RNA sequencing-based gene expression data in CRC was obtained from the Gene Expression Profiling Interactive Analysis (GEPIA) database for cancer genomics (24). The association between CDK9 expression and the survival of 605 patients with CRC, including 121 patients who died, was analysed using the Tumor Immune Estimation Resource (TIMER) database (25). The co-expression of proteins was analysed by the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database (26). For gene expression, P<0.01 was used as the cut-off. The Kaplan-Meier method with a log-rank test was used for estimating patient survival rates. P<0.05 was considered to indicate a statistically significant difference.

A total of 35 patients with CRC who were admitted and treated at Tianjin Medical University Cancer Institute and Hospital between May 2018 and September 2018 were enrolled randomly. The patients comprised 16 men and 19 women, aged 27–79 years. During surgical resection, specimen (both tumor and paired control tissues) were obtained from the treatment-naïve patients with CRC. The gene stability of BAT-25, BAT-26, NR-21, NR-24 and Mono-27 in the specimens from all patients was examined using next-generation sequencing (Hongzhong Precision Medicine). All patients had MSS CRC according to the definition of the National Cancer Institute (there was no ‘instability’ in the results of the five aforementioned loci) (27). All samples and clinical data were collected after ethical approval was granted by the Ethics Committee of Tianjin Medical University Cancer Institute and Hospital. Written informed consent was obtained from all patients for participation in the present study. The clinical features of the patients are presented in Table I. The records of all of the patients contained basic information, including Tumor-Node-Metastasis (TNM) stage, the degree of differentiation, lymph node metastasis and distant metastasis, according to the 2002 International Cancer Alliance TNM staging criteria (28).

Fresh colorectal tumor tissues and paired control tissues from patients with CRC were prepared by mechanical disruption, followed by digestion with 0.5 mg/ml collagenase type IV (cat. no. C5138; Sigma-Aldrich; Merck KGaA) in 10% FBS with 10 U/ml DNase I in RPMI-1640 medium (both from Gibco; Thermo Fisher Scientific, Inc.) at 37°C for 30 min. Digested tissues were incubated for 5 min at 37°C with EDTA (0.5 M) to prevent dendritic cell/T-cell aggregation and filtered through a 100-µm and a 40-µm filter. The isolation and culture of tissue/tumor-infiltrating cells were performed as previously described (29).

Colon and rectal tumor tissues, and paired normal tissues were were fixed with 4% formaldehyde for 24 h at room temperature (RT), embedded in paraffin and cut into 4-µm sections. The sections were dewaxed with xylene then hydrated through a graded series of ethanol (100, 95, 90, 80 and 70%) for 5 min in each percentage at RT. Slides were boiled in 10 mM sodium citrate buffer pH 6.0 at 95°C for 10 min and subsequently incubated in 3% hydrogen peroxide for 10 min. Sections were blocked with 100–400 µl 5% normal goat serum in TBS-Tween (cat. no. 5425; Cell Signaling Technology, Inc.) for 1 h at RT. To assess CDK9 expression in the MSS phenotype mCRC specimens, sections were incubated with a CDK9 antibody (dilution, 1:100; cat. no. 2316; Cell Signaling Technology, Inc.) overnight at 4°C, and were then incubated with a horseradish peroxidase-conjugated secondary antibody (dilution, 1:1,000; cat. no. ab6721; Abcam) at room temperature for 1 h. Following washing with PBS, the sections were incubated for 30 min at RT with streptavidin-biotin conjugated with horseradish peroxidase [dilution, 1:100; cat. no. KIT-0305; UltraSensitive™ SP (Mouse/Rabbit) IHC kit]. Subsequently, the slides were stained with 3,3′-diaminobenzidine (Fuzhou Maixin Biotech Co., Ltd.), and the nuclei were counterstained with haematoxylin for 5 min at RT. Morphometric analyses of the tumor and paired normal tissues were performed using an Olympus BX51 light microscope (magnification, ×20; Olympus Corporation). Images were obtained from 10 randomly selected areas.

Staining intensity was scored as follows: 0, Negative; 1, weakly positive (<25% of cells stained); and 2, positive (>25% of cells stained). Samples were subsequently grouped into 2 categories: Low expression (0 and 1) and high expression (2).

Tumor/paired control tissue-infiltrating cells (1×10⁶ cells) were incubated with human FITC anti-CD3 monoclonal antibody (mAb) (1:20; cat. no. 300452; BioLegend, Inc.), APC anti-CD8 mAb (1:20; cat. no. 344722; BioLegend, Inc.), phycoerythrin (PE) anti-CD39 mAb (1:20; cat. no. 328208; BioLegend, Inc.), APC/Cy7 anti-PD-1 mAb (1:20; cat. no. 329921; BioLegend, Inc.), and PE/Cy7 anti-T-cell immunoglobulin mucin family member 3 (Tim-3) mAb (1:20; cat. no. 345013; BioLegend, Inc.) in cell staining buffer (cat. no. 420201; BioLegend, Inc.) for 15 min at RT. After washing with PBS and centrifugation at 400 × g for 5 min at 4°C, 1×10⁶ cells were suspended in 300 µl cell staining buffer (cat. no. 420201; BioLegend, Inc.) and analysed on a BD FACSCalibur (BD Biosciences) flow cytometer. The data were analysed using FlowJo v10 software (FlowJo, LLC).

Statistical analysis was performed using GraphPad Prism v5 software (GraphPad Software, Inc.). Data are presented as the mean ± standard error of the mean of three repeats. A χ² test was used to evaluate the association between CDK9 expression and the clinicopathological parameters. A log-rank test was used to compare the survival curves of two groups. Cox regression analysis was performed for multivariate analysis of prognostic variables. A correlation between transcripts per million (TPM) of CDK9 and other genes was calculated for statistical significance, correlation co-efficient, and is represented using a scatter plot. One-way ANOVA followed by Tukey's multiple comparison test was used for multiple-group analyses. P<0.05 was considered to indicate a statistically significant difference.

To examine the association between CDK9 and prognosis in patients with colon and rectal cancer, the present study analysed the TIMER database, and found that high CDK9 expression was significantly associated with a shortened survival of patients with colon cancer (P=0.001; 445 cases total; 98 cases of mortality). The same trend was observed in patients with rectal cancer; however, this was not statistically significant (P=0.325; 160 cases total; 23 cases of mortality; Fig. 1A). As shown in Table II, CDK9 was a risk factor for survival in patients with colon cancer based on a Cox proportional hazard model analysis. However, Cox model analysis demonstrated that CDK9 did not affect rectal cancer progression (Table III). In addition, CDK9 expression had no significant effect on prognosis when the survival time was >3 years (data not shown). Therefore, CDK9 may serve an important role in the progression of advanced colon cancer. The GEPIA database was searched and CDK9 mRNA expression in stage III–IV patients was upregulated compared with that in stage II patients, and the expression level was positively associated with the clinical tumor stage in stage IIIA-IIIC patients although the association was not significant (Fig. 1B). These data suggested that CDK9 may promote lymph node metastasis in colon cancer.

By analysing RNA sequencing (RNA-seq) data obtained from GEPIA, and the pathological estimation data from human CRC tumors, the present study revealed that patients with high expression levels of CDK9 and fewer CD8⁺ T cells in the TME exhibited poorer prognoses (Figs. 1A and 2A). Next it was determined whether CDK9 affected infiltration of CD8⁺ T cells in CRC. Specimens were collected from 35 patients with stage III–IV MSS CRC, and the clinical and pathological characteristics of the patients are shown in Table III. CDK9 expression in these patients was detected by immunohistochemical staining. The results demonstrated that the CDK9-positive expression rate (a score of 2) in stage III right-sided colon cancer was 85.7% (6/7), which was significantly higher than that in paracancerous tissues, with a positivity rate of 28.6% (2/7; χ²=4.667; P=0.03; Fig. 2B and C). However, there was no significant difference identified between CDK9 expression in all cases of left-sided colon cancer and rectal cancer (3/12) compared with the corresponding right-sided cancer (4/12; Fig. 2C). CDK9 expression in stage IV colon cancer was much higher than that in the paracancerous region, regardless of the primary tumor site (Fig. 2D). Clinically, the survival of patients with stage III–IV right-sided colon cancer was significantly shorter than that of patients with left-sided colon cancer (30,31). Therefore, these results suggested that abnormally high CDK9 expression predicted poor prognosis, and this may be a significant factor in the prognosis of patients with left/right-sided colon cancer.

Additionally, cells were isolated from the fresh tumor samples and matched normal tissues of these patients. Flow cytometry was utilized to detect tumor infiltration by CD8⁺ T cells. According to the immunohistochemical staining results, patients with the MSS-phenotype CRC were divided into the CDK9 high- and low-expression groups, and the results of flow cytometry were analysed. The frequency of tumor-infiltrating CD8⁺ T cells in the CDK9 low-expression group was markedly higher than that in the CDK9 high-expression group (Fig. 2E and F). This result demonstrated that CDK9 inhibited the recruitment and infiltration of CD8⁺ T cells in the TME.

Subsequently, it was explored how CDK9 inhibited the recruitment and migration of CD8⁺ T cells to the TME by analysing RNA-seq data from CRC samples. As shown in Fig. 3A, CDK9 mRNA expression was positively correlated with the expression of C-X-C motif chemokine ligand 12 (CXCL12), C-C motif chemokine ligand 21 (CCL21) and atypical chemokine receptor 3 (ACKR3; also referred to as CXCR7), which are associated with lymphocyte migration. It has been reported that CXCL12 decreases the number of tumor-infiltrating natural killer (NK) cells and CD8⁺ T cells (32,33). These data indicated that CDK9 may be involved in CXCL12/CCL21/CXCR7-axis-mediated negative immune regulation in the TME. Since CDK9 was associated with the expression of numerous chemokines in the TME and affected the migration of immune cells, the present study investigated whether CDK9 mediated ‘editing’ of the TME, and the function and phenotype of immune cells. To explore this, the present study analysed the association between CDK9 and CD8⁺ T cell exhaustion-associated genes by analysing RNA-seq data from CRC. Notably, there was a significant positive correlation between the transcription of CDK9 and several genes that induce T lymphocyte exhaustion, including nuclear factor of activated T cells 1 (NFATC1), V-set immunoregulatory receptor, nuclear factor of activated T cells 5 (NFAT5) and ectonucleoside triphosphate diphosphohydrolase 1 (ENTPD1; also referred to as CD39; Fig. 3B). Additionally, the anti-inflammatory molecules transforming growth factor β1 (TGFB1) and SMAD4 were correlated with CDK9 expression in CRC tumors (Fig. 4A). Furthermore, CDK9 was co-expressed with TGFB1 and SMAD4 according to STRING database analysis (Fig. 4B). Interestingly, CDK9 was significantly positively correlated with BRAF mRNA expression, which serves an important role in the development of CRC (Fig. 4C). It was revealed that 26.2% (96/367) of patients with CRC had BRAF mutations, and their antitumor immunity was more inhibited than that of wild-type patients, based on analysis of the TIMER database. The decrease in tumor-infiltrating immune cells caused by the BRAF mutation was significantly different between colon and rectal cancer (Fig. 4D and E). Compared with wild-type rectal cancer (diploid/normal), the arm-level gain (affecting ≥50% of the chromosome) of BRAF led to decreased tumor-infiltrating CD8⁺ T cells (P<0.001), B cells (P<0.05) and dendritic cells (P<0.05) in colon cancer (Fig. 4D-F), whereas only tumor-infiltrating dendritic cells (P<0.01) were decreased in rectal cancer (Fig. 4G). These data indicated that CDK9 was involved in BRAF-mediated immunosuppression, which was affected by the site of the primary tumor. Overall, these data implied that CDK9 may contribute to CRC immune escape by promoting CD8⁺ T cell exhaustion and inhibiting the infiltration of numerous immune cell populations.

To confirm the association between CDK9 and tumor-infiltrating CD8⁺ T cell exhaustion, tumor-infiltrating cells were isolated from fresh CRC tissues and paired normal tissues from advanced CRC tumors, and the cells were co-cultured with antibodies and analysed by flow cytometry. According to the aforementioned results of the CDK9 immunohistochemistry, patients with MSS phenotype CRC were divided into two groups, a CDK9 high- and a CDK9 low-expression group, for statistical analysis. The results revealed that there were no significant differences identified in the CD8⁺PD-1⁺ T cell frequency between the CDK9^high group and the CDK9^low group (Fig. 5). Compared with the CDK9^low group, the frequency of CD8⁺Tim-3⁺ T cells was increased by 37.5% in the CDK9^high group (Fig. 5). Notably, the frequency of infiltrating CD8⁺CD39⁺ T cells in CDK9^high CRC tumors was increased 1.12-fold compared with that in CDK9^low tumors (Fig. 5). These data indicated that CDK9 was positively associated with tumor-infiltrating CD8⁺ T cell exhaustion, independent of the PD-1/PD-L1 signal, and may be used to evaluate the immune-type of the TME in patients with MSS mCRC.