Concentrations of ciprofloxacin in food defined as safe alter the gut microbiome and ciprofloxacin susceptibility in humans: an interventional clinical study

Abstract

It is unknown if the low residual concentrations of ciprofloxacin/enrofloxacin allowed in food could influence the human microbiome. The European Medicines Agency reports that it is safe for an average human to ingest 372 µg of enrofloxacin/ciprofloxacin daily. We randomized 30 individuals to either 372 µg of ciprofloxacin or placebo daily for 27 days in a 2:1 fashion and measured the effect on the gut microbiome and the susceptibility of Escherichia coli to ciprofloxacin. After 27 days the E. coli of the individuals in the ciprofloxacin, but not the placebo arm, had reduced ciprofloxacin susceptibility. The receipt of ciprofloxacin was also associated with a reduction in the number of E. coli colony forming units in the gut. The low concentration of antimicrobials allowed in food may have an impact on the composition of the gut microbiome and antimicrobial susceptibility.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-18714-3.

Introduction

Antimicrobial resistance (AMR) is an escalating global health concern ¹. Several recent studies have suggested that the low concentrations of antimicrobials permitted in food may play a role ^1,2. Gullberg et al. were among the first researchers to show that very low antimicrobial concentrations could select for AMR ³. They introduced the concept of the minimum selective concentration (MSC), the lowest concentration of an antimicrobial that can induce or select for resistance ^3,4. They found that the ciprofloxacin MSC for Escherichia coli (0.1 ng/mL) was 230-fold lower than the minimum inhibitory concentration (MIC). More recent in vitro studies have shown that ciprofloxacin concentrations as low as 1/1000th of the MIC can induce resistance in Neisseria gonorrhoeae⁵ and Neisseria subflava µg/L (MSCs-0.004 ng/mL and 0.06 ng/mL, respectively) ⁶. A systematic review of 62 studies found that sub-MIC concentrations of fluoroquinolones could induce fluoroquinolone resistance in a wide range of bacterial species via both target mutations and upregulated efflux pumps ⁷.

The concentrations of antimicrobials in food frequently exceed the legal maximum residue limit (MRL) allowed by regulatory authorities. A systematic review found that antimicrobial residue concentration exceeded MRLs in 52 out of the 73 studies reviewed ⁸. For example, a survey of fluoroquinolone concentration in beef in British supermarkets found that the median concentration of enrofloxacin in beef (533 mg/kg, IQR 235–1060) was over five-times higher than the European Medicines Authority (EMA) MRL (100 mg/kg) ².

Consuming food with high quinolone concentrations has also been associated with elevated urinary and fecal concentrations of quinolones in humans. A study from South Korea found high concentrations of enrofloxacin and ciprofloxacin in the urine of the general population, that was linked to the consumption of beef, chicken, and dairy products ⁹. Similarly, a study in northern China detected veterinary fluoroquinolones in the urine of 57% of healthy children ¹⁰. Another study in China found ciprofloxacin and enrofloxacin concentrations above the detection level in the feces of around half the population ¹¹. The median concentration of these antimicrobials was 20 µg/kg. Since enrofloxacin is used in food animals and not in humans, the likely source was the ingestion of veterinary antimicrobials in food ¹¹. These findings may explain why the intensity of fluoroquinolone use for food animals has been found to be positively associated with quinolone resistance in E. coli and other infections in humans ¹².

The EMA determines the acceptable daily ingestion (ADI) of medicinal compounds based on studies evaluating thresholds for microbiological and cellular toxicity ^13–15. The EMA fluoroquinolone ADIs are determined based on microbiological toxicity, established by evaluating the MICs of common human bacterial commensals based on estimated dose exposures in the human colon ^14,15. The ADIs are then used to calculate MRLs, which are the maximum concentrations allowed in food products based on average consumption patterns ^14,15. The calculation of current ADIs and MRLs has not assessed if these concentrations could induce AMR ^1,14.

The EMA defines the ADI of enrofloxacin as 372 µg for an average person weighing 60 kg (6.2 µg/kg) ¹⁴. Recent studies have found that one-tenth of this dose can select for de novo ciprofloxacin resistance in a Galleria mellonella model of chronic Klebsiella pneumoniae infection ¹⁶. More recently, it was found that one-tenth of the enrofloxacin ADI dose could induce enrofloxacin resistance in a murine model of chronic K. pneumoniae gut colonization when fed enrofloxacin daily ¹⁷. This study aimed to assess, for the first time, if ADI doses of ciprofloxacin could induce resistance in E. coli colonizing the human gastrointestinal tract.

We used ciprofloxacin instead of enrofloxacin for a number of reasons. First, enrofloxacin is not licensed for use in humans ¹⁸. Second, the ciprofloxacin and enrofloxacin MICs for most Enterobacteriaceae are very similar ¹⁹. Third, most ingested enrofloxacin is metabolized to ciprofloxacin in mammals ¹⁸. This means that the concentration of ciprofloxacin is typically higher than that of enrofloxacin in food animals fed enrofloxacin ¹⁸.

Results

A total of 36 participants were screened (Fig. 1). Three participants were excluded due to not being colonized with endogenous E. coli, one for not meeting the inclusion criteria and two for other reasons. Thirty participants were randomized after confirmed colonization with endogenous E. coli.

Fig. 1.

Flow diagram of study enrollment, randomization and follow up.

All enrolled participants completed the study. The participants’ adherence to the ciprofloxacin/placebo was good. Five participants missed one dose, two missed two doses and one participant missed 9 doses. A further participant in the intervention group discontinued the treatment 5 days before visit 3 due to abdominal cramps but remained in the study for follow-up. For three individuals, no E. coli colonies were detected in the visit 3 samples. All these individuals were in the ciprofloxacin arm.

The participants in the placebo arm (median age 67.5 years [IQR 64.3–74]) were older than those in the ciprofloxacin arm (median age 47.0 years [IQR 38.0–65.3]). Baseline characteristics were well balanced in terms of weight, sex ratio, diet and antibiotic usage between the two arms (Table 1).

Table 1.

Baseline characteristics per group.

Characteristic	Placebo	Ciprofloxacin	Overall
	n = 10	n = 20	n = 30
Male [n(%)]	6 (60.00%)	10 (50.00%)	16 (53.33%)
Female [n(%)]	4 (40.00%)	10 (50.00%)	14 (46.67%)
Age [median (Q1–Q3)]	67.50 (64.25–74.00)	47.00 (38.00–65.25)	63.00 (38.75–68.00)
Weight [median (Q1–Q3)]	74.25 (62.75–79.88)	76.50 (69.75–83.28)	75.50 (65.62–83.08)
Type of diet
Vegan	0 (0.00%)	0 (0.00%)	0 (0.00%)
Vegetarian	0 (0.00%)	4 (20.00%)	4 (13.33%)
Flexitarian	6 (60.00%)	10 (50.00%)	16 (53.33%)
Omnivorous	4 (40.00%)	6 (30.00%)	10 (33.33%)
Frequency meat eaten?
Never	0 (0.00%)	4 (20.00%)	4 (13.33%)
Less than once per month	1 (10.00%)	0 (0.00%)	1 (3.33%)
1–3 per month	0 (0.00%)	0 (0.00%)	0 (0.00%)
Once per week	2 (20.00%)	2 (10.00%)	4 (13.33%)
2–4 times per week	3 (30.00%)	7 (35.00%)	10 (33.33%)
5–6 times per week	1 (10.00%)	4 (20.00%)	5 (16.67%)
Every day	3 (30.00%)	3 (15.00%)	6 (20.00%)
More than one meal with meat per day	0 (0.00%)	0 (0.00%)	0 (0.00%)
Frequency dairy products consumed?
Never	0 (0.00%)	0 (0.00%)	0 (0.00%)
Less than once per month	0 (0.00%)	1 (5.00%)	1 (3.33%)
1–3 per month	0 (0.00%)	0 (0.00%)	0 (0.00%)
Once per week	0 (0.00%)	1 (5.00%)	1 (3.33%)
2–4 times per week	4 (40.00%)	1 ( 5.00%)	5 (16.67%)
5–6 times per week	0 (0.00%)	2 (10.00%)	2 (6.67%)
Every day	6 (60.00%)	15 (75.00%)	21 (70.00%)
How often antibiotics consumed in the prior 10 years?
Never	4 (40.00%)	2 (10.00%)	6 (20.00%)
Once	0 (0.00%)	5 (25.00%)	5 (16.67%)
2–3 times	3 (30.00%)	9 (45.00%)	12 (40.00%)
4–6 times	1 (10.00%)	2 (10.00%)	3 (10.00%)
More than 6 times	2 (20.00%)	2 (10.00%)	4 (13.33%)

Effect of low dose ciprofloxacin on E. coli MIC

PAP

In the primary analysis, the generalized additive model indicated that the E. coli ciprofloxacin MIC distribution was altered at visit 3 compared to visit 1 in the ciprofloxacin but not the placebo arm (Fig. 2; SFig. 1). The plot suggests a shift to the right (i.e. higher colony count at the higher ciprofloxacin concentrations) in the ciprofloxacin arm at visit 3. In addition, there was no decrease in colony counts with increasing concentrations of ciprofloxacin in this group, whereas for the other groups and visits, colony counts tended to decrease with increasing concentration of ciprofloxacin.

Fig. 2.

Generalized additive model of association between the expected number of colonies of E. coli (log scale) and the concentration of ciprofloxacin for the participants in the two study arms at baseline (visit 1) and study completion (visit 3).

5-colony MIC distribution

Mixed effects linear regression revealed no significant interaction effect of visit by arm on mean log₂ ciprofloxacin MIC (STable 1). There were four individuals who had E. coli detected at visit 1 but not visit 3. These individuals were all from the ciprofloxacin arm and their E. coli from visit 1 all had low ciprofloxacin MICs (range 0.008–0.064 mg/L) (SFig. 2,3). The ciprofloxacin MICs of all E. coli colonies increased from 0.047 mg/L (IQR 0.023–0.19 mg/L) at visit 1 to 0.19 mg/L at visit 3 (IQR 0.047–0.38 mg/L; P = 0.0003 [Wilcoxon Rank-Sum test]; SFig. 2B). There was no increase in the placebo arm (visit 1: 0.047 mg/L [IQR 0.032–0.064 mg/L], visit 3: 0.047 mg/L [IQR 0.032–0.22 mg/L]; SFig. 2A).

E. coli abundance

In the ciprofloxacin arm, there were fewer E. coli colonies detected at visit 3 (median 19 colony forming units [CFU]/100 µl, IQR 1–188) than visit 1 (median 326 CFU/100 µl, IQR 22–1905; P = 0.0065 [Wilcoxon Rank-Sum test]; Fig. 3). There was no difference in the number of E. coli colonies between visits in the placebo arm (Fig. 3).

Fig. 3.

Variations in the E. coli colony counts between the baseline (Visit 1) and after 27 days of ciprofloxacin (Visit 3) in the ciprofloxacin (A) and placebo (B) arms.

Sensitivity analysis

Sensitivity analyses including dilutions of the samples down to 1:1000 produced very similar conclusions using the same generalized additive model as before (STable 2; SFig. 4).

Integrative analyses of microbial communities: Taxonomic, functional and culture-based correlations

Due to resource constraints, sequencing was only performed on the visit 1 and 3 isolates from the ciprofloxacin arm (STable 3). The following microbiome/resistome results thus only pertain to the ciprofloxacin arm.

Taxonomic profiling and microbial community dynamics

There was a non-significant decline in Observed richness (p = 0.547), Shannon (p = 0.194), Simpson (p = 0.412) and inverse Simpson diversity (p = 0.412) indices between Visit 1 and Visit 3 (SFig. 5) in the ciprofloxacin arm. The MaAsLin2 analysis identified several microbial taxa (179 out of 344) significantly associated with the two visits (Fig. 4, SFig. 5,6,7, STable 4) in the ciprofloxacin arm. Most taxonomic changes, including the significant increase in Clostridium leptum and decrease in E. coli, were observed within the ciprofloxacin arm (Fig. 4). The shifts were observed across key phyla, including Firmicutes, Bacteroidetes, Proteobacteria, and Actinobacteria.

1. Increased taxa in Visit 3: A significant increase in relative abundance of Clostridium leptum (q = 0.0015), Streptococcus parasanguinis (q = 0.0036), Fusobacterium nucleatum (q = 0.0036), Eubacterium sp. CAG-252 (q = 0.0077), Flavonifractor plautii (q = 0.0110), Ruminococcus bromii (q = 0.0063), Lactobacillus sakei (q = 0.0155), Peptostreptococcus anaerobius (q = 0.0140) was observed (Fig. 4).

2. Decreased taxa in Visit 3: A significant decrease in the relative abundance of Escherichia coli (q = 0.0036), Bacteroides uniformis (q = 0.0036), Prevotella sp. CAG-279 (q = 0.0036), Campylobacter ureolyticus (q = 0.1987), Firmicutes bacterium CAG-24053–14 (q = 0.0076), Phascolarctobacterium succinatutens (q = 0.0155) and Streptococcus thermophilus (q = 0.1755) were identified (Fig. 4).

Fig. 4.

Distribution of top 30 significant bacterial taxa relative to the two visits in the ciprofloxacin arm; visit 1 (V1) and Visit 3 (V3). Individual box plots showing the distribution of relative abundance for each taxa by visits. The black asterisks depict FDR adjusted p (q) values of < 0.05 using MaAsLin2.

The Hallagram visualization revealed similar results and are provided in SFig. 8, 9, STable 5 and SBox 5.

Functional characterization of microbial communities

The MaAsLin2 analysis identified a total of 26 metabolic pathways that differed significantly between the two visits (q < 0.05; SFig. 9, 10). These pathways related to carbohydrate metabolism, nitrogen metabolism, lipid and vitamin biosynthesis (SBox 7). Two of the pathways downregulated at visit 3 were the PWY-6803 phosphatidylcholine acyl editing and the P108-PWY pyruvate fermentation to propanoate pathways.

Correlation between metagenomics versus culture data

Spearman’s rank correlation revealed a strong positive association between the colony counts of E. coli as ascertained by culture and the number of E. coli reads in the metagenomics data (ρ = 0.733, p = 1.69e−07; SFig. 10, 11).

Resistome diversity and abundance of antimicrobial resistance genes (ARG)

Analysis of the baseline (Visit 1) and follow-up (Visit 3) anorectal samples revealed no significant differences in ARG diversity (SBox 8, SFig. 12, 13, 14, 15).

Adverse events

All adverse events reported were gastrointestinal disorders: 4 in the ciprofloxacin arm (4/20; 20%, 95% Wilson CI: 8.07–41.60%) and 1 in the placebo arm (1/10; 10%, 95% Wilson CI: 1.79–40.42%). Diarrhea was the most common adverse event. There were no serious adverse events. There was one adverse event (abdominal cramps) which was considered related to the study drug and led to the discontinuation of the study drug.

Discussion

This was the first study to assess the effects of ADI doses of an antimicrobial in humans. Despite the small sample size, we found indications that the consumption of ADI doses of ciprofloxacin leads to reduced susceptibility of E. coli to ciprofloxacin and a distributional shift to the right (in other words there is a higher colony count at the higher ciprofloxacin concentrations for ciprofloxacin visit 3 but not for the other arm and visits). The metagenomic analyses also revealed a number of significant alterations in the abundance of various taxa and the activity of certain biochemical pathways after the receipt of the ciprofloxacin. In a number of individuals, the receipt of ciprofloxacin was associated with the loss of all E. coli detected via culture. All these individuals had E. coli with low ciprofloxacin MICs at their baseline visit (visit 1), suggesting that the receipt of low-dose ciprofloxacin may have eliminated their highly susceptible E. coli. The integrated taxonomic, functional, and culture-based approaches revealed changes in specific taxa and metabolic pathways between visits 1 and 3 in the ciprofloxacin recipients. However, we found no evidence of a difference in the relative abundance of ARGs between the two visits. Taken together, these results suggest that ADI doses of ciprofloxacin are associated with detectable effects in the gut microbiome and antimicrobial susceptibility of E. coli.

Microbial community shifts across visits

The significant decrease in Escherichia coli at visit 3, determined using culture, Maaslin2 and Halla suggests that low-dose ciprofloxacin has altered its abundance. E. coli is a predominant species among the facultative anaerobic bacteria of the gastrointestinal tract ²⁰, where it performs certain functions beneficial to the host such as the synthesis of vitamin K and colonization resistance ^21,22. Other key taxa known to be important for gut health that were less abundant following the receipt of ciprofloxacin include Bacteroides uniformis, Prevotella sp. CAG-279 and Firmicutes bacterium CAG-24053-14^23,24. These changes were accompanied by reciprocal increases in a number of obligate anaerobes at visit 3, such as Ruminococcus bromii and Clostridium leptum (q = 0.0015). These results are broadly commensurate with those from in-vitro chemostat studies. For example, Carman et al., found that low dose ciprofloxacin added to a chemostat model of the human gut reduced E. coli and Bacteroides fragilis counts in a dose dependent fashion ²⁵. In another ex vivo study, Kim et al., found similar effects of low dose enrofloxacin (0.1 to 150 µg/ml) on fecal suspensions of the gut microbiome of three individuals ²⁶. The abundance of Bacteroidetes (including Bacteroides uniformis) and Proteobacteria decreased following exposure to ciprofloxacin.

Functional implications of metabolic pathway shifts

In visit 3, the downregulation of the PWY-6803 phosphatidylcholine acyl editing pathway and P108-PWY pyruvate fermentation to propanoate was observed. PWY-6803 is critical for modifying the fatty acid composition of phosphatidylcholine (PC) (glycerophosphocholines), thereby influencing membrane fluidity, cellular signalling, and overall lipid homeostasis ²⁷. P108-PWY pathway is essential for converting pyruvate into propanoate, a short-chain fatty acid that supports colonic health and systemic energy regulation ²⁸. The simultaneous reduction of these two pathways suggests a significant shift in microbial metabolic function that may impact host metabolic homeostasis ²⁹. These findings are similar to those found in previous in-vitro analyses ²⁶.

This study has a number of limitations. Foremost amongst these are the small sample sizes in both arms. We also only evaluated a single dose of a single antimicrobial and we only assessed the effect on antimicrobial susceptibility after 27 days on a single species of bacteria. It is certainly possible that other species of bacteria (such as B. fragilis ²⁵) may be more sensitive to the effects of low dose antimicrobials. In the in vitro study by Carman et al., described above low dose ciprofloxacin resulted in an increase in the proportion of B. fragilis isolates resistant to ciprofloxacin from 0 to 95% ²⁵. No effect on resistance in E. coli was detected but the number of colonies isolated following the addition of ciprofloxacin was very low. A disadvantage of using endogenous E. coli is that the baseline MICs of participants may have been elevated by prior exposure to the antimicrobial of interest in food or elsewhere. If this relationship is saturated, then the experiment may fail to detect the effect of the antimicrobial on AMR. Low dose ciprofloxacin may also facilitate other adverse outcomes such as horizontal gene transfer and collateral resistance to other antimicrobials ²⁰.

Arguments could be made that the selected dose of ciprofloxacin we tested was too high or too low. The dosing schema could be considered too low because it only runs over 27 days versus a possible lifetime of exposures from antimicrobials in food. The schema also does not evaluate the scenario where food products contain multiple fluoroquinolones and other antimicrobials. The EMA guidelines do not specify a cumulative ADI of all fluoroquinolones. The dose we assessed could also be considered too high. It may be considered unlikely that an individual would be exposed to such a dose of fluoroquinolones on a daily basis. Furthermore, we did not evaluate the fluoroquinolone concentrations of the food ingested by the participants which means that the total fluoroquinolone dose ingested may have exceeded the ADI. Future studies could control antimicrobial concentrations in the food of study participants and assess the effects of lower doses of fluoroquinolones. Due to budgetary constraints, we were unable to sequence the visit 1 and 3 samples from the placebo group. This means that the microbiome changes we observed between visit 1 and 3 in the ciprofloxacin group may have been due to temporal changes due to some unmeasured factor. These microbiome changes are thus best interpreted as being hypothesis generating and require confirmation in larger trials that include sequencing of the control group. A temporal change in microbiome not related to the receipt of ciprofloxacin would not, however, explain the reduction of E. coli colonies in the ciprofloxacin but not the placebo arm between visits 1 and 3. The fact that this same decline in E. coli abundance between visits 1 and 3 in the ciprofloxacin arm was seen in the metagenomics analysis, suggests that the changes in microbiome as assessed by metagenomics was congruent with that seen in the phenotypic assessment. Moreover, we did not adjust for potential confounders such as age or baseline microbiome composition. The placebo group were older than the ciprofloxacin group. Although randomization was used to allocate participants to study arms, residual imbalances may still exist and could have influenced some of the observed outcomes. We cannot exclude the possibility that this difference may have explained some of the study’s findings.

We did not find that the receipt of ADI doses of ciprofloxacin was associated with an increase in ARGs. This may mean that these doses have little or no effect on the resistome. Alternatively, our tools for assessing the resistome may be too crude and lack the resolution to detect subtle or strain specific mutations. For example, the method we used to assess the resistome would not have been able to tell us if there was a large increase in the canonical fluoroquinolone-resistance-inducing gyrA and parC mutations in a specific species of gut bacteria or it may fail to capture non-canonical or novel resistance-conferring variants.

These limitations mean that the results are best viewed as being useful to guide future larger trials. Despite these limitations, we found that low dose ciprofloxacin resulted in substantive changes in bacterial composition, metabolic pathways and antimicrobial susceptibility. This is despite the dose of ciprofloxacin used being approximately 1000-fold lower than a typical therapeutic dose. In addition to increasing rates of AMR, there has also been a sharp drop in the gut bacterial diversity of Western populations over the past few decades ^30,31. These changes have been linked to a number of adverse health outcomes ³². Whilst the causes of these changes are likely to be multiple, our results are compatible with low-dose antimicrobials in food playing a role. As a result, our study findings provide additional impetus for further studies to more definitively delineate safe concentrations of antimicrobials in food.

Methods

Study design, setting, and participants

We performed a pilot single blind, single-center randomized, placebo-controlled trial (RCT) to compare the effect of an ADI dose of ciprofloxacin on the ciprofloxacin susceptibility of fecal E. coli. The study took place at the Clinical Trial Site of the Institute of Tropical Medicine (ITM) in Antwerp, Belgium. Healthy volunteers were recruited via emails and advertisements on social media. The inclusion and exclusion criteria are listed in SBox 1, 9.

Participants had to be confirmed (via culture of anorectal swabs) to be colonized with endogenous E. coli before randomization. The study was carried out in accordance with the CONSORT 2010 and other relevant guidelines.

Randomization

Subjects who met all the inclusion and none of the exclusion criteria were randomized with a 2:1 ratio to either daily low dose ciprofloxacin for 27 days (n = 20) or daily placebo for 27 days (n = 10). The randomization list was prepared by an independent sponsor biostatistician using R v4.3.0 (R Foundation for Statistical Computing, Vienna, Austria) and was not shared with the study team until the database was locked.

Study procedures

Three study visits were planned. During the baseline visit (Visit 1; day 0), we collected three anorectal Eswab samples, as well as data on previous antibiotic use, diet, concomitant medication and weight. A physical examination was performed if deemed necessary.

At visit 2 (day 2 − 1/+ 2), participants were classified based on presence or absence of E. coli detected in the anorectal swabs. Individuals in whom E. coli was not detected exited the study at this point. Those colonized (n = 30) were randomized (2:1) in a single blinded manner to receive either low dose ciprofloxacin or placebo. They received detailed instructions to take 5 ml of the study product ciprofloxacin/placebo, with or without food, once daily, for 27 days.

At visit 3 (day 30 ± 3) three anorectal Eswabs were taken. Data was collected on other antibiotic use, diet, concomitant medication, adverse events, medication compliance, and a physical examination was performed if necessary.

The anal swab was collected via a Standard Operating Procedure that stipulated that the physician or participant could take the swab by inserting the swab 5 cm into the anus and rotating it three times. The order in which the anorectal swabs were taken was recorded at the time of taking the swabs.

The method used to manufacture the ciprofloxacin and placebo syrup are provided in SBox 2. The methods used to obtain and process the anorectal swabs is provided in SBox 3.

Laboratory procedures

At the end of the study, the E. coli ciprofloxacin susceptibility was calculated for the anorectal swabs from visits 1 and 3 as follows:

E. coli culture and MIC determination

The skim milk aliquots were thawed to room temperature and 100 µl of a 1:10 dilution in PBS was then plated onto 9 Chromocult® Coliform Agar plates with the following concentrations of ciprofloxacin: no ciprofloxacin, 0.002 mg/L, 0.004 mg/L, 0.008 mg/L, 0.015 mg/L, 0.03 mg/L, 0.06 mg/L, 0.125 mg/L and 0.250 mg/L. In three participants, no E. coli had grown on any of the plates on either visit 1 or 3. For these three individuals, this procedure was repeated using undiluted samples. The plates were incubated at 36 °C, 5% CO₂ overnight, and the number of colonies were counted on each plate using an automatic colony counter SCAN 300® (Interscience, France), version 8.7.3.0. The number of colonies growing on the undiluted plates was then adjusted so that for all samples, we report the number of colonies growing per 10 µL of the undiluted sample.

Sensitivity analysis

For all samples where at least one plate had more than 500 colonies, 100µL from two additional dilutions of 1:100 and 1:1000 dilution in PBS were then plated onto 9 Chromocult® Coliform Agar plates with the same concentrations of ciprofloxacin as above. We used the results from the lowest dilution (most concentrated sample) where the highest colony count on any of the plates was less than 100. The number of colonies growing on the plates were adjusted for dilution factor so that once again for all samples, we report the number of colonies growing per 10 µL of the undiluted sample.

Outcomes

The plates were used to determine for each individual and each time point the following:

Population analysis profile (PAP)

The number of E. coli colonies growing on each of the 9 plates with differing concentrations of ciprofloxacin (0.250 mg/L, 0.125 mg/L 0.06 mg/L, 0.03 mg/L, 0.015 mg/L, 0.008 mg/L, 0.004 mg/L, 0.002 mg/L and 0 mg/L) was used as proxy for the distribution of E. coli MICs per individual and visit.

Five-colony MIC distribution (5C_MIC)

Five colonies were randomly selected from the plate without ciprofloxacin and had their MICs determined via an E-test (bioMérieux, Marcy l’Etoile, France) according to the manufacturer’s instructions. The species identity of each isolate was determined via MALDITOF-MS, on a MALDI Biotyper® Sirius IVD system. The MBT Compass IVD software and library (Bruker Daltonics, Bremen, Germany) was used. Only isolates that were confirmed as E. coli were used in subsequent analyses.

Statistical testing

Shift in PAP

This shift was evaluated with a generalized additive model (GAM) for count data (family = “poisson”). GAM models are a tool that can be used when there are nonlinear relationships among response (colony count) and predictor(s), in this case the log₂ ciprofloxacin concentration (mg/L). No ciprofloxacin (0.00 mg/L) was transformed to a value of 0.0009765625. This corresponds to − 10 on the log₂ scale. The correlated nature of the data (multiple observations per participant) was taken into account using random intercepts (results not shown). For interpretational purposes, a GAM model without random intercepts was generated. GAM models are usually interpreted visually without too much emphasis on the model estimates. The R package mgcv was used ²². The exact same model was used for the sensitivity analysis.

Shift in 5-colony MIC distribution

We used mixed effects linear regression with random intercept and slope in function of arm, visit, and their interaction to assess if the receipt of ciprofloxacin was associated with a shift in the mean log₂ ciprofloxacin MIC.

The secondary outcomes including impact on the microbiome are detailed in SBox 5.

The primary analysis was performed using the intention-to-treat (ITT) approach. There were no major protocol violations and thus no additional per protocol analysis was necessary. All computations were made using R, version 4.2.3 and STATA MP version 16.

Microbiome characterization

The microbiome characterization was carried out as described in SBox 1, 4. Raw shotgun metagenomic sequencing reads are available in the NCBI Sequence Read Archive (SRA) under BioProject accession number PRJNA1236935.

Sample size

This study was designed as a pilot study to guide future studies. It had a budget sufficient to enrol and follow-up of 30 participants, which determined the sample size.

Adverse events

All severe adverse events (SAEs) and a list of prespecified adverse events (AEs) occurring between visits 2 and 3 were collected at visit 3. The prespecified AEs included: diarrhea, nausea or vomiting, AEs leading to study drug discontinuation, and AEs considered related to the study drug by the investigator. The number and percentage of participants with AEs was calculated with a 95% Wilson confidence interval per arm.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

CK, TV, LP, DVDB, YVH, IDB, LV, BS, SA and SB contributed to the conceptualization, methodology and funding acquisition. SA, LvA and ZG conducted the laboratory work. All authors contributed to the investigation. EG, BJ and CK performed the formal analysis. SB performed the bioinformatic analyses. All authors contributed to the writing of the manuscript and approved the final version.

Funding

This work was supported by SOFI 2021 grant: "Preventing the Emergence of untreatable STIs via radical Prevention" (PRESTIP).

Data availability

Anonymized individual participant data and a data dictionary defining each field in the dataset can be shared on approval of a written request to the corresponding author and in agreement with ITM data sharing policy. Contact ckenyon@itg.be for data access. Raw shotgun metagenomic sequencing reads are available in the NCBI Sequence Read Archive (SRA) under BioProject accession number PRJNA1236935.

Declarations

Competing interests

The authors declare no competing interests.

Ethical clearance and trial registration

All participants provided written informed consent at baseline (visit 1). The Institutional Review Board of the ITM, the designated Belgian EC, and the Competent Authorities of Belgium (FAMPH) approved the trial. The study was registered in the EU Clinical Trials Register (EU CT number 2023-506205-18-00; https://euclinicaltrials.eu/search-for-clinical-trials/?lang=en&EUCT=2023-506205-18-00; Trial registration 23/02/2024).