Pancreatic cancer is commonly known as the "king of cancer".
In recent years, world-class celebrities who have died of pancreatic cancer include: world tenor Pavarotti, Nobel Laureate Staman...
Pancreatic cancer is found late, the treatment is limited, the efficacy is poor, the fatality rate is high, and the one-year survival rate is only 25%1.
There are two types of pancreatic tumors. The most common is pancreatic adenocarcinoma (PDAC), commonly referred to as "pancreatic cancer", which originates in the ducts of the pancreas. The other is a cell that originates in the pancreas that secretes hormones, that is, a pancreatic endocrine tumor.
The location of the pancreas is deep, the early symptoms of pancreatic cancer are not obvious, the cause of the disease is not clear... These all restrict the early detection of pancreatic cancer.
Carbohydrate antigen 19-9 (CA19-9) in the blood is currently the only FDA-approved marker for pancreatic cancer. However, CA19-9 levels are associated with a variety of other symptoms (e.g., biliary obstruction) and screening specificity for pancreatic cancer is low (0.75, 95% CI 0.72 to 0.86)2, and more efficient screening methods are urgently needed.
Recently, a study published in Gut found that the fecal microbiome can be used for pancreatic cancer screening with an accuracy of 0.84 (AUROC value, a model evaluation index) when used alone; when combined with blood CA19-9, the accuracy improved to 0.943. The study was conducted by German and Spanish scientists led by academic leader Peer Bork.
Previous studies have repeatedly revealed the association between microbes and pancreatic cancer. The pancreas itself is colonized by specific microorganisms, including common bacteria in the mouth and intestines4, 5. Mouse model studies have found that gut-derived bacteria promote the formation of pancreatic ductal carcinoma6. Other studies have found that specific microbes in the human mouth and gut are associated with the risk of pancreatic cancer development7.
Inspired by these studies, Bork's team systematically excavated stool, saliva, tumors, and paracous tissue samples based on 136 French subjects (57 patients with pancreatic cancer, 50 controls, 29 patients with chronic pancreatitis), and 76 German subjects (44 patients with pancreatic cancer, 32 controls) and systematically excavated the predictive effects of fecal and salivary microorganisms on pancreatic cancer using DNA sequencing techniques (Figure 1).
Figure 1: Study sample and data composition. rRNA：ribosomal RNA
The study's first question was to confirm the association of fecal microbial composition with pancreatic cancer. After excluding factors such as age and sex, the microbial composition of stool was significantly associated with pancreatic cancer (R2=0.01, P=0.5). Univariate testing found that nine bacterial species associated with pancreatic cancer (multiplex tests corrected for PVeillonella atypica), Fusobacterium nucleatum, and broad-habitat homogeneous Skadovia bacteria (Alloscardovia omnicolens) were enriched in the pancreatic cancer group.
Immediately after, the researchers built a classifier consisting of multiple fecal microorganisms to identify pancreatic cancer patients. In the multiple logistic regression (LASSO) model, 27 bacteria were selected (Figure 2), and their combination was accurate in discriminant for pancreatic cancer to ACHIEVE AUROC=0.84 (Figure 3). This classification model is denoted model 1 (model-1).
The 27 bacteria include bacteria enriched in the pancreatic cancer group: Methanobrevibacter smithii, Skadovia bacteria of the broad-habitat variant, atypical Veronella and Bacteroides finegoldii, as well as bacteria missing from the pancreatic cancer group: Faecalibacterium prausnitzii. Bacteroides coprocola, Bifidobacterium bifidum, and Romboutsia timonensis.
Figure 2: Distribution of 27 fecal microorganisms selected in Model 1 in the training set sample. PDAC: Pancreatic ductal carcinoma, commonly known as pancreatic cancer.
Bacteria lacking in the pancreatic cancer group are often associated with a variety of non-cancerous lesions8, which are not markers of pancreatic cancer and affect the model's specificity for pancreatic cancer. To this end, the researchers used the same method to select markers only from bacteria enriched in the pancreatic cancer group, and reconstructed a classification model, referred to as model 2 (model-2).
The results show that the accuracy of model 2 (AUROC= 0.71) is lower than that of model 1. This is due to the fact that Model 2 employs stricter conditions that include only bacteria enriched in the pancreatic cancer group, reducing the sensitivity of predictions.
Carbohydrate antigen CA19-9 is secreted by tumor tissue, and fecal microorganisms can be seen as environmental factors, and in the prediction of pancreatic cancer, blood CA19-9 levels are likely to complement fecal microbial levels. CA19-9 was tested in conjunction with the above two models, and the accuracy of both models was significantly improved: the AURORAC of model 1 was increased from 0.84 to 0.94, and the AURORAC of model 2 was increased from 0.71 to 0.89.
Figure 3: Model 1 and Model 2 predict the effects of models 1 and models 2 in the training population and the validation population. DE: German validated population; ES: French training population; TPR: true positive value.
More importantly, the staging of pancreatic cancer does not affect the predictive accuracy of the above classifier. In 25 patients with early stage (T1, T2) and 32 patients with advanced (T3, T4) pancreatic cancer, the predictor effect of the classifier on early and late stage pancreatic cancer was similar and there was no bias.
Now that you've built the classifier models, the next step is to verify the applicability and generalizability of these models. To do this, the researchers tested the classifier model in two different scenarios.
The first scenario was a disease-control group of 76 Germans (44 pancreatic cancer patients and 32 controls). The test results show that the accuracy of Model 1 and Model 2 is similar to that of the training set (French): model 1 has an AURORAC of 0.83 and model 2 has an AURORAC of 0.85. The accuracy was also significantly improved when tested in combination with blood CA19-9: the AURORAC of model 1 + CA19-9 was 0.91, and the AURORAC of model 2 + CA19-9 was 0.92.
Another scenario is 5792 stool metagenomic samples. The samples were from 25 published studies involving populations in 18 countries, including patients with a variety of diseases such as diabetes, colorectal cancer, breast cancer, liver disease and enteritis. At a false-positive expectation of 10%, Model 2 has a lower false positive rate than Model 1. The average false positive rate of model 2 is less than 5%, which is better than the training set result, while the average false positive rate of model 1 is 15%, which is worse than the training set result. This suggests that Model 2 has better specificity in predicting pancreatic cancer.
The pancreas is a secretory organ whose duct is connected to the duodenum to form a connecting channel that allows intestinal bacteria to "run" to the pancreatic duct. Based on this, the researchers hypothesized that there are also fecal bacteria associated with pancreatic cancer in the pancreas. By amplification of the gene of interest and fluorescence in situ hybridization (Figure 4), the researchers detected at least 13 fecal bacteria associated with pancreatic cancer in more than 25% of pancreatic tissue samples (tumor and non-tumor). Some bacteria are enriched in tumor tissue: Lactobacillus spp, mucophilin Akkermansia muciniphila, and Bacteroides spp. These results confirmed the presence of pancreatic cancer-specific microorganisms found in feces in pancreatic organs, in line with other research findings9.
Figure 4: Fluorescence in situ hybridization microscope diagram. Bacteroides are located inside the tumor tissue nucleus (top left); Bifidobacterium is located outside the tumor tissue nucleus (top right); Lactobacillus is located outside the non-tumor tissue nucleus (bottom left); Streptococcus is located outside the non-tumor tissue nucleus (bottom left); Veron's cocci is located outside the tumor tissue nucleus (bottom right).
In summary, the study not only developed a noninvasive early screening method for pancreatic cancer based on fecal microorganisms (Figure 5), but also confirmed the presence of specific microorganisms in pancreatic tumors, providing new ideas for the prevention, mechanism research and treatment of pancreatic cancer.
Figure 5: Graphical summary of the article
Prior to this, the fecal microbiome has been developed for testing for a variety of cancers, including colorectal cancer10 and stomach cancer11. Since fecal microbes can not only "flow" through various parts of the body12, but also affect multiple organs through metabolites13, it can be speculated that fecal microorganisms will continue to glow in the fight against cancer in humans.
bibliography:
1. Park W, Chawla A, O’Reilly EM. Pancreatic Cancer: A Review. JAMA. 2021;326(9):851-862. doi:10.1001/jama.2021.13027
2. Xing H, Wang J, Wang Y, et al. Diagnostic Value of CA 19-9 and Carcinoembryonic Antigen for Pancreatic Cancer: A Meta-Analysis. Gastroenterol Res Pract. 2018;2018:8704751. doi:10.1155/2018/8704751
3. Kartal E, Schmidt TSB, Molina-Montes E, et al. A faecal microbiota signature with high specificity for pancreatic cancer. Gut. 2022:gutjnl-2021-324755. doi:10.1136/gutjnl-2021-324755
4. Riquelme E, Zhang Y, Zhang L, et al. Tumor Microbiome Diversity and Composition Influence Pancreatic Cancer Outcomes. Cell. Aug 8 2019;178(4):795-806.e12. doi:10.1016/j.cell.2019.07.008
5. Gaiser RA, Halimi A, Alkharaan H, et al. Enrichment of oral microbiota in early cystic precursors to invasive pancreatic cancer. Gut. 2019;68(12):2186. doi:10.1136/gutjnl-2018-317458
6. Thomas RM, Gharaibeh RZ, Gauthier J, et al. Intestinal microbiota enhances pancreatic carcinogenesis in preclinical models. Carcinogenesis. Jul 30 2018;39(8):1068-1078. doi:10.1093/carcin/bgy073
7. Geller LT, Barzily-Rokni M, Danino T, et al. Potential role of intratumor bacteria in mediating tumor resistance to the chemotherapeutic drug gemcitabine. Science. Sep 15 2017;357(6356):1156-1160. doi:10.1126/science.aah5043
9. Nejman D, Livyatan I, Fuks G, et al. The human tumor microbiome is composed of tumor type–specific intracellular bacteria. Science. 2020;368(6494):973. doi:10.1126/science.aay9189
11. Zhou C-B, Pan S-Y, Jin P, et al. Fecal signatures of Streptococcus anginosus and Streptococcus constellatus for non-invasive screening and early warning of gastric cancer. Gastroenterology. 2022;doi:10.1053/j.gastro.2022.02.015
12. Schmidt TSB, Hayward MR, Coelho LP, et al. Extensive transmission of microbes along the gastrointestinal tract. eLife. 2019/02/12 2019;8:e42693. doi:10.7554/eLife.42693
Responsible editor 丨 Wang Xuening