Exocrine Pancreatic Function Modulates Plasma Metabolites Through Changes in Gut Microbiota Composition

Abstract

Purpose

Exocrine pancreatic function is critically involved in regulating the gut microbiota composition. At the same time, its impairment acutely affects human metabolism. How these 2 roles are connected is unknown. We studied how the exocrine pancreas contributes to metabolism via modulation of gut microbiota.

Design

Fecal samples were collected in 2226 participants of the population-based Study of Health in Pomerania (SHIP/SHIP-TREND) to determine exocrine pancreatic function (pancreatic elastase enzyme-linked immunosorbent assay) and intestinal microbiota profiles (16S ribosomal ribonucleic acid gene sequencing). Plasma metabolite levels were determined by mass spectrometry.

Results

Exocrine pancreatic function was associated with changes in the abundance of 28 taxa and, simultaneously, with those of 16 plasma metabolites. Mediation pathway analysis revealed that a significant component of how exocrine pancreatic function affects the blood metabolome is mediated via gut microbiota abundance changes, most prominently, circulating serotonin and lysophosphatidylcholines.

Conclusion

These results imply that the effect of exocrine pancreatic function on intestinal microbiota composition alters the availability of microbial-derived metabolites in the blood and thus directly contributes to the host metabolic changes associated with exocrine pancreatic dysfunction.

The plasma metabolome is closely intertwined with a wide range of metabolic and inflammatory disorders, which can strongly impair the quality of life for patients and shorten their life expectancy (1-3). A key determinant of the plasma metabolome and thus metabolic health is the gut microbiota (4). Intestinal microbiota contribute to host metabolism through various mechanisms, specifically via production of small molecules with signaling functions, such as short-chain fatty acids (5) or modulation of drug metabolites (6). Therefore, understanding the mutual dependencies between shifts in the gut microbiome, the plasma metabolome, and other host factors is of high priority to design new strategies for the prevention and treatment of metabolic and other diseases. Yet, comparatively little is known about how disturbances in host factors shape gut microbial metabolism, how this alters the availability of microbial-derived metabolites, and whether these metabolites contribute to the host’s metabolic homeostasis.

Deep phenotyping of population-based cohorts, covering circulating metabolites, gut microbial species, and detailed clinical characteristics now allows for the triangulation between those factors at scale. Previous examples included the triangulation between visceral obesity, gut microbiota, and plasma metabolites in twins (7). We have recently identified that exocrine pancreatic function has a far larger contribution to the gut microbial composition than any other host factor in individuals without known pancreatic disease (8). Exocrine pancreatic function is the master control mechanism for food digestion and nutrition. Its impairment leads to severe metabolic consequences (1). Here we have investigated whether the known role of exocrine pancreatic function in regulating blood metabolite composition depends on its other role of regulating intestinal microbiota composition—in other words, whether and how the intestinal microbiome contributes to the metabolic changes characteristic for exocrine pancreatic insufficiency. We integrated exocrine pancreatic function (fecal pancreatic elastase enzyme-linked immunosorbent assay), gut microbiota composition (16S ribosomal ribonucleic acid [rRNA] gene sequencing of fecal samples), and the plasma metabolome (mass spectrometry) to test whether metabolic consequences associated with impaired exocrine pancreatic function might be mediated through shifts in the gut microbiome among 2226 volunteers from 2 independent population-based studies, Study of Health in Pomerania (SHIP) and SHIP-TREND (Fig. 1) (9).

Figure 1.

Design idea of the present study. Triangles fulfilling all 3 criteria were selected for mediation analysis. Abbreviations: OLS: ordinary linear regression; q: q-value correcting p-values for multiple testing by controlling the false discovery rate.

Materials and Methods

Study population

The Study of Health in Pomerania is a population-based cohort study that consists of the 2 independent cohorts, SHIP (initial recruitment 1997-2001; n = 4308) and SHIP-TREND (initial recruitment 2008-2012, n = 4420) (9). For the current study, data from a second follow-up of the SHIP cohort (SHIP-2: 2008-2012; n = 2333) and the recruitment phase of SHIP-TREND were used. Overlapping data with stool pancreatic elastase measurements, fecal microbiota profiles based on 16S rRNA gene sequencing, and plasma metabolome data were available for 1409 individuals in SHIP-2 and 933 volunteers in SHIP-TREND. After exclusion of all data sets with a history of antibiotic treatment at the time of sample collection, a history of pancreatic disease (acute or chronic pancreatitis, pancreatic cancer, pancreatic enzyme replacement therapy), or with missing covariate data (body mass index [BMI], diabetes mellitus, alcohol consumption, smoking, dietary data, pancreatic disease history) at total of 1319 SHIP-2 and 907 SHIP-TREND data sets remained for analysis. All participants provided written informed consent prior to the inclusion into the study.

Body mass index was calculated as kilograms per square meter. Smoking was estimated as average amount of cigarettes smoked per day. For determination of average alcohol consumption in grams per day, a 30-day recall that captured the amount of different alcoholic beverages (beer, wine, and/or spirits) that had been consumed was evaluated. Diabetes mellitus was assumed in case of a history of diabetes mellitus in combination with current treatment (dietetic, oral medication, or insulin), a hemoglobin A1c ≥6.5%, or a serum glucose ≥11.1 mM. For evaluation of dietary profiles, a food frequency score as described elsewhere was calculated. In brief, food frequency intake data (ordinal scale) from 15 dietary items (meat; sausage; fish; boiled potatoes; pasta; rice; raw vegetables; boiled vegetables; fruits; whole grain, black, or crisp bread; oats and cornflakes; eggs; cake or cookies; sweets; and savory snacks) were assigned with 0 (unfavorable), 1 (normal), or 2 (recommended) points to calculate a score with a theoretical range from 0 (unhealthy diet) to 30 (healthy diet), (10,11).

Microbiota measurements

For determination of intestinal microbiota profiles, 16S rRNA gene sequencing was performed as described before in detail (8). In brief, the study participants collected their fecal samples in their home environment in a tube containing stabilizing ethylenediamine tetra-acetate buffer, which was sent to our laboratory by mail. After isolation of deoxyribonucleic acid, the samples were stored at −20°C and underwent 16S rRNA gene sequencing based on the V1-V2 regions of the 16S gene, which was performed on a MiSeq platform (Illumina, San Diego, CA, USA). MiSeq FastQ files were created with CASAVA 1.8.2 (https://support.illumina.com/sequencing/sequencing_software/casava). For amplicon data processing and assignment of taxonomy, the software package DADA2 (12) was used according to the recommended procedures by the authors (https://benjjneb.github.io/dada2/bigdata.html), adapted to the V1-V2 region. In brief, 5 bases were truncated from the 5’ end of the sequence of both reads. Forward and reverse reads were truncated to a length of 200 and 150 bases, respectively. Reads were excluded if they contained ambiguous bases, expected error rates >2, or originated from PhiX spike-in. Error profiles were deducted based on 1 million reads of the respective sequencing run. After dereplication, error correction, and merging of forward and reverse reads, the removeBimeraDenovo() function (consensus mode) was used to create amplicon sequence variant abundance tables for all samples and to identify and remove chimeric amplicon sequences. Taxonomy was finally assigned using a Bayesian classifier and the Ribosomal Database Project training set version 16. We determined the predicted genomic potential for the of the microbial “palmitoleate biosynthesis I pathway” with PICRUSt2 (13) using the standard workflow as described at https://github.com/picrust/picrust2/wiki/Workflow and amplicon sequence variants from the DADA2 pipeline as input. All samples were normalized to 10 000 16S rRNA-gene read counts for analysis.

Pancreatic elastase measurements

A monospecific pancreatic elastase enzyme-linked immunosorbent assay (BIOSERV Diagnostics GmbH, Germany) was used for determination of pancreatic elastase concentrations in fecal samples according to the manufacturer’s instructions (14).

Metabolomics measurements

Measurement of plasma metabolites for both studies were done using the AbsoluteIDQ p180 Kit (BIOCRATES LifeSciences AG, Innsbruck, Austria) on an AB SCIEX 5500 QTrap™ mass spectrometer (AB SCIEX, Darmstadt, Germany) with electrospray ionization combined with a HPLC system (Agilent 1260 Infinity Binary LC, Santa Clara, USA). Following quality control and data-processing, 177 metabolites were taken forward for statistical analyses.

Other laboratory parameter measurements

Total triglycerides, total cholesterol, serum glucose, white blood cell counts, and high sensitivity C-reactive protein were determined on a Dimension VISTA platform (Siemens Healthcare Diagnostics, Eschborn, Germany). Hemoglobin A1c concentrations were measured using high-performance liquid chromatography (Bio-Rad Diamat, Munich, Germany).

Statistical analysis

To investigate differences in the overall gut microbiota composition between SHIP-2 and SHIP-TREND individuals, we used permutational analysis of variance (vegan (15) function “adonis”; 1000 permutations) based on a Bray-Curtis dissimilarity (vegan function “vegdist”). We computed the alpha diversity scores Simpson diversity number (N2) and Shannon diversity index (H) using the “diversity” function (vegan). For specific taxon association analysis, microbiota data were cleared from zeros (treating them as missing values) to avoid biased regression due to zero-inflation. Log-transformation and standardization to a mean of zero and standard deviation of 1 was performed to allow comparable effect sizes. Standardization was done for each study separately. The same procedure was carried out for metabolite data. Linear regression models were used to test for significant associations of stool concentrations of elastase (exposure) with genus abundancies adjusting for the possible confounding factors age, sex, BMI, alcohol consumption, cigarettes per day, diabetes, a food frequency score, and sequencing batch (8,16). The same approach was used to test for significant associations between elastase and plasma metabolites. Having established elastase-associated microbiota and metabolite measures, linear regression analysis were performed to test for the association between microbiota measures (exposure) and plasma metabolites (outcome) using the same adjustment set. All of these analyses were performed for SHIP-TREND and SHIP-2 separately to account for the different demographics of participants and meta-analyzed afterward with a fixed effect model using the R package metafor. Only regression models with at least 100 observations were included in the present study. To control for multiple testing, the false discovery rate was controlled at a rate of 5% according to Benjamini-Hochberg. Mediation analyses were performed to test for the amount of effect from stool elastase on plasma metabolite levels mediated through differential abundancies in microbiota measures. The proportion mediated was estimated for triangles with concordant association directions using 500 bootstrap samples. Only triangles of elastase-microbiota-metabolites with an indirect effect passing a stringent Bonferroni-corrected significance threshold (P < 2.3E-3) are reported to present a mediation pathway. The same adjustment set as previously described was used further including a covariate indicating the study sample. All statistical analyses were performed using R version 3.5.1 (17).

Results

Characteristics of study sample

We included a total of 1319 participants from SHIP-2 and 907 SHIP-TREND participants into statistical analysis (Table 1). None of these individuals were under antibiotic treatment at the time of sample collection and none had a history of pancreatic disease. The overall gut microbiome structure in both cohorts was similar (Supplementary Figure 1, (18)) with more than 50% of the gut microbiota belonging to the taxa Bacteroides, Prevotella, Lachnospiraceae (unclassified family), Ruminococcaceae (unclassified family), Faecalibacterium, and Alistipes. Performing permutational analysis of variance based on a Bray-Curtis dissimilarity revealed small differences in the gut microbiota communities between both cohorts (r² = 0.002, F.Model 4.309, P < 0.001, adonis). When adjusting the gut microbiota for the possible confounding factors age, sex, BMI, alcohol consumption, cigarettes per day, diabetes, diet, and sequencing batch, no significant overall differences in the gut microbiota could be found between SHIP-2 and SHIP-TREND individuals (r² = 0.000, F.Model 0.951, P = 0.476, adonis). All following analyses were also adjusted for these possible confounders.

Table 1.

Phenotypic characteristics of the study samples

	SHIP-2 (n = 1319)	SHIP-TREND (n = 907)
Age (years)	60.0 (48.0-71.0)	51.0 (40.0-61.0)
Female sex (%)	49.5	56.2
BMI (kg/m2)	27.5 (25.0-30.8)	26.9 (24.1-30.0)
Diabetes mellitus (%)	11.8	1.4
HbA1c (%)	5.4 (5.0-5.8)	5.2 (4.8-5.5)
Serum glucose (mmol/L)	5.4 (5.0-5.9)	5.3 (4.9-5.7)
Total cholesterol (mmol/L)	5.4 (4.7-6.2)	5.5 (4.8-6.2)
Total triglycerides (mmol/L)	1.6 (1.1-2.3)	1.2 (0.9-1.8)
White blood cells (Gpt/L)	5.8 (4.9-6.9)	5.5 (4.7-6.4)
hs CRP (mg/L)	1.1 (0.7-2.3)	1.2 (0.6-2.5)
Pancreatic elastase (µg/g stool)	461.0 (323.0-578.0)	486.0 (339.0-594.5)
Smoking (cigarettes/day)	1.9	2.4
Alcohol consumption (g/day)	4.1 (1.1-12.0)	4.3 (1.3-10.8)
FFS	14.0 (12.0-17.0)	14.0 (12.0-17.0)

Continuous variables are given as median (first-third quartile). Binary variables are stated as percentages. Smoking is given as mean cigarettes smoked per day. All numbers were rounded to 1 decimal place. Data on total cholesterol and white blood cells were missing in 2 cases each. Data on high sensitivity c-reactive protein (hs CRP) was not available for 38 SHIP-TREND and 82 SHIP-2 participants.

Abbreviations: BMI, body mass index; FFS: food frequency score; HbA1c, hemoglobin A1c; hs CRP, high sensitivity c-reactive protein; n, number of cases; SHIP, Study of Health in Pomerania.

The exocrine pancreatic function associates with gut microbiota diversity

In a first step, we carried out a 3-stage statistical approach with overall indices of gut microbial diversity. We observed positive associations between fecal elastase levels and markers of alpha diversity (Shannon diversity index: q = 0.042; Simpson diversity number: q = 0.023) (Supplementary Table 1, (18)). Both diversity indices were also associated with a total of 5 metabolites, namely serotonin and spermidine (positively) as well as lysophosphatidylcholine (lysoPC) a C16:1, PC aa C32:1, and PC aa C40:5 (all inversely; Supplementary Table 2, (18)). We did not find evidence for a mediation of these associations by pancreatic elastase.

The exocrine pancreatic function associates with gut microbiota and plasma metabolites

In accordance with our previous study (8), pancreatic elastase concentrations were significantly associated with changes in the abundances of 28 microbial genera (20 positively, 8 inversely) (Fig. 2 and Supplementary Table 3, (18)). Associating pancreatic elastase concentrations with changes in plasma metabolite levels revealed 16 significant metabolite associations (6 positively, 10 inversely, Fig. 2 and Supplementary Table 4, (18)). Positive associations included the amino acids asparagine (q = 0.020), threonine (q = 0.049), the biogenic amines spermine (q = 0.010), spermidine (q = 0.012), and serotonin (q = 0.012), and the hydroxylated sphingomyelin a C14:1 (q = 0.049). The strongest inverse associations were seen with the lysoPC a C16:1 (q = 0.010) and further several acyl carnitine species from medium to long chain length partially carrying double bonds, PC aa C32:1 (q = 0.020) and PC aa C40:5 (q = 0.043) and acetylornithine (q = 0.043), as well as the sum of hexoses (q = 0.010). We refer to the associated microbiota and plasma metabolites as candidates for triangulation in the following.

Figure 2.

Associations of pancreatic elastase, gut microbiota, and plasma metabolites. The upper part of the figure displays all gut microbiota at genus level that were significantly associated with variation in fecal pancreatic elastase levels, whereas the lower part shows all significant associations of changes in plasma metabolites with pancreatic elastase. Metabolites found to be significantly associated with changes in gut microbiota are connected by either red (positive association) or blue (inverse association) lines with the corresponding taxon. Unclassified taxa at genus level: family (f), class (c), order (o), and phylum (p). Abbreviations: H1, hexose; PC, phosphatidylcholine; SM (OH), hydroxylated sphingomyelin.

We investigated measures of heterogeneity from meta-analyses results to identify effect estimates vastly differing between both cohorts and found only minor evidence for significantly different effect estimates. The associations of pancreatic elastase with plasma taurine and Flavonifractor levels were the only ones surviving multiple testing corrections, providing evidence that the association between pancreatic elastase and plasma taurine or Flavonifractor levels differed between both cohorts. This highlights the consistent effect of the presented associations even if both cohorts had different demographics.

Subsets of candidate microbiota and plasma metabolites are associated with each other

Association of pancreatic-associated microbiota with pancreatic-associated plasma metabolite levels generated 448 possible combinations with a sufficient number of observations. Out of these, 22 microbiota-metabolite pairs were significantly associated with each other (Fig. 2 and Supplementary Table 5, (18)). Ruminococcaceae, Clostridiales, and Clostridium_XlVa associated each with 4 metabolites including lysoPC a C16:1, PC aa C32:1, PC aa C40:5, and hydroxylated sphingomyelin C14:1 reflecting a consistent pattern with respect to the association with pancreatic elastase. This indicates a high probability of mediation. Oscillibacter and Collinsella were associated with PC aa C32:1 as well as lysoPC a C16:1. Bacteroides and Olsenella were both associated with plasma serotonin levels.

Higher plasma levels of the palmitoleate-containing lysoPC a C16:1 (but not PC aa C32:1) were associated with the predicted microbial potential for palmitoleate biosynthesis (P = 0.020). Correspondingly, the “palmitoleate biosynthesis I pathway” was inversely associated with increasing abundancies of Ruminococcaceae, Oscillibacter, Clostridiales, and Firmicutes, whereas higher abundances of Clostridium_XIVa and Flavonifractor were positively associated with the same pathway (Supplementary Table 6, (18)). Collinsella, however, was inversely associated with the “palmitoleate biosynthesis I pathway” in contrast to its positive association with lysoPC a C16:1, which might be explained by unobserved confounding.

The exocrine pancreatic function modulates plasma metabolites via changes in gut microbiota

We performed mediation analysis to test for the magnitude of the effect of pancreatic elastase on plasma metabolite levels mediated through differential abundancies in microbiota. Only triangles of elastase-microbiota-metabolites with an indirect effect passing a stringent Bonferroni-corrected significance threshold (P < 0.0023) were interpreted as representing a mediation pathway. Thus, we identified 10 significant triangles among exocrine pancreatic function, plasma metabolite profiles, and gut microbiota composition with an indirect effect out of 22 possible paths suggesting a partial mediation of the effect of exocrine pancreatic function on plasma metabolites through differential abundancies of microbiota (Fig. 3 and Supplementary Table 7, (18)). Those included a cross-linkage between Ruminococcaceae, Oscillibacter, and Collinsella with lysoPC a C16:1, PC aa C32:1, and PC aa C40:5 as well as between Bacteroides and serotonin.

Figure 3.

Mediation analysis between significantly associated plasma metabolites and gut microbiota. Heatmap of the proportion mediated by microbiotic measures on plasma metabolite levels significantly associated with stool elastase. Darker colors indicate a higher proportion mediated, and thick frames indicate a significant indirect effect after correction for multiple testing (P < 0.0023). Microbiota and metabolites were clustered based on association patterns. Shaded boxes indicate significant mediations in which case the proportion of effect mediation could not be estimated. Grey boxes indicated not tested associations. Unclassified taxa at genus level: family (f) and phylum (p). Abbreviations: PC, phosphatidylcholin; SM (OH), hydrodroxylated sphingomyelin.

Discussion

In the present study we integrated microbial and metabolomics profiling from 2 population-based cohorts to provide evidence that exocrine pancreatic function associates with changes in the plasma metabolite composition by modulating host microbiota, which may, in turn, influence exocrine pancreatic function. Our findings contribute to a better understanding of the complex interplay between the host and their microbial community for metabolic health.

Several plasma phosphatidylcholine (PC) lipid levels were not only associated with variation in exocrine pancreatic function, as would be expected, but also appeared to be mediated by the taxa Collinsella, Ruminococcaceae, and Oscillibacter. The increased abundance of palmitoleic acid-carrying PCs (PC aa C32:1 and lysoPC a 16:1) associated with reduced pancreatic elastase levels could provide a link between impaired exocrine pancreatic function, gut microbial dysbiosis (19), and altered host immunity. Palmitoleic acid, which can be of microbial origin, has been shown to inhibit pro-inflammatory cytokine responses (20), an important severity mechanism in pancreatitis (21,22). A putative increase, due to reduced exocrine pancreatic function, would represent a compensatory anti-inflammatory response—and one supported by gut microbial palmitoleic acid secretion. This would alleviate host inflammation triggered by microbial dysbiosis. However, this may come at the price of increased rates of malignancies as higher serum levels of palmitoleic acid have been associated with higher rates of cancer deaths in a Swedish large-scale cohort (23).

The gastrointestinal tract represents the most important source of serotonin within the human body and certain gut microbiota can stimulate its biosynthesis in enterochromaffin cells (24). It was shown in rodent models that serotonin, released by enterochromaffin cells, stimulates pancreatic secretion (25), which may explain the positive association of serotonin and fecal pancreatic elastase concentrations. We also found an association of Bacteroides and Olsenella with plasma serotonin levels as well as an at least partial mediation via Bacteroides. This indicates, again, an involvement of gut microbiota in serotonin metabolism. Our results may therefore imply that the effect of exocrine pancreatic function on intestinal microbiota composition is not a one-way street but a complex, multidirectional network that includes circulating metabolites. This fact significantly changes our understanding of the regulatory pathways that control nutrition, digestion, and metabolism and of the interlocked roles of its players: pancreatic exocrine function and the gut microbiome and metabolome as well as the circulating metabolic signature.

A number of plasma metabolites that were significantly associated with exocrine pancreatic function showed no significant association with abundancies of gut microbiota. The negative association of the sum of hexoses, which is basically a measure of plasma glucose metabolism, can be explained by a concomitant reduction of insulin-secretion in individuals with reduced exocrine pancreatic function. Additionally, pancreatic elastase has been shown to enter the circulation where it increases insulin secretion and insulin sensitivity (26); hence, its loss may further deteriorate glucose tolerance. Several acylcarnitine species were inversely associated with exocrine pancreatic function. Acylcarnitines are largely ingested via diet but also stem from endogenous biosynthesis (27) and are needed for energy generation through mitochondrial β-oxidation. Accumulation of acylcarnitines has been linked to insulin resistance (28), which can explain the increased acylcarnitine levels in case of reduced exocrine pancreatic function. Due to the large number of possible metabolite/microbiota species combinations, the current study may have been underpowered to detect more subtle effects that can underlie nonsignificant associations.

The alpha diversity indices, Shannon diversity index and Simpson diversity number, were positively associated with pancreatic elastase levels, in contrast to our previous study where we found no significant association with both diversity indices (8). The difference may be explained by the larger statistical power of the present analysis. In general, higher alpha diversity reflects more resilient and stable ecological communities (29) whereas lower diversity has been repeatedly associated with disease such as in obesity or chronic pancreatitis (30,31). Therefore, the positive association between pancreatic elastase levels and alpha diversity may reflect an intact regulation of the gut microbiome due to preserved exocrine pancreatic function.

The exact mechanisms that explain the association of exocrine pancreatic function with intestinal microbiota changes have still not been determined. On the one hand, reduced exocrine pancreatic function would result in a diminished secretion of digestive enzymes, thus, leading to an increased supply of undigested food components in the large intestine, which would promote the growth of specific bacterial taxa. On the other hand, the exocrine pancreas is known to secrete antimicrobial peptides such as cathelicidin-related antimicrobial peptide (32), which has been shown to play a pivotal role in control of the gut microbiota in rodents. However, in a previous study, we could not demonstrate a difference between human fecal cathelicidin concentrations in individuals with and without reduced exocrine pancreatic function (8). Therefore, there still remains a need for validation of the role of pancreatic antimicrobial peptides in humans.

We noted that some of the triangles between pancreatic function, microbiota abundancies, and plasma metabolites showed inconsistent effect estimates. For instance, Flavonifractor abundancies and the plasma levels of PC aa C32:1 were positively associated with pancreatic elastase levels. Plasma levels of PC aa C32:1, however, were inversely associated with pancreatic elastase levels contradicting the predicted relationship from the microbiota association. We have to assume that, despite the deep phenotyping, as yet unidentified confounding factors have contributed to these associations. Such findings make it clear that additional relationships between the 3 entities are still to be discovered and have to be tested in multiple individuals to reveal consistent triangles among host factors, gut microbial abundancies, and plasma metabolites.

While our study is distinguished by its large size and deep phenotyping across 2 large population-based cohorts, its cross-sectional design does not allow for causal interference, and even if we imply a causal direction for mediation analysis, longitudinal studies or experimental approaches will be needed to establish true causality. Identification of gut microbiota on the species level and of blood metabolites by unrestricted gas chromatography–mass spectrometry and liquid chromatography-mass spectrometry will reveal additional interlinkages between host and bacterial metabolism.

We show here the feasibility of a multi-step statistical approach to investigate the crosslinks between human and microbial metabolism using exocrine pancreatic function as an anchor phenotype. Whether the microbial and metabolic pathways affected by reduced exocrine pancreatic function can be manipulated via prebiotics, probiotics, or novel microbiota transfer techniques to improve the harmful dysbiosis associated with reduced exocrine pancreatic function or chronic pancreatitis and its metabolic deficiencies (33) will have to be tested in interventional trials.

Glossary

Abbreviations

BMI

body mass index

FFS

food frequency score

PC

phosphatidylcholin

SHIP

Study of Health in Pomerania

SM(OH)

hydrodroxylated sphingomyelin

Additional Information

Disclosure Summary: The authors have nothing to disclose.

Data Availability

Restrictions apply to the availability of some or all data generated or analyzed during this study to preserve patient confidentiality or because they were used under license. The corresponding author will on request detail the restrictions and any conditions under which access to some data may be provided.