https://orcid.org/0000-0003-0951-3593
Abstract
To evaluate an antibiotic inactivation strategy to protect the gut microbiome from antibiotic‐mediated damage.
SYN‐004 (ribaxamase) is an orally delivered beta‐lactamase intended to degrade penicillins and cephalosporins within the gastrointestinal tract to protect the microbiome. Pigs (20 kg, n = 10) were treated with ceftriaxone (CRO) (IV, 50 mg kg−1, SID) for 7 days and a cohort (n = 5) received ribaxamase (PO, 75 mg, QID) for 9 days beginning the day before antibiotic administration. Ceftriaxone serum levels were not statistically different in the antibiotic‐alone and antibiotic + ribaxamase groups, indicating ribaxamase did not alter systemic antibiotic levels. Whole‐genome metagenomic analyses of pig faecal DNA revealed that CRO caused significant changes to the gut microbiome and an increased frequency of antibiotic resistance genes. With ribaxamase, the gut microbiomes were not significantly different from pretreatment and antibiotic resistance gene frequency was not increased.
Ribaxamase mitigated CRO‐mediated gut microbiome dysbiosis and attenuated propagation of the antibiotic resistance genes in pigs.
Damage of the microbiome can lead to overgrowth of pathogenic organisms and antibiotic exposure can promote selection for antibiotic‐resistant micro‐organisms. Ribaxamase has the potential to become the first therapy designed to protect the gut microbiome from antibiotic‐mediated dysbiosis and reduce emergence of antibiotic resistance.
Introduction
The gut microbiome, defined as the collective genetic information of the commensal microbiota, comprises a complex ecosystem that works symbiotically within the body to aid host metabolism, immunity and maintenance of health (Hamady et al. 28; Carding et al. 12). Antibiotic exposure inherently causes dysbiosis, a perturbation of the number and composition of the microbiota that affects normal microbial balance. Antibiotic use can result in antibiotic‐associated diarrhoea (McFarland 50), promote selection of antibiotic‐resistant micro‐organisms (The Review on Antimicrobial Resistance 76; Francino 24), and lead to emergence of opportunistic pathogens, including Clostridium difficile (Dressman 20; Stevens et al. 69; Crowther and Wilcox 15; Theriot et al. 77). Antibiotic‐induced changes in microbiota composition can persist for months or years after cessation of antibiotic treatment (De La Cochetiere et al. 16; Jernberg et al. 34; Dethlefsen et al. 18; Dethlefsen and Relman 17). As dysbiosis has been reported to be associated with a broad range of physiological states, including allergies (Trompette et al. 79), asthma (Arrieta et al. 4), autism (Hsiao et al. 33; Buffington et al. 9), cancer (Garrett 25), diabetes (Livanos et al. 45; Pedersen et al. 57), recovery from neurological injury (Kigerl et al. 39), metabolic syndrome, and obesity (Tilg and Kaser 78; Sanmiguel et al. 64; Economopoulos et al. 23; Yang and Kweon 80), protection of the gut microbiome from unintended effects of antibiotics becomes increasingly urgent.
A novel strategy to protect the microbiome from antibiotic‐mediated dysbiosis is prophylactic use of a beta‐lactamase enzyme to degrade antibiotics in the proximal gastrointestinal (GI) tract before the colonic microbiota are harmed (Kaleko et al. 35). Beta‐lactamases are naturally occurring enzymes that confer antibiotic resistance by hydrolysing beta‐lactam antibiotics, the most widely used intravenous (IV) broad‐spectrum antimicrobials (Arlington Medical Resources 3), many of which are excreted via the bile into the intestinal tract at high concentrations (Maudgal et al. 47). In North America and Europe, ceftriaxone (CRO) is the most frequently prescribed IV beta‐lactam (Arlington Medical Resources 3) and has been linked to dysbiosis and increased risk of C. difficile infection (Slimings and Riley 66; Crowther and Wilcox 15; Zycinska et al. 81). Notably, oral delivery of a beta‐lactamase isolated from Bacillus licheniformis, the PenP protein (Neugebauer et al. 54), also called P1A (Harmoinen et al. 29), was shown in human clinical trials to effectively degrade IV ampicillin and piperacillin in the GI tract, preserve microbiome diversity, and prevent antibiotic‐associated diarrhoea (Pitout 58; Tarkkanen et al. 75). However, because it is a penicillinase, P1A has limited clinical application. SYN‐004 (ribaxamase), originally named SYN‐004 (Kaleko et al. 35), was engineered from the P1A enzyme to broaden its antibiotic degradation profile to include cephalosporins, such as CRO, while maintaining penicillinase activity (Kaleko et al. 35). Ribaxamase is intended for oral use with IV beta‐lactam antibiotics, including CRO, and is formulated into enteric‐coated pellets that release the enzyme into the upper small intestine at pH >5·5 (Kaleko et al. 35).
Efficacy studies with jejunal‐fistulated dogs treated with ribaxamase and/or IV CRO showed that CRO was eliminated from the intestine in the presence of ribaxamase and that ribaxamase remained biologically active for at least 8 hours, the duration of antibiotic release into the intestine (Kaleko et al. 35). Safety assessments in dogs showed that ribaxamase was well tolerated and, when delivered with CRO, did not interfere with antibiotic pharmacokinetics (Kokai‐Kun et al. 40). Based on these encouraging data, ribaxamase was advanced into human clinical testing. Phase 1 clinical studies demonstrated that ribaxamase was well tolerated (Roberts et al. 62), Phase 2a studies confirmed that ribaxamase degraded CRO in the human GI tract (Kokai‐Kun et al. 41), and a double‐blind, placebo‐controlled Phase 2b study designed to assess the ability of ribaxamase to prevent C. difficile‐associated disease and antibiotic‐associated diarrhoea by protecting the gut microbiome from CRO‐mediated changes is in progress (clinicaltrials.gov 14).
In addition to causing dysbiosis, antibiotics can lead to other deleterious consequences, such as promoting antibiotic resistance in micro‐organisms (The Review on Antimicrobial Resistance 76; Francino 24). The human gut microbiota is considered to be an established reservoir of antibiotic resistance that, with frequent antibiotic exposure, increases the potential for widespread acquisition and propagation of resistance (Francino 24). Ribaxamase is expected to reduce the selective pressure on the gut microbiota to lessen this risk. Indeed, clinical studies using the ribaxamase precursor, P1A (Harmoinen et al. 29, 30; Pitout 58; Tarkkanen et al. 75), given to patients being treated with IV ampicillin demonstrated that P1A reduced emergence of ampicillin‐resistant organisms in the faecal microbiome (Tarkkanen et al. 75), and the current ribaxamase Phase 2b clinical study is also evaluating the presence of antibiotic‐resistant micro‐organisms (Clinicaltrials.gov 14).
To assess the effect of antibiotics on the gut microbiome, a porcine model of antibiotic‐mediated microbiome disruption was developed. Pigs represent an excellent, nonprimate model of the human intestinal tract. Porcine and human GI tracts have analogous segments and digestion characteristics of the small intestine are similar (Dressman 20; Dressman and Yamada 21; Rowan et al. 63; Kararli 36). In addition, the porcine commensal colonic microbiome is comparable to that of humans and promotes GI immune system development (Hopwood and Hampson 32; Bauer et al. 6). Here, we demonstrate that ribaxamase protects the porcine gut microbiome from CRO‐mediated dysbiosis.
Materials and methods
Test article
Ribaxamase, a 29 kDa, engineered, recombinant protein was manufactured in Escherichia coli (Kaleko et al. 35). Ribaxamase was formulated for oral delivery by incorporation into Eudragit®‐coated sucrose pellets designed for release of active enzyme at pH 5·5 or greater (Kaleko et al. 35). The pellets, confirmed by electron microscopy to be uniform spheres of approximately 1 mm diameter, contained approximately 15% ribaxamase (Kaleko et al. 35). Hard capsules suitable for oral delivery were filled with the pellets for total ribaxamase content of 75 mg per capsule.
Animals and test article administration
Ten healthy male and female 2‐month‐old pigs, Sus scrofa domestica, Yorkshire cross, approximately 20 kg, were obtained from Archer Farms, Inc. (Darlington, MD). After arrival to the test site, Noble Life Sciences, Inc. (Sykesville, MD), animals were quarantined/acclimated for 5 days during which time the health status of each animal was evaluated daily. Animals were fed Southern States Non‐medicated Hog Feed (SSC‐25‐629001, Lot G6148) and were, therefore, never exposed to in‐feed antibiotics. At the end of quarantine, all animals were deemed healthy and were randomly divided into two groups. One group (n = 5) received CRO, 50 mg kg−1, IV, once a day for 7 days, and the other group (n = 5) received CRO plus ribaxamase (75 mg capsule, PO, four times a day). Ribaxamase administration was started the day before CRO treatment and continued for one day after CRO treatment, for a total of 9 days.
CRO was supplied as a powder (1 g per vial; WG Critical Care, 44567‐701‐25). Each vial was reconstituted with 4·5 ml of sterile water, creating 5 ml of injectable suspension (1 g 5 ml−1). CRO was administered through an IV catheter to animals under sedation. For sedation, a TELAZOL cocktail consisting of TELAZOL (50 mg ml−1), ketamine (250 mg) and xylazine (250 mg) was administered intramuscularly at a dose of 0·5–1·0 ml 50 lbs−1 to induce and maintain sedation. Once sedated, each pig received a total of 1 g of CRO administered slowly through the IV catheter, followed by a heparinized saline flush. CRO was delivered at the same time daily (12 pm). Ribaxamase was supplied as size 0 hard capsules filled with ribaxamase‐containing enteric‐coated pellets (Kaleko et al. 35). Each capsule contained 75 mg of ribaxamase. One ribaxamase capsule was administered orally, four times a day at the same time, 7:00 am, 12:00 pm, 5:00 pm and 10:00 pm to each pig. Animals were fed three times per day, at 7:00 am, 12:00 pm and 5:00 pm after test article administration. Animals had free access to water at all times.
Blood was collected on day 2 of CRO treatment. Blood was collected aseptically from the cranial vena cava of anesthetized animals. As blood collections required anesthetization of animals, only three blood draws were conducted within a 24 h period, at 1, 6, and 19 h after CRO delivery. At each time, approximately 9 ml of blood was collected and dispensed into a serum separator vacutainer tube. After coagulation, samples were centrifuged and the serum transferred to a cryovial and stored at −80°C until shipment to the analysis laboratory. Faecal samples were collected at four timepoints, two prior to study initiation (day −7 and day −4), one during treatment (at day 4) and one at the end of the antibiotic treatment period (day 8). Samples were collected fresh upon defecation and placed directly into the OMNIgene® GUT sample kit collection tubes (DNA Genotek, Ottawa, Canada). Faecal samples were stored at room temperature until shipped at the end of the study for analysis.
All animal procedures were conducted in accordance with principles and guidelines established by the Institutional Animal Care and Use Committee in accordance with the Animal Welfare Act at Noble Life Sciences, Inc. (Sykesville, MD). Noble Life Sciences is fully accredited by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC), has Office of Laboratory Animal (OLAW) assurance and is USDA licensed.
CRO serum measurement
Serum was analysed for CRO using a high‐performance liquid chromatography (HPLC) method (Owens et al. 56). Pig serum samples (100 μl) were deproteinated with acetonitrile (500 μl), centrifuged and the supernatants combined with 1·0 ml methylene chloride. Samples were centrifuged and 20 μl of the aqueous layer was injected into a NovaPak C18 reverse‐phase column (4 μm, 3·9 × 150 mm; Waters Corp, Milford, MA) and monitored for absorption at 254 nm. The mobile phase consisted of 0·01 mol l−1 phosphate buffer, pH 7·0, 0·01 mol l−1 tetrapentylammonium bromide and acetonitrile in a 75 : 25 ratio pumped at a flow rate of 1·2 ml min−1. Pooled naïve pig serum was used to prepare the standard curve with six points ranging from 0·5 to 50 μg ml−1. The assay was linear over a range of 0·5–50 μg ml−1 (R = 1·00). Interday coefficients of variation for the low (1·0 μg ml−1) and high (40 μg ml−1) quality control samples were 5·3 and 3·8% respectively. Peak height was used to integrate all the peaks. Sigma Plot was used to calculate drug concentrations and a −1 weighting factor was used. The HPLC assay was developed, validated and performed by the Center for Anti‐Infective Research and Development at Hartford Hospital (Hartford, CT). Statistical analyses were performed using a two‐tailed Student's t‐test.
Faecal DNA extraction, whole‐genome shotgun sequencing and metagenomic analyses
Total DNA was isolated from faecal specimens, using the MOBIO Power‐Soil® DNA Isolation Kit (Qiagen, Germantown, MD), following the manufacturer's instructions. Each DNA sample was normalized in 3–18 μl of nuclease‐free water for a final concentration of 0·5 ng μl−1 using the Biomek FX liquid handler (Beckman Coulter Life Sciences, Brea, CA). Libraries were constructed using the Nextera XT Library Prep Kit (Illumine, San Diego, CA). For each sample, an input of 0·5 ng was used in the tagmentation reaction, followed by 13 cycles of PCR amplification using Nextera i7 and i5 index primers and 2X KAPA master mix per the modified Nextera XT protocol. The PCR products were purified using 1·0X speed beads and eluted in 15 μl of nuclease‐free water. The final libraries were quantified by PicoGreen fluorometric assay (100X final dilution) and the concentrations were in the range of 0·1–4·0 ng μl−1. The libraries were pooled by adding an equimolar ratio of each based on concentration determined by PicoGreen, and loaded onto a high sensitivity (HS) chip run on the Caliper LabChipGX (Perkin Elmer, Waltham, MA). The base pair size reported was in the range of 301–680 bp. Samples were sequenced using a single Illumina HiSeq v3 flowcell by multiplexing eight libraries per lane targeting 25 million 100 bp reads per sample. The majority of the libraries achieved that target, however, 10 libraries had a low sequencing yield. These 10 samples were remade and the concentration of each library was verified using Kapa qPCR. The samples were resequenced and all achieved well over 10 million reads per sample.
Unassembled whole‐genome shotgun metagenomic sequencing reads were directly analysed using the CosmosID, Inc. (Rockville, MD) bioinformatics software package, as described (Hasan et al. 31; Lax et al. 44; Ottesen et al. 55; Ponnusamy et al. 59), to achieve bacterial identification to species, subspecies and/or strain level and quantification of micro‐organism relative abundance. The software utilizes curated genome databases (GenBook®) and a high‐performance data‐mining algorithm that rapidly disambiguates hundreds of millions of short reads of a metagenomic sequence into discrete micro‐organisms engendering the identified sequences, without the need for sequence assembly. The analysis algorithm has two separable comparators. The first consists of a precomputation phase and the second is a per sample computation. The input to the precomputation phase is a curated reference microbial database that contained, at the time of this analysis, 5000 genomes of which 3775 belonged to the Bacteroidetes, Firmicutes or Proteobacteria phyla common to the gut. Of these three phyla, there were a total of 1510 species. The CosmosID, Inc. curated reference microbial database is updated continually. The output of the precomputation phase is a whole‐genome phylogeny tree, together with sets of fixed length k‐mer fingerprints (biomarkers) that are uniquely identified with distinct nodes and leaves of the tree. These biomarkers encompass all regions of the genome, both genic and intergenic. The second per sample computational phase searches the hundreds of millions of short sequence reads against k‐mer fingerprints. The resulting statistics are analysed to give fine‐grain composition and relative abundance estimates at all nodes of the tree. Overall classification precision is maintained through aggregation statistics. The CosmosID, Inc. metagenomic data analysis software accurately identifies bacteria to species, subspecies and/or strain level without the need for the complete genome to be available for each identified organism. With this analysis method, the number of kmers matching each bacterial taxon can be considered a surrogate for the number of reads matched to a particular taxon. In this study, the number of kmers per sample that were assigned to bacterial taxa ranged from 160 000 to 2 500 000, with an average of approximately 600 000. Similarly, the community resistome, the collection of antibiotic resistance genes in the microbiome, was also identified using the CosmosID, Inc. bioinformatics software package to query unassembled sequence reads against the CosmosID curated antibiotic resistance gene database in an analogous manner to that described for bacterial species identification. Antibiotic resistance genes were identified based on percentage of gene coverage for each gene as a function of the gene‐specific read frequency in each sample. For each reference gene, sets of unique k‐mers that span the entire gene are interrogated through the data sets and the average frequency of all k‐mers are recorded. In this study, the number of kmers per sample that were assigned to antibiotic resistance genes ranged from 320 000 to 2 000 000, with an average of approximately 800 000. The CosmosID, Inc. antibiotic resistance database is organized as a phylogenetic tree, which avoids the potential problems of highly similar sequences affecting abundance estimates. In addition, this approach circumvents the need for read assembly for each gene.
For comparative analyses among the treatment groups, all data sets were subsampled to a fixed 10 million read depth to ensure uniform population diversity and reduce bias in the data analyses arising from variation in read depth. Analyses of the sequence data included generation of heat maps based on relative abundance of each micro‐organism (%) in each sample, using NMF R software package (Gaujoux and Seoighe 26). Likelihood ratio testing was performed using a parameterization of the Dirichlet‐Multinomial distribution developed for comparisons of whole genome shotgun metagenomic data sets (La Rosa et al. 42). Similarity index calculations were performed as described (Tarkkanen et al. 75) using the Pearson correlation and boxplots were computed using the ggplot2 R library (McGill et al. 51). Principal coordinate analysis (PCoA) was performed using the Bray–Curtis distance measure and clustered using the Partitioning Around Medoids algorithm (Kaufman and Rousseeuw 37). Resistome analysis was performed by identification of antibiotic resistance genes based on the percentage of gene coverage for each gene as a function of the gene‐specific read frequency in each sample. Statistical analyses were performed using a one‐tailed Student's t‐test.
Data availability
Faecal DNA metagenomics sequencing data are available in Sequence Read Archive (SRA) (https://submit.ncbi.nlm.nih.gov/subs/sra/), Accession SRP093227.
Results
Ribaxamase does not affect systemic CRO levels
Pigs were treated with IV CRO once daily for seven consecutive days. Ribaxamase was delivered orally four times per day, starting the day before CRO treatment for nine consecutive days. To verify that ribaxamase did not affect systemic CRO levels in the pigs, animals that received CRO alone and animals that received CRO+ribaxamase had blood drawn on day 2 of antibiotic treatment, which corresponds to day 3 of ribaxamase delivery. At the time of blood collection, animals had received a total of two doses of CRO, and for the CRO+ribaxamase cohort, 13 doses of ribaxamase. As blood collections required anesthetization of animals, only three blood draws were conducted within a 24‐h period, at time points of 1, 6, and 19 h after CRO delivery, and serum analysed for the presence of CRO. CRO serum levels were not statistically different for CRO‐alone and CRO+ribaxamase cohorts at 1 and 6 h (Fig. 1). The CRO levels at the 19‐h time point were below the limit of detection of the assay (0·5 μg ml−1), indicating that the CRO half‐life in pigs is less than the 8–10 h reported for humans (Rocephin; Genentech USA, Inc. 27). These data confirm that ribaxamase did not alter serum antibiotic levels in the pigs, similar to data obtained in the dog CRO pharmacokinetic studies (Kokai‐Kun et al. 41).
Ribaxamase does not affect systemic ceftriaxone (CRO) levels. CRO was measured in pig serum collected on day 2 of antibiotic treatment using a validated HPLC‐based assay. Pigs were treated with CRO alone (n = 5; black bars) or CRO+ribaxamase (n = 5; white bars). The data are displayed as mean and standard deviation. P values were obtained by comparing the CRO alone and the CRO+ribaxamase groups at each time point using a two‐tailed Student's t‐test. At 2 h, P = 0·763 and at 6 h, P = 0·079.
Ribaxamase attenuates CRO‐mediated gut dysbiosis
To assess CRO‐mediated changes to the microbiome of the pigs, faecal DNA was subjected to whole‐genome shotgun metagenomic analyses. Heatmaps of the bacterial taxa were constructed based on the relative abundance of each bacterial strain in each sample for each animal. Heatmaps were organized to allow comparison of the microbiomes of animals before and after treatment with CRO or CRO+ribaxamase (Fig. 2). Compared to pretreatment microbiomes (days −7 and −4), CRO resulted in the reduction and/or loss of specific bacterial species and overgrowth of other taxa. These antibiotic‐induced microbiome changes occurred rapidly and were apparent at the first postantibiotic treatment time point (day 4). In contrast, CRO administered along with ribaxamase showed fewer changes to the faecal microbiota. Notably, the total number of bacterial species detected in each faecal sample remained relatively constant at each time point in both treatment cohorts. Specifically, at pretreatment day −7, and −4, a total of 40 and 39 taxa, respectively, were detected in the CRO cohort, and 43 and 33, in the CRO+ribaxamase cohort. In the post‐treatment day 4 and 8 samples, 37 and 50 taxa, respectively, were detected in the CRO cohort, and 46 and 47 taxa, in the CRO+ribaxamase cohort. In the CRO cohort, decreased abundance was observed in the Ruminococcus, Clostridiales, Dorea and Coprococcus genera, with the bacterial species Faecalibacterium prausnitzii and Oxalobacter formigenes most affected. The relative abundances of F. prausnitzii and O. formigenes decreased 50% from day −4 to day 8 in the CRO cohort while no decrease was observed in the CRO+ribaxamase group. An increase was observed in the Bacteroides, Parabacteroides and Fusobacterium genera. Bacteroides vulgatus, Parabacteroides distasonis and Fusobacterium varium relative abundances increased from not detectable at day −7 and day −4 in CRO and CRO+ribaxamase cohorts to an average of 1·8% (detected in 3/5 pigs), 2·5% (detected in 5/5 pigs) and 2·2% (detected in 5/5 pigs), respectively, in the CRO cohort at day 8. In the presence of ribaxamase, B. vulgatus, P. distasonis and F. varium relative abundances were 0, 0·06 (detected in 1/5 pigs) and 0·04% (detected in 1/5 pigs), respectively, at day 8.
Heatmap analysis of the relative abundance of bacterial species present in the pig faecal microbiome. Pigs were treated with CRO (n = 5) or CRO+ribaxamase (n = 5). Faeces were collected prior to antibiotic treatment (days −7 and −4), and after treatment (days 4 and 8). Pig faecal DNA was subjected to whole‐genome shotgun metagenomic analyses to determine relative abundance of bacterial species in each sample. Data are displayed as the abundance of each bacterial species relative to all species in each faecal sample. Each row of the heat map represents an individual animal at the indicated time point. The bacterial taxa are displayed at the bottom of the figure, the treatment group and day of collection of the faecal sample on the left, and the animal numbers on the right (P1–P10). The red boxes display bacterial taxa diminished in the CRO cohort at days 4 and 8, and maintained in the CRO+ribaxamase cohort as compared to pretreatment days −7 and −4. The yellow boxes display bacterial taxa enriched in the CRO cohort at days 4 and 8 and their diminished abundance in the presence of ribaxamase (CRO+ribaxamase cohort), as compared to pretreatment days −7 and −4. The colour gradient key displays a linear scale of the relative abundance.
A likelihood ratio test employing parameterization of the Dirichlet‐Multinomial distribution (La Rosa et al. 42), was used to compare microbiomes of pretreatment (day −4) with post‐treatment (days 4 and 8) (Table 1). Comparison of microbiome populations of the CRO cohort yielded chi‐squared values of 185 (P < 0·0001) and 470 (P < 0·0001) post‐treatment (days 4 and 8, respectively), indicating that pre‐ and postantibiotic faecal microbiomes were significantly different. In contrast, the CRO+ribaxamase chi‐squared values were 63 (P = 0·17) for day 4 and 79 (P = 0·38) for day 8, indicating the two faecal sample populations were not significantly different. These results indicate that CRO is associated with a significant change in the microbiome, whereas the addition of ribaxamase protected against changes in the microbiota poplulation caused by CRO. Heatmap analyses yielded similar results (Fig. 2), with CRO‐mediated changes to the microbiome becoming apparent within 4 days of antibiotic treatment. Therefore, additional microbiome sequence analyses were performed comparing pretreatment day −4 with post‐treatment day 4 samples.
Microbiome profiles are not significantly different in the presence of ribaxamase. Microbiome population profiles prior to antibiotic treatment (day −4) were compared to those after antibiotic treatment (day 4 and day 8) using a likelihood ratio test (La Rosa et al. 42)
| Treatment group | Day | Chi‐squared | P value |
|---|---|---|---|
| Ceftriaxone | 4 | 185 | <0·0001 |
| 8 | 470 | <0·0001 | |
| Ceftriaxone+ribaxamase | 4 | 63 | 0·17 |
| 8 | 79 | 0·38 |
| Treatment group | Day | Chi‐squared | P value |
|---|---|---|---|
| Ceftriaxone | 4 | 185 | <0·0001 |
| 8 | 470 | <0·0001 | |
| Ceftriaxone+ribaxamase | 4 | 63 | 0·17 |
| 8 | 79 | 0·38 |
Microbiome profiles are not significantly different in the presence of ribaxamase. Microbiome population profiles prior to antibiotic treatment (day −4) were compared to those after antibiotic treatment (day 4 and day 8) using a likelihood ratio test (La Rosa et al. 42)
| Treatment group | Day | Chi‐squared | P value |
|---|---|---|---|
| Ceftriaxone | 4 | 185 | <0·0001 |
| 8 | 470 | <0·0001 | |
| Ceftriaxone+ribaxamase | 4 | 63 | 0·17 |
| 8 | 79 | 0·38 |
| Treatment group | Day | Chi‐squared | P value |
|---|---|---|---|
| Ceftriaxone | 4 | 185 | <0·0001 |
| 8 | 470 | <0·0001 | |
| Ceftriaxone+ribaxamase | 4 | 63 | 0·17 |
| 8 | 79 | 0·38 |
A comparison of relative abundance of strains of the faecal bacteria in the baseline samples (day −4) with day 4 treatment samples was performed using a similarity index (Tarkkanen et al. 75). Mean decrease in similarity index for the CRO cohort was 64% compared to 20% for the CRO+ribaxamase group (Fig. 3), indicating that the CRO+ribaxamase cohort retained 80% similarity to pretreatment while the CRO group showed 36% similarity. Therefore, CRO caused a greater divergence from the pretreatment microbiomes that was attenuated by the presence of ribaxamase.
Strain relative abundance analysis of faecal microbiomes. The strain relative abundance based on per cent similarity (Tarkkanen et al. 75) using the Pearson correlation was calculated for CRO (black box) and CRO+ribaxamase (white box) day 4 microbiomes compared to pretreatment day −4 microbiomes (grey line). The changes in similarity from day −4 (baseline) are indicated with the grey arrows.
PCoA, comparing pretreatment microbiomes (day −4) with those at day 4 for CRO and CRO+ribaxamase cohorts (Fig. 4), was performed. Distance between points indicates degree of difference in sample diversity with points close together more similar. CRO+ribaxamase samples clustered closely with pretreatment samples, while CRO samples formed two discrete groups, distinct from the single cluster containing pretreatment controls and CRO+ribaxamase samples. Therefore, bacterial populations of the control and CRO+ribaxamase groups were more similar to each other than to those of the CRO group. In contrast, CRO microbiomes diverged greatly from pretreatment populations and each other. Results of the microbiome analyses demonstrate that ribaxamase reduces CRO‐mediated impact on the microbiome.
Principal coordinates analysis of faecal microbiomes. Faecal microbiomes from each animal were analysed via principal coordinate analysis using the Bray–Curtis distance measure and clustered using the PAM algorithm (Kaufman and Rousseeuw 37). Component 1 and component 2 captured 54% of the variation between the samples. Pretreatment day −4, green circles; day 4 post‐treatment CRO, blue triangles; and CRO+ribaxamase, orange squares.
Ribaxamase reduces antibiotic resistance gene propagation
To determine if ribaxamase affected the propagation of antibiotic resistance, faecal DNA whole‐genome shotgun metagenomic data were analysed for the presence of antibiotic‐resistant genes, as a measure of the population of antibiotic‐resistant bacteria in the faecal microbiomes. Heatmaps of identified antibiotic resistance genes in the faecal microbiome of each animal compared before and after antibiotic treatment (Fig. 5a) showed CRO was associated with higher abundance of antibiotic resistance genes post‐treatment (day 4) compared with CRO+ribaxamase microbiomes. Many of the resistance genes detected at day 4 were encoded beta‐lactamases (Fig. 5b). Specifically, blaOXA genes encoding extended‐spectrum OXA class D beta‐lactamases (McArthur and Wright 48; Bush et al. 10), were present at high levels in two of five CRO animals at day 4 but absent or observed at lower frequencies in the CRO+ribaxamase cohort. Similarly, blaCTX‐M_82 or blaCFX_A4 genes, encoding extended spectrum class A serine beta‐lactamases (McArthur et al. 49), were detected at day 4 at high levels in two of the CRO animals. An additional beta‐lactamase gene, AmpC, encoding the extended spectrum cephalosporin‐resistant class C beta‐lactamases (McArthur et al. 49), was present in all but one animal prior to antibiotic treatment at day −7. From day −7 to day −4, AmpC levels increased in both cohorts prior to any antibiotic‐mediated selection mechanisms. By day 4, however, AmpC gene frequencies continued to increase in the CRO cohort, while, in the CRO+ribaxamase animals, AmpC decreased to levels similar to day −7 frequencies.
Heatmap analysis of the frequency of antibiotic resistance genes in the pig faecal microbiomes. Faecal microbiome metagenomic data were analysed for the presence of antibiotic resistance genes based on the percentage of gene coverage as a measure of the relative gene frequency in each sample. Each row of the heat map represents an individual animal at the indicated time point, day −7, day −4, or day 4. The antibiotic resistance genes are displayed at the bottom of the figure, the treatment group and day of collection of the faecal sample on the left, and the animal numbers on the right (P1–P10). (a) Heatmap displaying all antibiotic resistance genes found in the faecal microbiomes. (b) Heatmap displaying only beta‐lactamase genes. The black and yellow boxes indicate antibiotic resistance genes that are more prevalent in the CRO cohort compared to the CRO+ribaxamase group. The colour gradient key displays a linear scale of the per cent gene coverage as a measure of the relative gene frequency.
In addition to beta‐lactamases, other resistance genes displayed an increased frequency in response to antibiotic exposure. Many of these genes encode components of multidrug efflux transporter systems, systems that confer resistance to a broad range of antibiotics, including the beta‐lactams (Sun et al. 72). Several of these genes were selected for additional analysis, including: acrE, encoding a component of the AcrEF‐TolC multidrug efflux transporter system (Lau and Zgurskaya 43; McArthur et al. 49); baeR, encoding a response regulator of the MdtABC multidrug efflux transporter system (Nagakubo et al. 53; McArthur et al. 49); emrY, encoding a component of the EmrKY‐TolC multidrug efflux transporter system (Tanabe et al. 74; McArthur et al. 49); mdtD, encoding a component of the MdtABC multidrug efflux transporter system (Nagakubo et al. 53; McArthur et al. 49); and mdtN, encoding a multidrug resistance efflux pump from the major facilitator superfamily (Sulavik et al. 71; McArthur et al. 49). Two genes closely related to the beta‐lactamases, pbp2, encoding penicillin‐binding protein 2 (Bharat et al. 7), and pbp4, encoding penicillin‐binding protein 4 (Sun et al. 73) were also selected, in addition to the beta‐lactamase, AmpC, encoding the cephalosporin‐resistant Class C beta‐lactamases (McArthur et al. 49). For each selected gene, change in the mean of relative gene frequencies from day −4 to day 4 was compared in the CRO or CRO+ribaxamase cohorts (Fig. 6). Relative gene frequencies increased for each gene in the CRO cohort. In contrast, the gene frequencies decreased for each gene in the CRO+ribaxamase group, except pbp2, the levels of which increased only slightly above baseline. The greatest increase was observed for the mdtD gene, the levels of which doubled from day −4 to day 4 in the CRO cohort, while mdtD levels were reduced in the CRO+ribaxamase cohort.
Changes in the frequency of selected antibiotic resistance genes. The change in the relative frequency (mean) of the indicated antibiotic resistance genes for the CRO (black bars) or CRO+ribaxamase (white bars)‐treated animals from pretreatment day −4 compared to post‐treatment day 4 is displayed. A negative value indicates a reduction in frequency, a positive value indicates an increased frequency, and a zero value represents no change in gene frequency. The genes are listed on the horizontal axis: acrE, encodes a component of the AcrEF‐TolC multidrug efflux transporter (Lau and Zgurskaya 43; McArthur et al. 49); baeR; encodes a response regulator of the MdtABC multidrug efflux transporter system (Nagakubo et al. 53; McArthur et al. 49); emrY, encodes a component of the EmrKY‐TolC multidrug efflux transporter system (Tanabe et al. 74; McArthur et al. 49); mdtD, encodes a component of the MdtABC multidrug efflux transporter system (Nagakubo et al. 53; McArthur et al. 49); mdtN, encodes a multidrug resistance efflux pump from the major facilitator superfamily (Sulavik et al. 71; McArthur et al. 49); pbp2, encodes penicillin‐binding protein 2 (Bharat et al. 7), pbp4, encodes penicillin‐binding protein 4 (Sun et al. 73) and AmpC, encodes a class C beta‐lactamase (McArthur et al. 49).
Two antibiotic resistance genes that convey resistance to nonbeta‐lactam antibiotics, aminoglycosides and tetracycline, were selected for further analysis (Fig. 7). Aminoglycoside_strA (Scholz et al. 65) encodes an aminoglycoside phosphotransferase (McArthur et al. 49) and Tetracycline_tet39 (Agerso and Guardabassi 1) encodes a component of a tetracycline efflux pump (McArthur et al. 49). Neither gene was detected in the pretreatment samples (day −7). At day −4, both genes were detected in one of the animals (Pig 3). At day 4 after antibiotic treatment, both genes were detected at high levels in four of the five animals treated with CRO. In contrast, in the CRO+ribaxamase cohort, aminoglycoside_strA levels remained low or absent in all animals and only one animal in the CRO+ribaxamase cohort was identified with a high level of the tetracycline_tet39 gene. Aminoglycoside_strA and tetracycline_tet39 gene frequencies at day 4 for CRO and CRO+ribaxamase cohorts were both significantly different, P = 0·025 and P = 0·045 respectively. These data demonstrate that ribaxamase was associated with reduced propagation of antibiotic resistance genes in the faecal microbiomes in the porcine antibiotic‐mediated gut dysbiosis model.
Changes in the frequency of selected antibiotic resistance genes. The relative frequency of two selected antibiotic resistance genes, (a) Aminoglycoside_strA and (b) Tetracycline_tet39, are displayed from CRO (left panel in blue) or CRO+ribaxamase (right panels in orange)‐treated animals. Each dot represents the relative gene frequency in each animal's microbiome from pretreatment days −7, or −4, or post‐treatment day 4. The medians are displayed in each data set. P values were obtained by comparing CRO and CRO+ribaxamase groups at day 4 using a one‐tailed Student's t‐test. P = 0·025 for Aminoglycoside_strA and P = 0·045 for Tetracycline_tet39.
Discussion
The importance of the microbiome in health and disease is increasingly evident. Antibiotics, while life‐saving, have an unintended consequence of mediating changes to the pretreatment microbiomes, termed dysbiosis, and promoting emergence of antibiotic‐resistant micro‐organisms. Ribaxamase is the first of a class of agents representing a new treatment paradigm, namely, protection of the microbiome by antibiotic inactivation without interfering with infection control. Ribaxamase, intended for oral use with IV beta‐lactam antibiotics, protected the microbiome from antibiotic‐mediated dysbiosis and reduced the propagation of antibiotic resistance genes, without affecting the systemic antibiotic levels. Beta‐lactam antibiotics, including penicillins and cephalosporins, are commonly prescribed and are highly effective as broad‐spectrum antibiotics. Unfortunately, they are also associated with the emergence of opportunistic infections like C. difficile disease (Crowther and Wilcox 15; Smits et al. 67). Co‐administration of ribaxamase with beta‐lactams is anticipated to allow exploitation of anti‐infective efficacy without adverse GI effects associated with beta‐lactams.
In this study, a porcine model of antibiotic‐mediated dysbiosis was developed for the evaluation of microbiome protection. The use of whole‐genome shotgun metagenomic sequencing for faecal microbiome analyses allowed elucidation of species and strain‐level details not available from 16S ribosomal DNA sequencing. Microbiome analyses demonstrated that ribaxamase, when delivered with CRO, shielded the pig microbiota from antibiotic‐mediated damage. CRO‐mediated dysbiosis is manifested as a reduction and/or loss of bacterial species and outgrowth of other species. The number of bacterial taxa identified in each sample did not change with antibiotic treatment, presumably due to the loss of specific species being compensated by the emergence of others. However, CRO treatment caused a significant change in the composition of the microbiome. In contrast, with ribaxamase, CRO‐mediated changes were reduced and post‐treatment microbiomes were not significantly different from pretreatment microbiota populations. Notably, CRO exposure reduced the abundance of two key commensal species, F. prausnitzii and O. formigenes, by at least 50% between pretreatment day −4 and treatment day 8. With ribaxamase no decrease was observed. Low levels of both species have been associated with human disease; F. prausnitzii, an important butyrate‐producing bacterium, with Crohn's disease (Sokol et al. 68; Quevrain et al. 60) and allergic conditions (Melli et al. 52), and O. formigenes, an oxalate‐metabolizing bacterium, with higher prevalence of calcium oxalate kidney stones (Kaufman et al. 38). Outgrowth of several species, B. vulgatus, P. distasonis and F. varium, all associated with colitis in animal models and/or humans (Rath et al. 61; Allen‐Vercoe 2; Dziarski et al. 22), was observed after exposure to CRO but was attenuated in the presence of ribaxamase.
In addition to causing microbiome damage, antibiotic exposure fosters bacterial resistance (The Review on Antimicrobial Resistance 76; Francino 24), with worldwide overuse of antibiotics contributing to the global threat of multidrug‐resistant pathogenic organisms (The Review on Antimicrobial Resistance 76; Francino 24). Antibiotic stewardship programmes have been implemented that focus on the prudent use of antibiotics, including reducing administration of broad‐spectrum antibiotics and antibiotic cocktails, and promoting more precise, targeted bactericidal strategies (Centers for Disease Control and Prevention 13). In this study, ribaxamase was shown to reduce the emergence of antibiotic resistance genes in the pig faecal microbiome. Whole‐genome shotgun metagenomic sequencing provided relevant information on in the presence of genes conferring antibiotic resistance that would not have been discernible with 16S ribosomal DNA sequencing. Analyses revealed pigs treated with CRO displayed increased frequency of antibiotic resistance genes compared with animals receiving CRO+ribaxamase.
Antibiotic resistance genes enriched in microbiomes of pigs treated with CRO encoded beta‐lactamases, components of multidrug efflux transporter systems and the phosphotransferase aminoglycoside_strA (McArthur et al. 49). The beta‐lactamase genes included cfxA, CTX_M, OXA and AmpC, each of which represents an extended‐spectrum beta‐lactamase family that confers resistance to penicillins, cephalosporins and, in some cases, carbapenems (McArthur et al. 49). Interestingly, CRO exposure resulted in the expansion of a broad range of antibiotic resistance genes and not just those conferring resistance to the beta‐lactams. Similar to the results obtained in this study, in pigs treated with a growth‐promoting antibiotic cocktail containing chlortetracycline, sulfamethazine and penicillin, the abundance of antibiotic resistance genes was enhanced, including a gene encoding an aminoglycoside phosphotransferase, even though the animals were not exposed to an aminoglycoside (Looft et al. 46). Moreover, the control animals not exposed to antibiotics displayed a decrease in the frequency of antibiotic resistance genes (Looft et al. 46), a phenomenon also noted in the current study. These observations may be explained by the fact that multiple resistance genes are frequently carried together on transferrable plasmids (Carattoli 11) and aggregate on movable elements in response to selective pressure (Barlow 5). Both the aminoglycoside_strA (Scholz et al. 65) and tetracycline_tet39 (Agerso and Guardabassi 1) genes are found on plasmids. However, it is unlikely that these genes were on the same plasmid as all animals did not harbour both genes simultaneously and at the same frequencies. An alternative hypothesis is that the beta‐lactam antibiotic kills commensal species thus enabling the overgrowth of species resistant to a multitude of antibiotics (Britton and Young 8). For example, an analysis of vancomycin‐resistant enterococcus (VRE) colonization in patients treated with a panel of antibiotics revealed VRE propagation was enhanced by antibiotics targeting obligate anaerobes and was not solely a result of vancomycin treatment (Donskey et al. 19). Irrespective of the mechanisms of selection for antibiotic resistance, CRO treatment increased the abundance of antibiotic resistance genes, while ribaxamase attenuated this enrichment.
Initially, ribaxamase is intended for use with IV beta‐lactam antibiotics. However, the beta‐lactams are frequently administered in combination with a beta‐lactamase inhibitor. As expected for class A serine beta‐lactamases, ribaxamase, and its precursor, P1A, displayed reduced activity in the presence of high concentrations of sulbactam and tazobactam in vitro (Stiefel et al. 70; Kaleko et al. 35). Surprisingly, beta‐lactamase inhibitors did not affect P1A efficacy in clinical studies with piperacillin/tazobactam (Pitout 58) nor in dogs treated with piperacillin/tazobactam, ampicillin/sulbactam or amoxicillin/clavulanate (Pitout 58). These seemly contradictory observations are hypothesized to be the result of the different excretion kinetics of the antibiotic and inhibitor in vivo and/or sufficiently high beta‐lactamase enzyme concentrations in the intestine overcoming inhibition. Ribaxamase may also prove useful for patients receiving multiple antibiotics, either simultaneously with, prior to or following beta‐lactams. While it is clear that certain antibiotics carry higher risks for C. difficile infection (Crowther and Wilcox 15; Smits et al. 67; Zycinska et al. 81) and VRE (Donskey et al. 19), it is becoming apparent that total antibiotic exposure, including duration of therapy, and the dose utilized, are factors in secondary infection risk (Smits et al. 67) and in emergence and propagation of antibiotic‐resistant micro‐organisms (Centers for Disease Control and Prevention 13). Thus, ribaxamase has the potential to be efficacious with beta‐lactam/inhibitor combinations and, by reducing cumulative antibiotic exposure, lessening microbiome disruption in patients receiving multiple antibiotics.
The gut microbiome represents a fragile balance of micro‐organisms that can be disrupted by antibiotic use. Interference with the normal ecology of the gut can have a negative impact on health. The ultimate goal of the antibiotic inactivation strategy, as described here, is the preservation of the native, predisease microbiome. Ribaxamase is intended to mitigate the risk associated with beta‐lactam antibiotic treatment without affecting its well‐established, broad‐spectrum efficacy. The results of this study demonstrate that ribaxamase can mitigate gut microbiome disruption caused by the administration of an IV beta‐lactam antibiotic in a pig model of antibiotic‐mediated gut dysbiosis.
Acknowledgements
The authors acknowledge Jeffrey Riley, CEO of Synthetic Biologics, Inc. for providing support for the SYN‐004 (ribaxamase) technology and Dr John Monahan for guidance on the research program. We are grateful to Amy McDonald at Noble Life Sciences, Inc. for providing outstanding technical assistance with pig dosing and sample collection. We thank Drs Christian Furlan‐Freguia, Klaus Gottlieb and John Kokai‐Kun for critical review of the manuscript, and Dr Stephen Altieri and Leslie Marlow for disclosure review of the manuscript.
Conflict of Interest
The authors declare the following potential conflicts of interest with respect to the research, authorship and/or publication of this article: S.C., J.A.B., S.H. and M.K. are employees of Synthetic Biologics, Inc. R.C. is the founder of CosmosID, Inc., a fee‐for‐service provider engaged by Synthetic Biologics, Inc. P.S. and N.H. are employees of CosmosID, Inc. The authors disclose receipt of the following financial support for the research, authorship and/or publication of this article: This work was supported by Synthetic Biologics, Inc.
