Inammatory Conditions Promote Resistance to Immune Checkpoint Inhibitors in High Microsatellite Instability Colorectal Cancer

Inammatory conditions are common complications in colorectal cancer (CRC) and play signicant roles in tumor progression and immunosuppression. However, the inuence of inammatory conditions on the tumor response to immune checkpoint inhibitors (ICIs) remains unclear. We included a high microsatellite instability (MSI-H) CRC patient whose primary tumor progressed and liver metastasis regressed after Pembrolizumab treatment. An organoid-T cell coculture model demonstrated an inhibited local immune response instead of systemic immunosuppression. Single-cell RNA sequencing suggested that neutrophils suppress the immune microenvironment, mostly through CTLA-4-associated pathways. A cohort of 73 patients with MSI-H CRC who received ICIs were enrolled, among whom inammatory conditions were correlated with a poor tumor response. We demonstrated that inammatory conditions in MSI-H CRCs correlate with resistance to ICIs through neutrophil-associated immunosuppression. Additional CTLA-4 blockade may improve the sensitivity to PD-1 blockade. Clinically, inammatory conditions could predict a poor response to ICIs in MSI-H CRCs. sets of quantitative data that did not deviate from the Gaussian distribution. The Wilcoxon test was used for the comparison of paired quantitative data that did not deviate from the Gaussian distribution. For Student’s t test, the mean value is shown, and the standard deviation (SD) is displayed by the error bar (mean±SD). For the Mann-Whitney U rank-sum test, the median value is shown, and the range is displayed by the error bar. The Wald chi-square test was used to compare the differences in categorical parameters.Distributions of progression-free survival (PFS) and OS were determined using Kaplan-Meier methods. Univariate and multivariable Cox proportional hazards models were used to predict the outcomes of inuential factors. The Pearson or Spearman rank correlation test was used to measure the relationship between two variables. All P values were two-sided, and those <0.05 were considered statistically signicant. ROC curves were constructed to quantify the diagnostic performance of the prognostic factors for ICI response by assessing the respective areas under the curve (AUCs) with the 95% condence intervals (CIs).


Introduction
High microsatellite instability (MSI-H), highly correlated with DNA mismatch repair de ciency (dMMR), plays a prominent role in the tumorigenesis of colorectal cancer (CRC) 1,2 . dMMR and MSI-H are associated with high mutation burden, high tumor neoantigen load and dense in ltration of immune cells 3,4 . It has been well accepted that a dMMR/MSI-H status bene ts CRC patients receiving immune checkpoint inhibitors (ICIs), especially PD-1 blockade 5,6 . However, recent studies have reported that the immune status differs among dMMR/MSI-H CRCs, and over 50% of patients still experience resistance to ICIs with an unclari ed mechanism 6 . It has been revealed that both local and systemic in ammation have important roles in tumorigenesis, disease progression, and patient prognosis in various cancers [7][8][9] . Recently, studies found that in ammation is also associated with immunosuppression, and elevated in ammatory cells in the tumor microenvironment are associated with resistance to ICIs [10][11][12] . In addition, the in ammatory response is associated with alterations in peripheral blood leukocytes that can be captured by a high neutrophil-tolymphocyte ratio (NLR), which is also associated with poor long-term survival across all ICIs in patients with various soild tumors 13-15 . In ammatory conditions caused by obstruction or perforation are common complications in CRCs, while their signi cance in the response to ICIs remains unclari ed in MSI-H CRCs. Here, we investigated the impact of in ammatory conditions on dMMR/MSI-H CRC patients receiving ICIs.
In ammatory conditions are associated with resistance to ICIs in dMMR/MSI-H CRCs A 35-year-old woman with metastatic MSI-H descending colon cancer, who had peritonitis due to perforation of primary tumor and received transverse colostomy, was enrolled (Patient 1). The primary tumor and liver metastatic lesions regressed after 3 courses of Pembrolizumab in the combination of chemotherapy. After 5 courses of treatment, the patient had a fever, with increased white blood cell count and C-reactive protein concentration. She was considered having immune-related adverse reactions, and received only chemotherapy at the sixth course. After 6 courses of treatment, the primary tumor was diagnosed progressing with elevated tumor markers, and recurrent fever remained. The patient was then considered having localized infection in the perforating site and received another 3 courses of Pembrolizumab treatment combined with anti-infective therapy ( Supplementary Fig. 1). The primary tumor continued to progress. Meanwhile, the metastatic lesions continually regressed after 9 courses of treatment (Fig. 1A). The patient nally received resection of primary tumor and 15 courses of postoperative Pembrolizumab. The liver metastases were evaluated as complete remission and remained no evidence of diseases on last follow-up in July 2021.
Patient 1 achieved mixed response in primary tumor (progressive disease, PD) and metastatic tumor (partial response, PR) after 5 courses of Pembrolizumab combined with chemotherapy and one course of chemotherapy. Localized infection around the primary tumor was noticed and considered to be one of the potential reasons leading to the resistance of primary tumor. We therefore hypothesized that local in ammatory conditions could be associated with resistance to ICIs. To demonstrate this, a cohort of 73 dMMR/MSI-H CRC patients who received ICIs was retrospectively included (Table 1). We found that in ammatory conditions during ICI treatment were correlated with a higher ratio of stable disease (SD) and PD (87.5% vs. 33.3%, P < 0.001) and worse progression-free survival (PFS) (P = 0.001) ( Fig. 1B and C). Multivariate Cox proportional hazards regression indicated that in ammatory conditions were independently associated with the risk of progression (Table 2).   (Fig. 1D). Patients 1 and 2 had PD and SD, while Patient 3 was diagnosed with complete response (CR) after ICIs. Tumor organoids were cocultured with paired TILs or T cells from peripheral blood mononuclear cells (PBMCs). We observed that the apoptotic proportion of organoid cells was higher in the PBMC group in Patient 1 and was higher in the TIL group in Patient 3, while the proportions in Patient 2 were comparable between those two groups. In addition, Patient 1 and Patient 3 had comparable apoptotic proportions in the PBMC coculture group, yet the apoptotic proportion in the TIL group was the lowest in Patient 1 (Fig. 1E). These phenomena indicate an inhibited local immune response to tumors instead of systemic immunosuppression in Patient 1.
An inhibitory role of neutrophils in the tumor immune status is revealed by single-cell RNA sequencing (scRNA-seq) To investigate the component of the microenvironment in the primary tumor of Patient 1, scRNA-seq was conducted. The 4489 quali ed cells were divided into epithelial cells and immune cells ( Fig. 2A and B). The CNV landscape analysis showed that most of the epithelial cells (Epithelial-1 to Epithelial-6, Epithelial-8) were malignant tumor cells, which were characterized by frequent copy number loss and deletion in several chromosomes (Supplementary Fig. 2 and Supplementary Table). Among immune cells, CD8 + T cells were mainly identi ed as exhausted cells expressing PDCD1 and KLRD1 separately (CD8 TEX PDCD1 and CD8 TEX KLRD1). In addition, CD4 + T cells were mainly regulatory T cells (Treg cells) expressing FOXP3 (with or without CTLA-4 expression). In addition, 132 myeloid cells were identi ed, which were characterized by high expression of PTPRC, C1QB, AIF, and LYZ (Supplementary Fig. 3 and Supplementary Table). Furthermore, myeloid cells expressed FCGR2A, FCGR3A, FCGR3B, CD44 and CD55, which indicated that they are neutrophil-like cells (Fig. 2C) [16][17][18] .
To further investigate the interaction between immune cells, a cell-cell network analysis was conducted, suggesting that myeloid cells had a core role in the immune microenvironment, as they harbored the most connections with other cell types, especially with exhausted CD8 + T cells (Fig. 2D). To further investigate the interactions that occur in the ecosystem, a signi cant L-R pair was conducted to calculate the intensity of the interactions, which suggested crosstalk between myeloid cells and T cells via CTLA-4-CD80/CD86 (Fig. 2E). Other L-R pairs involving chemokines and cytokines between myeloid cells and T cells were also identi ed, including CCL5-CCR1, CCL5-CCR5, CCL3-CCR5, TNF-TNFRSF1B, IFNG-type II IFNR, and CSF1R-CSF1 ( Supplementary Fig. 4).
To investigate the role of the CTLA-4-CD80/CD86 axis in the exhaustion of T cells under in ammatory conditions, PD-1 neutralized murine T cells were stimulated in the presence or absence of neutrophils before coculture with CT26 cells. The apoptosis assays suggested that additional treatment with CTLA-4 neutralizing antibody for T cells augmented the apoptotic proportion of CT26 cells. More importantly, coculturing with neutrophils attenuated the cytotoxicity of T cells, while this effect could be rescued by CTLA-4 neutralization (Fig. 2F).
Elevated neutrophil in ltration is associated with a poor immune status and poor response to ICIs.
To investigate the in uence of neutrophils on MSI CRCs, The Cancer Genome Atlas (TCGA) data were analyzed using ImmunCellAI. We found that poor response to ICIs was associated with elevated neutrophil in ltration (Fig. 3A). Moreover, neutrophil in ltration negatively correlated with total in ltration scores (Fig. 3B) and the in ltration of cytotoxic T cells, Th1 cells, NK cells and B cells (Fig. 3C). To validate our ndings, data were further analyzed using The Cancer Immunome Atlas (TCIA) and Tumor Immune Estimation Resource (TIMER). We still found that elevated neutrophils were associated with lower immunophenoscores (IPS) and poor overall survival (OS) in MSI CRCs.
An elevated NLR is associated with a poor immune status and resistance to ICIs in dMMR/MSI-H CRC According to a previous study, the NLR shows a positive correlation with in ltrating neutrophils in tumors 19 . In a cohort of 142 surgically resected dMMR/MSI-H CRCs, we demonstrated that the CD8 + TILs in the IM and CN were signi cantly lower in patients with an NLR > 3 (Fig. 4A). To demonstrate that the NLR could predict the tumor response to ICIs, clinical data from the 73 samples were analyzed. We found that patients with in ammatory conditions had higher NLRs (Fig. 4B) and that patients with an NLR > 3 had a higher SD + PD ratio (57.6% vs. 35.0%, P = 0.025) and worse PFS (P = 0.022) after receiving ICIs ( Fig. 4C and D). To compare the e ciency between the NLR and in ammatory disease in predicting tumor response to ICIs, ROC curves were used. The AUC of having an in ammatory condition was 0.723 (P = 0.005, 95% CI, 0.572 ~ 0.874), while the AUC of an NLR > 3 was 0.642 (P = 0.071, 95% CI, 0.495 ~ 0.790). When two predictors were combined, the AUC was 0.758 (P = 0.001, 95% CI, 0.618 ~ 0.897) ( Supplementary Fig. 5).

Discussion
ICIs are very effective treatments for patients diagnosed with MSI-H CRCs, but 30-50% of patients present primary or secondary resistance to the treatment 5,20 . In the present study, we demonstrated that MSI-H CRC patients with in ammatory conditions have higher risks of resistance to ICIs through neutrophil-associated T cell exhaustion. In addition, both in ammatory conditions and a high NLR predict a poor response to ICIs, and the prognostic value was further increased when these two predictors were combined.
Prior to the current study, the in uence of neutrophils on immune status and ICI response has been revealed 21 . Tumor-associated neutrophils are generally considered to promote tumorigenesis among multiple tumor types, and there are also studies regarding neutrophils establishing a pre-metastatic niche for tumor cells [22][23][24][25] . In pancreatic cancer, it has been found that IL17-induced neutrophil extracellular traps could mediate resistance to ICIs 10 , and neutrophils in tumors effectively suppress normal T-cell immunity in gastric cancer 26 . In the current study, we have signi cantly expanded upon these previous observations. The current study demonstrates that both local in ammatory conditions and the NLR are associated with a poor immune status and a poor tumor response to ICIs, and the cell-cell network analysis in Patient 1 indicated that neutrophils play a prominent role in immunosuppression in MSI-H CRCs.
Although neutrophils can be activated by in ammatory conditions, our ndings indicate that neutrophils in tumors are functionally distinct from their peripheral counterparts, since different cytotoxicities between T cells in peripheral blood and TILs were observed. Apart from our study, Wang et al also revealed that tumor-in ltrating neutrophils exhibited an activated phenotype compared with normal activated peripheral cohorts 26 . In addition, multiple cell types in the tumor microenvironment could contribute to the pool of cytokines including G-CSF, GM-CSF, CXCR2 ligands and IL17, which activate neutrophils and educate other immune cells to be tumor-associated 26,27 . Therefore, tumor-in ltrating neutrophils are thought to contribute more to immunosuppression than peripheral neutrophils.
Studies have suggested that the induction of PD-L1 by in ammatory factors may be one of the most important factors affecting the therapeutic e ciency of PD-1 blockade 12,26 . However, the current study indicated that the CTLA-4-CD80/CD86 axis also participate in the interaction of neutrophils and T cells.
More signi cantly, an in vitro experiment demonstrated that CTLA-4 neutralization may reduce the inhibitory function of neutrophils on T cells. Depleting immunosuppressive tumor-associated myeloid cells is an attractive therapeutic approach to promote antitumor immune responses 28,29 . However, these inhibitors have provided minimal therapeutic bene ts in cancer patients as monotherapies, and whether these treatments could restore sensitivity to ICIs remains unclear. Because ICIs directed against PD-1 and CTLA-4 are highly effective in dMMR/MSI-H CRCs, and CTLA-4 blockade reduces immature myeloid cells in cancers 30 , blockade of the CTLA-4 axis may be appropriate for MSI-H CRC patients with in ammatory conditions.
CD80/CD86 appears to play a central role in in ammatory diseases 31,32 . As a response to hypoxia and speci c cytokines, myeloid cells express elevated levels of CTLA-4 ligands and other immune checkpoint inhibitors. CTLA-4 ligands such as B7 molecules are also highly expressed in dendritic cells 33 . In T cells, CTLA-4 and CD28 exist as homodimers capable of binding to CD80/CD86 via the same extracellular motif. Some researchers have claimed that upregulating CD80 in neutrophils could promote the activation of T cells by interacting with CD28 34 , while others have suggested that CTLA-4 has a substantially higher a nity and avidity, thereby outcompeting CD28 and simply preventing it from eliciting its stimulatory signals 30,35 . Together, these ndings suggest that the interaction of CTLA-4 and CD80/CD86 contributes most to T cell exhaustion under in ammatory conditions.
It has also been shown that elevated neutrophils in peripheral blood predict poor prognosis and resistance to ICIs in multiple cancers 36,37 . In the study of Fan et al., an NLR > 5 was associated with poor clinical response to anti-PD-1 therapy in patients with advanced gastric and colorectal cancers 38 . In metastatic MSI-H CRC, the pan-immune-in ammation value, calculated using routine blood test data, is a strong predictor of outcome in patients receiving ICIs 39 . In the current study, an NLR > 3 was also identi ed as a predictor of poor ICI response. The detection of neutrophil frequency, NLR, or neutrophilreleasing factors in patients' serum is easy, inexpensive, and applicable. Additionally, in cancer patients without in ammatory conditions, an elevated NLR is associated with worse immune status and poor ICI response 40 . Since the AUC of the in ammatory condition combined with the NLR was larger than that of single factors, we consider that in MSI-H CRCs, patients with in ammatory conditions could have a poor tumor response to ICIs. Among patients without in ammatory conditions, an NLR > 3 could be a promising predictor for a poor response to ICIs.
In conclusion, the current study demonstrates that in ammatory conditions in MSI-H CRCs correlate with resistance to ICIs through neutrophil-associated immunosuppression, and additional CTLA-4 blockade may potentially improve the sensitivity among those patients. Clinically, both in ammatory conditions and an NLR > 3 could predict a poor tumor response to ICIs in MSI-H CRCs.

Patient inclusion and follow-up
Patients with MSI-H CRC from Sun Yat-sen University Cancer Center (SYSUCC, Guangzhou, China) who received ICI treatment were enrolled. The exclusion criteria were as follows: 1) receiving only postoperative ICIs after radical surgery; 2) receiving less than 2 courses of ICIs. Finally, 73 patients were enrolled. In addition, dMMR/MSI-H CRC patients who received surgical treatment with su cient tumors for counting tumor-in ltrating lymphocytes (TILs) were enrolled as previously described 41 . Follow-up data, blood test, CT scanning, and determination of responses to ICIs were collected from the tracking system. Tumor responses were determined as complete response (CR), partial response (PR), stable disease (SD) and progressive disease (PD

Preparation and treatment of human T cells Paired peripheral blood mononuclear cell (PBMC)-derived T cells and TILs from 3 MSI-H CRC patients (Patient 1, Patient 2 and Patient 3) were isolated using a human pan-T cell isolation kit according to the manufacturer's instructions (Miltenyi Biotec). Human T cells were cultured in Human
ImmunoCult-XF T Cell Expansion medium (Stem Cell) with penicillin (100 U/ml) and streptomycin (100 µg/ml) at 37°C with 5% CO 2 . Cells were prestimulated with IL-2 (200 U/ml, Peprotech), anti-CD3 (Peprotech) and anti-CD28 (Peprotech) for 48 hours in the presence of paired CRC organoid cells.

Organoid-T cell coculture
To evaluate the cytotoxicity of T cells, organoids were dissociated into single cells and plated (5×10 4 per well) in a 24-well plate in the absence of Matrigel 24 hours before coculture. Pretreated T cells (1×10 6 ) were added to each plate. After 6 hours, tumor cells were obtained. Using an Annexin V Apoptosis Detection Kit (Dojindo), the cells were resuspended in phosphate buffered saline (PBS) for cell apoptosis analysis using a Beckman CytoFLEX FCM (Beckman Coulter). The proportion of apoptosis was calculated using FlowJo V10 (BD). Each experiment contained 3 replicates and was repeated three times.

Preparation and treatment of murine neutrophils and lymphocytes
Neutrophils from female BALB/c mice (8 weeks old) were derived from the bone marrow using the mouse Neutrophil Isolation Kit (Miltenyi Biotec), and were cultured in RPMI 1640 supplemented with 10% FBS, penicillin (100 U/ml), and streptomycin (100 µg/ml) at 37°C with 5% CO2. Neutrophils were stimulated with mouse recombinant GM-CSF (Peprotech) for 24 hours in the presence of CT26 cells.
T cells for CT26-T cell coculture were isolated from the spleens BALB/c mice with mouse pan-T cell isolation kits ( Single-cell RNA-sequencing (scRNA-seq) The scRNA-seq was conducted using the primary tumor of Patient 1. An scRNA-seq library was prepared using the DNBelab C4 system as previously described 42 . Brie y, the single-cell suspension was transformed into the scRNA-seq library of barcodes through the steps of droplet encapsulation, emulsi cation and fragmentation, mRNA capture bead collection, reverse transcription, cDNA ampli cation and puri cation. An indexed sequencing library was constructed according to the manufacturer's instructions. The sequencing library was quanti ed using the Qubit SSDNA Assay Kit (Thermo). DIPSEQ T1 was used for sequencing libraries at the National Gene Bank (CNGB, BGI-SHENZHEN, Shenzhen, China). The read structure was paired with Read 1 and Read 2. Read 1 contained 30 bases, including 10 base pair (bp) cell barcode 1, 10 bp cell barcode 2 and 10 bp unique molecular identi er (UMI), and Read 2 contained 100 transcriptional base sequences and a 10 bp sample index. The FASTQ raw data were converted to a Cell Ranger-speci c FASTQ le using an in-house Perl script. These FASTQ les were then processed separately using a modi ed version of the Cell Ranger count pipeline, which aligned cDNA reads with the GRCH38 human reference using STAR software (v2.5.3) 43 . The mapped reads were then ltered out for valid cell barcodes and UMIs to generate a gene-cell matrix for downstream analysis.

Unsupervised clustering and cell type annotation
Cell clustering was conducted by the Seurat (v3.1) 44 package in RStudio. Genes expressed in less than 3 cells were ltered out, and cells with fewer than 500 or more than 10,000 genes were excluded. The 3 libraries were then integrated using the "Merge" functions, and the batch effects were checked if the cells were separately distributed with the "DimPlot" function. Then, the integrated data are scaled to calculate the principal component analysis. The rst 30 PCs were used to construct the SNN network, and the graph-based clustering method Louvain algorithm was used to identify the cell clusters with a resolution of 0.6. Finally, UMAP is used to visualize the clustering results in two-dimensional space. To annotate each cluster as a speci c cell type, we selected some classic markers of immune cells and epithelial cells.
The cell types were annotated using a violin diagram. The CNV of each epithelial cell was estimated by the inferCNV package 45 , using immune cells as a normal control. The epithelial cell clusters with abnormal CNV patterns were annotated as malignant tumor cells.

Cell-cell interaction analysis
To analyze cell-to-cell interactions, we used CellPhoneDB 46  Immunohistochemistry (IHC) analysis and lymphocyte counting IHC staining was conducted as previously described 41 . All specimens were prepared as 4 μm FFPE

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.