Dynamic changes in Antibiotic Resistance Genes and Gut Microbiota after H. Pylori Eradication Therapies

Background: Short-term antibiotics exposure is associated with alterations in microbiota and antibiotic resistance genes (ARGs) in the human gut. While antibiotics are critical in the successful eradication of Helicobacter pylori, the short-term and long-term impacts on the composition and quantity of antibiotics resistance genes after H. pylori eradication is unclear. This study used whole genome shotgun metagenomic of stool samples to characterize the gut microbiota and ARGs, before and after H. pylori eradication therapy. Results: Forty-four H. pylori-infected patients were recruited including 21 treatment naïve patients who received clarithromycin-based triple therapy (CLA group) and 23 patients who failed previous therapies, in which 10 received levooxacin-based quadruple therapy [LEVO group] and 13 received other combinations [OTHER group] in the current study. Stool samples were collected at baseline (before current treatment), 6-week and 6-month after eradication therapy. At baseline, there was only a slight difference among the three groups on ARGs and gut microbiota. After eradication therapy, there was a transient but signicant increase in gut ARGs 6-week post-therapy, among which the LEVO group had the most signicant ARGs alteration compared to other two groups. For treatment naïve patients, those with higher ARG richness and ErmF abundance were prone to fail CLA eradication. For gut microbiota, the bacteria richness decreased at 6-week and there was a signicant difference in microbiota community among the three groups at 6-week. Conclusions: Our ndings demonstrated the dynamic alterations in gut microbiota and ARGs induced by different eradication therapies, which could inuence the choices of antibiotics in eradication therapy.


Background
Successful eradication of H. pylori by the combination of antibiotics lowers the subsequent risk of gastric cancer and peptic ulcer disease [1,2]. However, due to the widespread use of antibiotics, antibiotic resistance and emergence of drug-resistance bacteria has become one of the largest global health threats [3]. In particular, increasing antibiotic resistance to clarithromycin has reduced the effectiveness of standard clarithromycin-based triple therapy (STT) in H. pylori eradication therapy [4]. Hence, bismuth quadruple or non-bismuth quadruple, concomitant therapies with three antibiotics have been recommended in areas with background prevalence of clarithromycin resistance higher than 15% according to the Maastricht V/Florence Report, Toronto Consensus, and American College of Gastroenterology (AGG) [5][6][7]. In patients who have failed rst-line eradication therapy, alternative therapies such as levo oxacin-containing and other combinations, including bismuth or rifabutin, are suggested.
However, a growing body of evidence suggests that even a short course of antibiotics may lead to disruption of the balance in the gut microbiota, or dysbiosis [8][9][10][11]. It is increasingly recognized that altered gut microbiome could be associated with various gastrointestinal diseases, metabolomic disorders like diabetes mellitus and even central nervous systems (CNS) disorders [12][13][14]. With the wide availability of advanced DNA sequencing technologies, whole genome sequencing (WGS) could provide a deeper analysis of microbial diversity [15][16][17]. However, most studies on the consequences of H. pylori eradication on gut microbiota are based on the 16S rRNA metagenomic analysis, which was less effective in resolving microbial species.
Metagenomic studies have shown that antibiotic resistance genes (ARGs) are widespread in natural environments and the human gut serves as one of the major reservoirs [18][19][20]. Further analysis showed that the ARGs in human gut microbiota could differ across populations [21]. While macrolide resistance genes could overrepresent after short-term clarithromycin exposure [22], various antibiotics may exert diverse effects on microbiota composition and even antibiotic resistance [23]. It remains unclear how ARG composition changes as a result of H. pylori eradication therapies.
To address these questions, we have conducted shotgun metagenomic sequencing to investigate the impacts of short-term antibiotics exposure related to different H. pylori eradication therapies, including patients with prior treatment failures, on gut antibiotic resistance genes (ARGs) alteration. Moreover, we determined the dynamic changes in gut microbiota composition after different H. pylori eradication regimes.

Patient and HP eradication therapies
This was a prospective study including adult patients (> 18 years), who were diagnosed to have active H. pylori infection, including both treatment naïve and previous treatment failure. Patients were recruited in the Ulcer Clinic of the Queen Mary Hospital of Hong Kong. As a routine clinical practice, H. pylori-infected and treatment naïve patients were given conventional clarithromycin-containing triple therapy (CLA group; esomeprazole 20mg, amoxicillin 1g, clarithromycin 500mg, all given twice a day) for at least one week.
For patients who had failed previous H. pylori eradication therapies, they were given either levo oxacincontaining therapy (LEVO group; esomeprazole 20mg twice a day, levo oxacin 750mg daily, tetracycline 500mg three times a day, and metronidazole 400mg three times a day for at least one week) or other combinations (OTHER group; esomeprazole 20mg twice a day, amoxicillin 1g twice a day and rifabutin 600mg daily for 2 weeks; esomeprazole 20mg twice a day, bismuth 262mg three times a day, metronidazole 400mg three times a day and tetracycline 500mg three times a day for at least 10 days; esomeprazole 20mg twice a day, moxi oxacin 400mg daily, nitrofurantoin 100mg three times a day and ursodeoxycholic acid 750mg daily for one week) depending on their previous treatment regimens.
Fecal sample collection, library preparation, and sequencing Fecal samples were collected from patients at three different time points. First stool samples were collected before current H. pylori eradication therapy, whereas second stool samples were collected 6 weeks after the completion of anti-H. pylori therapy. A follow-up urea breath test (C 13 -UBT) was arranged to document treatment outcome at 6-week. A third stool sample was collected 6 months after the eradication therapy. All fecal samples were stored in the OMNIgene⋅Gut collection kit (DNA Genotek, Ottawa, Canada) before transferred for storage at -20°C. Library preparation was conducted according to the protocol of the HAPA Hyper Prep Kit (KR0961-V1.14), which had been recently veri ed in our pilot study to have the best yield for ARG detection [24]. Subsequently, metagenomic sequencing was performed on the Illumina NovaSeq 6000 System (Illumina, USA; Paired-end; read length, 2 x 150 bp) at the Centre of PanorOmic Science (CPOS) of the University of Hong Kong.
Metagenomic data pre-processing FastQC v0.11.9 was used for quality control of the sequencing data. Low-quality reads were removed, and low-quality bases at the 3' or 5' end of the reads were trimmed. Human sequence contaminations were ltered by BBMap v38.86 [25] with parameter minid (minimum alignment identity) = 0.95. After quality control and ltering, each sample contains an average of 54

ARG identi cation and abundance normalization
A command-line version of RGI v5.1.1 (The Resistance Gene Identi er) [27] together with the latest reference database were downloaded from the Comprehensive Antibiotic Resistance Database [28] (CARD) for identifying ARGs from contigs using default parameters. RGI predicted open reading frames (ORFs) from input contigs using Prodigal, and then the translated protein sequences from these identi ed ORFs were aligned to the known ARGs reference sequences in the CARD. RGI reported ARGs under three criteria: Perfect (totally matched to the reference sequences or mutations in the CARD), Strict (more exible, to ensure the detection of a functional ARG), and Loose (outside cut-offs or partial hits for the detection of potential novel ARG). In this study, Perfect and Strict ARGs with identity higher than 90% were considered as detected unique ARGs. The number of ARGs detected in each sample was the number of unique ARG names generated by RGI output. The drug class, ARG subtypes, and resistance mechanisms were identi ed according to the antibiotic resistance ontology classi cation provided by the CARD.
The abundance of ARGs was represented by RPKM (reads per kilobase of reference sequence per million sample reads). For each sample, all reads are mapped to the reference protein sequences of reported ARGs by ShortBRED [29]. Read alignment software USEARCH [30] was called by ShortBRED with pair-end reads treated as separated reads. More speci cally, usearch_local command is used with 95% identity and 95% read length coverage (by nucleotide length). ShortBred normalizes the number of reads mapped to each ARG by the total number of reads in the sample and the length of the reference sequences (all lengths are calculated based on nucleotide), and report RPKM value for each ARG in each sample. The sum of the abundance (RPKM) of each ARG resistant to the same class of drug was de ned as the ARG class's relative abundance. The number of changed ARGs was de ned as the difference between the number of observed unique ARGs between two time points.

Microbiota taxonomic pro ling and diversity analysis
Taxonomic pro ling of the microbiome was performed on quali ed FASTQ les using MetaPhlAn2.0 [31], which mapped raw reads to the database for the detection of the presence and read coverage of cladespeci c makers and estimate the relative abundance. The relative abundance of taxonomic levels of species, genera, and phylum was extracted from MetaPhlAn2.0 output for downstream microbiota analysis. Alpha diversity estimated the richness and evenness of each sample while the beta diversity estimated the microbiome dissimilarity between samples. The alpha diversity indices (Shannon index, evenness) in each sample was calculated based on the relative abundance of each species using the vegan package in R [32]. Species richness was de ned as the number of species detected in each sample. For estimation of microbiota community (beta diversity), Bray-Curtis distance was calculated based on phyloseq [33] package in R based on the relative abundance of species level.

Statistical analysis
All statistical analysis was performed in the R software v4.0.3 or GraphPad Prism 9.0 unless otherwise stated. Baseline characteristics were expressed as mean ± SEM for continuous data and n/N (%) for proportional data. For the comparison of change of observed number of ARGs, Student's t-tests were applied. For taxonomic pro ling and potential ARGs, the Wilcoxon signed-rank test or Mann-Whitney U test was used for paired and unpaired samples respectively for metagenomic data, and the Benjamini-Hochberg procedure [34] was used to decrease the false discovery rate (FDR) when multiple comparisons were applied. Principal coordinates analysis (PCoA) of beta-diversity of gut microbiota between or within groups was visualized based on Bray-Curtis distance matrices. Permutational multivariate analysis of variance (PERMANOVA) was then performed to compare microbiota community dissimilarity using analysis of variance using distance matrices (ADONIS) for 999 permutations.

Results
Patient's characteristics A total of 44 H. pylori-infected patients were enrolled in the study, including 21 patients in the CLA group, 10 patients in the LEVO group, and 13 patients in the OTHER group. The average number of prior failed eradication therapies in the CLA, LEVO, and OTHER group was 0, 1.1, and 2.5, respectively. Other baseline characteristics were similar among the three groups (Table S1). A total of 121 stool samples were collected. All patients completed the rst (baseline) and second (6-week post-therapy) stool sample collection, while 14 patients (66.7%) in the CLA group, 8 patients (80%) in the LEVO group, and 11 patients (84.6%) in the OTHER group had completed the third (6-month post-therapy) stool sample collection.

Baseline ARG differences
At baseline, the abundance of tetM (p = 0.005) was more abundant in the LEVO group when compared with the CLA group. In the OTHER group, the relative abundance of ErmF (p = 0.015) was more abundant, while the abundance of tetA(46) (p = 0.03) and Klebsiella pneumonia KpnH (p = 0.006) were less abundant compared with the CLA group (Table 1). There was however no signi cant difference in the ARG classes and ARG richness among the three groups ( Figure S1).  1B).
To better understand the changes in gut ARGs, we further examined the unique ARGs in different treatment groups. In the CLA group, the abundance of ErmF (MLS resistance genes, p = 0.0004) and tetO (tetracycline resistance genes, p = 0.005) were signi cantly increased at 6-week compared to baseline level ( Fig. 2A). In the OTHER group, only ErmB (MLS resistance genes, p = 0.04) was signi cantly decreased at 6-week compared with baseline (Fig. 2B). The LEVO group had the highest number of signi cantly altered ARGs among the three groups. A total of 28 unique ARGs were signi cantly increased at 6-week compared to the baseline level (Fig. 2C). Over half of the altered ARGs (19/28) belonged to multidrug resistance genes, the rest of them (9/28) were aminocoumarin, aminoglycoside, beta-lactam, nitroimidazole, and rifamycin resistance genes. However, there was no signi cant difference in the abundance of unique ARGs at 6-month compared with baseline in all three treatment groups.
ARG difference between successfully eradicated and failed patients in the CLA group In the CLA group, there was a signi cant difference in the ARG richness and unique ARGs between those who were successfully eradicated and failed eradication therapy. The number of observed unique ARGs was signi cantly higher in the failed patients as compared with those successfully eradicated at baseline (p = 0.019) and 6-month (p = 0.015), but not at 6-week (p = 0.29) after the eradication therapy ( Figure  S3A). The Shannon index also showed that the alpha diversity of ARGs was signi cantly higher in failed patients (p = 0.015) at 6-month ( Figure S3B).
In addition to the ARG diversity difference, the expression of unique ARGs was also differentially enriched. At baseline, the abundance of ErmF (MLS resistance genes) was signi cantly higher in patients who failed eradication therapy ( Figure S4A). At 6-month, 25 unique ARGs were highly abundant in failed patients compared with successfully eradicated patients ( Figure S4B).

Gut microbiota changes after H. pylori eradication therapies
We also performed taxonomy analysis of the gut microbiota to explore the microbiota composition and diversity alteration in the three groups after eradication therapies. At baseline, the species richness was signi cantly lower in the OTHER group (p = 0.021) which had an average of two prior treatment failures, when compared with the CLA group (treatment naïve, Figure S5). At the phylum level, Bacteroidetes was the most abundant phylum, followed by Firmicutes, Actinobacteria, and Proteobacteria phylum, which account for over 98% of the microbiota (Figure S6A). At the genus level, Ruthenibacterium (p = 0.012) and Phascolarctobacterium (p = 0.001) were less abundant in the LEVO group, while Eggerthella, Gordonibacter, Anaerostipes, and Parasutterella were more abundant (p = 0.004, p = 0.011, p = 0.019, p = 0.027, respectively) as compared with the CLA group (Table 1). There was however no signi cant difference in beta diversity among the three groups at baseline (Fig. 3A).
The diversity of microbiota was altered after eradication therapy. There was a signi cant difference in beta diversity among the three groups at 6-week (PERMANOVA, p = 0.002; Fig. 3B) but not at 6-month (PERMANOVA, p = 0.323; Fig. 3C). No signi cant difference in beta diversity was observed within treatment groups ( Figure S7). Compared with baseline, the species richness index (alpha diversity) was signi cantly decreased in the CLA group (p = 0.016), the LEVO group (p = 0.046), and the OTHER group (p = 0.001) 6-week after the eradication therapy (Fig. 3D). The Shannon index (alpha diversity) was also signi cantly decreased at 6-week in the CLA group (p = 0.042) and the OTHER group (p = 0.0009), but not in the LEVO group (p = 0.77) (Fig. 3E). However, compared with baseline, alpha diversity (both richness and Shannon index) was restored 6-month after the eradication therapy ( Fig. 3D and 3E).
At the phylum level, only Firmicutes was observed to signi cantly changed at 6-week after the eradication therapy, while other dominant phylum had no signi cant difference at either 6-week or 6-month ( Figure  S6B). The average relative abundance of Firmicutes increased in the CLA group (29.53-37.60%, p = 0.032), had an increasing trend in the LEVO group (27.17-37.38%, p = 0.074), but decreased in the OTHER group (29.22-15.96%, p = 0.004) at 6-week. All restored to baseline levels at 6-month (Fig. 4A).

Discussion
In this study, we have performed a comprehensive review of the dynamic changes of human gut microbiota and ARGs in patients who underwent various H. pylori eradication therapies using the whole genome sequencing method. Overall, we observed a transient but dramatic shift in ARG richness, ARG class level, and unique ARGs at 6-week after the eradication, especially in patients treated with levo oxacin-based therapy. All of these changes on ARGs were restored at 6-month after eradication therapy.
In addition to the change of ARGs, we also found signi cant changes in gut microbiota. The alpha diversity signi cantly decreased at 6-week in all three treatment groups, which all restored at 6-month.
The microbiota community structure (beta diversity) was signi cantly separated at 6-week after different eradication therapies among the three groups. Moreover, microbiota difference was observed at 6-week after the treatment at different taxonomy levels and partially restored at 6-month. For treatment naïve patients, those with higher ARG richness and ErmF gene abundance at baseline were prone to failure of clarithromycin-based triple therapy, and there was also high diversity and abundance of unique ARGs 6month after the eradication therapy.
Our results showed that several ARGs (MLS, tetracycline, and multidrug resistance genes) were differentially present in the LEVO and the OTHER group compared to the CLA group (treatment naïve) at baseline. The species diversity was lower in the OTHER group and there were several differentially abundant genera compared with the CLA group. These ndings may indicate that after previously failed H. pylori eradication therapies, there was emergence of antibiotic resistance, especially for clarithromycin and tetracycline, which was associated with reconstructions of gut microbiota. Moreover, we found that those who failed eradication therapy in the CLA group had higher number of unique ARGs and abundance of ErmF (MLS resistance genes) at baseline. In our patients, we found that the tetracycline (tetQ, tetO, and tetW) and MLS (ErmF and ErmB) resistance genes were highly abundant and prevalent in human fecal samples ( Figure S8). Notably, it was also found that the vancomycin resistance genes (VanRG and VanRA) were also highly prevalent and abundant in some populations such as Danish, Spanish, and Chinese individuals [21].
One of the major issues related to H. pylori eradication therapy was antimicrobial resistance. Prior antibiotic treatment can induce multidrug resistance (MDR) [35], which is increasing worldwide and could hamper the success rate of conventional H. pylori eradication therapy [36]. The effect of different antibiotics used in various H. pylori eradication therapies on gut ARGs however remains unclear. In this study, we showed a transient increase in ErmF (macrolide resistance) after exposure to clarithromycin and amoxicillin. Besides, the relative abundance of the ErmF gene was overexpressed at baseline in patients who failed multiple previous treatments compared with those treatment naïve patients. A recent Russian study revealed that the ErmB, CFX group (beta-lactam), and tetQ genes were increased after clarithromycin-based quadruple eradication therapy [22]. Consistently, a small study reported that the macrolide resistance gene ErmB gene increased immediately after treatment and remained at a high level four years after clarithromycin-containing triple therapy using the 16S rRNA sequencing method [11]. It would be interesting to explore in future studies whether baseline scarriage of ErmF or ErmB gene can be a predictor of treatment failure.
Escherichia coli are commensals, commonly found in the lower part of the intestine and usually harmless, while virulent isolates are associated with diarrhea and colitis [37,38]. A Japanese study showed that uoroquinolone consumption was closely associated with E. coli resistance [39]. Our study showed that the relative abundance of E. coli signi cantly increased 6 weeks after the levo oxacin-based eradication therapy. In addition, resistant E. coli carried numerous genes that confer resistance to betalactam (Escherichia coli ampH beta-lactam), aminoglycoside (acrD), fosfomycin (GlpT), sulfonamide (sul1), phenicol (catS) [40], multidrug (Escherichia coli EF-Tu mutant, acrA, mdfA, soxR) [41][42][43]. E. coli could also gain high-level resistance to various antibiotics after levo oxacin and tetracycline exposure by inducing the e ux system during the bio lms formation, especially the emrY/K (tetracycline resistance) and evgS/A (multidrug resistance) pumps [44]. Consistent with the previous study, the relative abundance of those ARGs together with other multidrug resistance genes signi cantly increased 6-week after levo oxacin-containing therapy compared with baseline.
One recent systematic review reported a remarkable rise in the resistance rate of levo oxacin from 17-27% from 2006 to 2015 in the Asia-Paci c region and this may affect the e cacy of levo oxacincontaining therapies [45]. In our study, levo oxacin-based therapy has the lowest eradication success rate as second-line therapy. Therefore, alternative second-line therapy like bismuth quadruple therapy (PPI, bismuth subsalicylate, metronidazole, and tetracycline) should be considered. Bismuth salt has a synergistic effect with antibiotics and confer no resistance [46]. Recent data also con rmed that 14-day bismuth combining quadruple therapy is a highly effective (over 90% cure rate) and safe second-line option in patients with previous treatment failure [47]. Thus, bismuth quadruple therapy may be preferred in view of the post-treatment ARG pro les.
Previous studies have shown that the consumption of antibiotics leads to immediate [22,48], short-term[8, 49,50] and long-term [50][51][52] alterations in the human gut microbiota. However, most of the studies used 16S rRNA sequencing methods and few of them had employed the more detailed metagenomic sequencing methods [22]. In our study, we found that the relative abundance of Firmicutes increased at 6week in the CLA and LEVO group but signi cantly decreased in the OTHER group. Notably, it was reported that the relative abundance of Bacteroidetes and Firmicutes phylum decreased signi cantly whereas Proteobacteria increased immediately after the eradication therapy. All phylum restored to baseline level at 8-week [8,22,50,53,54]. The relative abundance of Firmicutes phylum was also found to be decreased 4-week after clarithromycin-based triple therapy in a recent study [49]. In contrast, another study showed an opposite trend of Firmicutes phylum, which increased 6-week after bismuth quadruple therapy [51]. It thus appears that different antibiotics may cause distinct short-term effects on the gut microbiota. The bacteria from Firmicutes phylum can ferment carbohydrates into a variety of short-chain fatty acids (SCFAs), which can increase the intestinal barrier function [55]. The dramatic alteration of Firmicutes phylum after the eradication therapy may seriously affected gut ecosystem and the in ammation recovery process.
Signi cant changes at the genus level were also identi ed and most of which were involved in the production of SCFAs. Previous studies have shown that some butyrate-producing bacteria, the Lachnoclostridium, Roseburia, Eubacteria hallii, Erysipelatoclostridium[56, 57] displayed protective effects by generating butyrate (SCFAs) to suppress chronic intestinal in ammation [58]. Besides, Bacteroides spp., which produce acetate and propionate (SCAFs), could also protect against gut in ammation [57]. As expected, these bene cial bacteria were found to be enriched 6-week after different eradication therapies, especially, Lachnoclostridium genus was found to have high-level 6-month after eradication therapy, which was similar to previous ndings that Lachnoclostridium enriched 26 weeks after bismuth quadruple therapy [51]. Taken together, those data demonstrate the temporary microbiota perturbation caused by antibiotic exposure and potential long-term protection of bismuth-containing eradication therapy, which was related to the recovery of gut in ammation.
Our study has several strengths. Other than looking at the ARG richness, we demonstrated dynamic changes in antimicrobial classes and unique ARGs under different eradication antibiotics exposure. Besides, we used shotgun whole genome sequencing (WGS) which had better detection of bacterial species, deeper sequencing depth, and identi cation of potential ARGs compared with the 16S rRNA sequencing method [16]. This study also included both patients who were naïve to and had failed previous eradication therapies to examine difference at baseline and after treatment.
There are some limitations of this study. First, as there are three different treatment groups with samples collected at three different time points, each group had a relatively small sample size which may hinder the application of the results. Second, this study only focused on the changes of gut microbiota and ARG from the metagenomics level, further studies should evaluate the metabolomics and proteomics for more integrated analysis, which can further reveal the consequences of receiving different H. pylori eradication therapies. Third, gut microbiota could change with lifestyle modi cations, for instance, after exercise training and dietary intervention [59]. Since no dietary intervention or physical exercise was enforced in this study, the impacts of these lifestyle changes could not be analyzed. However, as most changes restored to baseline at 6-month, suggesting these factors are unlikely to be playing a signi cant role. Finally, considering the regional variations in antibiotic consumption and resistance, the applicability of our ndings may need further validation in a more diverse population, particularly in an area with high background antibiotics resistance.

Conclusion
In summary, this study uncovers a transient alteration in the gut microbiome and ARGs after various H. pylori eradication therapies, and most of which resolved after 6 months. However, the use of tetracycline and uoroquinolone-containing therapies in retreatment was more likely to induce multidrug resistance genes and had a greater impact on the ARGs. Our ndings provide new insights into the perturbations of gut microbiota and ARGs associated with various H. pylori eradication therapies, which may facilitate future clinical treatment strategies particularly in patients with treatment failure.

Competing interests
The authors declare no competing interests.  Signi cantly altered ARGs 6-week after the eradication therapy in different treatment groups.
Differentially enriched unique ARGs in CLA (A), OTHER (B) and LEVO (C) 6-weeks after the eradication therapy. For ARG abundance box plots, the boxes color coated by red denote baseline abundance, while those color coated by blue denote 6-week abundance. The horizontal box lines represent the median, the boxes extend from the rst to the third quartile (25th to 75th percentiles). *p < 0.05, **p < 0.01 and ***p < 0.001 by Wilcoxon signed-rank test with false discovery rate (FDR) correction (only those ARGs with adjusted p <0.05 are shown). ARGs names that are too long have been abbreviated (rpoB: Bi dobacterium adolescentis rpoB mutants conferring resistance to rifampicin; ampH: Escherichia coli ampH bate-lactamase; EF-Tu mutation: Escherchia coli EF-Tu mutants conferring resistance to Pulvomycin; marR: Escherchia coli marR mutant conferring antibiotic resistance; soxR: Escherchia coli soxR with mutation conferring antibiotic resistance) Figure 3 Microbiota diversity (alpha and beta diversity) alteration after eradication therapies (A-C) Principal Coordinates Analysis (PCoA) from Bray-Curtis distance between three treatment groups at (A) baseline, (B) 6 weeks after the eradication therapy, and (C) 6 months afterward. (D-E) Boxplots showing alpha diversity indices (Richness, Shannon index) in three different treatment groups. The horizontal box lines represent the rst quartile, the median, and the third quartile. Whiskers denote the range of the values within 1.5 times the interquartile range (IQR) from the rst and third quartiles, respectively. *p < 0.05, **p < 0.01 and ***p < 0.001 by Wilcoxon signed-rank test and Mann-Whitney test for paired and unpaired samples. Figure 4