses. For in-vivo studies, we utilized our previously established PM mouse model20,31. In this tumor model we used MSI cells MC38 in C57BL/6 mice, while MSS CT26 cells that we used in BALB/c mice could not be used due to low tolerability of HIPEC conditions via this mice strain. Nevertheless, compared to other controls, HIPEC treatment had an impact on growth of PM lesions, which was due to accumulation of significantly higher functional CD8 + T cells within PM lesions. These results suggest that HIPEC treatment can mount antigen-specific CD8 + T cells response against PM.

CD8 + T-cells are known to control tumor growth in the primary tumor22,32,33 particularly from MSI type colorectal cancer34,35. In addition data is available from hematogenous metastasis, where the CD8 + T-cell infiltration in liver oligometastasis from colorectal cancer was analyzed and a higher CD8+/CD3 + T-cell ratio correlates with a significant longer recurrence-free and overall survival27. In the peritoneal cavity however, the role of CD8 + T cells is poorly described. Our observations suggest the critical role of CD8 + T cells also in the peritoneum. However, the presence of CD8 + T cells within PM lesions only partially explains favorable outcomes and the functional state of CD8 + T cells is likely an additional decisive factor36,37. CD8 + T cell activation can be enhanced either via immunotherapies or by modulating the immunosuppressive tumor microenvironment17. In the context of this study, we did not thoroughly explore the role of other cell types. For example, one study compared immune cell infiltrations between non-paired primary tumors and peritoneal metastasis7. This study found more NK cells in PM lesions whereas the primary tumor contained primarily CD8 + T-cells. A profound analysis of the PM-microenvironment may help to understand how the specific components interact and possibly attenuate the immune reaction in the peritoneum.

Our study observed that HIPEC can induce CD8 + T cell mediated tumor control in the mouse model. However, the molecular mechanism of HIPEC induced T-cell immunity within PM lesions remains to be explored. Induction of immunogenic cell death or a boost of a pre-existing immune reaction by cytotoxic drugs are potential mechanisms19. A recent study has shown that Mitomycin C in combination with hyperthermia triggers an immune reaction via Hsp 90 in a subcutaneous tumor mouse model38. While in-depth mechanisms remain to be elucidated, our data indicates that HIPEC does not only act through drug mediated tumor cytotoxicity but is able to induce immunogenic changes. This better explains the impressive impact on survival, observed in selected patients, and highlights the need for research with a different perspective. So far, cytoreductive surgery is seen as a purely tumor ablative procedure and drugs for intraperitoneal treatment are selected based on their cytotoxic profile. Increasing cytotoxicity however, may not improve the effect as observed by several clinical studies39 but come at the price of increased postoperative complications9. In conclusion, our data highlights that the presence of CD8 + T cells within PM lesions correlates with prolonged survival of human patients. In addition, we show that heated chemotherapies induce immunogenic changes on cancer cells enhancing CD8 + T cells mediated immunity. Even though our PM mouse model shows, that induction of CD8 + T cells is essential to for the efficacy of HIPEC. We are not able to confirm it in a large number of patients, which is certainly a limitation of the findings, requiring systemic collection of patient samples before and after HIPEC.

Overall, we conclude that induction of CD8 + T-cell immunity may be associated with improved survival rates observed after multimodal treatment, including HIPEC. This study opens the door for further experimental and clinical research toward an immunomodulating role of locoregional intraperitoneal therapies.