Up to four biannual administrations of mass azithromycin treatment are associated with modest changes in the gut microbiota of rural Malawian children

Community-level mass treatment with azithromycin has been associated with a mortality benefit in children. However, antibiotic exposures result in disruption of the gut microbiota and repeated exposures may reduce recovery of the gut flora. We conducted a nested cohort study to examine associations between mass drug administration (MDA) with azithromycin and the gut microbiota of rural Malawian children aged between 1-59 months. Fecal samples were collected from the children prior to treatment and 6 months after two or four biannual rounds of azithromycin treatment. DNA was extracted from fecal samples and V4-16S rRNA sequencing used to characterize the gut microbiota. Firmicutes, Bacteroidetes, Proteobacteria and Actinobacteria were the dominant phyla while Faecalibacterium and Bifidobacterium were the most prevalent genera. There were no associations between azithromycin treatment and changes in alpha diversity, however, four biannual rounds of treatment were associated with increased abundance of Prevotella. The lack of significant changes in gut microbiota after four biannual treatments supports the use of mass azithromycin treatment to reduce mortality in children living in lowand middle-income settings. Introduction Azithromycin is a broad-spectrum, macrolide antibiotic characterized by a long intraand extra-cellular half-life. Its use is indicated in the treatment of atypical pneumonia, skin and soft tissue infections and sexually transmitted infections. The World Health Organization (WHO) recommends mass azithromycin treatment at the community level as one of the key strategies for the elimination of trachoma as a public health problem. Studies of mass azithromycin distribution for trachoma control in endemic areas indicate that mass treatment has secondary effects, which include reductions in child morbidity and mortality. Recently, a multi-site trial, conducted in Malawi, Tanzania and Niger reported lower mortality rates in children (under 5 years of age) who received mass azithromycin treatment compared to children who received a placebo. The specific mechanism through which azithromycin reduced mortality in children is not understood, however, several studies have reported a reduction in the community burden of nasopharyngeal carriage of Streptococcus pneumoniae8-10 and a reduction in the Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 6 October 2020 doi:10.20944/preprints202010.0138.v1


Introduction
Azithromycin is a broad-spectrum, macrolide antibiotic characterized by a long intra-and extra-cellular half-life. Its use is indicated in the treatment of atypical pneumonia, skin and soft tissue infections and sexually transmitted infections 1

. The World Health
Organization (WHO) recommends mass azithromycin treatment at the community level as one of the key strategies for the elimination of trachoma as a public health problem 2 .
Studies of mass azithromycin distribution for trachoma control in endemic areas indicate that mass treatment has secondary effects, which include reductions in child morbidity and mortality [3][4][5][6] . Recently, a multi-site trial, conducted in Malawi, Tanzania and Niger reported lower mortality rates in children (under 5 years of age) who received mass azithromycin treatment compared to children who received a placebo 7 . The specific mechanism through which azithromycin reduced mortality in children is not understood, however, several studies have reported a reduction in the community burden of nasopharyngeal carriage of Streptococcus pneumoniae [8][9][10] and a reduction in the abundance of Campylobacter spp. in the gut 11 . Additionally, reduced risks of diarrhea and acute lower respiratory infections related to azithromycin mass drug administration (MDA) for trachoma control have been reported in Tanzanian children 4,5 . Thus, azithromycin may reduce morbidity and mortality by reducing carriage of pathogenic bacteria.
Available evidence indicates that azithromycin treatment causes alterations in the gut microbiota that can be measured in the weeks immediately following treatment. Recent randomized, placebo-controlled trials of Parker et al. 12 and Wei et al. 13 characterized the intestinal microbiota of Indian and Danish infants respectively, at baseline and 14 days after a 3-day course of azithromycin and reported changes in intestinal microbiota. Both studies reported a decrease in alpha diversity in fecal samples of treated children compared to those who received placebo. Additionally, Parker et al. 12 reported a decrease in relative abundance of Proteobacteria and Verrucomicrobia while Wei et al. 13 reported a decrease in the abundance of Actinobacteria after treatment in fecal samples of treated children compared to placebo. Alterations in the gut microbiota have also been reported more than 6 months after exposure to antibiotics, suggesting that the shortterm changes may persist for a longer period of time. An observational study by Korpela et al. 14 , which characterized the fecal microbiota of Finnish children aged 2-7 years with varying durations of exposure to azithromycin, clarithromycin, penicillin or no antibiotic exposure over a 24-months span, showed a decrease in microbial richness in children treated with macrolides compared to those treated with penicillin or without exposure to antibiotics. Additionally, a significant decrease in the relative abundance of Actinobacteria and an increased abundance of Bacteroidetes and Proteobacteria was found in children who had used macrolide antibiotics within 6 months of sample donation. However, Wei et al. 13 reported no significant differences in alpha diversity and taxonomic composition between the treatment and control groups 4 years after treatment. To date, it is not clear how long the azithromycin-related changes in the gut microbiota persist. A study investigating the intestinal microbiota of three individuals over a period of 10 months covering two courses of the antibiotic ciprofloxacin reported loss of diversity and a change in community composition that occurred within 3 to 4 days of initiating a course 15 . Approximately 1 week after the end of each treatment course, the return to a pre-treatment state was still incomplete.
In the current study, we investigated the association between azithromycin treatment and changes in the fecal microbiota of rural Malawi children 6 months after 2 or 4 biannual azithromycin distributions. We explored the diversity and composition of the intestinal microbiota in children who received azithromycin versus children who received placebo.   A retrospective nested cohort study was then conducted within the framework provided by the three cross-sectional surveys described above. Of the recruited children who provided fecal samples at the baseline cross-sectional survey, 121 were also sampled, by chance, at an additional survey thereby providing the potential to generate longitudinal pairs of samples (before and after treatment). Baseline samples from 103 such children had sufficient volume remaining for 16S rRNA gene sequence analysis and were therefore included in the nested cohort study (Figure 1). Fifty-four of these children were sampled at the baseline and 12-month surveys and 55 children were sampled at the baseline and 24-month surveys (Figure 1).

Fecal sample collection
Fecal samples from participating children were collected by their mothers or guardians, who were provided with sterile fecal collection bottles (Wheaton, UK) and given verbal instruction, by a nurse, on how to collect the sample. Fecal samples were returned to the field team as soon as possible after collection. These samples were put in a cool box with ice packs until the end of the day (not more than 8 hours after collection) and were brought to the laboratory where they were stored at -80°C.

V4 16S rRNA gene sequencing
Samples were thawed at room temperature and 250mg amounts weighed into sterile Eppendorf tubes. Total, genomic DNA was then extracted using the PowerSoil DNA Isolation Kit (MO BIO Laboratories Inc, Carlsbad, CA, now a part of Qiagen, Germany).

Sequence processing
FastQ files, containing raw sequence data, were processed in QIIME 2 software (version 2018.6) 19 . Barcoded sequences were demultiplexed using the demux function in QIIME 2. Poor quality reads were filtered out and chimeras were removed.
The 16S rRNA sequences were clustered de novo into OTUs at ≥ 97% identity.
Taxonomy was assigned to the OTUs using a naïve Bayes classifier pre-trained on the SILVA 16S database 20 . To exclude spurious OTUs, only bacterial OTUs identified to the genus level, with sequences more than 0.005% of the total number of sequences 21 and a frequency of more than 0.01% in any sample were retained in the analyses.

Statistical analysis
All statistical analyses were performed in R 22 . Shannon (H) and Simpson (D) diversity indices were calculated using the phyloseq package 23 . Differences in the distribution of parametric data between groups were tested using Student's t-test or ANOVA while the Information and consent forms were translated into local languages (Yao and Chichewa) prior to their approval by the local ethics committee. Consent was first obtained at the community level through discussions with the village chief and community elders who then indicated the willingness, or unwillingness, of the community to participate through verbal consent. Written, informed consent (by thumbprint or signature) was then obtained from the parent or guardian of each child before recruitment. During the consenting process, all parents and guardians were informed of their freedom to withdraw their child from the study at any time without giving reason for doing so. All subjects gave their informed consent for inclusion before they participated in the study.
The study was conducted in accordance with the Declaration of Helsinki.

Baseline demographics
A baseline survey of prevalence of carriage of macrolide resistant enteropathogens enrolled 1090 children, however, of these, only 709 (65%) returned fecal samples. Of these 709 children, 103 (9%) were included in the nested cohort study that examined the fecal microbiota using 16S rRNA sequence analyses. One baseline sample consistently failed to amplify and could not be sequenced; 102 samples were retained for analysis after quality filtering. The age and sex of children whose fecal samples were included in the analysis of the nested cohort compared to all children enrolled in the cross-sectional survey are shown in Table 1. The proportion of male children was comparable between those whose fecal samples were included in this nested study and all enrolled participants. However, children whose fecal samples were included in the nested cohort were younger compared to all enrolled participants.  Figure 2B).

Major phyla. The proportion of total number of reads for each phylum in a sample
represents phylum abundance after rarefaction to 1000 reads. B. Major genera. The proportion of total number of reads for each genus in a sample represents genus abundance after rarefaction to 1000 reads. The stacked bar plot only shows the 10 most abundant genera. Genera with relative abundance < 1% were grouped as "Other".

Definition of datasets and baseline demographics of the datasets
Samples from 54 children who were sampled at baseline and at the 12-month survey, which was conducted 6 months after 2 rounds of biannual treatment, were included in the final analysis. The data from these samples are referred to as the "BL vs 2MDA dataset". Samples from 55 children who participated in baseline sampling and sampling at the 24-month survey, conducted 6 months after 4 rounds of biannual treatment, were included in the final analysis. This data set is referred to as the "BL vs 4MDA dataset".
Within the BL vs 2MDA dataset, 30 children received placebo and 24 received azithromycin treatment while within the BL vs 4MDA dataset, 30 children received placebo and 25 received azithromycin treatment (Figure 1). Age and sex of children were comparable between azithromycin and placebo arms in both data sets (Table 2).

No change in alpha diversity measures observed following 2 and 4 treatment rounds
The BL vs 2MDA dataset was used to explore changes in microbiota diversity 6 months after 2 rounds of azithromycin MDA. At the 12-month survey, no differences in alpha diversity, measured by Shannon and Simpson diversity indices, were found between the azithromycin and placebo treatment arms (Table 3). Similarly, there was no difference in Shannon or Simpson diversity index distribution between baseline and 12-month fecal samples within the azithromycin arm ( Table 4).
The BL vs 4MDA dataset was used to assess the effect of up to 4 rounds of biannual azithromycin treatment on fecal microbiota diversity 6 months after the last treatment round. A comparison of fecal microbiota diversity between azithromycin and placebo arms after 4 rounds of azithromycin had been distributed showed no differences in Shannon and Simpson diversity indices ( Table 3)

. A comparison of Shannon and
Simpson diversity indices between fecal samples collected baseline and 24-month surveys in the azithromycin group in the unadjusted analysis showed differences between the two time-points but no differences were observed after adjusting for age and sex ( Table 4).   Generalized linear mixed models, adjusted for age, sex and time since last treatment with the participant as a random effect, were then used to assess the longitudinal differences in the relative abundance of individual genera by treatment arm. The analyses showed no significant differences in the relative abundance of Streptococcus between the baseline and 24-month surveys in the azithromycin arm. There were also no significant differences in the relative abundance of Bifidobacterium between the baseline and 24-month surveys in both the azithromycin and placebo arms. However, azithromycin treatment was associated with increased relative abundance of Prevotella at the 24-month survey; samples collected at the 24-month survey in the azithromycin arm had 30% increased odds of relative abundance of Prevotella compared to samples collected at the baseline survey within the same azithromycin arm (Table 5). No longitudinal differences were shown in the placebo arm in the adjusted analysis ( Table   5).

Discussion
We examined associations between 2 or 4 biannual azithromycin treatment rounds and changes in intestinal microbiota diversity and composition in children resident in treated communities. Azithromycin administration was not associated with changes in alpha diversity but was weakly associated with changes in gut microbiota composition after 4 biannual treatment rounds.
The lack of an effect of azithromycin treatment on microbiota diversity is consistent with data by Wei et al. 13 , although the follow-up period in the present study was shorter. In the study by Wei et al. 13 , Danish children aged 12-36 months were prescribed a 3-day course of azithromycin or placebo and the gut microbiota was characterized in fecal samples collected from each child at 4 years of follow-up. The study reported no significant differences in alpha diversity (measured by observed richness and Shannon index) between the treatment and placebo groups.
Conflicting results on alpha diversity have been recently reported by the MORDOR trial in Niger 24 . A reduction in inverse Simpson and Shannon diversity indices in children who received azithromycin compared to children who received placebo was detected 6 months after 2 biannual treatment rounds. The discrepancy in the findings between that study and our own could be attributed to differences in sequencing methods. In the present study, we performed V4-16S rRNA sequencing while the study in Niger performed whole genome sequencing, which has higher resolution to detect differences.
Four biannual azithromycin treatment rounds in the present study were weakly associated with increased abundance of Prevotella, detectable 6 months after the last treatment round. While this finding has not previously been reported, it is consistent with a recent study that reported an increased abundance of Bacteroidetes, the phylum to which Prevotella belongs, in children who received macrolides within the 6 months preceding sample collection 14 . Prevotella is a gram-negative commensal bacterium found at mucosal sites of the respiratory tract, gut, and oral cavity. Reduced abundance of this bacterium has been associated with Crohn's disease in pediatric patients 25 .
Therefore, a trend towards increased abundance of Prevotella 6 months after 4 biannual azithromycin treatment rounds may suggest potential long-term beneficial effects of mass azithromycin treatment to the gut, although further studies from additional study sites would be needed to validate this finding.
The absence of large-scale changes in the gut microbiota 6 months after 2 or 4 biannual azithromycin treatment rounds suggests that any effects of azithromycin mass treatment on the gut microbiota may last days or weeks after treatment but not months. Our observation that there were modest changes in gut microbiota composition 6 months after 4 biannual rounds of azithromycin treatment but not 6 months after 2 biannual rounds suggests that the long-term effects of azithromycin MDA on the gut microbiota may be dependent on the number of treatments.
Our study has several limitations. The study was reliant on samples collected in larger cross-sectional surveys and participation in those surveys was low. At the baseline survey, 1090 children were enrolled and provided with stool sample collection kits.
However, only 709 (65%) returned stool samples. Similar participation rates were seen at the 12-and 24-month surveys. This may have resulted in selection bias, particularly if families of children who participated had different health-seeking behaviors than those families of children who chose not to participate. Furthermore, the samples included in the present nested cohort analyses were selected based on the availability of longitudinal pairs (baseline and 12 months or baseline and 24-month), which may also be prone to selection bias. Another limitation was that not all of the participants who were sampled received antibiotic as scheduled at each treatment round. While it is unusual to achieve 100% treatment coverage in MDA programs, having participants miss treatment might have contributed to the magnitude of the effect of azithromycin on the gut microbiota reported by this study.
Funding for this work was provided by the Bill and Melinda Gates Foundation, under grant numbers OPP1066930 and OP1032340.