Data were retrospectively collected from 228 patients with LARC who received intensity-modulated radiation therapy (IMRT) with concurrent capecitabine treatment followed by surgery at Peking University Cancer Hospital (Beijing, China) between December 2008 and June 2015. The IMRT regimen consisted of 22 fractions of 2.3 Gy (gross tumor volume) and 1.9 Gy (clinical target volume), which has been described previously (19). Surgery was recommended ≥8 weeks after the completion of radiation. Adjuvant chemotherapy was also routinely recommended to the patients. Each enrolled patient satisfied the following criteria: i) Cancerous lesion located within 10 cm from the anal verge; ii) cancer staged as T3-4 or any T and N+ by endorectal ultrasonography (7th AJCC cancer staging system) (20), pelvic magnetic resonance imaging (MRI) or computed tomography; iii) presence of distant metastases excluded by imaging examinations; iv) preoperative radiotherapy of 50.6 Gy/22 fractions; and v) radical surgery following TME. Patients with the following characteristics were excluded from the present study: i) Previous chemotherapy or pelvic radiation; ii) previous history (within 5 years) of malignant tumor; and iii) presence of unresectable cancer. All patients signed the consent forms before treatment. The patient information and samples were gathered from the hospital surgical database. The blocks of formalin-fixed paraffin-embedded (FFPE) samples from each patient in this surgical database were stored at The Department of Pathology at Peking University Cancer Hospital. The slices of FFPE samples were stored at −80°C for long term and −20°C for short term storage. The present retrospective study was approved by The Ethics Committee of Peking University Cancer Hospital (approval no. 2021KT93). Informed consent for inclusion in the present study was waived by The Ethics Committee due to the retrospective nature of the present study.

FFPE tissue sections (5 µM) were prepared and stained with hematoxylin (5 min) and eosin (2 min) at room temperature for histological evaluation. For IHC, the sections were deparaffinized in a xylene and an ethanol gradient at room temperature. Antigen retrieval was performed with citrate buffer pH 6.0 at 95°C for 10 min, followed by incubating with an endogenous peroxidase blocker (cat. no. ZLI-9310; ZSGB-Bio) for 10 min at room temperature. The sections were then washed with PBST (0.1% Tween-20) for 5 min three times and incubated with 5% goat serum (cat. no. ZLI-9021; ZSGB-Bio) for 1 h at room temperature. The sections were incubated with the following primary antibodies at 4°C overnight: chemokine (CXC motif) ligand 8 (CXCL8; 1:500; cat. no. 27095-1-AP; Proteintech Group, Inc.) and CMTM4 (1:500; cat. no. HPA014704; Sigma-Aldrich; Merck KGaA). Following washing with PBST (0.1% Tween-20) three times, the slices were incubated with the secondary antibody (undiluted; cat. no. SAP-9100; ZSGB-Bio) at room temperature for 40 min. Diaminobenzidine (Dako; Agilent Technologies, Inc) substrate was used to observe staining, and the sections were re-stained with hematoxylin for 5 min at room temperature (Beyotime Institute of Biotechnology). A bright field microscope (Leica Microsystems, Inc.) was used to capture images. The H score is a reliable method that can effectively quantify protein expression by two independent pathologists (21). The H score was calculated as follows: (Percentage of weak staining) + (Percentage of moderate staining) + (Percentage of strong staining) within the target region, with scores ranging from 0 to 300 (22). The of CMTM4 and CXCL8 had different expression pattern, therefore the cut-off value is different. H scores of 0–120 were defined as the CXCL8 low group and H scores of 121–300 as the CMTM4 high group. H scores of 0–150 were defined as the CMTM4 low group and H scores of 151–300 as the CXCL8 high group.

The human colon cancer LoVo cell line purchased from the American Type Culture Collection (cat. no. CCL-229) was cultured in a hot cell incubator (5% CO₂, 37°C) in RPMI-1640 (Invitrogen; Thermo Fisher Scientific, Inc.) supplemented with 10% bovine serum (cat. no. 164210; Procell Life Science & Technology Co., Ltd.) and 100 U/ml streptomycin. When the cells reached 80–90% confluency, pancreatin digestion was carried out and the cells were resuspended in fresh culture medium. The consumables used for cell culture were sterilized by high pressure and the cells were regularly examined for mycoplasma contamination by reverse transcription-quantitative PCR (RT-qPCR).

The lentivirus targeting human CMTM4 and an shRNA scramble sequence (negative control; NC) were generated and synthesized by GenePharma Co., Ltd. The short hairpin (sh)RNAs used in the present study were as follows: shCM4-3, GAAAUUGCUGCCGUGAUAUTT; shCM4-6, GCAUAUGCAGUGAACACAUTT; and sh NC, UUCUCCGAACGUGUCACGUTT. Briefly, the shRNAs targeting human CMTM4 were cloned into a 3rd generation lentiviral transfer plasmid, pLV-eGFP (vector backbone, pLenti-MP2), generating a lenti-shCMTM4 construct for expression knockdown. The virus was added at a volume ratio of 1:100 and at a titer of 1×10⁸ TU/ml into LoVo cells, while simultaneously adding polybrene at a volume ratio of 1:500 at 37°C and 5% CO₂. Fresh and complete culture medium was replaced within 24 h from infection, and the infection efficiency was detected using western blotting and immunofluorescence after 72 h.

Adenovirus carrying the human CMTM4 gene and the empty adenovirus were packaged by GenePharma Co., Ltd. LoVo cells were infected with the adenovirus at a multiplicity of infection of 100 for 72 h. All other steps were as described above.

The LoVo cells were infected with the indicated lentivirus for 72 h before irradiation. X-ray irradiation was performed using an X-ray generator (EDGE™ Radiosurgery system; Varian Medical Systems, Inc.) with gantry 0°, collimator 0°, field 30×30 cm, energy 6 Mv with extra-fine 2.5 mm MLC leaves. Indicated doses were shown in each experiment.

A total of 2×10⁶ cells with the indicated treatments were harvested and lysed using RIPA buffer containing 50 mM Tris-HCl pH 7.4, 150 mM NaCl, 50 mM NaF, 1 mM EDTA, 1 mM DTT, 1% TritonX-100, 0.1% sodium dodecyl sulphate (SDS) and 1X protease inhibitor cocktail. After a 20-min incubation on ice, the cell lysates were centrifuged at 13,800 × g for 20 min at 4°C, and the supernatants were recovered. Protein samples (20 µg) were resolved by 12% SDS-polyacrylamide gel electrophoresis and blotted onto polyvinylidene fluoride membrane (cat. no. 88520; Thermo Fisher Scientific, Inc.), which was blocked in 5% skim milk in TBST (Tris-buffered saline containing 0.1% Tween-20) at room temperature for 45 min. The membrane was then probed with rabbit anti-CMTM4 (1:500; cat. no. HPA014704; Sigma-Aldrich; Merck KGaA) and mouse anti-GAPDH (1:1,000; cat. no. TA-082519; OriGene Technologies, Inc.) antibodies overnight at 4°C. Following washing with TBST, the membrane was incubated with the HRP-anti-mouse (1:2,000; cat. no. ab6789; Abcam), HRP-anti-rabbit (1:2,000; cat. no. ab6721; Abcam) secondary antibody at room temperature for 45 min. The band strips were visualized using enhanced chemiluminescence detection system (cat. no. 34580; Thermo Fisher Scientific, Inc). and the images were captured using chemiluminescence imaging systems (Azure Biosystems, Inc.).

The cells were digested and counted, and 800 cells from each group were resuspended in fresh culture medium. After 10–14 days of cell culture, the cells were fixed at room temperature with precooled methanol for 30 min, the methanol was discarded, the cells were washed with PBS three times and then incubated with 1% crystal violet dye solution for 30 min at room temperature. The cells were then washed with PBS three times and dried in a fume hood. The bottom of the cell plate was scanned, and the number of clusters with >50 cells were counted using ImageJ software (National Institutes of Health; V48.1).

Migration (cat. no. 3422; Corning, Inc.) and invasion assays (cat. no. 354480; Corning, Inc.) were performed using 8-µm pore size plates with a filter insert. The invasion inserts precoated with diluted Matrigel, were preheated at 37°C for 1 h before the invasion experiments. Briefly, 2×10⁵ cells resuspended in 200 µl medium without serum were inoculated into the upper chamber of each well; 800 µl medium containing 10% FBS was then added to the lower chamber. The cells were allowed to migrate for 36–48 h at 37°C. The rotating pores were fixed in 100% methanol and dyed in 0.1% crystal violet solution at room temperature for 30 min. The cells in the upper chamber were then removed with absorbent cotton. The polycarbonate membrane was then removed and sealed on a glass slide with resin, and the cells penetrating the underside of the membrane were counted in four randomly selected visual fields.

The TIMER database (https://cistrome.shinyapps.io/timer) is a comprehensive online analysis software for investigating the gene expression and immune cell infiltration in different cancer types (23). The TIMER database contains 166 rectum adenocarcinoma (READ) tumor samples and 10 normal samples; 457 colon adenocarcinoma (COAD) tumor samples and 41 normal samples. TIMER was used to explore the correlation between CMTM4 and the levels of immune cell infiltration. The association of CMTM4 expression with gene marker of different immune cell including CD4⁺ T cells, CD8⁺ T cells, B cells, macrophages, neutrophils and monocytes were further investigated (24). P≤0.05 was considered to indicate a statistically significant difference.

To use high-throughput methods to analyze gene expression patterns under different experimental conditions, microarray analysis was performed. Total RNA was extracted from the indicated samples using Qiagen RNeasy kit (cat. no. 74104; Qiagen GmbH). The mRNA library construction and RNA-seq analysis were constructed and performed by Shenzhen BGI Co., Ltd. using the Illumina Genome Analyzer platform. The differentially expresses genes with false discovery rate<0.01, fold change >1.5 were determined. Principal component analysis was performed using the ‘stats’ package and plotted with the ‘ggplot2’ package in R (version 3.5) (25). Gene Set Enrichment Analysis (GSEA) was performed using the GSEA software (Broad Institute) as previously described (26). DAVID analysis was performed for transcription factor enrichment as previously described (27,28).

Total RNA was extracted from cell lines and FFPE tissues using TRIzol^® reagent (cat. no. 10057821; Thermo Fisher Scientific, Inc.) and RNeasy FFPE Kit (cat. no. 73504; Qiagen GmbH), respectively, according to the manufacturer's instructions. cDNA was synthesized using a GoScript™ reverse transcription system (cat. no. A5001; Promega Corporation) according to the manufacturer's instructions. Each assay was tested in duplicate. The expression of CMTM4 and CXCL8 were assessed by SYBR GREEN Mixture (ROX reference dye; cat. no. QPK-201; Toyobo Co., Ltd.). For the RT-qPCR experiments of cell lines and FFPE, GAPDH served as the internal control. Relative mRNA expression was calculated using 2^−ΔΔCq (29). The thermocycling conditions were as follows: Pre-denatured at 95°C for 5 min, 40 cycles at 95°C for 10 sec, 60°C for 20 sec and 72°C for 20 sec. The primers used are listed in Table SI.

Data was analyzed using GraphPad Prism 8.3 (Dotmatics) and SPSS 25 (IBM Corp.). Comparison of multiple groups was performed using one-way ANOVA and Tukey's post hoc test. Comparisons of CMTM4 or CXCL8 expression from pre-operative with post-operative tissues were performed using paired t-test. Comparison of the RT-qPCR results between the CMTM4 control and overexpression/shRNA groups was performed using unpaired t-test. The results are presented as the mean ± standard deviation from at least three independent experiments. The optimal cut-off value and Kaplan-Meier curves of CMTM4 and CXCL8 in the survival analysis were determined with the ‘survival’ (R version 3.7.0) and ‘survminer’ package (R version 4.3.2; http://www.R-project.org). Following cut-off value determination, patients were stratified into high- and low-expression groups for each biomarker. Disease-free survival (DFS) and overall survival (OS) were evaluated using Kaplan-Meier curves to illustrate differences in survival distributions, with group comparisons performed via the log-rank test. The univariable and multivariable Cox proportional hazards regression models were applied to assess the prognostic significance of CMTM4, adjusting for relevant clinical covariates where appropriate. For all statistical tests, including the log-rank comparisons and Cox regression analyses, a single-sided P<0.05 was considered to indicate a statistically significant difference.

In total, 228 consecutive patients with rectal cancer (152 men and 76 women) were included in the present study. The median age of the patients was 57 years (range, 25–80 years), the median follow-up time was 54.7 months and 73 (32.0%) patients experienced recurrence or metastasis. Overall, 21.8% of patients reached pathological complete response [pCR; no observed adenocarcinoma cells in the surgical resection specimen, pathological stage after nCRT (yp)T0N0M0 (30)] following nCRT. Downstaging (ypT0-2N0M0) occurred in 134 patients (58.8%). Additional post-operative characteristics and distribution of relevant parameters are listed in Table I. Patients with pN+, pT3 or nerve invasion were significantly associated with a poorer overall survival (OS; P=0.001, P=0.042 and P<0.001, respectively) and DFS (P<0.001, P=0.001 and P=0.008, respectively) (Table II).

CMTM4 staining was conducted on pre- and post-operative tissues from patients with LARC. CMTM4 was localized in the cell membrane and cytoplasm in patients with LARC and chemoradiotherapy did not change the CMTM4 localization (Fig. 1A). To confirm the IHC observations, five samples with high and three with low CMTM4 expression were selected for analysis, and DNA was extracted from these samples. The average relative level of CMTM4 in the IHC high group was 1.477 and in the IHC low group was 0.487 (P=0.015; Fig. 1B). The results of the RT-qPCR experiments indicated that the protein expression detected by IHC was consistent with the mRNA expression level. Therefore, the expression detected by IHC was reliable in the downstream analysis. CMTM4 exhibited lower expression in tumor tissues compared with adjacent normal tissues in the postoperative samples (P<0.0001; Fig. S1A). Lower CMTM4 expression in pre-operative tissues was significantly associated with improved DFS and OS (P=0.029 and P=0.048, respectively; Fig. 1C). CMTM4 expression in LARC tissues following TME surgery was not associated with OS or DFS (P=0.18 and P=0.168, respectively; Fig. S1B). The multivariate model was employed to evaluate the comprehensive prognostic value of features obtained from the univariate analysis. CMTM4 was an independent prognostic factor of DFS [hazard ratio (HR), 1.759; 95% confidence interval (CI), 1.037–2.984; P=0.036] in patients with LARC and nerve invasion and pN+ in post-nCRT was an independent prognostic factor of OS and DFS in patients with LARC (Table III). The changes in CMTM4 expression were compared between pre-nCRT and post-nCRT and radiation therapy significantly increased CMTM4 expression (P=0.031; Fig. 1D). Further analysis indicated that CMTM4 expression in the pCR group was lower than that in the non-pCR groups (P=0.041; Fig. 1E). In summary, CMTM4, a potential new biomarker for patients with LARC before nCRT, was negatively associated with chemoradiotherapy response and prognosis.

To further investigate the role of CMTM4 in chemoradiotherapy, a CMTM4 knockdown cell line was established using lentivirus shRNA. The knockdown efficiency was verified by immunofluorescence, western blotting and RT-qPCR (Fig. 2A-C). Compared with the NC cells, CMTM4 knockdown significantly increased cell proliferation, migration and invasion (Fig. 2D-F). Following radiation exposure (2 and 5 Gy), there was no significant difference between the NC and CMTM4 knockdown groups in terms of colony formation ability (Fig. 2D). However, compared with the NC group, interfering with CMTM4 knockdown significantly decreased the cell migration and invasion under IR treatment (Fig. 2E and F). These in vitro experiments were consistent with the findings in the clinical data; CMTM4 expression induced radiotherapy resistance in colon cancer cells.

RNA samples were extracted from LoVo-NC, LoVo-shCM4-3 and LoVo-shCM4-6 cells, and microarray analysis was performed to examine the effect of CMTM4 on the gene expression profiles (Fig. 3A). Compared with the LoVo-NC group, the LoVo-shCM4-3 group had 184 upregulated and 220 downregulated genes, while the LoVo-shCM4-6 group had 232 upregulated and 455 downregulated genes (Fig. 3B). The expression of 105 genes was mutually altered by CMTM4 knockdown, including 38 upregulated and 67 downregulated genes. CXCL8 was shown to be upregulated by CMTM4 knockdown in the RNA-seq analysis (Fig. 3C and Table SII). Gene ontology function analysis indicated that CMTM4 knockdown altered different pathways involved in ‘Metabolism’, ‘Genetic Information Processing’, ‘Environmental Information Processing’, ‘Cellular Processes’, ‘Organismal Systems and ‘Human Diseases’ (Fig. 3D). These 105 genes were shown to participate in the KEGG pathways associated with different physiological and pathological processes including ‘Metabolism’, ‘Genetic Information Processing’, ‘Cellular Processes’ and ‘Human Diseases’. The profiles of the representative top 19 KEGG enrichment pathways (P<0.05) are listed in Table SIII, such as ‘Aminoacyl-tRNA biosynthesis’ in translation, ‘FoxO signaling pathway’, ‘PI3K-Akt signaling pathway’ in signal transduction, ‘Mitophagy’ in transport and catabolism, ‘Apoptosis’ in cell growth and death, ‘NOD-like receptor signaling pathway’ in immune system.

nCRT typically triggers metabolism, inflammation and an immune system response (31,32). Therefore, the TIMER database was used to explore the association between CMTM4 and immune cell infiltration in READ. In the READ dataset, a significant although weak correlation was observed between CMTM4 expression and immune cell infiltration in B cells (ρ=0.231, P=6.13×10⁻⁰³), CD8⁺ T cells (ρ=0.214, P=1.13×10⁻⁰²) but not with purity (ρ=−0.105, P=2.16×10⁻⁰¹), CD4⁺ T cells (ρ=0.033, P=7.03×10⁻⁰¹), macrophages (ρ=0.071, P=4.09×10⁻⁰¹), neutrophils (ρ=0.064, P=4.57×10⁻⁰¹) and dendritic cells (ρ=0.056, P=5.15×10⁻⁰¹) (Fig. S2). The TIMER database was also used to further evaluate the association between CMTM4 expression and immune marker sets in READ and colon adenocarcinoma (COAD). The association between CMTM4 and immune cell markers of B cells, monocytes, tumor-associated macrophages (TAMs), M1 macrophages, M2 macrophages, neutrophils, natural killer cells, dendritic cells, general T cells and CD8⁺ T cells were examined. Potential CMTM4-related immune gene markers in READ were selected as follows: Cor>0.15 and P<0.05, which were found only in the READ dataset and not in COAD. As shown in Table IV, with or without tumor purity adjustment, negative weak correlations were observed between CMTM4 and IL-10 in TAMs (ρ=−0.229, P=2.97×10⁻⁰³; ρ=−0.21, P=1.30×10⁻⁰²), cluster of differentiation 33 (CD33; ρ=−0.186, P=1.66×10⁻⁰²; ρ=−0.180, P=3.41×10⁻⁰²) in neutrophils and transforming growth factor β1 in T follicular helper (Tfh) cells (ρ=−0.221, P=4.31×10⁻⁰³; ρ=−0.275, P=1.04×10⁻⁰³). Without tumor purity adjustment, CMTM4 expression was weakly negatively correlated with CXCL8 (ρ=−0.159, P=4.10×10⁻⁰²), IL13 in T helper (Th)2 (ρ=−0.186, P=1.64×10⁻⁰²) and granzyme B in exhausted T cells (GZMB; ρ=−0.158, P=4.15×10⁻⁰²). Taking tumor purity into consideration, CD66b in neutrophils (ρ=−0.257, P=2.27×10⁻⁰³) and programmed cell death 1 in T cell exhaustion (ρ=−0.172, P=4.24×10⁻⁰²) were negatively correlated with CMTM4 expression in READ. BDCA-4, otherwise known as neuropilin-1, (ρ=0.169, P=2.95×10⁻⁰²; ρ=0.271, P=1.24×10⁻⁰³) in dendritic cells, STAT1 in Th1 cells (ρ=0.178, P=2.15×10⁻⁰²; ρ=0.253, P=2.6×10⁻⁰³) and B-cell lymphoma 6 (ρ=0.191, P=1.36×10⁻⁰²; ρ=0.219, P=9.46×10⁻⁰³) were weakly positively correlated with CMTM4 expression in READ with or without tumor purity adjustment. Without tumor purity adjustment, CMTM4 expression was only weakly positively correlated with that of IL-21 (ρ=0.154; P=4.77×10⁻⁰²) in Tfh cells. The association between CMTM4 and other gene markers, as well as B cells, monocytes, TAMs, M1 macrophages, M2 macrophages, nature killer cells, dendritic cells and CD8 cells are shown in Table SIV. V-set and immunoglobulin domain containing 4 and membrane spanning 4-domains A4A in M2 macrophages, HLA-DPB1, HLA-DQB1, HLA-DRA and HLA-DPA1 in dendritic cells and CD3D in general T cells were related to CMTM4 expression in both COAD and READ. Meanwhile, STAT3 in T helper 17 and STAT5B in regulatory T cells were positively correlated with CMTM4 expression in both COAD and READ.

To verify the RNA-seq and TIMER database results, the CMTM4 overexpression and CMTM4 shRNA knockdown cell lines were constructed. CMTM4 knockdown by shRNA significantly altered CD66b, CXCL8, STAT1, PD-1 and GZMB levels (Fig. 4A). However, CMTM4 overexpression did not change the mRNA level of STAT1 and GZMB (Fig. 4B). There was a negative association between CMTM4 and CD66b in the TIMER database (Table IV), but the mRNA level of CD66b was positively associated with CMTM4 in the RT-qPCR experiments (Fig. 4A and B). Therefore, CD66b has been excluded from the further analysis and CXCL8 was specifically chosen as CMTM4-regulated gene of interest. We hypothesized that CMTM4 negatively regulates CXCL8.

Next, CXCL8 expression in the surgical tissues from patients with LARC (n=56) were detected using IHC. The depth of immune cell infiltration may exhibit different functions in antitumor activity. The location of CXCL8 expression was classified into intratumor and tumor invasive margins (Fig. S3). Both the intratumor CXCL8 expression or the tumor invasive margins in the CMTM4 negative group were higher than those in the CMTM4 positive group (P=0.012 and P=0.049, respectively; Fig. 4C and D). This indicated that the location of CXCL8 did not affect the negative association between CXCL8 and CMTM4. Additionally, 5 post-nCRT tissues with high CMTM4 expression and 3 tissues with low CMTM4 expression were also collected for RT-qPCR analysis. The average relative expression of CXCL8 in the CMTM4 high group was significantly lower than the expression in the CMTM4 low group (0.075 vs. 0.646; P=0.031; Fig. S4A). The negative association between CMTM4 and CXCL8 mRNA expression was also verified. Furthermore, Kaplan-Meier (K-M) analysis results showed that CXCL8 expression in tumor margins was linked to a shorter DFS (95% CI, 1.189–6.166; P=0.018; Fig. 4E) but did not significantly affect the OS (P=0.383; Fig. 4F). There were no significant associations between CXCL8 expression in the intratumor region and survival time (DFS, P=0.546; OS, P=0.973; Fig. S4B and C). In addition, univariate analysis showed that clinicopathological characteristics such as nerve invasion, pathological tumor (pT) stage and node (N) stage indicated a poorer DFS and OS (Table V). Multivariate analysis showed that CXCL8 expression in the tumor invasive margin regions and pathological N stage were both independent DFS predictors in patients with LARC (95% CI, 1.053–10.514, P=0.041; 95% CI 1.586–10.750, P=0.004; Table VI). Therefore, CMTM4 was negatively associated with the neutrophil-related cytokine, CXCL8, in LARC tissues and CXCL8 expression in surgical tissues may be an independent prognostic factor in patients with LARC treated with nCRT.