Colorectal Cancer‐Derived Small Extracellular Vesicles Promote Tumor Immune Evasion by Upregulating PD‐L1 Expression in Tumor‐Associated Macrophages

Abstract Tumor‐associated macrophages (TAMs) are one of the most abundant cell types in colorectal cancer (CRC) tumor microenvironment (TME). Recent studies observed complicated “cross‐talks” between cancer cells and macrophages in TME. However, the underlying mechanisms are still poorly elucidated. Here, PD‐L1 levels are very low in CRC cells but highly abundant in TAMs, and a specific PD‐L1+CD206+ macrophage subpopulation are identified, which is induced by tumor cells and associated with a poor prognosis. Mechanistic investigations reveal that CRC cells can secrete small extracellular vesicles (sEVs) taken up by macrophages that induce M2 like polarization and PD‐L1 expression, resulting in increased PD‐L1+CD206+ macrophage abundance and decreased T cell activity in CRC TME. sEV‐derived miR‐21‐5p and miR‐200a are identified as key signaling molecules mediating the regulatory effects of CRC on macrophages. Further studies reveal that CRC‐derived miR‐21‐5p and miR‐200a synergistically induces macrophage M2 like polarization and PD‐L1 expression by regulating the PTEN/AKT and SCOS1/STAT1 pathways, resulting in decreased CD8+ T cell activity and increased tumor growth. This study suggests that inhibiting the secretion of specific sEV‐miRNAs from CRC and targeting PD‐L1 in TAMs may serve as novel methods for CRC treatment as well as a sensitization method for anti‐PD‐L1 therapy in CRC.


RNA isolation and qRT-PCR
Total RNA was extracted from cells or sEVs using RNAiso Plus (TaKaRa) and then reverse transcribed into cDNA using the PrimeScript II 1st Strand Synthesis Kit (TaKaRa). qRT-PCR was performed on a ViiA7 real-time PCR system using UltraSYBR Mixture (CWBio). Relative gene expression levels were normalized to those of ACTB and calculated utilizing the 2 -ΔΔt method.

Identification of PD-L1+ CD206+ TAM subgroup
Firstly, for single cell sequencing data, according to the cell annotation information uploaded by Lee et al., the gene expression data of myeloid cells are extracted from the expression profile matrix. Subclustering was performed for the four major cell types defined in the initial clustering using the graph-based algorithm in Seurat. For each cluster subset, differentially expressed genes were selected with a mean expression between 0.0125 and 3 and a dispersion of more than 0.5 using the FindVariableGenes function. According to the differentially expressed genes among these different clusters of macrophages, corresponding molecular characteristic maps were drawn. Principal component analyses were performed using these genes; the number of principal components was differentially selected from the knee point of the scree plot for each cell type to accommodate different population complexities.
Resolutions from 0.2 to 1.6 were explored for better subcluster representation.
Secondly, for bulk RNA sequencing data from TCGA and GEO datasets, the CIBERSORT algorithm was applied to quantify the absolute amount of 22 infiltrating immune cell types in each sample. CIBERSORT employs deconvolution of bulk gene expression data and a sophisticated algorithm for in silico quantification based on a leukocyte gene signature matrix, termed LM22, which contains 547 genes that distinguish 22 human hematopoietic cell phenotypes. Among the 22 infiltrating immune cell types, there are three macrophages types (M0, M1, and M2).
CIBERSORT algorithm indicates the presence of M2 macrophage infiltration in tumor tissues, which will be used for further analyses. Based on the gene expression of PD-L1 and CD206 in these tumor samples, we calculated the ratio of them and grouped these tumors into corresponding high or low groups (PD-L1/CD206 ratio: SPP1/CD68 > 1, high group; PD-L1/CD206 < 1, low group).

Enzyme linked immunosorbent assay (ELISA)
ELISA reagents (duo-set kit) for human IL-2 were purchased from ExCell Bio (China), and this assay was performed according to the manufacturer's instructions.
Absorbance was measured with a wavelength correction (A450 nm) using a microplate reader.

Generation of miRNA "sponge"
For the miR-21-5p or miR-200a knockdown plasmid, a sequence encoding a string of specific antisense sequences to miR-21-5p or miR-200a (sponge) was designed as proposed by a previous study (Ma et al., 2011). The sponge construct contained three copies of an antisense sequence for each miRNA. Then, the sponge construct was cloned into the same vector and co-expressed with GFP under the same promoter control. Figure S1. IL-10 and IL-12 levels in TAMs from CRC tissues and freshly isolated human monocytes.

Supplementary Figures
A, B) IL-10 and IL-12 levels were detected in CRC TAM and PBM by ELISA. C, D) IL-10 and IL-12 levels were detected in the culture supernatants of macrophages cocultured with SW620 or NCM460 cells by ELISA. * P < 0.05, ** P < 0.01, *** P < 0.001.    A) Primary macrophages were isolated from fresh CRC tissues and subjected to the detection CD206 and PD-L1 by immunofluorescence. B) Double-stained flow cytometry analyses of CD206 and PD-L1 for the macrophages cocultured with CRC cells. Primary CRC TAMs were isolated to stain CD206 or PD-L1, respectively, to help to clearly define the corresponding "positive" or "negative" cells as well as to designate quadrant gates. Staining of cells with IgG