Fecal Microbiota Functional Gene Effects Related to Single-Dose Antibiotic Treatment of Travelers’ Diarrhea

Abstract

Background

Travelers’ diarrhea (TD) is common among military personnel deployed to tropical and subtropical regions. It remains unclear how TD and subsequent antibiotic treatment impact the resident microflora within the gut, especially given increased prevalence of antibiotic resistance among enteric pathogens and acquisition of multidrug-resistant organisms. We examined functional properties of the fecal microflora in response to TD, along with subsequent antibiotic treatment.

Methods

Fecal samples from US and UK military service members deployed to Djibouti, Kenya, and Honduras who presented with acute watery diarrhea were collected. A sample was collected at acute presentation to the clinic (day 0, before antibiotics), as well as 7 and/or 21 days following a single dose of antibiotics (azithromycin [500 mg], levofloxacin [500 mg], or rifaximin [1650 mg], all with loperamide). Each stool sample underwent culture and TaqMan reverse transcription polymerase chain reaction analyses for pathogen and antibiotic resistance gene detection. Purified DNA from each sample was analyzed using the HumiChip3.1 functional gene array.

Results

In total, 108 day 1 samples, 50 day 7 samples, and 94 day 21 samples were available for analysis from 119 subjects. Geographic location and disease severity were associated with distinct functional compositions of fecal samples. There were no overt functional differences between pre- and postantibiotic treatment samples, nor was there increased acquisition of antibiotic resistance determinants for any of the antibiotic regimens.

Conclusions

These results indicate that single-dose antibiotic regimens may not drastically alter the functional or antibiotic resistance composition of fecal microflora, which should inform clinical practice guidelines and antimicrobial stewardship.

Clinical Trials Registration Number

NCT01618591.

When properly administered, antibiotics can effectively treat most common bacterial infections, including travelers’ diarrhea (TD). Nevertheless, antibiotics can have devastating effects on the normal gut microflora that have been linked to negative health outcomes, including infection susceptibility, symptom exacerbation, and health consequences associated with multidrug-resistant organism (MDRO) acquisition [1]. These concerns have led researchers and clinicians to reconsider treatments strategies for TD, which occurs in 10%–40% of travelers visiting less-developed countries [2] and has an average attack rate of 30 cases per 100 person-months among deployed military personnel [3]. The current guidelines for moderate to severe TD recommend single-dose antibiotic therapy (3-day course when symptoms persist for >24 hours) [4, 5]. Epidemiological studies have associated antibiotic use during travel with MDRO acquisition in up to 75% of travelers returning from the Asian subcontinent [6]. To limit antibiotic exposure and simplify treatment regimens, a clinical trial (TrEAT TD) was conducted among deployed military personnel with TD; it assessed the efficacy of 3 different single-dose antibiotic regimens [7]. The findings demonstrated that single-dose azithromycin, levofloxacin, or rifaximin was sufficient to clinically cure most patients (96%) within 72 hours; however, the impact of single-dose antibiotics on fecal microflora and MDRO acquisition is poorly understood.

Fecal microbiome studies primarily utilize marker gene amplification (16S rRNA) and high-throughput sequencing to assess bacterial community composition. While this method addresses “what’s there?”, it provides little insight into “what’s going on?” To study the functional capabilities of bacterial communities, functional gene arrays (FGAs) can overcome many disadvantages associated with high-throughput sequencing methods, including sampling errors and species abundance inconsistencies [8]. The HuMiChip FGA was developed to specifically characterize the functional capabilities of human microbiomes; it contains ~134 000 probes targeting microbial genes involved in >157 functional categories pertaining to host–microbiome interactions [9]. Herein, we utilized HuMiChip FGA in conjunction with a customized TaqMan Array Card (TAC) to characterize functional fecal microbiome compositions of TrEAT TD patients before and after single-dose antibiotics and to assess antibiotic resistance acquisition, primarily in extended-spectrum ß-lactamase (ESBL) genes [10].

METHODS

Study Design

This study utilized archived stool specimens from deployed US and UK military personnel who participated in the TrEAT TD clinical trial [7]. All TrEAT TD participants were active-duty US or UK military, ≥18 years old, and deployed to Afghanistan (UK), Djibouti (US), Kenya (UK), or Honduras (USA). Participants who presented with acute watery diarrhea (AWD) had initial stool samples collected (day 0) and were randomized into 1 of 3 single-dose antibiotic treatment groups: rifaximin (1650 mg), azithromycin (500 mg), or levofloxacin (500 mg). An adjunct dose of loperamide (4 mg) was given to all participants within each treatment arm with the antibiotic, followed by an additional 2 mg after each unformed stool (self-administered, not to exceed 16 mg/d for 2 days). A second stool sample and third stool sample were collected 7 and 21 days postantibiotic treatment. All stool samples were immediately frozen and stored at –70ºC.

Patient Consent

Data were collected from subjects who provided authorization for collection and analysis of their data through informed consent and HIPAA authorization processes and in accordance with the Helsinki Declaration of 1975 (revised in 2008). All samples were obtained with approval from the Uniformed Services University Institutional Review Board (IRB), UK Ministry of Defence Research Ethics Committee, and Kenyan Medical Research Institute’s IRB. The TrEAT TD trial is registered at ClinicalTrials.gov (NCT01618591).

Genomic DNA Extraction

Genomic DNA (gDNA) was purified from stool samples (200 mg) using the QIAamp DNA Stool Mini Kit (Qiagen, Valencia, CA, USA) [9]. A frozen aliquot (200 mg) of stool sample was added to a 2-mL screw-cap vial containing 300 mg of 0.1-mm glass beads (Sigma, St. Lous, MO, USA) and kept on ice until 1.4 mL of ASL buffer from the QIAamp kit was added. Samples were immediately bead-beaten (45 seconds, speed 6.5) using a FastPrep machine (Bio 101, Morgan Irvine, CA, USA) and then heated for 5 minutes at 95ºC. Subsequent steps were to the manufacturer’s recommendation following the pathogen detection protocol. gDNA was quality-checked (A260/A280 > 1.7, A260/A230 > 1.8) using a NanoDrop 1000 spectrophotometer (Thermo Scientific) and quantified using the Pico-Green kit (Invitrogen, Carlsbad, CA, USA). All gDNA extractions were performed in a blinded manner without access to identifiable subject data. gDNA samples were stored at –20ºC.

HuMiChip 3.1 Analysis

HuMiChip 3.1 array contains 29 467 strain-specific probes targeting 2063 microbial strain/species and 133 924 sequence- and group-specific probes targeting 157 functional gene families involved in various metabolic pathways and host–microbiome interaction processes [9]. One microgram of purified gDNA from stool samples was labeled, hybridized to the HuMiChip array, and imaged. Probe spots were discarded if they met any of the following specifications: signal intensity <200 or <1.3 times the background signal, signal-to-noise ratio <2, or coefficient of variation of the background <0.8.

TaqMan Array Card Analysis

TAC assays for detection of various enteropathogens and ESBL genes in purified gDNA from stool samples were performed [10]. TaqMan probe targets were considered positive if the C_T value was <35. In addition to the 19 enteropathogens, probes were designed for detection of 6 ESBL genes: kpc, ndm, ctx, shv, cmy, and tem. TAC was performed on day 0 and day 21 samples only.

Statistical Analyses

Normalized HuMiChip data and TAC data were imported, processed, and visualized using R, version 3.5.2 (R Core Team). Principal component analysis (PCA) was used to visualize functional gene differences between sample groups. Significant differences were determined using the multiresponse permutation procedure (MRPP), and permutational multivariate analysis of variance was performed using distance matrices (Adonis, using Bray-Curtis dissimilarity matrices with 99 permutations). The relative abundance of various functional gene categories was computed by dividing the total signal intensity of all probes within a specific group by the total signal intensity of all functional probes within the HuMiChip array and compared between groups using response ratio analysis [11].

RESULTS

Study Population

Our study population was a subsampling of 326 deployed military personnel with AWD enrolled in the TrEAT TD clinical trial (Table 1) [7]. Based on sample availability, 119 were included in our study, with representation from each treatment arm (azithromycin, 42; levofloxacin, 46; rifaximin, 31). To obviate selection bias, we ensured that demographic and baseline clinical characteristics were comparable across treatment arms and representative of the TrEAT TD study population (Supplementary Table 1). Study subjects were young (median age, 28 years), male (91.6%), White (79.8%), and a member of the US (54.6%) or UK military (45.4%). All patients reported AWD symptoms for ≥24 hours, with a mean maximum number of loose or liquid stools in a 24-hour period of 5.8. For most patients (71.4%), AWD symptoms resulted in decreased activity level/ability or inability to function in their primary duty assignment.

Table 1.

Demographics and Baseline Clinical Characteristics for Patients who Submitted Samples for HuMiChip and TAC Analyses^a

Characteristic, No. (%)	Azithromycin (500 mg) n = 42	Levofloxacin (500 mg) n = 46	Rifaximin (1650 mg) n = 31	Total n = 119
Age, median (IQR), y	28 (23.5–33.75)	28 (22.5–33)	30 (23–38)	28 (23–34)
Male	41 (97.6)	40 (87)	28 (90.3)	109 (91.6)
Race
White	38 (90.5)	31 (67.4)	26 (83.9)	95 (79.8)
Black	3 (7.1)	11 (23.9)	2 (6.5)	16 (13.4)
Other	1 (2.4)	4 (8.7)	3 (9.7)	8 (6.7)
US military	20 (47.6)	25 (54.3)	20 (64.5)	65 (54.6)
Deployment countryb
Kenya	22 (52.4)	21 (45.7)	11 (35.5)	54 (45.4)
Djibouti	9 (21.4)	11 (23.9)	9 (29)	29 (24.4)
Honduras	11 (26.2)	14 (30.4)	11 (35.5)	36 (30.3)
Baseline clinical characteristics
Maximum LLS in 24 h (all cases), mean (SD), No. of stools	5.6 (2.6)	5.7 (3.2)	6.2 (3.3)	5.8 (3)
Total LLS before presentation, mean (SD), No. of stools	9.1 (7.2)	9.1 (8)	9.1 (6.1)	9.1 (7.2)
Total LLS in 8 h before presentation, mean (SD), No. of stools	3 (1.6)	3.1 (1.5)	3.8 (3)	3.3 (2.1)
Impact of illness on activity level/ability to function in primary duty assignment, No. (%)
Normal	16 (38.1)	13 (28.3)	5 (16.1)	34 (28.6)
Decreased (≤50%)	21 (50)	16 (34.8)	20 (64.5)	57 (47.9)
Decreased (>50%)	5 (11.9)	14 (30.4)	6 (19.4)	25 (21)
Complete inability to function	0 (0)	3 (6.5)	0 (0)	3 (2.5)

Abbreviations: IQR, interquartile range; LLS, loose or liquid stools; TAC, TaqMan Array Card.

^aOnly patients who had acute presentation samples (day 0) matched to postantibiotic samples (day 7 and/or 21) were included.

^bMilitary personnel stationed in Kenya were UK military; Djibouti- and Honduras-stationed personnel were US military.

Thirty-nine patients (32.8%) provided a HuMiChip-analyzed stool sample at days 0, 7, and 21 (Supplementary Figure 1). An additional 45 (37.8%) patients provided day 0 and day 21 matched samples, while a small subset of patients provided either day 0 and day 7 samples (4, 3.4%) or day 7 and day 21 samples (6, 5.0%). Twenty-five patients had only 1 HuMiChip-analyzed stool sample; 20 for day 0, 1 for day 7, and 4 for day 21. In addition, 105 (88.2%) patients provided matched stool samples for TAC analysis at days 0 and 21.

Epidemiological Associations With Stool Functional Composition at Illness Presentation

In conjunction with patient metadata, HuMiChip array data were analyzed to determine if samples were functionally diverse or fairly homogenous upon acute presentation (before treatment). PCA analysis of all functional probes revealed distinct clustering of samples by country, with pronounced separation between samples from Kenya and Djibouti (Figure 1A). In-depth investigation revealed numerous gene functional categories that differed in relative abundance between samples from Kenya and Djibouti, with the relative abundance of genes associated with antibiotic resistance being higher in Kenyan samples (Figure 1B).

Figure 1.

Functional composition differences within the fecal microbiota at acute presentation. Principal component analysis (PCA) and response ratios (RRs) were calculated using all functional probes within the HuMiChip2 array. A, PCA analysis of day 0 samples grouped by country of deployment. B, RRs of significant changes in relative abundance of functional subcategories listed along the right side of the plot between the Kenya and Djibouti samples. The various subcategories are grouped into their parental gene category listed along the top of each subplot. RRs with 95% CIs that do not overlap the vertical dashed line at 0 were considered significantly different between groups. RRs to the right of the dashed line represent a higher relative abundance in Kenya samples, and RRs to the left of the dashed line represent a higher relative abundance in Djibouti samples. C, PCA analysis of day 0 samples grouped by disease impact on activity level. D, RRs of significant changes in relative abundance of functional subcategories between disease-impacted and nonimpacted samples. RRs to the right of the dashed line represent a higher relative abundance in “no impact” samples, and RRs to the left of the dashed line represent a higher relative abundance in “impacted” samples. Grouping by country of deployment and disease impact yielded significantly distinct clusters (multiresponse permutation procedure and Adonis P < .01). Ellipses within the PCA plots encompass 75% of points within groups.

There were distinct differences in functional gene composition measured in stool based on disease severity, as measured by illness impact on activity level (Figure 1C). Patients with moderate–severe disease (significant impact on activity) had a higher abundance of probes associated with L-phenylalanine and L-tyrosine synthesis, pantothenate and CoA, and peptidoglycan synthesis. Conversely, patients with mild disease (no activity impact) had more probes associated with glycine and glutamine amino acid synthesis, cysteine metabolism, biotin, sucrose, and putrescine/spermidine (Figure 1D). There were no differences in stool functional gene composition with regard to race and gender (data not shown).

Single-Dose Antibiotic Treatment Impact on Stool Functional Gene Composition

To assess the impact of single-dose antibiotic regimens on fecal functional microbiota, acute presentation samples (day 0) were compared with patient-matched postantibiotic samples (day 7 and/or 21). PCA analysis of functional probes did not result in significant clustering by time point (Supplementary Figure 2A). Furthermore, the number of probes detected and the Shannon diversity index did not significantly differ at different time points (Table 2). Probe abundance comparisons at the gene category level indicated differences in probes associated with amino acid metabolism/transport/synthesis, central carbon metabolism, cofactor biosynthesis, complex carbohydrates, and organic acids (Supplementary Figure 2B & C).

Table 2.

HuMiChip Diversity Metrics

	Day 0	Day 7	Day 21
Probe countsa	8748.9 ± 2355.5	8662.6 ± 2498.2	8499.9 ± 2172.2
Azithromycin	9278.7 ± 2453.6	7983.9 ± 2842.5	8561.4 ± 1542.0
Levofloxacin	8431.5 ± 2322.7	9215.8 ± 1290.8	8897.2 ± 2203.4
Rifaximin	8496.2 ± 2393.6	8674.9 ± 3660.9	7543.2 ± 2943.5
Shannon Indexa	9.03 ± 0.28	9.01 ± 0.33	9.00 ± 0.31
Azithromycin	9.09 ± 0.29	8.91 ± 0.40	9.03 ± 0.18
Levofloxacin	9.00 ± 0.28	9.11 ± 0.15	9.04 ± 0.35
Rifaxim	9.00 ± 0.30	8.98 ± 0.44	8.86 ± 0.38

^aMean ± SD.

Functional comparisons were performed within each antibiotic treatment group, and there were no major differences in overall functional composition using PCA (Figure 2A–C) and alpha diversity metrics (Table 2) of HuMiChip functional probes. While treatment group–specific alterations in probe abundances between acute and convalescent samples were detected (Figures 2D and E), no antibiotic treatment group had a more profound effect on the stool microbiota functional composition than others.

Figure 2.

Functional differences in the fecal microbiota between pre- and postantibiotic treatment samples by treatment group. Principal component analysis (PCA) and response ratios (RRs) were calculated using all functional probes within the HuMiChip2 array. PCA analysis of (A) day 0, (B) day 7, and (C) day 21. Ellipses within the PCA plots encompass 75% of points within groups (multiresponse permutation procedure and Adonis P > .05). RRs of significant changes in relative abundance of functional subcategories listed along the side of the plot between (D) day 0 and day 7 samples and (E) day 0 and day 21 samples. The various subcategories are grouped into their parental gene category listed along the top of each subplot. RRs with 95% CIs that do not overlap the vertical dashed line at 0 were considered significantly different between groups. RRs to the left of the dashed line represent a higher relative abundance in day 0 samples, and RRs to the right of the dashed line represent a higher relative abundance in (D) day 7 or (E) day 21 samples. PCA and RR colors correspond to the specific antibiotic treatment group. Only matched samples were considered in the analysis (must have provided a sample at each time point compared in the plot). Abbreviations: AZI, azithromycin; LEV, levofloxacin; RIF, rifaximin.

Stool Functional Composition Based on Predominant Etiologies at Presentation

While not feasible to evaluate the effect of specific etiologies on fecal functional gene composition due to number constraints, acute samples were grouped into either bacterial or viral etiologies. After excluding co-infections with viral and bacterial pathogens, patients largely had bacterial-associated AWD as determined by culture (47 of 53) or TAC (53 of 60) (see [7] for culture methods). Six patients by culture and 7 patients by TAC were determined to have viral-associated AWD (2 in common between detection methods). PCA analysis of all functional probes did not reveal major differences in stool functional composition (Supplementary Figure 3A & B); however, response ratio analysis did indicate relative abundance changes in various probe functional categories, including amino acid synthesis and metabolism, cofactor biosynthesis, feeder pathways to glycolysis, glycan and central carbon synthesis and metabolism, and complex carbohydrates (Supplementary Figure 3C & D).

Markers of Antibiotic Resistance Did Not Correlate With Single-Dose Antibiotic Treatment

To determine if single-dose antibiotics resulted in increased prevalence of antibiotic resistance determinants in the stool, we extracted all HuMiChip probes associated with antibiotic resistance and compared their abundance before (day 0) and after antibiotic treatment (day 21). For all treatment arms, no antibiotic resistance gene families were significantly altered in prevalence or abundance between the acute and convalescent samples (Figure 3; see Supplementary Table 2 for antibiotic resistance probe categories).

Figure 3.

Comparison of gene categories associated with antibiotic resistance in acute vs convalescent stool samples. Response ratios (RRs) showing no changes in the relative abundance of antibiotic resistance gene categories between day 0 and day 21 samples. The various antibiotic resistance gene categories are listed along the left of each plot. A detailed description of each category is given in Supplementary Table 1. RRs with 95% CIs that do not overlap the vertical dashed line at 0 were considered significantly different between groups. RRs to the left of the dashed line represent a higher relative abundance in day 0 samples, and RRs to the right of the dashed line represent a higher relative abundance in day 21 samples. Colors correspond to the specific antibiotic treatment group. Only matched samples were considered in the analysis (must have provided a sample at each time point compared in the plot). Abbreviations: AZI, azithromycin; LEV, levofloxacin; RIF, rifaximin.

The custom TAC included probes for 6 ESBL genes (kpc, ndm, cmy, ctx, shv, tem) to determine if ESBL prevalence increased after the single-dose antibiotic regimen (Supplementary Table 3). The kpc and ndm genes were not detected in any stool samples. Furthermore, cmy was only detected in 1 stool sample (0.95% of all samples) within the azithromycin treatment arm at day 21. The ctx gene had a slightly increased prevalence at day 21 compared with day 0 (8.57% to 15.24%), while shv had a slightly decreased prevalence from days 0 to 21 (24.76% to 18.10%) for all treatment arms. Similarly, high levels of tem were detected in day 0 and day 21 samples (85.54% and 86.75%, respectively).

DISCUSSION

Since their inception, antibiotics have revolutionized how we treat and prevent infections and continue to be a mainstay in management. Along with stewardship efforts to maximize antibiotic life cycles and ensure that effective drugs are available for current and emerging infections, efforts must also be put toward understanding the off-target health impacts of antibiotic use, including fecal microbiome dysbiosis and resistance acquisition. Our findings, in conjunction with the TrEAT TD clinical trial [7], suggest that single-dose antibiotic regimens are not only sufficient for TD treatment, but have minimal functional impact on the fecal microflora and acquisition of resistance microbial resident flora. To the best of our knowledge, this is the first study to analyze the functional properties of the fecal microbiota before and after single-dose antibiotic therapy.

Our study investigated 3 antibiotics, azithromycin, levofloxacin, and rifaximin, all of which have previously been shown to impact the composition and/or functionality of gut microflora in multidose studies to varying degrees [12–14]. Recently, levofloxacin prophylaxis was shown to minimally alter the host fecal microflora compared with other broad-spectrum beta-lactam antibiotics [13]. Conversely, in a study among antibiotic-naïve children in Niger, single-dose azithromycin compared with placebo was associated with a notable decrease in α- and γ-diversity at 5 days postantibiotic, although β-diversity was not significantly reduced [15]. Similarly, in a randomized placebo-controlled trial among children between 6 and 59 months in Burkino Faso, a 5-day course of azithromycin resulted in lower α-diversity compared with placebo, whereas courses of amoxicillin and cotrimoxazole did not differ [16]. Reports also have found that treatment of irritable bowel syndrome and TD with rifaximin does not appear to result in severe fecal flora perturbations, nor is it associated with acquisition of antibiotic-resistant bacteria [17, 18]. While not statistically significant, we noted that fecal diversity levels at days 7 and 21 were consistently higher in the levofloxacin group compared with azithromycin and rifaximin. This may support a hypothesis that single-dose levofloxacin may have a differential effect on fecal community dynamics during antibiotic treatment for gastrointestinal infections. Little is known about differential microbiota dynamics of single- vs multidose antibiotics in the context of acute enteric infections. In a controlled human infection challenge model of enterotoxigenic Escherichia coli (common TD pathogen), acute diarrheal illness was associated with decline in fecal microbiome diversity, though the study size was small (n = 5) and disentangling from antibiotic effects is difficult [19]. The observed “noneffects” for azithromycin and rifaximin may also result from single-dose rather than multidose administration. Future randomized controlled trials comparing single-dose and multidose regimens with TD are needed to understand different microbiota effects.

While microbiota functional diversity did not dramatically change after single-dose antibiotic treatment, there were subtle changes in certain gene functional categories. Specifically, probes associated with amino acid synthesis, transport, and metabolism were differentially abundant at pre- and postantibiotic time points. Previous research has demonstrated that amino acids have multiregulatory properties in the gut, resulting in abundance fluctuations in response to physiological stressors, including infection and antibiotic treatment [20, 21]. In our study, we observed amino acid–associated probes that were both more and less abundant at acute presentation (day 0) than convalescent (day 7 or day 21). Thus, it is hard to conclude the exact role of amino acid genes in response to TD or single-dose antibiotics other than to ensure homeostatic levels within the gut. Nevertheless, the finding of higher glutamine synthesis in our study is consistent with its positive role in reducing consequences of infectious diarrhea [22]. Glycine has also been described to have gastrointestinal tissue damage–protective effects in humans [23]. Moreover, exogenously produced vitamins are known to play a role in gastrointestinal health, and our observation of higher levels of biotin synthesis among individuals with lesser impact of infectious diarrhea is consistent given its anti-inflammatory effects and importance in gut microbial composition (diversity) [24]. Together, these data support the potential importance of amino acids and other microbial metabolites in reducing adverse health effects of intestinal infections.

The gene category associated with thiamine (vitamin B1) biosynthesis was the only one consistently more abundant in acute samples (day 0) compared with convalescence (days 7 or 21). Thiamine is an essential nutrient for humans and must be obtained either via diet or endogenous production by gut microflora. Overabundance of thiamine probes during acute presentation may indicate a microbiome composition with susceptibility to enteric infections or an infection-induced mechanism to drive thiamine production [25]. These data suggest that the role of thiamine in TD and/or antibiotic therapy warrants further investigation.

Two epidemiological factors that resulted in unique functional compositions at acute presentation were geographic location and disease severity. The observed country-specific functional fecal compositions in our study were not surprising given the numerous reports showing patient location as a major driver of microbiome composition [26, 27]. We noted substantial separation between US and UK military personnel in Djibouti and Kenya, respectively, suggesting geographic location or dietary factors related to cultural differences as prominent influencers of fecal microbiome functional composition. These data agree with results from the American Gut Project that indicated alpha-diversity differences between US and UK populations [28]. However, US service members stationed in Honduras did not cluster solely with the Djibouti patients, indicating a potential functional composition effect due to local duty station.

Fecal microbiome functional composition was distinguishable between those who had mild vs moderate–severe TD. This link between gut microbiota and illness severity has been previously observed in numerous human diseases, including irritable bowel disease, malaria, and coronary artery disease [29–31]. Notably, we detected increased abundance of sucrose, biotin, and polyamine genes in individuals with mild disease. Abundance of these functional categories has been previously associated with a “healthy” gut flora [24, 32] and may contribute to modulation of TD symptom severity.

Finally, we wanted to determine if HuMiChip data could distinguish between bacterial and viral TD. Delineating between these 2 etiological classes can vastly alter treatment strategies and avoid unnecessary antibiotic administration. While some functional categories were significantly different between the 2 classes, the overall functional compositions were indistinguishable. This may be a limitation of the probe set contained within the HuMiChip array; however, it may also suggest that there is no functional difference between bacterial and viral TD. Previous research has shown that the composition of fecal microflora in travelers with diarrhea is similar to healthy travelers, regardless of etiology [33]. Continued research is needed to inform diagnostic strategies for delineation of viral vs bacterial etiologies in acute enteric infections.

Our study is not without limitations. First, we do not have access to stool samples from patients before TD development (ie, healthy controls) as this was a post hoc analysis of a clinical trial aimed at AWD disease resolution [7]. Given that most patients (96%) were clinically cured by 3 days post-treatment, day 21 stool samples should provide a proxy for healthy controls [34]; however, it can take longer for post-TD/antibiotic homeostasis to return, so day 21 samples may not fully approximate a postinfectious healthy state. It is important to note that all TrEAT TD participants had AWD (dysentery or febrile diarrhea were excluded), and we lacked travelers within the Asian continent, which is known to have high ESBL acquisition rates. Thus, these results may not apply to nonwatery diarrhea TD patients nor be globally generalizable. Second, HuMiChip technology has a finite set of probes and may not capture all functional genes within a sample. Furthermore, the probes target microbial genomic DNA and may not reflect expression levels of gene targets. Despite these probe limitations, HuMiChip and its closely related GeoChip technologies have documented success in characterizing functional composition of microbial communities and will continue to be implemented in future studies [9, 35, 36]. Our study was limited to collection and interpretation of fecal specimens. It is known that acute TD infections largely occur in the small intestine, and in the case of rifaximin, colon bioavailability is limited. Thus, studies that could evaluate dynamics in small intestine space may be more informative, though challenging to conduct in a clinical TD setting. Finally, there were limitations with respect to our ascertainment of ESBL, which relied on molecular detection methods. To deem an ESBL or pathogen as “present” in a sample, cycle thresholds must be set within the TAC. With guidance from previous literature and a recent validation study using this platform, we opted for a cycle threshold of 35 [10, 37]. However, PCR-based discrimination of clinical antimicrobial resistance is not yet a certain science, so these data should be interpreted with caution. Furthermore, we limited our ESBL targets to 6 genes, which are only a subset of the possible genes associated with the ESBL phenotype. Selection of these genes was based on previous research demonstrating their preponderance in ESBL-producing Enterobacteriaceae at the time of TAC card construction [6]t and the limitation of targets allowable on our custom-made TAC card, which was primarily designed to detect travel-associated enteropathogen identification.

Overall, he benefits of single-dose antibiotic therapies over multidose, multiday regimens are becoming more apparent, including simplicity, with likely increased adherence and similar or increased efficacy [7, 38, 39]. Antibiotics are known to be among the most potent modulators of gut microflora, which can permit colonization by opportunistic pathogens and select for antibiotic resistance determinants [40]. Our findings demonstrate that single-dose azithromycin, levofloxacin, or rifaximin has minimal impact on the functional composition of the fecal microbiome and antibiotic resistance acquisition. Taken together, the efficacy of single-dose therapy for TD in the TrEAT TD study [7] and the minimal impacts on fecal functional composition, as shown herein, continue to support single-dose therapy as recommended TD treatment. However, to fully understand regimen effects, we strongly urge future studies to examine associations between antibiotic use and impacts on bowel microflora in single-dose vs multidose regimens (with and without loperamide).

Supplementary Data

Supplementary materials are available at Open Forum Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Acknowledgments

Financial support.  This study (IDCRP-065) was supported by the Infectious Disease Clinical Research Program (IDCRP), a Department of Defense (DoD) program executed by the Uniformed Services University of the Health Sciences (USUHS) through a cooperative agreement with The Henry M. Jackson Foundation for the Advancement of Military Medicine, Inc (HJF). This project has been funded in whole, or in part, with federal funds from the Defense Health Program (Award DHP_6.7_15_C2_I_15_J9_1319), the Bureau of Medicine and Surgery to the Uniformed Services University of the Health Sciences (USU Grant Agreement HU0001-11-1-0022; USU Project No. G187V2), the National Institute of Allergy and Infectious Diseases, National Institutes of Health (NIH), under Inter-Agency Agreement Y1-AI-5072, and the Defense Health Program, US Department of Defense, under award HU0001190002.

Disclaimer.  The contents of this publication are the sole responsibility of the authors and do not necessarily reflect the views, opinions, or policies of Uniformed Services University of the Health Sciences (USU), the National Institutes of Health or the Department of Health and Human Services, the Department of Defense (DoD), the Departments of the Army, Navy, or Air Force, The Henry M. Jackson Foundation for the Advancement of Military Medicine, Inc., or the UK Ministry of Defence. Mention of trade names, commercial products, or organizations does not imply endorsement by the US Government.

Potential conflicts of interest.  The authors do not declare any conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Author contributions.  Study design: J.F., A.R.M., D.R.T., M.S.R. Data collection: J.D.V.N., M.T., B.S., M.P.S., P.C., J.F., D.R.T., M.S.R. Data analysis and interpretation: R.C.J., J.D.V.N., A.R.M., D.R.T., M.S.R. Drafting the manuscript: R.C.J., D.R.T., M.S.R. All authors have reviewed and approve the drafted manuscript.