Characterization of the salivary microbiome in patients with pancreatic cancer

Abstract

Clinical manifestations of pancreatic cancer often do not occur until the cancer has undergone metastasis, resulting in a very low survival rate. In this study, we investigated whether salivary bacterial profiles might provide useful biomarkers for early detection of pancreatic cancer. Using high-throughput sequencing of bacterial small subunit ribosomal RNA (16S rRNA) gene, we characterized the salivary microbiota of patients with pancreatic cancer and compared them to healthy patients and patients with other diseases, including pancreatic disease, non-pancreatic digestive disease/cancer and non-digestive disease/cancer. A total of 146 patients were enrolled at the UCSD Moores Cancer Center where saliva and demographic data were collected from each patient. Of these, we analyzed the salivary microbiome of 108 patients: 8 had been diagnosed with pancreatic cancer, 78 with other diseases and 22 were classified as non-diseased (healthy) controls. Bacterial 16S rRNA sequences were amplified directly from salivary DNA extractions and subjected to high-throughput sequencing (HTS). Several bacterial genera differed in abundance in patients with pancreatic cancer. We found a significantly higher ratio of Leptotrichia to Porphyromonas in the saliva of patients with pancreatic cancer than in the saliva of healthy patients or those with other disease (Kruskal-Wallis Test; P < 0.001). Leptotrichia abundances were confirmed using real-time qPCR with Leptotrichia specific primers. Similar to previous studies, we found lower relative abundances of Neisseria and Aggregatibacter in the saliva of pancreatic cancer patients, though these results were not significant at the P < 0.05 level (K-W Test; P = 0.07 and P = 0.09 respectively). However, the relative abundances of other previously identified bacterial biomarkers, e.g., Streptococcus mitis and Granulicatella adiacens, were not significantly different in the saliva of pancreatic cancer patients. Overall, this study supports the hypothesis that bacteria abundance profiles in saliva are useful biomarkers for pancreatic cancer though much larger patient studies are needed to verify their predictive utility.

Keywords: 16S rRNA; Early detection biomarker; High-throughput sequencing; Pancreatic cancer; Salivary microbiome.

Conflict of interest statement

The authors declare there are no competing interests.

Figures

Figure 1. Relative abundance of phyla identified in patient saliva summarized by diagnosis group.

Relative abundance of phyla in oral communities from 108 study patients summarized by diagnosis group (H, healthy control; O, other disease; and P, pancreatic cancer).

Figure 2. Mean relative abundances of particular genera in pancreatic cancer patients (P) compared to healthy (H) and other disease (O) patient groups.

Relative abundances of genera in oral communities from 108 patients. Arrows point to specific genera that showed interesting trends across diagnosis groups.

Figure 3. Abundance ratio of Leptotrichia to Porphyromonas between different patient categories.

Each symbol represents the ratio of Leptotrichia Oral Taxon 221 and Leptotrichia hongkongenesis to Porphyromonas for an individual patient (n = 108). Patients are grouped into 3 different categories depending on their diagnosis: healthy control (H), other diseases (including cancer) (O), and pancreatic cancer (P). Horizontal bar and error bars represent the mean and SEM, respectively. ∗∗∗p < 0.001 (Kruskal–Wallis test followed by Dunn’s multiple-comparison test).

Figure 4. Correlation between Leptotrichia abundance from 16S rRNA sequences and from real-time qPCR.

Cross validation of total Leptotrichia OTU abundance using real-time qPCR. After using 16S rRNA as a reference gene for normalization of the levels of Leptotrichia genus, data was normalized by fold change to three healthy controls with relatively low Leptotrichia OTU abundance. Each symbol represents a patient: P = 6, and O = 12. Leptotrichia OTU abundance was correlated with qPCR fold change according to Pearson’s correlation (r = 0.903).