Antibiotic Treatment Leads to Fecal Escherichia coli and Coliphage Expansion in Severely Malnourished Diarrhea Patients.

Malnutrition predisposes to diarrhea and diarrhea adversely affects the nutritional status creating a vicious cycle. The role of the gut microbiome in malnutrition is an active research area. Parenteral antibiotics are recommended by the World Health Organization in hospitalized pediatric patients with severe acute malnutrition (SAM) presenting signs of infections. Stool microbiota data for such patients are, however, lacking. To fill this gap, we studied the stool microbiota in 19 SAM patients from Bangladesh hospitalized with acute diarrhea (AD) and compared it with that of matched 20 healthy control subjects (HC) (Supplementary Table 1). SAM-AD patients were treated with parenterally administered gentamycin and ampicillin, whereas HC received no antibiotics for at least a month before sample collection.

Supplementary Table 1

Baseline Characteristics of SAM-AD Case and Matched HC

	HC	SAM-AD	P value
N	20	19
Age child, mo	13.0 (10.8 to 16.0)	13.0 (9.5 to 18.5)	.989
Age mother, y	25.5 (22.0 to 30.5)	24.0 (22.0 to 27.5)	.411
Weight, kg	8 (8 to 9)	5.9 (5.1 to 6.7)	2.93e-07
Height, cm	73.5 (69.8 to 77.4)	68.0 (64.5 to 70.7)	.0064
Mid arm circumference, cm	13 (13 to 14)	12 (11 to 12)	3.27e-05
Weight for age z score	-1 (-1 to -1)	-4 (-5 to -4)	1.01e-07
Height for age z score	-1 (-2 to -1)	-3 (-4 to -3)	4.35e-06
Body mass index	16 (15 to 16)	13 (12 to 13)	1.18e-07
Weight for height z score	-1 (-1 to -0)	-3 (-4 to -3)	1.01e-07
Body mass index z score	-0 (-1 to 0)	-3.4 (-4.2 to -2.8)	1.37e-07
Mid arm circumference z score	-1 (-1 to -1)	-3 (-3 to -3)	2.34e-05
Rectal temperature, °C	36.5 (36.0 to 36.7)	37.2 (37.0 to 37.2)	1.08e-07
Pulse, min-1	110.0 (100.0 to 120.0)	132.0 (130.0 to 136.0)	7.23e-08
Respiration rate, min-1	30.0 (30.0 to 32.0)	36.0 (35.0 to 36.0)	4.46e-07
Vomiting, d-1	0.0 (0.0 to 0.0)	0.0 (0.0 to 2.0)	.00345
Duration of diarrhea, d		4 (4 to 4)
Stool frequency, d-1	2.5 (2.5 to 2.5)	5.0 (3.5 to 7.0)	6.15e-06
Systolic blood pressure, mm Hg	90.0 (90.0 to 90.0)	90.0 (90.0 to 90.0)	.0166
Diastolic blood pressure, mm Hg	60.0 (60.0 to 60.0)	60.0 (60.0 to 60.0)	.101
Exclusive breastfeeding, mo	6.0 (5.8 to 6.0)	6.0 (2.5 to 6.0)	.0578
Number of siblings	1.5 (1.0 to 3.0)	2.0 (1.0 to 2.0)	.541

NOTE. Values are medians (interquartile range: first, third quartile). P values are calculated by a 2-sided Mann-Whitney test. Categorical variables were compared by chi-square test.

16S rRNA and metagenome sequencing showed a marked increase of Escherichia and Klebsiella abundances in SAM-AD over HC (Figures 1A, 1C, and 2A), but not of Streptococcus (Figure 1B). Compared with HC, SAM-AD showed a reduced microbiota diversity (Figure 1D) and a decrease in Prevotella, Blautia, Ruminococcus, Faecalibacterium, Megamonas, and Bifidobacterium (Figures 1A and 2A). SAM-AD patients showed a 10-fold-lower 16S copy number of stool bacteria than HC (Figure 1E), which was partially compensated by a 2-fold higher stool frequency.

Figure 1

Stool microbiota analysis by 16S rRNA gene sequencing. (A) Bubble plot for 20 HC subjects and 18 SAM-AD cases at genus level. Box plots for Streptococcus- (B) and Escherichia- (C) attributed sequences, alpha-diversity (D), and log10 copy numbers of 16S rRNA genes per gram stool (E).

Figure 2

Stool microbiome analysis by metagenome sequencing. (A) Taxonomical attribution of sequences from 9 HC subjects and 18 SAM-AD cases to the indicated bacteria and viruses. (B) Attribution of the listed virulence factor genes to cases and control subjects given as counts per million genes. Abundance of sugar and sugar derivate–digesting genes (C, D) and antibiotic-resistance genes (E) in counts per million. (F) Abundance of indicated phage sequences expressed as percentage of total attributed sequences normalized for genome size by MetaPhlAn2. (G) Correlation between the abundance of the reads attributed to Escherichia phage phAPEC8 at hospitalization and change in abundance of Escherichia over a period of 1.5 days estimated by16S rRNA sequencing.

Rotavirus was the dominant pathogen (Supplementary Table 2) in SAM-AD, contradicting reports on protection from rotavirus diarrhea by malnutrition. All other pathogens (Escherichia coli in 7, Cryptosporidium in 2, Vibrio cholerae in 1, norovirus in 1 patient), except in 1 patient with adenovirus, were associated with copathogens. Salmonella was not detected in any SAM-AD patient.

Supplementary Table 2

Pathogen Detection in Stools of SAM-AD Patients

Patient ID	TaqMan	Pathogen taxa	Escherichia coli pathogens	Virulence factors
506	Adenovirus (8)	Adenovirus (67)	neg	neg
511	Cryptosporidium (3)	neg	neg	neg
503	Ascaris (3)Rotavirus (3)	neg	EAEC-aaiC (0.01)	EAEC-aaiC (154)
516	neg	Shigella (1)	neg	neg
518	Vibrio cholerae (7)EPEC-bfp (5)EPEC-eae (4)EIEC-ipaH (3)	V cholerae (3)	neg	EPEC -bfp (36)
517	Cryptosporidium (15)EPEC -bfp (6)EPEC-eae (6)EIEC-ipaH (6)EAEC-aaiC (5)	neg	neg	neg
513	NA	neg	EAEC-aaiC (0.01) EAEC-aatA (0.01)	EAEC-aaiC (120) EAEC-aatA (79)
515	Norovirus (5)	neg	EAEC-aaiC (0.01)	EAEC-aaiC (123) EAEC-aatA (78)
501	Rotavirus (10)	neg	neg	neg
502	Rotavirus (5)EAEC-aaiC (3)	neg	EAEC-aaiC (0.02)	EAEC-aaiC (275)
504	Rotavirus (12)	neg	neg	neg
505	Rotavirus (8)EAEC-aatA (3)	neg	EAEC-aaiC (0.01) EAEC-aatA (0.02)	EAEC-aaiC (202) EAEC-aatA (62)
507	Rotavirus (8)	Adenovirus (3) Aeromonas (0.2)	neg	neg
508	Rotavirus (6)	neg	neg	neg
510	Rotavirus (6)	neg	ETEC-lt (0.03)	ETEC-lt (346)
514	Rotavirus (11)	Adenovirus (0.1)	neg	neg
519	Rotavirus (10)	NA	NA	NA
509	neg	Shigella (0.5)	neg	neg
512	Cryptosporidium (9)EIEC-ipaH (3)	Shigella (2) Cryptosporidium (0.5)	neg	EAEC-aatA (114) EIEC-ipaH (36)

NOTE. TaqMan, results of detection of 19 pathogens with TaqMan array card (difference to threshold expressed as cycle threshold), E coli pathotypes and the detected virulence factors are indicated, neg, no pathogen detection; pathogen taxa, percentage of taxons determined in metagenome sequencing of the indicated pathogen, neg, <0.1% of taxa; E coli pathogens, the indicated virulence genes of the specified E coli pathotype with % of identified E coli genes, neg, <0.01 of genes; virulence factors, gene read number corrected per million reads and length of target gene coverage.

NA, the corresponding sample was not investigated by metagenome sequencing.

Compared with HC, virulence factor genes were increased in SAM-AD for various pathogenic Enterobacteriaceae (uropathogenic, enterohemorrhagic, and enteroaggregative E coli, Shigella, Salmonella, and Yersinia) (Figure 2B). The top 10 most significant pathway changes in SAM-AD over HC (Supplementary Table 3) included increases in D-glucarate and D-galactarate degradation genes (Figure 2D). In addition, SAM-AD showed more antibiotic resistance genes than HC (Figure 2E), mostly E coli (63%) and Klebsiella (32%) associated.

Supplementary Table 3

Top 10 Pathways Significantly Enriched in SAM-AD Over HC in Stool Metagenome Data

Description	HC	SAM-AD	Q
Superpathway of l-arginine and l-ornithine degradation	13	105	0.0002
Superpathway of l-arginine, putrescine, and 4-aminobutanoate degradation	13	105	0.0002
D-glucarate degradation I	21	161	0.0004
Phytol degradation	45	391	0.0004
D-galactarate degradation I	20	120	0.0007
Superpathway of D-glucarate and D-galactarate degradation	20	120	0.0007
Methylphosphonate degradation I	11	94	0.0007
Superpathway of fermentation	48	179	0.0012
Phytate degradation I	28	218	0.0012
NAD/NADP-NADH/NADPH mitochondrial interconversion	32	177	0.0012

NOTE. Median counts per million reads as assessed by 1-sided Mann-Whitney U test corrected for multiple testing by the Benjamini-Hochberg procedure.

NAD, nicotinamide adenine dinucleotide; NADH, reduced nicotinamide adenine dinucleotide; NADP, nicotinamide adenine dinucleotide phosphate; NADPH, reduced nicotinamide adenine dinucleotide phosphate.

Escherichia phage followed by Vibrio phage DNA was increased in SAM-AD over HC (Figures 2A and 2F). The expansion of coliphages in SAM-AD was likely a consequence of increased abundance of bacterial host cells. Other mechanisms could play a role, such as increased accessibility or modified physiology of the bacterial host cells, for example as a consequence of immune system response to bacteria. Among SAM-AD patients, the abundance of sequences attributed to Escherichia phage phAPEC8 was negatively correlated with the abundance of its host (Figure 2A, SparCC [1000 bootstraps]: -0.52; N = 18; P = .008). It is unclear, however, whether coliphage expansion could lead to a collapse of E coli population because high abundance of phage at enrollment was not associated with a greater decrease of E coli abundance over a period of approximately 1 day (Figure 2G). Longer time series are necessary to determine whether bacteriophages could indeed control the expansion of host bacteria in the gut. Previous attempt to treat E coli–associated AD with a mixture of T4 bacteriophages had failed to demonstrate clinical benefit; however, the E coli dominance was much more pronounced in antibiotic-treated SAM-AD patients of the present study than in children with AD.

A marked increase of fecal E coli abundance at the expense of bifidobacteria was also described in European newborns not suffering from diarrhea but treated parenterally with ampicillin and gentamicin for suspected sepsis. Postantibiotics expansion of E coli and Salmonella typhimurium was also observed in mice model where it was shown to be a consequence of streptomycin-induced production of galactarate and glucarate in host’s cecum. This host-dependent mechanism may have contributed to the observed expansion of E coli, although the main driver was likely the high levels of antibiotic resistance displayed by E coli in Bangladesh. We think that the treatment with antibiotics rather than malnutrition and diarrhea was the main cause of the observed microbiota alteration, because Bangladeshi children with AD showed an increased abundance of commensal streptococci over control subjects, whereas children with SAM displayed a shift to a less mature fecal microbiota composition but not a marked E coli expansion.

Antibiotic-induced Enterobacteriaceae expansion studied in mice has been shown to be involved in the disruption of the symbiosis between colonocytes and obligate anaerobic butyrate producers, resulting in a vicious cycle whereby colonocyte metabolism is subverted to permit the outgrowth of oxygen-tolerant, nitrate-dependent Enterobacteriaceae. It is known that antibiotic treatment in humans may lead to diarrhea even in a presumed absence of obligate pathogens (antibiotic-associated diarrhea), but the microbiota of pediatric antibiotic-associated diarrhea has not been studied.

Currently, there is no evidence from humans that the antibiotic-induced expansion of normally commensal Enterobacteriaceae could be detrimental. However, the observations from animal models suggest that this is a possibility that should be investigated.

Supplementary Table 4

The Main Variables That Were Evaluated in the Study, Stratified By Sex

	HC		SAM-AD
	Female	Male	Female	Male
16S rRNA sequencing
N	12	8	2	16
Bifidobacterium (proportion of reads)	25.0 (6.62–47.1)	24.9 (18.8–55.6)	0.024 (0.02–0.02)	0.405 (0.05–6.07)
Escherichia (proportion of reads)	1.46 (0.71–9.76)	6.32 1.95–24.1)	62.6 (46.6–78.6)	57.0 (20.8–73.8)
Diversity (Faith index)	4.007 (2.86–4.37)	3.308 (3.08–3.51)	1.360 (1.28–1.44)	1.306 (1.06–2.19)
Shotgun metagenomics
N	12	8	3	16
Antibiotic resistance genes (counts per million reads)	955 (767–1213)	608 (530–1326)	4117 (3063–4140)	3978 (2115–6495)
D-galactarate degradation pathway (counts per million reads)	13.6 (6.94–30.50)	20.3 (5.65–25.30)	85.7 (78.80–119.01)	127 (70.90–186)
Phage sequences (percentage of total sequences, normalized for genome size)	0.469 (0.00–2.30)	0.000 (0.00–0.00)	73.0 (49.3–85.4)	0.794 (0.04–15.8)

NOTE. Median and interquartile range are displayed.