Gut Microbiome Profile After Pancreatectomy in Infants With Congenital Hyperinsulinism

Abstract

Objectives:

To characterize gut microbiome profiles of infants with congenital hyperinsulinism (HI) who underwent near-total or partial pancreatectomy for hypoglycemia management, as compared to healthy controls.

Methods:

Prospective observational cohort. Subjects were infants (0-6 months) with HI who underwent removal of pancreatic tissue for management of intractable hypoglycemia February 2017—February 2018 at The Children’s Hospital of Philadelphia. Fecal samples were collected postoperatively, on full enteral nutrition. The gut microbiome of HI subjects was analyzed and compared to age-matched samples from healthy infants.

Results:

Seven subjects with ≥50% pancreatectomy and six with <50% pancreatectomy were included. Alpha (within-sample) diversity was lowest among infants with ≥50% pancreatectomy (Richness: False discovery rate = 0.003; Shannon Index: False discovery rate = 0.01). Beta (between-sample) diversity (Bray-Curtis dissimilarity: P = 0.02; Jaccard distance: P = 0.001) differed across groups (≥ or < 50% pancreatectomy, controls). Bifidobacteria and Klebsiella species were least abundant among infants with ≥50% pancreatectomy but did not differ between infants with <50% pancreatectomy and historical controls.

Conclusions:

Infants with HI who underwent ≥50% pancreatectomy differed from age-matched infants in gut microbiome profile, while those with <50% pancreatectomy more closely resembled control profiles. The durability of this difference should be investigated.

Introduction

The gut microbiome has been linked to many aspects of health and disease, including glucose metabolism.¹ Pancreatic acinar cells regulate gut microbiota composition and innate immunity through secretion of antimicrobial protein,² which make up approximately 10% of proteins in pancreatic juice.³ Although endocrine^4,5 and exocrine⁶ insufficiency are anticipated outcomes after pancreatectomy, particularly after more extensive resection, it is unknown whether the gut microbiome is also altered or differs by extent of pancreatectomy.

Due to the gut microbiome’s associations with later health outcomes,^7–9 the impact of an altered gut microbiome may be particularly important in the first year of life.¹⁰ Although the gut microbiome of healthy infants has been well-described,¹¹ the gut microbiome of infants after pancreatic resection has not been characterized. At our institution, approximately 30 patients per year undergo pancreatic surgery for management of congenital hyperinsulinism (HI).¹² Hyperinsulinism is the most common cause of persistent hypoglycemia in infants and children and can be histologically classified as focal or diffuse.¹³ Surgery to remove localized areas of abnormal tissue can be curative for focal HI,¹³ while near-total pancreatectomy may be required in diffuse HI if euglycemia cannot be maintained with medical therapy alone.¹² In a prospective cohort study, we leveraged this unique patient population to characterize the gut microbiome profiles of infants who underwent partial or more substantial pancreatic resections versus healthy control infants.

Materials and Methods

Participants and sample collection

Subjects prospectively enrolled for this study (HI cohort) included infants with confirmed diagnosis of HI (based on clinical criteria as previously described¹⁴) who underwent pancreatectomy while younger than 6 months of age at The Children’s Hospital of Philadelphia (CHOP) between February 2017 and February 2018 while 6 months or younger, whose only antibiotic exposure perioperatively was a standard 24-hour treatment with cefalozin. Demographic and clinical data were collected through medical record review. After pancreatectomy and upon return of bowel function (typically 5-7 days post-operatively), total parenteral nutrition was transitioned to full enteral nutrition at a rate of one-third per day. Two to three days post-feed advancement, up to 4 fecal samples were collected per subject while hospitalized, ranging from 2-3 days to 5-7 days post-full enteral nutrition. Aliquoted samples were stored at −80C.

An age-matched cohort was assembled using fecal samples from the Infant Growth and Microbiome Study (IGRAM cohort), a prospective, longitudinal cohort study of infant growth in the first 2 years of life. The IGRAM study included healthy infants born at ≥37 weeks of gestational age to African American mothers who delivered at the Hospital of the University of Pennsylvania and who had no maternal or fetal adverse outcomes.^15,16 Both studies received approval from the Institutional Review Boards of the respective institutions.

Microbiome analysis

Microbiome analysis was completed by the CHOP Microbiome Center. Bacterial DNA was isolated from approximately 200 mg of stool. The PSP® Spin Stool DNA Plus Kit (STRATEC Molecular GmbH, Berlin, Germany) was used and isolated DNA was quantified using the Quant-iT™ PicoGreen™ dsDNA Assay Kit (Invitrogen™, Eugene, Ore.). Genomic DNA was prepared for shotgun metagenomic sequencing using the Nextera® XT DNA Library Preparation Kit (Illumina®, San Diego, Calif.). DNA extraction blanks and DNA-free water were included as negative control samples to assess environmental and reagent contamination. Laboratory-generated mock communities consisting of DNA from Vibrio campbellii, Cryptococcus diffluens and lambda phage were included as positive control samples. DNA sequencing was carried out on an Illumina® HiSeq 2500 instrument, generating 125bp paired-end sequence reads. Reads were quality-filtered and trimmed to remove adapter sequences using Trimmomatic v. 0.36.¹⁷ Reads aligning to the human host genome (version hg38) were removed using BWA v. 0.7.17-r1188.¹⁸ Taxonomic annotations were generated by Kraken v. 1.0¹⁹ using the standard database with all complete bacterial, archaeal, and viral genomes in NCBI RefSeq.

Statistical analysis

The HI cohort was divided into two subgroups based on the median percent of pancreatic resection (50%) to evaluate the microbiome profile after more-limited versus more-extensive pancreatectomy. Alpha diversity was assessed by the expected number of observed species (out of rarefying sample size of 10,000) and Shannon index. The association between alpha diversity and study group (< or ≥50% pancreatectomy or IGRAM cohort) was evaluated using a mixed model with random intercept for subject. Beta diversity (similarity between samples) was assessed by Bray-Curtis and Jaccard distances. The null hypothesis of no difference in centroid positions of study group was tested using PERMANOVA, and pairwise comparison was conducted to identify differences between groups. To test the association between log10 relative abundance of taxa and study group, a linear mixed model with subject random effects was used. Taxa were tested if the relative abundance in any sample exceeded 1%. False discovery rate (FDR) adjusted p-values were based on the Benjamini-Hochberg method.²⁰ Analyses were conducted using R v. 3.6.2.²¹

Results

Subject Characteristics

Thirteen subjects (7 female) with a total of 31 samples were included in the HI cohort and were age- and sex-matched with 27 samples from 13 subjects (7 female) from the IGRAM study. The majority of HI subjects were of Non-Hispanic White race/ethnicity (n = 9), and 1 as Non-Hispanic Black. The IGRAM cohort, by design, consisted of infants born to African American mothers. Two subjects from the HI cohort were born late pre-term at 35- and 36-weeks gestational age; the remainder (11/13) were born at ≥37 weeks gestational age. Vaginal delivery was the most common delivery mode among the HI cohort (10/12 with delivery mode recorded), similar to the IGRAM cohort (10/13).

Twelve of the 13 HI subjects had inactivating mutations of the genes encoding the ATP-sensitive potassium channel in the pancreatic beta cell (11 ABCC8, 1 KCNJ11); one had no identified mutation (Table 1). Median age at pancreatectomy among HI subjects was 52 days (interquartile range 36-92) and ranged from 27-190 days. Age at fecal samples ranged from 38 to 206 days among HI subjects and 30 to 362 among IGRAM subjects.

Table 1.

HI cohort characteristics (n = 13 subjects)

Female, n (%)	7 (54)
Race/ethnicity, n (%)
Non-Hispanic White	9 (69)
Non-Hispanic Black	1 (8)
Hispanic	3 (23)
Gestational age, weeks (median, IQR)	39.3 (38.3-40.0)
Birthweight, kg (median, IQR)	3.56 (3.32-4.22)
Large for gestational age by weight (>90th percentile), n (%)	4 (30.8)
Delivery mode, n (%)
Vaginal	10 (77)
C-section	2 (15)
Unrecorded	1 (8)
Mutation, n (%)
ABCC8	11 (84.6)
KCNJ11	1 (7.7)
None detected	1 (7.7)
Age at pancreatectomy, days (median, IQR)	52 (36-92)
Percent pancreatectomy, n (%)
2%	1 (7.7)
5%	3 (23.1)
15%	1 (7.7)
35%	1 (7.7)
50%	6 (46.2)
98%	1 (7.7)
Feeding post-pancreatectomy, n (%)
Exclusively breastfed	3 (23.1)
Exclusively formula fed	6 (46.2)
Mixed breastmilk, formula, and/or food	4 (30.8)
On dextrose post-pancreatectomy, n (%)	3 (23.1)

IQR, interquartile range

Six HI subjects had 2-35% of pancreatic tissue removed due to focal hyperinsulinism, while 6 had approximately 50% and 1 had 98% resected. Post-pancreatectomy, 3 HI subjects required ongoing management with continuous intragastric dextrose²² due to persistent hypoglycemia. Four HI subjects had no breastmilk exposure after pancreatectomy (formula-fed), while the remainder had some (n = 4) or exclusive (n = 3) breastmilk exposure (Table 1). Exclusive breast-feeding did not differ between pancreatectomy groups (17% versus 29% for <50% versus ≥ 50% pancreatectomy, respectively, P = 0.6). Six IGRAM subjects were exclusively formula-fed at the time of the closest study visit to fecal sample, while the remainder received both formula and breastmilk (n = 6) or were exclusively breastfed (n = 1). Exclusive breast-feeding did not differ between the HI cohort and IGRAM cohort (P = 0.3).

Gut microbiota composition in HI subjects post-pancreatectomy versus healthy controls

Alpha diversity as measured by both richness and Shannon Index was lower in fecal samples from infants with ≥ 50% pancreatectomy versus control (P = 0.001/FDR = 0.003 and P = 0.004/FDR = 0.01, respectively). In addition, the Shannon Index was lower among infants with ≥ 50% pancreatectomy than among those with <50% pancreatectomy (p=0.02/FDR = 0.04). Beta diversity, as measured by both Bray-Curtis dissimilarity (Figure 1a) and Jaccard distance (Figure 1b), differed by group (P = 0.02 and P = 0.001, respectively). In pairwise comparison, beta diversity of each HI group differed significantly from controls (P < 0.05 for each comparison by Bray-Curtis and Jaccard distances), but did not differ by extent of pancreatectomy (<50% or ≥ 50%).

Figure 1.

Beta diversity: Principal coordinates analysis (PCoA) derived from (a) Bray-Curtis and (b) Jaccard distance among samples of the 3 groups (control, <50% and ≥50% pancreatectomy).

We compared the relative abundance of 109 taxa that had an abundance of at least 1% in any sample. These included several Bifidobacterium species, including B. breve, B. adolescentis, B. breve, and B. angulatum, as well as Klebsiella pneumoniae. The abundance of B. breve was higher among infants with <50% pancreatectomy than those with ≥ 50% pancreatectomy (P = 0.0006/FDR = 0.02) but did not differ between either HI group and control (Figure 2a). In contrast, Klebsiella pneumoniae was significantly less abundant in infants with ≥ 50% pancreatectomy than either controls or infants with <50% pancreatectomy (P = 0.0005/FDR = 0.02 and P = 0.001/FDR = 0.02, respectively; Figure 2b).

Figure 2.

Abundance of Bifidobacterium breve (a) or Klebsiella pneumoniae (b) by group (control, <50% and ≥50% pancreatectomy).

Discussion

In this prospective observational study of infants with HI who underwent pancreatectomy, we found that fecal samples from infants who underwent more-extensive pancreatectomy had lower bacterial diversity than healthy historical controls as well as a lower abundance of Bifidobacteria species and Klebsiella pneumoniae than samples from infants with less-extensive pancreatic resection. Our study is the first to characterize the gut microbiome in a sample of pediatric patients post-pancreatectomy, as well as the first to describe the gut microbiome of infants with congenital hyperinsulinism. Additional strengths include the use of age-matched control samples in order to account for expected changes in the gut microbiome with age in infancy, as well as a consistent hospital setting with a single surgeon and standard peri-operative antibiotic and feeding advancement approach.

Our finding of lower diversity among patients post-pancreatectomy as compared to healthy controls is in line with previous findings of adult patients who underwent pancreaticoduodenectomy.²³ Although differences in antibiotic exposure may partly explain differences between the IGRAM cohort and HI cohort, both pancreatectomy groups received the same antibiotic exposure yet differed in alpha diversity. Greater gut microbiome diversity has generally been considered a marker of good health in adults.²⁴ In contrast, infancy is marked by initially low diversity that increases with age.²⁵ Whether the differences in diversity persist over time should be further investigated.

In addition to lower overall diversity, the lower abundance of both Bifidobacterium breve and Klebsiella pneumoniae in infants with ≥ 50% as compared to <50% pancreatectomy is notable. Lower Bifidobacterium²⁶ and higher Klebsiella pneumoniae²⁷ abundance have both been associated with necrotizing enterocolitis in infants. The impact of the balance of this differential abundance may depend on additional important risk factors, including prematurity. In our HI cohort, which consisted of only late pre-term or term infants, none developed necrotizing enterocolitis post-pancreatectomy. Notably, treatment with Bifidobacterium strains has been associated with improved metabolic health and glucose regulation (reduced hyperglycemia) in rats.²⁸ Whether Bifidobacterium may also contribute to reduced hypoglycemia is unknown.

We acknowledge several limitations to this study, including the relatively small sample size and short time frame of assessment post-pancreatectomy. As fecal samples prior to pancreatectomy were unavailable, it was not possible to directly compare within-subject change in microbiome profile pre- and post-pancreatectomy. In addition, although samples were age-matched for the HI cohort and controls, the control cohort differed in exposures beyond surgery, including prolonged hospitalization and antibiotic exposure. Finally, whether the identified differences in gut microbiome profiles were associated with differences in short- or long-term health outcomes is unknown.

In conclusion, infants who underwent extensive pancreatectomy due to HI had lower gut microbiome diversity and lower abundance of bacterial strains that have been associated with both decreased and increased risk of necrotizing enterocolitis. The persistence of these differences and potential long-term clinical impact should be further investigated.