Stemming from the large intestine (colon) or rectum, colorectal cancer (CRC) ranks third in men and ranks second in women in terms of global cancer incidence (1). Notably, in Eastern Asia, rectal cancer incidence rates rank among the highest (1). Rectal cancer patients without distant metastasis are usually managed by a standardized surgical technique termed total mesorectal excision (2). To reduce the risk of locoregional recurrence, the use of multimodal approach such as neoadjuvant concurrent chemoradiotherapy (CCRT) is recommended before surgical management for rectal cancer patients with stage T3/T4 or node-positive (N1/N2) disease (2). However, despite identical tumor histology, individual patient response to neoadjuvant CCRT can range from complete remission to disease progression (3). Identification of predictive factors such as tumor regression (4) has been introduced for decision-making according to preoperative staging information, but a comprehensive molecular characterization that is used to predict individual therapy response remains lacking.

The intestinal epithelium is continuously regenerated (every 3–5 days) by a population of stem cells (5) that differentiate into multiple epithelial cell subsets which operate in concert to maintain the intestinal barrier and provide host defense. Among these epithelial cell subsets, goblet cells are well known for their secretion of the glycosylated mucin 2 (MUC2) protein, which creates the protective mucus barrier covering the epithelium (6). Muc2-deficient mice have been indicated to develop spontaneous colitis, an inflammatory bowel disease, caused by direct contact between the bacteria and the colonic epithelium (7). Low MUC2 expression was also correlated with poor survival in CRC patients without receiving radiation or chemotherapy (8, 9). However, our previous study has suggested that MUC2 overexpression is an unfavorable predictive and prognostic factor for rectal adenocarcinoma patients receiving neoadjuvant CCRT (10). These observations imply that the roles of mucin in tumor progression and CCRT efficacy may be different. In addition, poorly differentiated CRC with high proliferative and metastatic capacity is correlated with worse patient survival (11). Interestingly, it has been reported that, in the radio-resistant rectal cancer cells, most of the differently expressed genes were associated with cell-cell adhesion and regulation of epithelial cell differentiation (12). Whether epithelial cell differentiation can function as a barrier to defend against radiation penetration remains an open question. Accordingly, the role of epithelial cell differentiation in the efficacy of CCRT in rectal cancer patients and the underlying molecular mechanisms deserve further investigation.

Using a transcriptome dataset, we focused on differentially expressed genes in relation to epithelial cell differentiation and then identified that the dehydrogenase/reductase 9 (DHRS9) gene level was the most considerably upregulated among CCRT nonresponders in rectal cancer. The human DHRS9 gene, which maps to chromosome 2q31.1, encodes an enzyme of the short-chain dehydrogenase/reductase (SDR) family. DHRS9, also known as retinol dehydrogenase 15 (RDH15), can catalyze the oxidation of retinol to retinaldehyde and is found specifically in the intestine and featured mainly by mucin production according to the analysis of single cell type expression cluster (https://www.proteinatlas.org/ENSG00000073737-DHRS9). DHRS9 is also implicated in the biosynthesis of all-trans-retinoic acid (atRA), which is the most active retinoid metabolite and contributes to the suppression of CRC progression through cell growth inhibition and cell differentiation induction (13). Moreover, low DHRS9 expression has been indicated to be correlated with poor prognosis in CRC patients, but those who received any anti-cancer therapy were excluded (14). Consequently, the current study intended to connect DHRS9 expression to CCRT efficacy and patient survival and illuminate the role of DHRS9 in rectal cancer patients undergoing neoadjuvant CCRT.

All-trans-retinoic acid (atRA), an active metabolite of vitamin A (retinol), regulates gene expression through binding to the nuclear receptors and is applied as a classical differentiation therapy for acute promyelocytic leukemia (APL) (17). Since atRA has been reported to exert tumor-suppressive functions in various cancer types (18), DHRS9, implicated in the biosynthesis of atRA, may be considered to have antitumor activities. Actually, DHRS9 downregulation has been correlated with poor survival in oral squamous cell carcinoma (19) and colorectal cancer (14) patients, but those who received any anti-cancer therapy were excluded in these two studies. Moreover, in pancreatic cancer, DHRS9 overexpression has been correlated with poor prognosis (20), implying that DHRS9 may play an oncogenic role based on distinct contexts. Interestingly, in the present study, we identified DHRS9 as the most significantly upregulated gene in relation to epithelial cell differentiation among CCRT-resistant rectal cancer patients. In our rectal cancer cohort, we also demonstrated that high DHRS9 immunoexpression is remarkably linked to poor therapeutic response to CCRT and inferior patient survival, suggesting that epithelial cell differentiation may play a role in reducing CCRT efficacy in patients with rectal cancer.

To identify the unrevealed functions of DHRS9 in rectal cancer, we carried out a gene coexpression analysis and found that the most significant GO terms positively correlated with DHRS9 were keratan sulfate biosynthetic process and UDP-galactose:beta-N-acetylglucosamine beta-1,3-galactosyltransferase activity (Figures 4A,B). Glycosaminoglycans (GAGs) are linear polysaccharides with highly negative charges and display distinct functions as determined by their molecular weight and sulfation degree. In spite of their complicated structure, the backbone of these glycans is generally simply composed of repeating disaccharide units containing alternating hexosamines (glucosamine or galactosamine) and uronic acids (glucuronic or iduronic acid) (21). Based on the combination of different amino sugars and uronic acids, GAGs are categorized into four primary groups, including keratan sulfate (KS), which alternates between N-acetylglucosamine (GlcNAc) and galactose (Gal) and does not contain uronic acids (22). GAGs, the major macromolecules in the extracellular matrix (ECM), may function as a protective barrier that impedes drug delivery to the tumor cells (23). Additionally, GAGs can also play a key role in cell signaling and modulate abundant biological functions (24). Recent evidence suggests that GAGs may be involved in atRA-induced neural differentiation (25) and cancer stem cell formation and therapeutic resistance (26). Keratan sulfate is also regarded as a glycan marker expressed by stem cells (27), which is suggested to confer drug resistance in CRC patients (28). Nevertheless, the correlations among DHRS9 expression, GAGs, especially keratan sulfate, and CCRT resistance in rectal cancer patients need to be further investigated.

Proteoglycans (PGs) are made up of a core protein and at least one covalently attached GAG side chain. Keratan sulfate is the newest GAG based on an evolutionary perspective but the least understood. On the basis of linkage structure utilized to attach to PG core proteins, internal structural organization, and tissue distribution, keratan sulfate can be classified into three types, namely KS type I (corneal KS, N-linked), II (skeletal KS, O-linked), and III (brain KS, O-linked) (29). Especially, KS type II attaches to a threonine or serine (Thr/Ser) residue on the core protein via an O-linked mucin-type structure (core 2, GalNAc-Thr/Ser). During the biosynthesis of KS, several glycosyltransferases and sulfotransferases act to add Gal or GlcNAc to an acceptor residue for chain elongation and undergo sulfation at carbon position 6 (C6) on Gal-GlcNAc either individually or collectively, respectively (30). The chain length and sulfation degree (charge heterogeneity) of KS increase with the age of the connective tissues where it is enriched in and their pathological status, including tumor development. Intriguingly, we observed that the B3GNT6 and B3GNT7 genes (Supplementary Figures S2A–B), which encode N-acetylglucosaminyltransferase enzymes, and the carbohydrate sulfotransferase 5 (CHST5) gene (Supplementary Figure S2C) were positively correlated with DHRS9.

B3GNT6 supports the formation of an O-linked mucin-type core 3 structure (31), and B3GNT7 is responsible for elongation of KS chains (32). In addition, CHST5 is a sulfotransferase that can transfer sulfate to O-linked glycans of mucin-type glycoproteins (33). Actually, the GalNAc residue of mucin-type glycoproteins can also serve as an acceptor molecule for the addition of Gal or GlcNAc that can also be sulfated and contribute to minimally sulfated KS type II-like mucins (34). As a key component of the ECM, mucins create a gel-like epithelial barrier thought to protect the gut lumen from external stress and microbial infection. However, aberrant mucin synthesis may also function as a barrier to drug penetration and cytotoxic T cell infiltration (35). MUC2, the major intestinal mucin, has been suggested to carry immunoregulatory signals to favor tumor growth in the large intestine (36), and our previous study also demonstrated that overexpression of MUC2 is linked to CCRT resistance and poor prognosis in patients with rectal adenocarcinoma (10). Notably, we also identified that the MUC2 gene was positively correlated with DHRS9 in gene coexpression analysis (Supplementary Figure S2D). In addition, brush border membrane was the most prominent term positively correlated with DHRS9 in terms of cellular components (Figure 4C). The epithelium of the villus is composed primarily of absorptive enterocytes and mucin-secreting goblet cells. Since the brush border is formed with microvilli on the apical surface of the enterocytes, whether DHRS9 can promote mucin synthesis through enterocyte–goblet cell interaction to defend against CCRT penetration deserves further analysis.

To predict the tumor response to preoperative CCRT, the tumor regression grade system is commonly used based on pathological features observed in surgical specimens. However, before surgery, there is currently no precise tool to predict CCRT effectiveness. Benefitting from the advancement of sequencing technologies, only small piece of tissue is required to obtain genetic information. To exploit our findings in the clinical practice, rectal cancer biopsies could be collected in the course of colonoscopic examination before receiving neoadjuvant CCRT. Afterwards, these biopsies could be used to conduct RNA sequencing or apply array platform and guide treatment more accurately according to the biomarkers such as high DHRS9 level detected.

The intestinal epithelium is composed of multiple well-differentiated epithelial cell subsets which work in concert to maintain the intestinal barrier and provide host defense. However, aberrant epithelial cell differentiation may be associated with therapy resistance. In this study, we identified that DHRS9 is the most significantly upregulated gene related to epithelial cell differentiation among CCRT-resistant rectal cancer patients. We also demonstrated that high DHRS9 immunoexpression is considerably associated with an advanced disease stage, CCRT resistance, and inferior prognosis in patients with rectal adenocarcinoma. Additionally, we also linked DHRS9 with unrevealed functions, such as keratan sulfate and mucin synthesis which may be implicated in CCRT resistance. Collectively, DHRS9 expression may assist decision-making for rectal cancer patients who underwent neoadjuvant CCRT.