Pancreatic cancer, particularly pancreatic ductal adenocarcinoma (PDAC), is one of the deadliest cancer, with a 5-year overall survival rate of 11% (Siegel et al., 2022). The high lethality of PDAC is attributed to both late diagnosis and limited therapeutic options. This is because symptoms are often nonspecific and only become apparent in the advanced stages of the disease, when tumors may already be locally non-resectable or have metastasized. Surgery is the only potential curative treatment, but this method is only possible in the early stage. Combinatorial chemotherapy remains the standard of care for PDAC patients, but most patients present with an advanced disease characterized by both inherent and rapidly acquired chemoresistance to current anticancer treatments (Zeng et al., 2019).

In the past decade, microbiome studies have explored the association of pancreatic cancer with the human oral and gut microbiomes (Farrell et al., 2012; Ren et al., 2017; Fan et al., 2018; Matsukawa et al., 2021; Kartal et al., 2022). A study identified 30 gut and 18 oral species significantly linked to pancreatic cancer, exhibiting AUCs ranging from 0.78 to 0.82 (Nagata et al., 2022). In addition, epidemiological investigations have revealed a connection between periodontitis and an elevated risk of pancreatic cancer development (Michaud et al., 2007). Notably, Neisseria elongata and Porphyromonas gingivalis in saliva have been associated with an increased risk of pancreatic cancer development (Fan et al., 2018). Despite the pancreas not being part of the alimentary canal, the suggested translocation of microbiota from the gut to the pancreas is thought to occur through the sphincter of Oddi (Ansari et al., 2023). Studies using mouse models strongly suggested profound associations of gut microbiome with pancreatic cancer (Pushalkar et al., 2018; Thomas and Jobin, 2020). Moreover, studies suggested that microbiome ablation with antibiotics in mouse models improved tumor immune surveillance and improved responses to PD-1 blockade (Pushalkar et al., 2018). On the one hand, antibiotic use may lead to microbiota dysbiosis, potentially fostering chemoresistance and influencing treatment outcomes (Routy et al., 2018). On the other hand, much evidence suggested that the gut microbiota plays a role in determining resistance to various anticancer treatments, encompassing conventional chemotherapy, immunotherapy, radiotherapy, and surgery (Garajova et al., 2021). Several studies have highlighted the microbiome’s involvement in drug resistance within gastrointestinal cancers, including esophageal and pancreatic cancers (Yang et al., 2009; Wilkinson et al., 2018; Garajova et al., 2021). An increasing number of studies has shown that gut microbiota may impact resistance to commonly used anticancer drugs such as irinotecan, oxaliplatin, cyclophosphamide, 5-fluorouracil, gemcitabine, and anthracyclines (Bloemen et al., 2010; Iida et al., 2013; Sivan et al., 2015; Goubet et al., 2018). For instance, Bacteroides spp., residing in the gastrointestinal tract, expedite the conversion of sorivudine (a synthetic thymidine analogue used as an antiviral agent) into bromovinyluracil (BVU), an intermediate product that inhibits 5-FU degradation by the enzyme dihydropyrimidine dehydrogenase (Nakayama et al., 1997; Chae et al., 2020). However, the gut microbial antibiotic resistance genes profiling of pancreatic cancer patients has never reported compared to healthy cohort. Therefore, enhancing our understanding of the gut microbiota and its interactions with anticancer drugs will empower us to devise innovative treatment strategies, thereby contributing to treatments of cancer patients.

In this study, we addressed the gut microbial antibiotic resistance genes profile using the metagenomic data from online public pancreatic cancer cohorts. By studying the types, abundance, copy number, and affiliated microorganisms of antibiotic resistance genes in the gut microbes of pancreatic cancer patients, we describe the panorama of antibiotic resistance genes in the gut microbes of patients with the disease. We aimed to find out the main types of antibiotic resistance genes and the microorganisms. This will provide a theoretical basis for more accurate assessment of the patient’s condition and adjustment of the treatment plan for pancreatic cancer in the future.

Antibiotic resistance has been a globally health concern. It affects the therapeutic efficacy of diseases. Traditional methods of antibiotic resistance testing are mainly based on minimal inhibitory concentration (MIC) test of cultured isolates of microorganisms. The gut contains thousands of microbial species, most of which are uncultured or difficult to culture. It is difficult to detect gut antibiotic resistance using traditional methods because of their large microbial community. Moreover, detecting antibiotic resistance of isolates does not reflect the overall of antibiotic resistance in gut microbes. The whole genome sequencing-based metagenomic technology can detect most of the DNA fragments in one sample and thus can detect the antibiotic resistance genes from gut. It can reveal the overall of antibiotic drug resistance to a certain extent by this method. Pancreatic cancer is one of the common malignant tumors in the digestive tract and is known as the “king of cancers” in the field of oncology. Studies in this field of microbiology have also been widely reported, but the gut microbial antibiotic resistance profile of pancreatic cancer are rarely reported. Here, we investigated the gut microbial antibiotic resistance profiles of pancreatic cancer using two online public data cohorts with oral and fecal samples, which can be used to validate the results against each other. In addition, we included a cohort of healthy individuals that also contained oral and fecal samples as a control. We found that the two pancreatic cancer cohorts had an overlap of more than 85% in abundance of top 300 species for both oral and fecal samples, suggesting that the two cohorts were well reproducible and could be used for validation of the results in subsequent analyses.

While numerous anticancer therapeutic regimens have achieved clinical success, the enduring challenges of heterogeneous response and resistance to chemotherapy and immunotherapy remain the hallmarks of cancer therapy. Recent findings have unveiled a correlation between the microbiota and the development of chemoresistance (Choy et al., 2018). Hence, integrating microbiome-modulating regimens (such as antibiotics, probiotics, and dietary interventions) with anticancer treatment could offer innovative therapeutic approaches for cancers associated with dysbiosis. Several studies have indicated the significant impact of gut microbiota on drug metabolism and encompassing anticancer drugs (Ma et al., 2019). This process may result in heightening or diminishing drug activity, coupled with variations in toxicity levels. Consequently, manipulating the microbial network through interventions like fecal transplantation or probiotics emerges as a promising strategy that may enhance the treatment efficacy for cancer patients. Ongoing clinical trials are exploring the potential of the microbiome to enhance the management of pancreatic cancer. In our study, genera such as Prevotella and Fusobacterium were significantly enriched in the pancreatic cancer cohort, compared with healthy cohort in oral samples. Based on a recent study about oral microbiota in pancreatic cancer, patients had a higher amount of Prevotella compared with the healthy control group (Wei et al., 2020). Fan et al. found that Fusobacterium and several bacteria might be a protective factor in oral samples of pancreatic cancer (Fan et al., 2018). Previous studies have reported that vancomycin-resistant genes (vanT gene and vanY gene) were identified from Streptococcus, Veillonella, and Prevotella. Kwack et al. revealed the biological mechanisms of vancomycin tolerance in the oral commensal bacterium Streptococcus anginosus by whole genome and RNA sequencing (Kwack et al., 2022). Li et al. tested the vancomycin resistance in Veillonella strains from oral of healthy adults by agar dilution method (Li et al., 2022). In the fecal samples, Prevotella, Streptococcus, and Klebsiella were the differential abundant microbes in the pancreatic cancer cohort compared with healthy cohort (Kabwe et al., 2022). The genus Streptococcus has been studied for its association with cancer progression and treatment outcomes over an extended period (Sobocki et al., 2021). Matsukawa et al. found that Klebsiella, Streptococcus, and other several microbiotas can be treated as prognostic factors for pancreatic cancer (Matsukawa et al., 2021). In fecal samples of pancreatic cancer, vancomycin-resistant genes (vanT gene, vanW gene, and vanY gene) were also found in some popular gut microbes (such as Streptococcus, Blautia, and Ruminococcus) (Stogios and Savchenko, 2020). An earlier study showed that vanD and vanG-Like gene clusters were located in Ruminococcus species, which isolated from human bowel flora (Domingo et al., 2007). The vanB gene has also been found in a vancomycin-resistant isolate of Streptococcus bovis isolated from a stool swab (Poyart et al., 1997). The above findings suggest the importance of vancomycin-resistant microbial communities in pancreatic cancer.

Two strategies are currently used to study the antibiotic resistance gene profiles of gut microbes: CARD database mapping vs. genome assembly and RGI prediction. Both strategies have their advantages: CARD database mapping strategy is more sensitive and can detect more antibiotic resistance genes, which is mainly used to calculate the abundance of those genes. The disadvantages of this strategy are that it may be affected by the length of the reads, which leads to a low prediction accuracy, and the range of antibiotic resistance genes detected may be limited by the content of the database. Another method for antibiotic resistance gene prediction is genome assembly and RGI prediction, which is a strategy based on the annotation of assembled genome sequences and is mainly used to detect the presence or absence of antibiotic resistance genes and the copy numbers. It offers the benefit of relatively high accuracy and the potential for uncovering new antibiotic resistance genes. However, a drawback is the inability to quantify their abundance. In this study, we compared these two prediction strategies. It was found that CARD database mapping strategy detected more types of antibiotic resistance genes than genome assembly and RGI prediction in both oral and fecal samples. In terms of the average number of antibiotic resistance genes of each sample, the two strategies did not differ much in oral and fecal samples. In terms of frequency, for both strategies, the genes with high frequency in oral and fecal samples were from the vancomycin and tetracycline resistance gene families, which shows that the results predicted by the two methods are in good agreement. However, there are some differences between the two strategies. According to the results of the heatmap and the Wilcox test, genome assembly and RGI prediction identified significant differences in some vancomycin-resistant genes between the pancreatic cancer patient cohort and the healthy cohort. However, this distinction was not observed with the CARD database mapping strategy. In summary, we found that the CARD database mapping strategy was superior in sensitivity in the prediction of the types of antibiotic resistance genes, whereas genome assembly and RGI prediction was better able to detect anomalies in copy number between groups.

There are still some shortcomings in this study. First, because we used publicly available online data, we lacked information on oral and fecal pairings; thus, we were unable to compare the antibiotic resistance genes between oral and fecal samples in a single individual. Second, the online public data also lacked information on antibiotic use and treatment of the patients, so we were unable to map antibiotic resistance genes to patients’ clinical information. Finally, the resistance of some microbes requires isolation by bacteria culture and measure by MIC. We hope that future research may complement the shortcomings.

We conducted an extensive evaluation of antibiotic resistance genes using both oral and fecal samples sourcing from two publicly available online datasets of pancreatic cancer populations. We found a high degree of data concordance between the two cohorts, which can therefore be used for cross-sectional comparisons. Meanwhile, we used two strategies to predict antibiotic resistance genes and compared the advantages and disadvantages of these two approaches. Based on genome assembly and RGI prediction strategy, we found that four vancomycin-resistant genes and one tetracycline-resistant gene differed significantly in copy number between the pancreatic cancer Cohort I and the healthy cohort. We also constructed microbe-antibiotic resistance gene networks and found that most of the hub nodes in the networks were antibiotic resistance genes. We hope that our research will contribute to adjuvant therapy for pancreatic cancer patients by designing new antibiotic regimens that can be targeted to alter the abundance of dysbiotic microbiota in the future.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary Material.

The studies involving humans were approved by PUMCH Institutional Review Board. The studies were conducted in accordance with the local legislation and institutional requirements. Written informed consent for participation was not required from the participants or the participants’ legal guardians/next of kin in accordance with the national legislation and institutional requirements.

XL: Data curation, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing – original draft, Writing – review & editing. KL: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Resources, Software, Validation, Visualization, Writing – original draft, Writing – review & editing. YY: Formal analysis, Methodology, Resources, Software, Writing – original draft. DC: Methodology, Project administration, Software, Supervision, Writing – original draft. XX: Formal analysis, Investigation, Methodology, Resources, Supervision, Writing – review & editing. ZH: Data curation, Investigation, Project administration, Resources, Software, Supervision, Visualization, Writing – original draft, Writing – review & editing. WW: Data curation, Funding acquisition, Investigation, Project administration, Resources, Supervision, Writing – original draft, Writing – review & editing.