https://orcid.org/0000-0001-7575-0453
Abstract
Increasing evidence implicates microbiome involvement in the development and progression of pancreatic ductal adenocarcinoma (PDAC). Studies suggest that reflux of gut or oral microbiota can lead to colonization in the pancreas, resulting in dysbiosis that culminates in release of microbial toxins and metabolites that potentiate an inflammatory response and increase susceptibility to PDAC. Moreover, microbe-derived metabolites can exert direct effector functions on precursors and cancer cells, as well as other cell types, to either promote or attenuate tumor development and modulate treatment response.
The occurrence of microbial metabolites in biofluids thereby enables risk assessment and prognostication of PDAC, as well as having potential for design of interception strategies. In this review, we first highlight the relevance of the microbiome for progression of precancerous lesions in the pancreas and, using liquid chromatography–mass spectrometry, provide supporting evidence that microbe-derived metabolites manifest in pancreatic cystic fluid and are associated with malignant progression of intraductal papillary mucinous neoplasm(s). We secondly summarize the biomarker potential of microbe-derived metabolite signatures for (a) identifying individuals at high risk of developing or harboring PDAC and (b) predicting response to treatment and disease outcomes.
The microbiome-derived metabolome holds considerable promise for risk assessment and prognostication of PDAC.
Introduction
Pancreatic ductal adenocarcinoma (PDAC) is the third most common cause of cancer-related death in the United States with a 5-year survival rate currently at approximately 10%. The detection of PDAC at earlier, resectable stages has a profoundly favorable impact on prognosis, with 5-year survival rates as high as approximately 25% to 30% being achievable in major treatment centers, increasing to 30% to 60% for tumors <2 cm, and as high as 75% for lesions under 10 mm in size (1).
Given the low incidence of PDAC in the general population (10 per 100 000), the difficulty of generating a cost-effective screening test with sufficiently high positive predictive value (PPV), the high cost of abdominal imaging-based screening, and the emotional toll of a false-positive result, screening for PDAC in average-risk individuals is currently not recommended (2). High-risk individuals, such as those with inherited risk or individuals with a history of chronic pancreatitis or new-onset diabetes (NOD) may benefit from surveillance and screening (2). However, high-risk persons only account for approximately 10% of all individuals who ultimately receive a PDAC diagnosis. Consequently, there is an unmet clinical need to establish biomarkers for identifying individuals who are at high risk of PDAC that would benefit from surveillance and screening to enable earlier detection and potential interception of disease.
Emerging evidence strongly implicates the microbiota in the development of PDAC. Anatomically, the pancreas is connected to the oral cavity, esophagus, and stomach upwards through the pancreatic duct, downwards to the duodenum, and is adjacent to the common bile duct. This creates the possibility of bacterial reflux from the upper gastrointestinal tract through the pancreatic duct and eventually into the pancreatic parenchyma through the large/little papilla. Microbiota from distal sites, e.g., the colon, may migrate through alternative pathways (3). These changes may culminate in microbial dysbiosis and release of microbial-derived toxins and metabolites that can potentiate inflammation and increase susceptibility to various diseases, including PDAC, as well as contribute to tumor progression and patient survivorship (3–5). The occurrence of microbial metabolites manifest in biofluids may thus provide an opportunity for liquid biopsy biomarkers in the context of PDAC.
The goal of this review is to summarize current knowledge of microbial-derived metabolite signatures in biofluids in PDAC and to specifically highlight their imminently translatable utility for assessing individual risk of developing or harboring PDAC as well as for predicting disease outcomes or response to treatment.
Microbiome, Microbial-Associated Metabolite Signatures, and Malignant Progression of Precursor Lesions of the Pancreas
Several studies have documented definitive roles for the intratumor microbiome in modulating tumor immunophenotype and thereby affecting individual response to treatment (5–8). Emerging data directly links microbial dysbiosis as a contributor to malignant progression of precursor lesions by promoting immune suppression and potentiating inflammation (3, 9).
A bona fide precursor of PDAC is intraductal papillary mucinous neoplasm (IPMN). Gaiser and colleagues investigated the intracystic pancreatic microbiome in a cohort of patients undergoing surgery for suspected pancreatic cystic neoplasms (PCNs) (9). Intracystic bacterial 16S DNA copy number was found to be significantly higher in IPMNs with high-grade (HG) dysplasia or with an associated PDAC (hereon referred to as IPMN/PDAC) compared to non-IPMN PCNs. Bacterial networks and linear discriminant analysis effect size (LEfSe) analyses revealed enrichment of oral bacteria taxa including Fusobacterium nucleatum and Granulicatella adiacens in cyst fluid from patients with HG IPMN. Halimi and colleagues similarly reported Gammaproteobacteria, particularly Klebsiella pneumoniae, G. adiacens, and Enterococcus faecalis, to be prominent in the cystic fluid of patients with IPMN. Co-culture experiments demonstrated that K. pneumoniae and E. faecalis bacteria were able to colonize pancreatic cells and induce phosphorylation of gamma-H2A histone family member X (γH2A.X), reflective of activation of a DNA damage response (10). Evaluation of microbes at the single-cell level in PDAC tumors and normal pancreas tissues similarly revealed high intratumoral and interindividual heterogeneity with respect to the extent of intracellular microbial colonization in tumors that was positively correlated with higher activated T cells infiltration and lower representation of regulatory immune cell phenotypes (11). These findings are consistent with a seminal study by Nejman and colleagues who also reported that microbes reside intracellularly within cancer cells as well as immune cells and that the extent of bacterial colonization is distinct among tumor types (12).
These studies present an intriguing question as to whether intracellular microbes might modify the metabolic phenotype of cells within the nascent tumor microenvironment and ultimately promote malignant progression. To this end, metabolomic analyses of cystic fluid and plasma samples from patients with IPMN identified elevated abundances of microbe-associated metabolites including trimethylamine-N-oxide (TMAO), the polyamine cadaverine (CAD), carboxylic acids, and conjugated bile acids to be associated with HG dysplasia or IPMN/PDAC (13). Moreover, microbe-derived TMAO, as well as CAD and indole (byproducts of microbial-mediated metabolism of lysine and tryptophan, respectively) clustered with Enterobacteriaceae, Granulicatella, Klebsiella, Stenotrophomonas, Streptococcus, Haemophilus, and Fusobacterium, which represent the phyla of Bacillota, Fusobacteriota, and Pseudomonadota (13).
To confirm occurrence of microbial metabolites in cystic fluid, our group has performed metabolomic analyses using liquid chromatography–mass spectrometry (LC-MS: see online Supplemental Materials for additional details) on a cohort of 87 patients with IPMN (68 with low-grade [LG] dysplasia; 12 with HG dysplasia; 7 with IPMN/PDAC). These studies resulted in a total of 20 quantified metabolites of microbial origin. Compared to patients with LG IPMN, TMAO, uremic toxins indoxyl sulfate (IS), cresyl sulfate (p-CS), and cresyl glucuronide (p-CG), and alpha-N-phenylacetyl-L-glutamine (PAGln), as well as the secondary bile acid, ursodeoxycholic acid (UDCA), were found to be highly elevated in cystic fluid of patients with IPMN/PDAC; indole and derivatives, indoleacryclic acid (IAA) and 3-methyl-2-oxindole (MOI), were highest in patients with HG IPMN (Fig. 1A). Using the Metabolomics Data Explorer database (14), we found that microbe-derived metabolites associated with HG IPMN or IPMN/PDAC were annotated to phyla of Actinomycetota, Bacillota, Bacterioidota, and Pseudomonadota (Fig. 1B) (14, 15).
Association between cystic fluid microbial-derived metabolites and malignant progression of IPMN. (A), Scatter plot depicting area under the receiver operating characteristic curves and corresponding 95% CIs of individual microbial-derived metabolites for distinguishing HG + IPMN/PDAC (n = 19), IPMN/PDAC (n = 7), and HG (n = 12) from LG IPMN (n = 68). (B), Box and whisker plots illustrating relative abundances (z-scores) of microbial-derived metabolites in cystic fluid of patients with IPMN. Abbreviations: AcCad, acetylcadaverine; AUC, area under the Receiver Operating Characteristic Curve; CI, confidence interval; DiAcCad, diacetylcadaverine; LG, low-grade dysplasia; HG, high-grade dysplasia; IPMN, intraductal papillary mucinous neoplasm; PDAC, pancreatic ductal adenocarcinoma.
TMAO is an oxidation product of trimethylamine (TMA) that is produced by the metabolism of choline, betaine, and L-carnitine by microbiota (16, 17). TMAO exerts pro-inflammatory effects through activation of the nuclear factor kappa B (NF-κB) pathway and the nucleotide-binding domain, leucine-rich–containing family, pyrin domain-containing-3 (NLRP3) inflammasome through interaction with the thioredoxin-interactive protein (TXNIP) (18–21), and persistence of TMAO may prolong tissue level inflammation and increase susceptibility to various diseases, including cancer.
Interestingly, Mirji and colleagues demonstrated in a syngeneic orthotopic mouse model of PDAC, that intraperitoneal administration of TMAO induces an immunostimulatory phenotype in tumor-associated macrophages (TAMs) that, in turn, potentiates effector T cell-mediated cancer killing in a type-I interferon (IFN)-dependent manner (6). Another study reported that TMAO exerts antitumor effects by activating the protein kinase RNA-like endoplasmic reticulum kinase (PERK), a marker of endoplasmatic reticulum stress, resulting in pyroptosis of tumor cells and enhanced antitumor activity of cluster of differentiation (CD) 8+ T cells (22). In addition to the demonstrated antitumor effects of TMAO, other microbial-associated metabolites are known to interact with immune cell subtypes within the tumor microenvironment.
For instance, indole and its associated derivatives, such as IAA, can activate the aryl hydrocarbon receptor (AhR) (23–27), which can result in immune suppression (28–30). In this regard, a recent study by Hezaveh and colleagues demonstrated that activation of AhR in TAMs suppresses antitumor immunity and that blunting of AhR activation polarizes TAMs to an pro-inflammatory state that is accompanied by increased tumor infiltration of effector/memory (CD62LnegCD44hi) CD8+ T cells and attenuated tumor growth (28).
Macrophage AhR activity was dependent on Lactobacillus-mediated metabolism of tryptophan to yield IAA, indole-3-aldehyde (IAld), and indole-3-lactate (I3L). To this effect, dietary restriction of tryptophan reduced TAM AhR activity and promoted intratumoral accumulation of tumor necrosis factor-alfa (TNFα)+ IFNγ+ CD8+ T cells, which was found to be reversable via supplementation with IAA, IAld, or I3L (28). IS, p-CS, and p-CG have also been reported to promote immune dysfunction in macrophages (25, 27, 31) as well as in T-helper 1 cell (Th1)-type cells (32). These changes are notable given that macrophages play key roles in the development and progression of precursor lesions to malignancy (33–36). Thus, elevation of microbial-derived metabolites, including TMAO and various indole derivatives, may reflect the dynamic alterations in the host immune response that accompany transition from premalignancy through more advanced disease. Moreover, microbe-derived signatures may similarly manifest in the peripheral circulation, providing an opportunity to develop biomarkers for risk assessment for developing or harboring PDAC.
Microbe-Derived Metabolite Signatures and Pancreatic Cancer Risk
The interactions between diverse microbial metabolites and cell and molecular constituents of the nascent tumor microenvironment may lead to dynamic changes in the host response that enable disease progression (Fig. 2). Importantly, these changes in microbial composition and associated metabolites may be manifest in the peripheral circulation (Table 1), providing an opportunity for assessment of risk of developing or harboring PDAC.
Association between microbiome-metabolome and malignant progression of precursor lesions of the pancreas. Translocation of gut- or oral-derived microbes into pancreas results in dysbiosis and release of diverse microbe-derived metabolites that promote inflammation and modulate the host immune response and increase susceptibility to developing PDAC and potential progression of precursor lesions to PDAC. Microbial metabolites such as TMAO, uremic toxins indoxyl sulfate, cresyl sulfate, cresyl glucuronide, and α-N-phenylacetyl-L-glutamine as well as other indole derivatives, secondary bile acids, and short chain fatty acids (SCFAs) can manifest in biofluids (e.g., cystic fluid, blood) and inform upon risk of developing or harboring PDAC. Created with BioRender.com. Abbreviations: AhR, aryl hydrocarbon receptor; IFNAR, interferon-α/β receptor; IFNγ, interferon-gamma; IPMN, intraductal papillary mucinous neoplasm; MΦ, macrophage; NLRP3, nucleotide-binding domain, leucine-rich–containing family, pyrin domain-containing-3; NF-κB1, nuclear factor kappa b; NK, natural killer; PDAC, pancreatic ductal adenocarcinoma; PERK, protein kinase RNA-like endoplasmic reticulum kinase; SCFA, short-chain fatty acid; TAM, tumor-associated macrophage; Th1, T-helper 1 cell; TMAO, trimethylamine-N-oxide; TNFα, tumor necrosis factor-alpha; Treg, regulatory T cell; TUDCA, tauroursodeoxycholic acid.
Association between microbial-derived metabolite levels in biofluids and pancreatic cancer.a
| Study | Specimen type | Platform | No. of specimens | Reported findings |
|---|---|---|---|---|
| Löser et al. (37) | Serum | LC-MS/MS | 20 PDAC cases, 30 healthy controls, 40 patients with benign GI disease | Elevated cadaverine levels (nM (SE)) in serum of PDAC cases (0.76 (0.09)) compared to healthy controls (0.32 (0.07)) or patients with benign disease (0.47 (0.05)) |
| Urine | Elevated cadaverine levels (nmol/mg of creatine (SE)) in urine of PDAC cases (10.7 (2.31)) and patients with benign disease (12.18 (1.8)) compared to healthy controls (3.4 (0.5)) | |||
| Tissue | 15 resected PDAC tissues | Elevated cadaverine levels in carcinoma compared to unaffected pancreas areas (20.9 (3.6) vs 7.2 (2.4 nmol/mg of DNA)) | ||
| Nissinen et al. (38) | Urine | LC-MS/MS | 68 PDAC cases, 7 cases with PLP, 36 AP cases, 18 CP cases, and 53 healthy controls | Acetylated cadaverine (median umol/g creatinine) was elevated in urine of patients with PDAC/PLP (1.4), AP (2.5), or CP (2.9) compared to healthy controls (0.54). |
| Luo et al. (40) | Plasma | LC-MS | 60 PDAC cases and 60 healthy controls | 4.7- and 4.2-fold change increase in TUDCA and TDCA in PDAC cases compared to controls (P < 0.001), respectively. |
| Zhao et al. (41) | Serum | LC-MS; LC-MS/MS; GC-MS | 80 PDAC cases, 36 patients with benign disease, and 48 healthy controls | 1.45-fold change increase in indoxyl sulfate (P: 0.011) in PDAC cases compared to controls |
| Cao et al. (42) | Serum | LC-MS | 28 stage I PDAC cases and 62 healthy controls | 5.76-fold increase in indole-3-propanoate (P: 0.029) in PDAC cases compared to controls |
| Guo et al. (43) | Serum | LC-MS | 59 unresectable and 63 resectable PDAC cases | >1.5 fold-change increase in indoleacrylic acid, indole-3-propanoate, and TUDCA in unresectable PDAC cases |
| Huang et al. (47) | Serum | LC-MS/MS | 129 prediagnostic sera collected within 12.5 years of a PDAC diagnosis and 258 control sera from participants in the Shanghani Cohort Study; 58 prediagnostic sera collected within 6.8 years of a PDAC diagnosis and 104 control sera from participants in the Singapore Chinese Health Study | In the Shanghai Cohort, compared to the lower quartile reference group, individuals with circulating TMAO in the highest quartile had a 2.81 (95% CI, 1.37–5.76) higher odds of developing PDAC. In the Singapore Cohort, the odds ratio for TMAO was 1.42 (95% CI, 0.50–4.04) |
| Morgell et al. (13) | Plasma | LC-MS | 10 IPMN/PDAC cases, 10 patients with HG IPMN, 20 patients with LG IPMN | Elevated levels of TMAO in patients with HG IPMN or IPMN/PDAC. TMAO yielded an AUC of 0.82 (95% CI, 0.65–0.94) for distinguishing PDAC from noncancerous controls |
| Michálková et al. (45) | Plasma | NMR | 41 PDAC cases, 31 patients with type 2 diabetes, 59 patients with RODM, and 28 healthy controls | Short-chain fatty propionic acid distinguished PDAC cases from patients with long-term type 2 diabetes (AUC: 0.70) |
| Michálková et al. (46) | Plasma | NMR | 88 PDAC cases, 32 patients with type 2 diabetes, 59 patients with RODM, and 28 healthy controls | The short-chain fatty acid propoanoate was identified to be elevated in plasma of PDAC cases compared to either healthy controls or patients with type 2 diabetes. |
| He et al. (39) | Serum | LC-MS | 30 PDAC cases and 30 control NOD patients | TUDCA yielded an AUC of 0.826 (95% CI, 0.720–0.949) for detecting PDAC among NOD patients |
| Fahrmann et al. (54) | Plasma | LC-MS | Cohort No. 1: 20 PDAC cases, 10 patients with CP, and 70 healthy controls; Cohort No. 2: 9 IPMN/PDAC cases and 51 patients with benign IPMN; validation set: 39 early-stage PDAC cases and 82 matched healthy controls | A 5-marker metabolite panel consisting of diacetylspermine, acetylspermidine, 2 lysophosphatidylcholines, and a microbial-derived indole derivative yielded an AUC of 0.892 (95% CI, 0.828–0.956) for detection of early-stage PDAC the an independent validation set |
| Irajizad et al. (55) | Serum | LC-MS | 172 prediagnostic sera collected within 5 years of a PDAC diagnosis and 863 non-case controls from the PLCO cohort | A 3-marker metabolite panel (TMAO, indoleacrylic acid, and another indole derivative) yielded an AUC of 0.64 (95% CI, 0.53–0.76) for identifying individuals who went on to receive a PDAC diagnosis within 5 years of blood draw |
| Mayerle et al. (44) | Plasma/serum | LC-MS/MS; GC-MS | 271 PDAC cases, 282 patients with CP, 100 patients with liver cirrhosis, 261 healthy controls or individuals with nonpancreatic diseases | Indoxy sulfate was elevated in plasma of patients with CP |
| Study | Specimen type | Platform | No. of specimens | Reported findings |
|---|---|---|---|---|
| Löser et al. (37) | Serum | LC-MS/MS | 20 PDAC cases, 30 healthy controls, 40 patients with benign GI disease | Elevated cadaverine levels (nM (SE)) in serum of PDAC cases (0.76 (0.09)) compared to healthy controls (0.32 (0.07)) or patients with benign disease (0.47 (0.05)) |
| Urine | Elevated cadaverine levels (nmol/mg of creatine (SE)) in urine of PDAC cases (10.7 (2.31)) and patients with benign disease (12.18 (1.8)) compared to healthy controls (3.4 (0.5)) | |||
| Tissue | 15 resected PDAC tissues | Elevated cadaverine levels in carcinoma compared to unaffected pancreas areas (20.9 (3.6) vs 7.2 (2.4 nmol/mg of DNA)) | ||
| Nissinen et al. (38) | Urine | LC-MS/MS | 68 PDAC cases, 7 cases with PLP, 36 AP cases, 18 CP cases, and 53 healthy controls | Acetylated cadaverine (median umol/g creatinine) was elevated in urine of patients with PDAC/PLP (1.4), AP (2.5), or CP (2.9) compared to healthy controls (0.54). |
| Luo et al. (40) | Plasma | LC-MS | 60 PDAC cases and 60 healthy controls | 4.7- and 4.2-fold change increase in TUDCA and TDCA in PDAC cases compared to controls (P < 0.001), respectively. |
| Zhao et al. (41) | Serum | LC-MS; LC-MS/MS; GC-MS | 80 PDAC cases, 36 patients with benign disease, and 48 healthy controls | 1.45-fold change increase in indoxyl sulfate (P: 0.011) in PDAC cases compared to controls |
| Cao et al. (42) | Serum | LC-MS | 28 stage I PDAC cases and 62 healthy controls | 5.76-fold increase in indole-3-propanoate (P: 0.029) in PDAC cases compared to controls |
| Guo et al. (43) | Serum | LC-MS | 59 unresectable and 63 resectable PDAC cases | >1.5 fold-change increase in indoleacrylic acid, indole-3-propanoate, and TUDCA in unresectable PDAC cases |
| Huang et al. (47) | Serum | LC-MS/MS | 129 prediagnostic sera collected within 12.5 years of a PDAC diagnosis and 258 control sera from participants in the Shanghani Cohort Study; 58 prediagnostic sera collected within 6.8 years of a PDAC diagnosis and 104 control sera from participants in the Singapore Chinese Health Study | In the Shanghai Cohort, compared to the lower quartile reference group, individuals with circulating TMAO in the highest quartile had a 2.81 (95% CI, 1.37–5.76) higher odds of developing PDAC. In the Singapore Cohort, the odds ratio for TMAO was 1.42 (95% CI, 0.50–4.04) |
| Morgell et al. (13) | Plasma | LC-MS | 10 IPMN/PDAC cases, 10 patients with HG IPMN, 20 patients with LG IPMN | Elevated levels of TMAO in patients with HG IPMN or IPMN/PDAC. TMAO yielded an AUC of 0.82 (95% CI, 0.65–0.94) for distinguishing PDAC from noncancerous controls |
| Michálková et al. (45) | Plasma | NMR | 41 PDAC cases, 31 patients with type 2 diabetes, 59 patients with RODM, and 28 healthy controls | Short-chain fatty propionic acid distinguished PDAC cases from patients with long-term type 2 diabetes (AUC: 0.70) |
| Michálková et al. (46) | Plasma | NMR | 88 PDAC cases, 32 patients with type 2 diabetes, 59 patients with RODM, and 28 healthy controls | The short-chain fatty acid propoanoate was identified to be elevated in plasma of PDAC cases compared to either healthy controls or patients with type 2 diabetes. |
| He et al. (39) | Serum | LC-MS | 30 PDAC cases and 30 control NOD patients | TUDCA yielded an AUC of 0.826 (95% CI, 0.720–0.949) for detecting PDAC among NOD patients |
| Fahrmann et al. (54) | Plasma | LC-MS | Cohort No. 1: 20 PDAC cases, 10 patients with CP, and 70 healthy controls; Cohort No. 2: 9 IPMN/PDAC cases and 51 patients with benign IPMN; validation set: 39 early-stage PDAC cases and 82 matched healthy controls | A 5-marker metabolite panel consisting of diacetylspermine, acetylspermidine, 2 lysophosphatidylcholines, and a microbial-derived indole derivative yielded an AUC of 0.892 (95% CI, 0.828–0.956) for detection of early-stage PDAC the an independent validation set |
| Irajizad et al. (55) | Serum | LC-MS | 172 prediagnostic sera collected within 5 years of a PDAC diagnosis and 863 non-case controls from the PLCO cohort | A 3-marker metabolite panel (TMAO, indoleacrylic acid, and another indole derivative) yielded an AUC of 0.64 (95% CI, 0.53–0.76) for identifying individuals who went on to receive a PDAC diagnosis within 5 years of blood draw |
| Mayerle et al. (44) | Plasma/serum | LC-MS/MS; GC-MS | 271 PDAC cases, 282 patients with CP, 100 patients with liver cirrhosis, 261 healthy controls or individuals with nonpancreatic diseases | Indoxy sulfate was elevated in plasma of patients with CP |
aAbbreviations: AP, acute pancreatitis; CP, chronic pancreatitis; GC, gas chromatography; GI, gastrointestinal; HG, high-grade; IPMN, intraductal papillary mucinous neoplasms; LC, liquid chromatography; LG, low-grade; MS, mass spectrometry; MS/MS, tandem mass spectrometry; NOD, new onset diabetes; PDAC, pancreatic ductal adenocarcinoma; PLCO, Prostate, Lung, Colorectal, and Ovarian; PLP, premalignant lesions of the pancreas; RODM, recently diagnosed diabetes mellitus; TDCA, taurodeoxycholate; TMAO, trimethylamine N-oxide; TUDCA, tauroursodeoxycholic acid.
Association between microbial-derived metabolite levels in biofluids and pancreatic cancer.a
| Study | Specimen type | Platform | No. of specimens | Reported findings |
|---|---|---|---|---|
| Löser et al. (37) | Serum | LC-MS/MS | 20 PDAC cases, 30 healthy controls, 40 patients with benign GI disease | Elevated cadaverine levels (nM (SE)) in serum of PDAC cases (0.76 (0.09)) compared to healthy controls (0.32 (0.07)) or patients with benign disease (0.47 (0.05)) |
| Urine | Elevated cadaverine levels (nmol/mg of creatine (SE)) in urine of PDAC cases (10.7 (2.31)) and patients with benign disease (12.18 (1.8)) compared to healthy controls (3.4 (0.5)) | |||
| Tissue | 15 resected PDAC tissues | Elevated cadaverine levels in carcinoma compared to unaffected pancreas areas (20.9 (3.6) vs 7.2 (2.4 nmol/mg of DNA)) | ||
| Nissinen et al. (38) | Urine | LC-MS/MS | 68 PDAC cases, 7 cases with PLP, 36 AP cases, 18 CP cases, and 53 healthy controls | Acetylated cadaverine (median umol/g creatinine) was elevated in urine of patients with PDAC/PLP (1.4), AP (2.5), or CP (2.9) compared to healthy controls (0.54). |
| Luo et al. (40) | Plasma | LC-MS | 60 PDAC cases and 60 healthy controls | 4.7- and 4.2-fold change increase in TUDCA and TDCA in PDAC cases compared to controls (P < 0.001), respectively. |
| Zhao et al. (41) | Serum | LC-MS; LC-MS/MS; GC-MS | 80 PDAC cases, 36 patients with benign disease, and 48 healthy controls | 1.45-fold change increase in indoxyl sulfate (P: 0.011) in PDAC cases compared to controls |
| Cao et al. (42) | Serum | LC-MS | 28 stage I PDAC cases and 62 healthy controls | 5.76-fold increase in indole-3-propanoate (P: 0.029) in PDAC cases compared to controls |
| Guo et al. (43) | Serum | LC-MS | 59 unresectable and 63 resectable PDAC cases | >1.5 fold-change increase in indoleacrylic acid, indole-3-propanoate, and TUDCA in unresectable PDAC cases |
| Huang et al. (47) | Serum | LC-MS/MS | 129 prediagnostic sera collected within 12.5 years of a PDAC diagnosis and 258 control sera from participants in the Shanghani Cohort Study; 58 prediagnostic sera collected within 6.8 years of a PDAC diagnosis and 104 control sera from participants in the Singapore Chinese Health Study | In the Shanghai Cohort, compared to the lower quartile reference group, individuals with circulating TMAO in the highest quartile had a 2.81 (95% CI, 1.37–5.76) higher odds of developing PDAC. In the Singapore Cohort, the odds ratio for TMAO was 1.42 (95% CI, 0.50–4.04) |
| Morgell et al. (13) | Plasma | LC-MS | 10 IPMN/PDAC cases, 10 patients with HG IPMN, 20 patients with LG IPMN | Elevated levels of TMAO in patients with HG IPMN or IPMN/PDAC. TMAO yielded an AUC of 0.82 (95% CI, 0.65–0.94) for distinguishing PDAC from noncancerous controls |
| Michálková et al. (45) | Plasma | NMR | 41 PDAC cases, 31 patients with type 2 diabetes, 59 patients with RODM, and 28 healthy controls | Short-chain fatty propionic acid distinguished PDAC cases from patients with long-term type 2 diabetes (AUC: 0.70) |
| Michálková et al. (46) | Plasma | NMR | 88 PDAC cases, 32 patients with type 2 diabetes, 59 patients with RODM, and 28 healthy controls | The short-chain fatty acid propoanoate was identified to be elevated in plasma of PDAC cases compared to either healthy controls or patients with type 2 diabetes. |
| He et al. (39) | Serum | LC-MS | 30 PDAC cases and 30 control NOD patients | TUDCA yielded an AUC of 0.826 (95% CI, 0.720–0.949) for detecting PDAC among NOD patients |
| Fahrmann et al. (54) | Plasma | LC-MS | Cohort No. 1: 20 PDAC cases, 10 patients with CP, and 70 healthy controls; Cohort No. 2: 9 IPMN/PDAC cases and 51 patients with benign IPMN; validation set: 39 early-stage PDAC cases and 82 matched healthy controls | A 5-marker metabolite panel consisting of diacetylspermine, acetylspermidine, 2 lysophosphatidylcholines, and a microbial-derived indole derivative yielded an AUC of 0.892 (95% CI, 0.828–0.956) for detection of early-stage PDAC the an independent validation set |
| Irajizad et al. (55) | Serum | LC-MS | 172 prediagnostic sera collected within 5 years of a PDAC diagnosis and 863 non-case controls from the PLCO cohort | A 3-marker metabolite panel (TMAO, indoleacrylic acid, and another indole derivative) yielded an AUC of 0.64 (95% CI, 0.53–0.76) for identifying individuals who went on to receive a PDAC diagnosis within 5 years of blood draw |
| Mayerle et al. (44) | Plasma/serum | LC-MS/MS; GC-MS | 271 PDAC cases, 282 patients with CP, 100 patients with liver cirrhosis, 261 healthy controls or individuals with nonpancreatic diseases | Indoxy sulfate was elevated in plasma of patients with CP |
| Study | Specimen type | Platform | No. of specimens | Reported findings |
|---|---|---|---|---|
| Löser et al. (37) | Serum | LC-MS/MS | 20 PDAC cases, 30 healthy controls, 40 patients with benign GI disease | Elevated cadaverine levels (nM (SE)) in serum of PDAC cases (0.76 (0.09)) compared to healthy controls (0.32 (0.07)) or patients with benign disease (0.47 (0.05)) |
| Urine | Elevated cadaverine levels (nmol/mg of creatine (SE)) in urine of PDAC cases (10.7 (2.31)) and patients with benign disease (12.18 (1.8)) compared to healthy controls (3.4 (0.5)) | |||
| Tissue | 15 resected PDAC tissues | Elevated cadaverine levels in carcinoma compared to unaffected pancreas areas (20.9 (3.6) vs 7.2 (2.4 nmol/mg of DNA)) | ||
| Nissinen et al. (38) | Urine | LC-MS/MS | 68 PDAC cases, 7 cases with PLP, 36 AP cases, 18 CP cases, and 53 healthy controls | Acetylated cadaverine (median umol/g creatinine) was elevated in urine of patients with PDAC/PLP (1.4), AP (2.5), or CP (2.9) compared to healthy controls (0.54). |
| Luo et al. (40) | Plasma | LC-MS | 60 PDAC cases and 60 healthy controls | 4.7- and 4.2-fold change increase in TUDCA and TDCA in PDAC cases compared to controls (P < 0.001), respectively. |
| Zhao et al. (41) | Serum | LC-MS; LC-MS/MS; GC-MS | 80 PDAC cases, 36 patients with benign disease, and 48 healthy controls | 1.45-fold change increase in indoxyl sulfate (P: 0.011) in PDAC cases compared to controls |
| Cao et al. (42) | Serum | LC-MS | 28 stage I PDAC cases and 62 healthy controls | 5.76-fold increase in indole-3-propanoate (P: 0.029) in PDAC cases compared to controls |
| Guo et al. (43) | Serum | LC-MS | 59 unresectable and 63 resectable PDAC cases | >1.5 fold-change increase in indoleacrylic acid, indole-3-propanoate, and TUDCA in unresectable PDAC cases |
| Huang et al. (47) | Serum | LC-MS/MS | 129 prediagnostic sera collected within 12.5 years of a PDAC diagnosis and 258 control sera from participants in the Shanghani Cohort Study; 58 prediagnostic sera collected within 6.8 years of a PDAC diagnosis and 104 control sera from participants in the Singapore Chinese Health Study | In the Shanghai Cohort, compared to the lower quartile reference group, individuals with circulating TMAO in the highest quartile had a 2.81 (95% CI, 1.37–5.76) higher odds of developing PDAC. In the Singapore Cohort, the odds ratio for TMAO was 1.42 (95% CI, 0.50–4.04) |
| Morgell et al. (13) | Plasma | LC-MS | 10 IPMN/PDAC cases, 10 patients with HG IPMN, 20 patients with LG IPMN | Elevated levels of TMAO in patients with HG IPMN or IPMN/PDAC. TMAO yielded an AUC of 0.82 (95% CI, 0.65–0.94) for distinguishing PDAC from noncancerous controls |
| Michálková et al. (45) | Plasma | NMR | 41 PDAC cases, 31 patients with type 2 diabetes, 59 patients with RODM, and 28 healthy controls | Short-chain fatty propionic acid distinguished PDAC cases from patients with long-term type 2 diabetes (AUC: 0.70) |
| Michálková et al. (46) | Plasma | NMR | 88 PDAC cases, 32 patients with type 2 diabetes, 59 patients with RODM, and 28 healthy controls | The short-chain fatty acid propoanoate was identified to be elevated in plasma of PDAC cases compared to either healthy controls or patients with type 2 diabetes. |
| He et al. (39) | Serum | LC-MS | 30 PDAC cases and 30 control NOD patients | TUDCA yielded an AUC of 0.826 (95% CI, 0.720–0.949) for detecting PDAC among NOD patients |
| Fahrmann et al. (54) | Plasma | LC-MS | Cohort No. 1: 20 PDAC cases, 10 patients with CP, and 70 healthy controls; Cohort No. 2: 9 IPMN/PDAC cases and 51 patients with benign IPMN; validation set: 39 early-stage PDAC cases and 82 matched healthy controls | A 5-marker metabolite panel consisting of diacetylspermine, acetylspermidine, 2 lysophosphatidylcholines, and a microbial-derived indole derivative yielded an AUC of 0.892 (95% CI, 0.828–0.956) for detection of early-stage PDAC the an independent validation set |
| Irajizad et al. (55) | Serum | LC-MS | 172 prediagnostic sera collected within 5 years of a PDAC diagnosis and 863 non-case controls from the PLCO cohort | A 3-marker metabolite panel (TMAO, indoleacrylic acid, and another indole derivative) yielded an AUC of 0.64 (95% CI, 0.53–0.76) for identifying individuals who went on to receive a PDAC diagnosis within 5 years of blood draw |
| Mayerle et al. (44) | Plasma/serum | LC-MS/MS; GC-MS | 271 PDAC cases, 282 patients with CP, 100 patients with liver cirrhosis, 261 healthy controls or individuals with nonpancreatic diseases | Indoxy sulfate was elevated in plasma of patients with CP |
aAbbreviations: AP, acute pancreatitis; CP, chronic pancreatitis; GC, gas chromatography; GI, gastrointestinal; HG, high-grade; IPMN, intraductal papillary mucinous neoplasms; LC, liquid chromatography; LG, low-grade; MS, mass spectrometry; MS/MS, tandem mass spectrometry; NOD, new onset diabetes; PDAC, pancreatic ductal adenocarcinoma; PLCO, Prostate, Lung, Colorectal, and Ovarian; PLP, premalignant lesions of the pancreas; RODM, recently diagnosed diabetes mellitus; TDCA, taurodeoxycholate; TMAO, trimethylamine N-oxide; TUDCA, tauroursodeoxycholic acid.
In this regard, early studies by Löser and colleagues demonstrated CAD, a polyamine produced from the decarboxylation of lysine via bacterial enzymes, constitutive lysine decarboxylase (LdcC) and lysine decarboxylase 1 (CadA), to be significantly elevated in serum of PDAC cases compared to patients with benign pancreas disease or healthy controls (37). Urinary CAD was also found to be elevated in patients with PDAC or with benign pancreas disease compared to healthy controls. Evaluation of CAD levels in histologically unaffected pancreas tissue and tumor areas from patients undergoing surgical resection of PDAC demonstrated CAD to be highly elevated in cancerous tissue (37). Another study found that urinary levels of acetylated CAD were elevated in patients with pre-malignant and malignant lesions in the pancreas compared to healthy controls (38).
Serum metabolomic analyses of patients with NOD, also referred to as type-3 diabetes mellitus (DM), revealed the secondary bile acid tauroursodeoxycholic acid (TUDCA) to be highly elevated in NOD patients with PDAC compared to non-PDAC NOD controls, with a corresponding Area Under the Receiver Operating Characteristic Curve (AUC) estimate of 0.826 (95% CI, 0.720–0.949) (39). These findings are consistent with another study that similarly found TUDCA as well as taurodeoxycholate (TDCA) to be elevated in plasma of PDAC cases (40). In addition to secondary bile acids, increased circulating levels of indoles, including indole-3-propionate, IAA, and IS, as well as short-chain fatty acids have also been associated with presence of PDAC and other inflammatory conditions of the pancreas (41–46).
A study of serum methionine-related metabolites identified elevated levels of TMAO to be associated with pancreatic cancer risk. Specifically, individuals in the highest quartile of circulating TMAO had 2.81-fold (95% CI, 1.37–5.76) higher odds of developing PDAC compared to the lowest quartile reference group (47). A similar trend was observed in an independent case-control specimen set (47). Elevated plasma levels of TMAO have also been associated with risk of gastric (48), colorectal (49), and liver cancers (50), suggesting broader relevance for risk assessment of gastrointestinal cancers. However, another study did not find a statistically significant association between elevated circulating levels of TMAO and colorectal cancer (CRC), which may be attributed to differences in diet amongst study subjects (51). Specifically, TMAO is generated from the metabolism of choline and carnitine by intestinal microbiota. Animal and animal products contain high amounts of choline and carnitine. Studies by Koeth and colleagues demonstrated that omnivorous human subjects produced more TMAO following ingestion of carnitine than did vegans or vegetarians (52). The Study With Appetizing Plantfood-Meat Eating Alternatives Trial (SWAP-MEAT) similarly demonstrated that patients who switched from animal products to plant-based products significantly reduced circulating levels of TMAO (53).
Previously, our group applied a training and testing approach to develop and validate a plasma metabolite panel for early detection of PDAC. A 5-marker metabolite panel (5MP) consisting of diacetylspermine, acetylspermidine, 2 lysophosphatidylcholines, and a microbial-derived indole-derivative was developed for prediction of PDAC. In the blinded validation set, the 5MP yielded an AUC of 0.892 (95% CI, 0.828–0.956) for detection of early-stage PDAC (54). We have since also evaluated the association between circulating microbial-derived metabolites and risk of PDAC in the Prostate, Lung, Colorectal, and Ovarian (PLCO) study cohort. A panel consisting of 3 microbial-derived metabolites, TMAO, IAA, and another indole derivative, was developed to predict 5-year risk of PDAC. The panel yielded an AUC of 0.64 (95% CI, 0.53–0.76) for identifying individuals who went on to receive a PDAC diagnosis within 5 years of blood draw, and an AUC of 0.70 (95% CI, 0.50–0.90) when considering cases diagnosed within 2 to 5 years of blood draw. Notably, inclusion of 5 host-derived metabolites combined with the 3-marker microbial panel yielded further improvements in 5-year risk prediction of PDAC, with an AUC of 0.79 (95% CI, 0.71–0.88) (55).
Intersection between the Microbiome and Microbe-Derived Metabolites with Response to Therapy and Survivorship of PDAC Patients
Microbial-metabolites and chemotherapy
In addition to playing a role in disease progression, recent studies have also implicated the microbiome as a key determinant of the efficacy of chemotherapy. This is exemplified by the finding that bacteria can metabolize gemcitabine (2′,2′-difluorodeoxycytidine), a frequently used first-line chemotherapeutic drug for PDAC, to the inactive derivative 2′,2′-difluorodeoxyuridine via bacterial cytidine deaminase (56). Conversely, therapeutic agents may also directly impact the microbiome and associated metabolism (57).
While the relationship between the microbiome and response to treatment has been largely explored through 16S and metagenomic analyses (58), recent evidence supports the idea that microbial-derived metabolites can also directly influence response to treatment (Fig. 3). To this end, Guenther and colleagues evaluated whether lipopolysaccharides (LPS) predict response to adjuvant gemcitabine (aGC) in PDAC. LPS levels were detected in tumors from 86 (22.9%) of 376 patients. Among patients receiving aGC, LPS positivity was associated with worse disease-free survival (8.3 vs 13.7 months; log-rank P: 0.002) and poor overall survival (21.7 vs 28.5 months, log-rank P: 0.001) (59). Notably, while PDAC patients have been found to exhibit a higher abundance of LPS-producing bacteria, short-chain fatty acid (SCFA)-producing bacteria tend to be lower, with levels of propionate and butyrate being reduced in the fecal content of PDAC patients compared to patients with autoimmune pancreatitis or healthy individuals (60, 61), suggesting a protective effect of SCFAs. To this end, co-treatment of PDAC cell lines with the SCFA butyrate and gemcitabine resulted in superior cancer killing effects compared to either treatment alone. In a xenograft model of PDAC, administration of butyrate reduced stromatogenesis, improved intestinal integrity, and increased representation of SCFA-producing bacteria (62). Treatment of PDAC-tumor bearing mice with probiotics led to an enrichment in butyrate-producing bacteria, including Eubacteriaceae, Ruthenibacterium, Faecalicatena, Pseudobutyrivibrio, and Roseburia. While combination treatment of probiotic with gemcitabine did not yield appreciable differences in antitumor effect compared to gemcitabine treatment alone, probiotics reduced gemcitabine-associated toxicity, such as restoration of red blood cell counts and platelet numbers, and increased epithelial regeneration in intestinal mucosa (63).
Microbial-metabolites and response to chemo- and immunotherapy in the context of PDAC. Loss of bacteria that produce metabolites, e.g., SCFAs butyrate, propionate, and pentanoate, with anticancer effects, coupled with increases in pathogenic bacteria that release immunosuppressive metabolites, e.g., LPS, or metabolites, such as queusoine, that can elicit an antioxidant response in cancer cells to protect against oxidative stress enables tumor progression. In the context of treatment, bacteria can also directly metabolize chemotherapeutic agents, such as gemcitabine, to inactive forms. Indole-3-lactate, derived via dietary changes or FMT, can be oxidized by neutrophil-derived myeloperoxidase (MPO) resulting in the generation of 3-methyl-2-oxindole (MOI) that, in combination with gemcitabine, induces reactive oxygen species (ROS) buildup in cancer cells and subsequent cancer cell death. Supplementation with probiotics to increase butyrate-producing bacteria can reduce gemcitabine-associated toxicity. Microbe-derived trimethylamine (TMA) and its subsequent conversion to trimethylamine N-oxide (TMAO) induces an immunostimulatory phenotype in tumor-associated macrophages that potentiates effector T cell-mediated cancer killing in a type-1 interferon (IFN)-dependent manner (see Fig. 2), which can be enhanced when used in combination with immune checkpoint blockade inhibitors. Created with BioRender.com. Abbreviations: CAR, chimeric antigen receptor; CD25, cluster of differentiation 25; CTL, cytotoxic T cells; CutC, copper(I)-thiophene-2-carboxylate; FMT, fecal microbiota transplant; IFNγ, interferon-gamma; LPS, lipopolysaccharide; MOI, 3-methyl-2-oxindole; MPO, myeloperoxidase; PD-1, programmed death protein 1; PDAC, pancreatic ductal adenocarcinoma; PD-L1, programmed cell death ligand 1; PRDX1, peroxiredoxin-1; ROS, reactive oxygen species; SCFA, short-chain fatty acid; Tex, exhausted T cells; TIM-3, T-cell immunoglobulin and mucin-domain containing-3; TLR4, toll-like receptor 4; TMA, trimethylamine; TMAO, trimethylamine-N-oxide; TNFα, tumor necrosis factor-alfa.
Tintelnot and colleagues found that indole-3-acetate (3-IAA) was elevated in plasma of patients with PDAC who responded to gemcitabine and nab-paclitaxel (GnP) or FOLFIRINOX treatment (5). Fecal microbiota transplantation, short-term dietary manipulation of tryptophan, and oral 3-IAA administration was demonstrated to increase the efficacy of chemotherapy in humanized gnotobiotic mouse models of PDAC. Mechanistically, the authors demonstrated that 3-IAA can be oxidized by neutrophil-derived myeloperoxidase (MPO) resulting in buildup of toxic metabolic byproducts such MOI. Increased MOI, in combination with chemotherapy, resulted in the accumulation of reactive oxygen species (ROS), reduction in ROS-degrading enzymes glutathione peroxidase 3 and glutathione peroxidase 7, and subsequent cell death in PDAC cells (5). Another study similarly found 3-IAA levels tended be elevated in PDAC patients who responded (stable disease + partial response) to gemcitabine + abraxane compared to those that did not (fold change 1.72; variable importance in the project [VIP] score: 1.61; P: 0.054) (64).
C57BL6 mice fed a high-fat diet exhibited enrichment of queuosine-producing bacteria compared to mice on a lean (control) diet. PDAC tumors (KPC001) implanted into mice on a high-fat diet were less sensitive to gemcitabine/paclitaxel therapy compared to mice on control diet. Treatment of MIA-PaCa2 and S2VP10 PDAC cells with a queuosine-precursor increased cancer cell proliferation and protected against paclitaxel-induced cell death. Mechanistically, queuosine induced expression of peroxiredoxin 1 (PRDX1), thus attenuating therapy-induced oxidative stress (65).
Microbial-metabolites and immunotherapy
In the context of immunotherapy, recent studies have demonstrated that microbe-derived metabolites, including TMAO (6), LPS (8), and SCFAs (7), can directly impact immune cell function and enhance or attenuate efficacy of immunotherapies (Fig. 3).
To this end, combination treatment of TMAO plus immune checkpoint blockade inhibitors, anti-programmed death protein 1(PD-1) or anti-T-cell immunoglobulin and mucin domain-containing-3 (TIM-3), significantly reduced tumor burden and improved survival of PDAC-tumor bearing mice compared to either treatment modality alone. Moreover, bacteria containing copper(I)-thiophene-2-carboxylate (CutC), an enzyme that generates the TMAO precursor TMA, correlated with long-term survival in patients with PDAC (6).
Gut-derived LPS has been shown to deposit into PDAC tumors, with extent of accumulation being associated with increased intestinal permeability. Increased tumoral LPS levels correlated with marked increases in tumor-infiltrating CD3+ CD8+ PD-1+ TIM-3+ T-lymphocytes. These changes were met with a concomitant increase in programmed cell death ligand 1 (PD-L1), suggesting that LPS may promote tumor immune suppression. Mechanistically, LPS upregulates PD-L1 through the toll-like receptor 4 (TLR4)/MYD88 innate immune signal transduction adaptor (MyD88)/protein kinase B (AKT)/NF-κB signaling pathway. Co-treatment of PDAC-bearing mice with LPS and anti-PD-L1 resulted in significantly attenuated tumor growth compared to anti-PD-L1 treatment alone (8).
In contrast to LPS, SCFAs may potentiate efficacy of adoptive cell therapy. Using an ovalbumin (OVA)-expressing Panc02 PDAC xenograft model, Luu and colleagues demonstrated that administration of antigen-specific cytotoxic T cells (CTLs) pretreated with the SCFA pentanoate led to a marked attenuation of tumor growth compared to mice treated with control CTLs. Functionally, pentanoate-treated CTLs had higher IFNγ production within the draining lymph nodes and spleen of recipient mice. Moreover, pentanoate-treated CD8+ T cells had higher proliferation capacity compared to control CTLs, which was linked to pentanoate-induced CD25 upregulation and prolonged interleukin 2 (IL-2)-mediated signal transducer and activator of transcription 5 (STAT5) activation. Adoptive transfer of pentanoate pre-treated receptor tyrosine kinase-like orphan receptor 1 (ROR1)-chimeric antigen receptor (CAR) T cells into ROR1-expressing Panc02 tumor-bearing mice resulted an elevated frequency of IFNγ+TNFα+ CAR T cells in tumors compared to the respective control, which was met with a robust anticancer response. Despite enhanced efficacy of adoptive T cell-based therapies, in vivo administration of pentanoate did not enhance antitumor immunity, nor did it enhance anti-PD-1 therapy (7).
Perspectives
Compelling evidence highlights the microbiome-derived metabolome as an abundant source of biomarkers for individualized risk assessment of pancreatic cancer and for predicting response to treatment and associated clinical outcomes. We have demonstrated discrete microbial-derived metabolite signatures in cystic fluid that associate with malignant progression of IPMN. Our prior studies have also led to the identification of novel blood-based metabolite signatures, based in part on microbial-metabolites, for 5-year risk of PDAC to inform the need for surveillance and screening for earlier detection of disease where patients are more likely to benefit from curative-intent treatment. Moreover, such tests may also identify individuals who may best benefit from cancer preventive strategies, such as a vaccine.
There are several notable considerations with respect to the current state of the field. Metabolomics analyses have generally been focused on measurement of host-derived metabolites. Moreover, biochemical conversions of metabolites (e.g., tryptophan) and other exogenous compounds by bacteria lead to complex derivatives that are not readily available in most databases, which precludes broader assessment of microbe-derived metabolites in biofluids. It is also important to consider sources of variation, such as diet (52, 53), that may independently influence the microbiome and associated metabolome regardless of disease status. While the abovementioned studies support utility of circulating microbial-derived metabolites for predicting risk of developing or harboring PDAC, several have also been shown to be associated with increased risk of other cancer types, particularly gastrointestinal (GI) cancers, which is an important consideration for clinical management of high-risk individuals. Consequently, utility of microbial-derived metabolite signatures for risk assessment should ideally be coupled with additional data (e.g., patient characteristics) and other pertinent cancer detection tests (e.g., imaging modality or other biomarkers). The stability and reliability of microbial-metabolite signatures for cancer risk prediction also warrant further investigation.
It is important to note that there is also an active debate regarding the stringency and significance of the cancer-associated microbiome, which has been mainly attributed to 3 aspects: (a) contamination in microbiome studies; (b) microbiome signal amplification during data processing; and (c) distinguishing association from causality. Contamination may occur due to storage conditions, and sample preparation and processing, as well as from contaminants derived from molecular biology reagent kits. For instance, Eisenhofer and colleagues reported that contaminant DNA and cross-contamination are frequent in microbiome studies and that over 60 common contaminant taxa have been identified in DNA extraction blank controls and no-template controls across multiple studies. To mitigate the impacts of contaminant DNA and reduce “noise”, authors developed a Minimum Standards Checklist for Performing/Reviewing Low Microbial Biomass Microbiome Studies (“RIDE”) (66). In our view, similar criteria must be implemented for studies evaluating the microbiome-metabolome. In addition to contaminants, false-positive findings may also be attributed to computational artifacts. Recently, Gihawi and colleagues re-analyzed data from a large-scale study (67) that reported strong correlations between microbial organisms and 33 different cancer types. Concern was raised regarding errors in the genome database and the associated computational methods that may have led to a considerable number of false-positive findings as well as errors in transformation of the raw data that created artificial signatures for distinguishing cancer types (68). Chrisman and colleagues also reported that Y-chromosome fragments not on the human reference genome are commonly misassigned to bacterial reference genomes (69). The above studies highlight a necessity for rigorously standardized protocols for identifying and removing contamination in whole-genome sequencing as well as metagenomics studies. In the context of the microbial-metabolome, stress should be placed on validation and authentication of microbial-metabolites in addition to mitigating contaminants that may arise from sample collection, storage, and processing. Use of external databases may also lead to improved discrimination between host- and microbial-derived metabolites. In this regard, Shuo Han and colleagues created the Metabolomics Data Explorer database. The database reports the metabolic profiles of 178 gut microorganism strains established in diverse biological fluids from gnotobiotic and conventionally colonized mice that were traced back to the metabolomic profiles of cultured bacteria (14).
The underlying biological implications of the microbiome and its association with cancer development and progression also warrant further investigation to delineate association from causation. In this regard, emerging evidence indicates that microbial-derived metabolites are more than bystanders, with capacity to exert diverse effector functions on cancer cells as well as other cell types that either promote or attenuate tumor development and that influence anticancer efficacy of chemotherapy and immunotherapy as described in the above sections. This provides an additional opportunity for novel treatment strategies, including antibiotic or probiotic treatment, as well as dietary interventions that may potentiate therapeutic efficacy and/or reduce treatment-associated toxicity. Nevertheless, most studies to date have been limited to analyses of a single metabolite or subset of metabolites (e.g., LPS, TMAO, indoles, or SCFAs) and their relation to tumor development and progression, which does not consider dynamic interactions of other microbe-derived metabolites and the host response. To fully maximize therapeutic potential, further deconvolution of the complex role(s) of microbe-derived metabolites on tumorigenesis is needed and may be aided by the assessment of microbial metabolites in biofluids or other biological materials as well as the use of innovative model systems and approaches. For instance, Shah and colleagues developed an in vitro microfluidics-based ecosystem called HuMix, which uses a nanoporous membrane to separate microbiome and cancer cells into different compartments, thus mimicking an intact epithelium barrier (70). The separate outflow collecting system enables researchers to investigate the source of metabolites and their direct effector functions (70). Several other co-culture systems have also been described for studying interactions between the microbiome and cancer (71), allowing improved understanding of the microbiome–cancer axis, thus moving away from association to causality. In vivo model systems whereby specific microbes are eliminated via selective antibiotics or reintroduced (e.g., fecal microbiota transplantation [FMT]) may provide additional means for evaluating the direct impact of the microbiota on tumor development and progression. The strength of this approach is exemplified by Riquelme and colleagues who demonstrated that FMT from short-term survivor (STS) and long-term survivor (LTS) groups resulted in pronounced differences in the tumor microbiome and host response in the Kras, p53, and Cre (KPC) mouse model of PDAC. Specifically, FMT from STSs accelerated tumor growth compared to mice with FMT from LTSs (4).
Whether microbial-derived metabolite signatures are a consequence of dysbiosis resulting from cancer or directly influence cancer development and progression is a point of consideration and it is likely that both scenarios are true. For instance, a tumor in the pancreatic head associates with loss of appetite and the inability of the pancreatic gland to functional normally, contributing to malnutrition that can directly impact gut microflora and, therefore, the microbiome-metabolome (72). Yet, studies by our group and those of others have also demonstrated that changes in circulating microbial-derived metabolites, such as TMAO, occur several years prior to a clinical diagnosis of PDAC (47, 55). While this does not rule out the presence of an occult tumor that may impair organ function, it does suggest that early dysbiosis is associated with increased risk of disease rather than a consequence of disease. Mechanistically, TMAO potentiates inflammation that may result in increased susceptibility of disease (55). In contrast, TMAO elicits a more favorable response to immunotherapy and enhances antitumor immunity (6). This paradox is akin to nuclear factor erythroid 2-related factor 2 (NRF2) activation, which has been shown to exert protective roles against carcinogenesis and cancer development, yet enables disease progression once a tumor has been established (73).
In conclusion, the microbiome-metabolome holds considerable promise for risk assessment and prognostication of PDAC and remains a largely untapped area of exploration. Concentrated efforts to fully maximize the potential clinical benefits of the microbiome-metabolome are warranted.
Supplemental Material
Supplemental material is available at Clinical Chemistry online.
Nonstandard Abbreviations
PDAC, pancreatic ductal adenocarcinoma; NOD, new-onset diabetes; IPMN, intraductal papillary mucinous neoplasm; HG, high-grade; TMAO, trimethylamine-N-oxide; CAD, cadaverine; IAA, indoleacryclic acid; TAM, tumor-associated macrophages; IFN, interferon; CD, cluster of differentiation; AhR, aryl hydrocarbon receptor; AUC, area under the curve; LPS, lipopolysaccharides; SCFA, short-chain fatty acid; 3-IAA, indole-3-acetate; PD-L1, programmed cell death ligand 1; CTL, cytotoxic T cell; FMT, fecal microbiota transplantation.
Author Contributions
The corresponding author takes full responsibility that all authors on this publication have met the following required criteria of eligibility for authorship: (a) significant contributions to the conception and design, acquisition of data, or analysis and interpretation of data; (b) drafting or revising the article for intellectual content; (c) final approval of the published article; and (d) agreement to be accountable for all aspects of the article thus ensuring that questions related to the accuracy or integrity of any part of the article are appropriately investigated and resolved. Nobody who qualifies for authorship has been omitted from the list.
Ricardo León-Letelier (Visualization-Supporting, Writing—original draft-Supporting), Rongzhang Dou (Writing—review & editing-Supporting), Jody Vykoukal (Writing—review & editing-Supporting), Michele Yip-Schneider (Resources-Equal, Writing—review & editing-Supporting), Anirban Maitra (Resources-Supporting, Writing—review & editing-Supporting), ehsan irajizad (Writing—review & editing-Supporting), Ranran Wu (Data curation-Supporting, Formal analysis-Supporting, Methodology-Supporting), Jennifer Dennison (Methodology-Supporting, Project administration-Supporting, Writing—review & editing-Supporting), Kim-An Do (Writing—review & editing-Supporting), Jianjun Zhang (Resources-Supporting, Writing—review & editing-Supporting), C. Max Schmidt (Funding acquisition-Lead, Resources-Lead, Writing—review & editing-Supporting), Samir Hanash (Resources-Supporting, Writing—review & editing-Supporting), and Johannes Fahrmann (Conceptualization-Lead, Data curation-Lead, Formal analysis-Lead, Investigation-Lead, Methodology-Lead, Supervision-Lead, Validation-Lead, Visualization-Lead, Writing—original draft-Lead).
Authors’ Disclosures or Potential Conflicts of Interest
Upon manuscript submission, all authors completed the author disclosure form.
Research Funding
Supported in part through the generous philanthropic contributions to The University of Texas MD Anderson Cancer Center Moon Shots Program. M.T. Yip-Schneider, J. Zhang, C.M. Schmidt, S. Hanash, and J.F. Fahrmann are supported by NCI grant U01CA239522. A. Maitra is supported by the Sheikh Khalifa bin Zayed Foundation and the MD Anderson Pancreatic Cancer Moonshot.
Disclosures
A. Maitra is listed as an inventor on a patent that has been licensed by Johns Hopkins University to Thrive Earlier Detection and serves as a consultant for Tezcat Biosciences. S. Hanash is a guest editor for Clinical Chemistry, ADLM.
