Establishment of cancer organoid cultures from MSS and MSI colorectal cancer patients.
To study the mechanisms of resistance/sensitivity to immunotherapy, we examined a cohort of CRC patients with varying microsatellite stability status. We assessed mutational profiles, microsatellite stability status and tumor mutational burden (TMB) in 123 CRC specimens using oncological genomic profiling (TruSight Oncology 500, Illumina) at our institution in 2021–2022 (Supplementary Table 1). In accordance with current case studies, 10 out of the 123 examined patients presented a molecular profile of MSI tumors.
To identify potential markers of resistance of CRC to immunotherapy, we initiated PDOs cultures from 10 CRCs displaying different microsatellite stability status and available matched tissues and blood (MSS = 6; MSI = 4) (Fig. 1).
Surgically resected primary tumors were characterized for immune cells subsets by immunohistochemical analysis (Fig. 1A). As expected, MSI tumors have an overall higher number of infiltrating CD3+ T-cells, CD8+ T-cells and CD68+ cells compared to MSS tumors (Fig. 1B).
However, within MSS tumors, there exists diversity in the extent of immune infiltration, with certain tumors showing level of immune infiltration similar to MSI tumors, from here on named MSS type II tumors, whereas others showing classical MSS features, from here on named MSS type I tumors.
To pinpoint markers able to capture the immune infiltration diversity observed among CRC tumors, we analyzed three PDOs models representative of each identified group (namely PDO1 as MSI model, PDO2 as MSS type 1 model, and PDO3 as MSS type 2 model). These PDOs and the matched primary tumors were characterized through immunohistochemical staining for the following CRC carcinogenesis markers: Caudal Type Homeobox Transcription Factor 2 (CDX2), Cytokeratin 20 (CK20), Leucine Rich Repeat Containing G Protein-Coupled Receptor 5 (LGR5), and for proliferative marker Ki-67 (Fig. 2A).
To confirm that organoids retained the mutational profile of the original tumor, Whole-Exome Sequencing (WES) analysis was conducted on PDOs and their respective primary tissues (Fig. 2B). As expected, this analysis demonstrated that MSI CRC exhibited a higher TMB as opposed to the other models, in both organoid and tumor tissue specimens (Fig. 2B). Consistent with the previously described IHC analysis, transcriptomic and immunodeconvolution analyses demonstrated that MSS type II CRC tissue has a tumor infiltrating immune component very similar to that of the MSI tumor (Fig. 2C).
To further assess the immune distinctive features of tumors, we used the MCP-counter algorithm on transcriptomic data to calculate the cytotoxicity score, which represents the immune system's ability, primarily through T-cells, to recognize and eliminate tumor cells. Of note, the two selected primary MSS tumors, chosen for their different tumor infiltrating immune components, have very different cytotoxicity scores. Indeed, the MSS Type II tumor closely resembles the MSI tumor, exhibiting with both higher cytotoxicity scores and CD8+ T-cells infiltration compared to the MSS Type I tumor (Fig. 2D).
Altogether, these data demonstrate that the MSS tumors may present intragroup heterogeneous phenotypes and some of them could share phenotypical and molecular characteristics common to MSI tumors, suggesting a potential sensitivity to Immune Checkpoint Inhibitors (ICIs).
Ex-vivo CRC patient-derived immunity-organoid interaction platform highlighted putative cancer-specific markers of resistance to immunotherapy.
To investigate the markers of resistance to immunotherapy in a complex system that recapitulates the main tumor and immune cell components, we developed an immune-organoid interaction platform with patient-matched T-cells, MDSC conditional medium (MDSCCM), and PDOs in absence or in presence of anti-PD-1 (pembrolizumab).
We verified that PDOs were able to maintain accurate feature alignment with the original tissue through immunostaining and WES (as previously described in Fig. 2A and B), including the PD-L1 expression (Supplementary Figure S1).
Then, T-cells (CD8+ and CD4+) isolated from PBMCs, were expanded and characterized for activation markers (CD25, HLA-DR), for PD-1 expression (CD279) and for their ability to activate and proliferate (Supplementary Figure S2A and B).
Freshly isolated MDSCs were cultured in vitro and conditioned media (MDSCCM) were recovered for the subsequent steps. MDSCs from individual patient were characterized by RT-PCR for TGF-β and IL-10 expression (Supplementary Figure S2C), for myeloid markers ARG1, NOS2, IDO1 and CCL4 (Supplementary Figure S2D) and for the ability to suppress the activation of T-cells isolated from a healthy donor (Supplementary Figure S2E). Consistent with the previous discussed results, MDSCCM from MSS CRC patients showed different immunosuppressive ability, with the MDSCs isolated from CRC3 patient exhibiting an immunosuppressive ability similar to that of the MSI CRC patient and with a less extent compared to that of the CRC2 patient (Supplementary Figure S2E).
Finally, the organoids were engaged with T-cells, activated, educated, and stained in the presence or absence of the immunosuppressive stimuli secreted by MDSCs (MDSCCM) and the ICI pembrolizumab (Fig. 3).
To measure apoptosis induction, the organoids were previously labelled with CellEvent™ Caspase-3/7 Green ReadyProbes™ and analyzed by time-lapse live microscopy. As expected, pembrolizumab do not affect T-cell recognition in the (MSS type I) PDO2 platform. On the contrary, the drug enhanced T-cell recognition and cytotoxicity activity not only in the (MSI) PDO1 platform, but also in the (MSS type II) PDO3 one, confirming the previous cytotoxicity score data (Fig. 3A and B).
Despite MDSCs exhibit an immunosuppressive ability on T cells from healthy donors, the addition of MDSCCM did not yield any statistically significant impact on the interaction between autologous T-cells and tumors in all assessed scenarios, but only an inhibitory trend of apoptosis in MSI and MSS type 2 patients.
To further confirm these results, Granzyme B (GZMB) release was evaluated by Luminex XMAP technology (Fig. 3C) and correlated with the observed apoptosis induction. These findings not only confirm the increase in pembrolizumab-induced T-cell apoptosis, as expected in MSI and MSS type 2 patients compared to MSS type 1 patients, but also appear to validate the immunosuppressive ability of MDSCs, albeit less pronounced in this type of platform.
Transcriptome Analysis reveals cancer-specific markers to ICIs treatment.
We performed RNAseq analysis to identify the CST markers associated with CRC resistance to ICIs (pembrolizumab) treatment, we analyzed differentially expressed genes between responder and non-responder PDOs (Fig. 4A and B). This analysis highlighted up-regulated (n = 371) and down-regulated (n = 347) common genes in responders PDOs (regardless microsatellite stability status) compared to the non-responder model PDO2 (Fig. 4C).
Gene Set Enrichment Analysis (GSEA) on the common downregulated genes (n = 347) identified signature shared with other gastrointestinal tumors, such as cholangiocarcinoma, gastric cancer and colorectal adenoma (Fig. 4D and E).
Gene Set Enrichment Analysis (GSEA) conducted on the commonly downregulated genes (n = 347) revealed a specific signature shared with other gastrointestinal tumors, such as cholangiocarcinoma, gastric cancer and colorectal adenoma, with REG4 serving as the hub gene (Fig. 4D and E).
To establish the inverse correlation between the identified signature and immune infiltration, we focused on patients with elevated tumor purity from the TCGA database of colorectal cancer (CRC) patients. Our investigation unveiled a resistant oncogenic signature (comprising REG4, CTSE, MUC1, TFF2, LCN2) that exhibits an inverse relationship with cytotoxicity and the immune infiltration deconvolution score (T- and NK-cells) (see Fig. 4F).
Overall, these data suggested an association between the identified signature and immunotherapy resistance.
REG4 and Mucin protein expression associates with low levels of cytotoxic CD8 + T-cells.
To translate the identified signature applicable in a clinical context, we examined the relationship between the expression levels of the key hub genes (REG4, MUC1, and MUC5AC) and the immune infiltration, as well as T-cell cytotoxicity, in colorectal cancer tissues from which we derived PDOs (Fig. 4D).
Multiplex IF analysis confirmed a low to absent expression of hub genes in CRC1 and CRC3, as expected, while CRC2 exhibited high levels in accordance with transcriptomics analyses (Figure S3A). Accordingly, we found higher amount of cytotoxic CD8+ T-cells (CD8+/GZMB+) and lower immune-suppressive Treg (CD4+/FOXP3+) in both CRC1 and CRC3 models respect to the CRC2 one (Figure S3B).
These findings suggests the role of the identified hub genes as CST markers associated with impaired infiltration of cytotoxic T-cells.
Validation of the identified CST markers in a cohort of MSI CRC patients treated with immunotherapy.
To validate our data in a clinically relevant patient cohort currently undergoing immunotherapy, we examined CST markers association to immunotherapy resistance. We screened by multiplex IF the expression of REG4, MUC1 and MUC5AC in a cohort of MSI CRC patients (n = 9) who underwent immunotherapy-based treatment and achieved either a complete response (CR; n = 3), a partial response (PR; n = 3), or experienced disease progression (PD; n = 3) (detailed clinical information is provided in Supplementary Table S2)
Identified CST markers correlated with RECIST response rate in adenocarcinoma CRC patients (Fig. 5A) as well as in mucinous adenocarcinoma CRC (Supplementary Figure S4A, C).
CRC samples from this patient cohort was also analyzed for immune infiltration components (Fig. 5B and Supplementary Figure S4B, C). Specifically, the tumor tissue of PD and PR patients exhibited a higher amount of regulatory immunosuppressive T-cells (CD4+, FOXP3+) compared to tumors from CR patients. Conversely, activated T-cells (CD8+, GZMB+) were more abundant in the tumors of CR patients compared to those with PR and PD.
It is worth noting that MSI tumors of patients who achieved a complete pathological response showed low to absent expression of CST markers.
This suggested that CST markers could potentially serve as indicator for immunotherapy response in the clinically setting of MSI patients.