Repurposing tebipenem pivoxil as alternative therapy for severe gastrointestinal infections caused by extensively-drug resistant (XDR) Shigella spp.

Diarrhoea remains one of the leading causes of childhood mortality globally. Recent epidemiological studies conducted in low-middle income countries (LMICs) identied Shigella spp. as the rst and second most predominant agent of dysentery and moderate diarrhoea, respectively. Antimicrobial therapy is often necessary for Shigella infections; however, we are reaching a crisis point with ecacious antimicrobials. The rapid emergence of resistance against existing antimicrobials in Shigella spp. poses a serious global health problem. Here, aiming to identify alternative antimicrobial chemicals with activity against multi-drug resistant (MDR) Shigella, we initiated a collaborative academia-industry drug discovery project, applying high throughput phenotypic screening across broad chemical diversity. We identied several suitable compounds with antibacterial activity against Shigella. These compounds included the oral carbapenem tebipenem, which was found to be highly potent against broadly susceptible Shigella and contemporary MDR variants. Additional in vitro screening demonstrated that tebipenem had activity against a wide range of other non-Shigella enteric bacteria. Cognisant of the risk for the development of resistance against monotherapy, we identied synergistic behaviour of two different drug combinations incorporating tebipenem. The orally bioavailable prodrug (tebipenem pivoxil) effectively cleared the gut of infecting organisms when administered in physiological doses to Shigella-infected mice and gnotobiotic piglets. Our data highlight the utility of broad compound screening for tackling the emerging antimicrobial resistance crisis and shows that tebipenem pivoxil (licenced for paediatric respiratory tract infections in Japan) could be repurposed as an effective treatment for severe diarrhoea caused by MDR Shigella and other enteric pathogens in LMICs. Daltonic). After the “MALDI” section was obtained a consecutive section for histology was taken. Prior to matrix coating samples for MALDI were optical scanned in a PathScanner. The matrix deposition was conducted in a sprayer ImagePrep System (Bruker Daltonic) which use The matric solution was α-CHCA (α-Cyano-4-hydroxycinnamic dissolved in 70:30 methanol/water (v/v) containing 0.2 % TFA at 7 mg/mL as the matrix solution. MALDI MSI data were obtained using a UltraeXtream TOF/TOF (Bruker


Introduction
Humanity is currently in the midst of a global antimicrobial resistance (AMR) crisis 1 . Many of the antimicrobials that we have historically used to treat and/or control infections caused by common bacterial pathogens have lost or are rapidly losing their e cacy. This situation has become worryingly common for a host of diseases and pathogens, but arguably the issue is epitomised by enteric (diarrhoeal) infections that arise in low and middle-income countries (LMICs). Annually, diarrhoea accounts for approximately 1.3 million deaths globally, with the preponderance of these occurring in young children residing in LMICs 2 . Due to a lack of other interventions, antimicrobials are a key tool in limiting the impact of these infections, but as AMR increases these infections are likely to have an even greater impact on the most vulnerable individuals.
Diarrhoeal disease is notoriously complex to disaggregate as a wide range of pathogens can trigger the syndrome; but of all the potential aetiological agents, bacteria belonging to genus Shigella pose the greatest AMR threat. Recent estimates suggest that Shigella are associated with ~ 125 million diarrhoeal episodes and 200,000 deaths annually 3 . Additionally, Shigella (and enterotoxigenic Escherichia coli (ETEC)) were found to be the most predominant bacterial pathogens in the stools of paediatric diarrhoeal patients in South Asia and sub-Saharan Africa 4 . This study also found that Shigella were the most prevalent pathogen in diarrhoeal children aged 2 to 5 years; a reanalysis of stool samples from this study using detailed molecular approaches suggested that burden of Shigella may actual be double that of previous estimates 5 .
Clinical evidence strongly supports the use of antimicrobials to treat severe Shigella associated diarrhoea 6 . Appropriate antimicrobials reduce the duration of fever and diarrhoea associated with Shigella by 1-2 days, which in turn treatment limits the severity of symptoms and the risk of further complications. Additionally, the shedding of the pathogen in the stools is prevented by antimicrobials, which then reduces onward transmission. As there is currently no licenced vaccine for Shigella, antimicrobials are one of the only reliable mechanisms available for disease control. However, Shigella are highly adept at acquiring multi-drug resistance (MDR) plasmids from other member of the Enterobacteriaceae and resistance mutations are commonly associated with "successful" lineages. AMR reduces treatment e cacy and limits treatment options. Consequently, there has been an "arms race" between the pathogen and antimicrobials, with the WHO ultimately recommending cipro oxacin, azithromycin, pivmecillinam, and ceftriaxone as the empirical drug choices for Shigella infections. Inevitably, resistance to these antimicrobials has emerged and increasingly resistant organisms, such as extensively-drug resistant (XDR) Shigella sonnei in Vietnam 7 , are spreading across LMICs 8 , in travellers returning from LMICs, and in men-who-have-sex-with-men (MSM) 9 . Accordingly, the US-CDC has declared AMR Shigella a serious global health threat requiring new interventions and the WHO has placed them on the global priority list of antibiotic-resistant bacteria to guide research, discovery, and development of new antibiotics 10 .
After the genomics revolution of the late 1990s, the research community embarked on a series of targetbased drug discovery programmes to identify novel antimicrobial compounds. This strategy, because of lack of translation from target-activity to whole-cell activity, produced largely disappointing results 11 . Conversely, phenotypic antimicrobial screens have proven to yield a higher hit rate than target-based approaches with several successful examples emerging in infectious diseases 12,13 . Aiming to identify novel compounds with activity against MDR Shigella, we established an academic-industrial collaboration 14 , creating a platform to perform in vitro phenotypic screening of > 1.5 million compounds against antimicrobial susceptible and MDR Shigella. We interrogated a collecting of compounds spanning broad chemical diversity, comprised of a combination of naive chemistry, compounds originating from previous antibacterial drug discovery programs, and currently marketed compounds.
Using this approach, we identi ed several novel options, including an oral carbapenem, that may be accelerated into further testing to treat infections caused by MDR Shigella and other gastrointestinal pathogens.

High throughput screening of compound collections
We conducted an antibacterial screening campaign, incorporating both GSK proprietary and nonproprietary compounds, to identify new chemical entities (NCE), repurposing and repositioning opportunities that could be targeted towards treating MDR Shigella infections. The chemical diversity included the GSK screening collection (~ 1.7 million compounds of unbiased chemical diversity) and ~ 4,000 compounds with known antibacterial pedigree from the GSK antibacterial programs (chemical diversity selected by MIC < 1µM against E. coli) and recognised, commercially available antimicrobials. A owchart of the screening procedure is shown in Fig. 1A. The compounds were tested at a concentration of 10µM for their ability to inhibit S. exneri 2457T growth in liquid medium using resazurin reduction as a surrogate of bacterial viability. Compounds ful lling the speci ed hit criteria, were progressed for dose response for Shigella 2457T growth inhibition. Ultimately, we identi ed 2,269 compounds in seven classes (bacterial topoisomerase inhibitors (BTIs), inhibitors of UDP-3-O-(R-3-hydroxymyristoyl)-Nacetylglucosamine deacetylase (LpxCs), microcin PDI, Met-tRNA inhibitors, β-lactams, uoroquinolones, and pleuromutilins) that inhibited Shigella 2457T growth with IC50 ≤ 1µM (Fig. 1). Two classes of antibacterial compounds exhibit activity against MDR Shigella The 2,269 selected active compounds identi ed via high throughput screening were advanced for evaluation against two clinical MDR Shigella isolates originating in Vietnam, S. exneri EG478 and S. sonnei 02-1181; the antimicrobial susceptibility pro les of these organisms are shown in Table S1. Among all chemical classes, compounds that retained activity in the clinical isolates belonged to only three compound classes: β-lactams, BTIs, and LpxCs. The majority (~ 90%) of BTI compounds had a signi cant loss of potency when tested against the two clinical MDR Shigella isolates compared to the original laboratory strain (Fig. 1B). This pattern was comparable for both the LpxCs and carbapenems, with compounds generally being more active against the laboratory strain than the clinical isolates.
Given the broad availability and acceptability of carbapenems globally, efforts to determine their potential to be progressed as treatment against shigellosis were prioritized. Meropenem, tebipenem, and ertapenem exhibited the greatest bactericidal activity against the two MDR Shigella clinical isolates and were prioritized for evaluation ( Fig. 1 and Table S2). Tebipenem was speci cally selected for further characterization as the pivoxil prodrug has good oral bioavailability and is commercially available in Japan 15 . Given the route of administration and the potential for accumulation in the gastrointestinal tract, we reasoned that tebipenem may be the most appropriate identi ed compound that could be quickly accelerated for treating MDR/XDR Shigella infections. Tebipenem as broad-spectrum antimicrobial for treating Shigella infections Under direct comparison in in vitro assays, we found that tebipenem inhibited the growth of the ATCC Shigella exneri with a potency that was comparable to currently available rst line antimicrobials in broadly susceptible organisms, including both cipro oxacin and ceftriaxone (Table 1). Given the propensity of Shigella spp. to horizontally acquire AMR genes, we assessed the activity of tebipenem against a collection of 82 contemporary clinical Shigella isolates (sonnei and exneri) from Vietnam and 53 (a range of species) from Kenya. The compound exhibited consistent in vitro activity against all clinical isolates, with MIC values ranging from 0.04 to 0.3µM ( Fig. 2A). Time kill assays of tebipenem performed in three of these Vietnamese clinical isolates (two S. exneri and one S. sonnei) further con rmed potent antibacterial activity of tebipenem at 2-4x MIC over 24 hours of exposure; population rebounds started to occur after six hours exposure with 0.5-1x MICs in all three isolates (Fig. 2B).  Fig. 2A). The frequency of generating spontaneous tebipenem resistance mutations was evaluated in vitro in Shigella exneri 2457T and determined to be in a low range compared to other antimicrobials (< 1x10 − 9 ). Ceftriaxone, meropenem, and ertapenem require intramuscular or intravenous administration, thereby restricting their use for severe infections in outpatients setting. These in vitro data indicate that tebipenem administered as a prodrug may be a practical alternative for oral treatment of severe Shigella associated diarrhoea in locations with endemic MDR and uoroquinolone resistant organisms.
As diarrhoea can be caused by multiple bacterial pathogens, we further assessed the activity of tebipenem against a range of alternative enteric bacterial pathogens using a further panel of 203 contemporary clinical isolates (Salmonella, E. coli, and Campylobacter). Tebipenem exhibited high level antibacterial activity against all non-Shigella enteric pathogens with MIC values ranging from 0.016 to 2 µg/ml, with a single Salmonella Enteritidis having an MIC of 4 µg/ml ( Fig. 2A) To protect the e cacy of the drug, potentially enhance its activity, and to mitigate the risk of resistance, we explored possible synergistic combinations between tebipenem and the current WHO approved drugs for shigellosis; azithromycin, cipro oxacin, ceftriaxone, and mecillinam. We additionally progressed a previous observation by combining tebipenem with a LpxC inhibitor (PF-5081090) 16 . Initially, these ve combinations were tested against laboratory isolates of Shigella, Salmonella, and E. coli (Sf2457T, St14028, and EcDH5a, respectively) (Fig. 3A). Tebipenem exhibited partial synergy with the LpxC inhibitor, reducing the Tebipenem MIC > 10-fold in all three organisms. Similarly, tebipenem exhibited partial synergy with azithromycin against S. exneri and S. Typhimurium, but this combination was antagonistic against E. coli. In combination with cipro oxacin, ceftriaxone, and mecillinam, tebipenem exhibited indifference or partial synergy with all three organisms. This potential synergism suggests that a combination of these drugs may be formulated to protect the e cacy of tebipenem by potentially limiting the emergence of spontaneous or acquired resistance to either of the antimicrobials. Consequently, we further assessed the potential for synergism between tebipenem with the LpxC inhibitor and also tebipenem in combination with azithromycin against a panel of alternative Gram negative MDR clinical isolates. The combination of tebipenem/LpxC inhibitor exhibited indifference or partial synergy in vitro against all the tested enteric bacteria (Salmonella spp., Shigella spp., K. pneumoniae, and E. coli; Fig. 3B).
A combination of tebipenem with azithromycin was partially synergistic in vitro against a range of invasive and non-invasive Salmonella isolates (Fig. 3B).

Tebipenem e cacy and pharmacokinetic properties in animal models of Shigella infection
Tebipenem is currently only approved in Japan for the treatment of respiratory tract infections. Aiming to con rm the potential of repurposing tebipenem for the treatment of enteric infections through in vivo experimentation we progressed the drug for pharmacokinetic and e cacy studies in animal models of shigellosis. The doses selected for e cacy studies were extrapolated from approved doses in children (4 and 6 mg/kg bid); equivalent doses in terms of systemic exposure were selected.
The pharmacokinetics of tebipenem was characterized in mice with intravenous and oral doses of tebipenem and tebipenem pivoxil (the oral ester prodrug of tebipenem), respectively. We found that systemic concentration of active drug was equivalent after doses of 39 mg/kg of subcutaneous tebipenem and 50 mg/kg of oral tebipenem pivoxil over the course of a six-hour experiment ( Fig. 4C and Table 2). The blood concentrations of subcutaneous tebipenem and oral gavage tebipenem pivoxil prodrug were found to be equivalent, with a comparable Cmax (71 and 62 µg/mL, respectively) and systemic exposure (AUC 0 − t , 80 and 94 µg·h/mL, respectively). The oral bioavailability of tebipenem when dosed as tebipenem pivoxil pro-drug was approximately 13-fold greater in mice than oral tebipenem as a free base; 39 mg/kg tebipenem free base orally had comparable bioavailability after 3 hours as 5 mg/kg tebipenem free base given intravenously (Table 2 and Fig. 4C). For the speci c treatment of enteric infections, oral administration of tebipenem pivoxil would theoretically facilitate release of the active ingredient at the site of infection. Therefore, we conducted qualitative MALDI-TOF analysis to assess the distribution of tebipenem pivoxil and (hydrolysed) tebipenem free base upon oral administration of tebipenem pivoxil in the intestinal tract of mice.
Monitorization of pro-drug and drug at 15 and 60 minutes ( Fig. 4A and 4B, respectively) post administration indicated that tebipenem pivoxil could be detected in the intestinal lumen, while tebipenem free base could be detected both in the lumen and within gut enterocytes. These data suggest that tebipenem pivoxil is only partly hydrolysed in the intestinal lumen, but hydrolysed immediately after penetration into enterocytes lining the apical surface of the gut, making it ideal for the treatment of invasive gut bacteria.
To determine if an increased systemic concentration of active drug by orally dosing the pivoxil prodrug resulted in superior performance, we compared the e cacy of the tebipenem pivoxil and tebipenem free base after challenging mice with Shigella. Oral tebipenem pivoxil resulted in a 2.3-log reduction of Shigella in the small intestine compared to no antimicrobial 24 hours post-infection. The quantity of Shigella in the small intestine in mice treated with oral tebipenem pivoxil was signi cantly lower at the same time point than tebipenem free base (1.2-log reduction, p ≤ 0.05) (Fig. 4D). We observed a similar in reduction in Shigella CFU the large intestine, but the difference in the e cacy between tebipenem pivoxil and tebipenem free base (with respect to reduction in Shigella CFU) was not statistically signi cant (data not shown). Notably, neither formulation reduced the bacterial load in the gut to below detectable limits.
Aiming to exploit a more physiological Shigella infection model we also measured the e cacy of tebipenem pivoxil in gnotobiotic piglets using a 50mg/Kg dose of tebipenem pivoxil (equivalent to the licenced 6 mg/kg dose in Japan). The tebipenem Cmax in the piglet was 8µg/ml and equivalent to those reported in a paediatric population 17 ; over the course of six hours the concentration of tebipenem declined to 1 µg/ml (Fig. 5A). Tebipenem pivoxil reduced the frequency of diarrheal stools in infected piglets, performing in an equivalent fashion to the cipro oxacin control. Explicitly, tebipenem pivoxil was highly e cacious in reducing Shigella in the gut, with a single dose of tebipenem pivoxil 24 hours after infection reducing the CFU/g Shigella in stool from 1x10 7 to zero within 24 hours, again performing equally well as cipro oxacin (Fig. 5B). We lastly performed necropsy and histopathology of the large intestine in tebipenem pivoxil-and placebo-treated piglets in order to compare the presence of intestinal lesions, in ammation, and the presence of intracellular Shigella. The histopathology showed a marked reduction in intestinal lesions, and in ammation between the placebo-treated group and the group who received tebipenem pivoxil; no viable Shigella could be observed in section from the treated large intestine ( Fig. 5C and 5D).

Discussion
AMR poses one of the greatest challenges in global health; therefore, new treatments and approaches are urgently needed to treat infections caused by antimicrobial resistant organisms. Here, we exploited a drug screening strategy to identify compounds that may have utility against highly antimicrobial resistant enteric pathogens, speci cally MDR/XDR Shigella. Ultimately, we identi ed and progressed tebipenem pivoxil (a licenced oral carbapenem) to pre-clinical testing for Shigella infections. Tebipenem pivoxil has many suitable characteristics for the treatment of infections caused by MDR and XDR Gram-negative bacteria; it is broad spectrum, it is rapidly bactericidal, is already used in paediatric clinical practice, and it can be administered orally, making it highly suited for the treatment of gastrointestinal infections in the community. Despite these positives, there are potential issues surrounding the widespread use of this drug, most notably the potential to generate resistance against carbapenems, a very important last resort class of antibacterial compounds.
We initiated the project by screening a large compound library against a single Shigella isolate grown extracellularly. Notably, none of the chemicals meeting the desired criteria (inhibition of bacterial growth with a physicochemical and safety pro le suitable for downstream developability) represented novel chemical classes. All hits were already known to possess antibacterial properties and belonged to either the GSK antibacterial subset or were commercially available. The fact that no new modes of action were identi ed after screening 1.7 million compounds via extracellular in vitro screening is concerning, suggesting we are close to exhausting the chemical space for antibacterial agents. However, innovative phenotypic assays, which interrogate other stages of the bacterial infection, such as intracellular activity or invasion blockade, may offer new opportunities for developing narrower spectrum drugs. Using the selected approach, we opted to focus on a de ned set of compounds with known antibacterial activity against E. coli, allowing us to identify two main chemical classes that additionally exhibited activity against clinical Shigella isolates. Speci cally, we found that carbapenems had high potency against both laboratory Shigella strains and clinical isolates of MDR Shigella from highly endemic LMICs. Consequently, tebipenem was selected as a lead candidate for further pre-clinical characterization due to the great potency and the availability of a prodrug which allows oral administration.
Tebipenem had enhanced activity (with respect to cipro oxacin) against a panel of 81 Shigella clinical isolates from Southeast Asia, an observation in line with the high prevalence of cipro oxacin resistance in many Shigella sonnei and a high proportion of Gram-negative bacterial pathogens across Southeast Asia. Tebipenem additionally exhibited high level of potency against a broad panel of > 300 clinically relevant MDR enteric pathogens. These data outline the potential use of this antimicrobial for the treatment of enteric infections of suspected bacterial origin without the necessity of an accompanying diagnostic test, which is a common practice in many LMICs. It is noteworthy that carbapenem resistance is emerging in some Gram-negative organisms in Asia and has spread to other regions. This phenomenon is partly associated with the spread of the NDM-1 gene, which encodes a betametalloprotease. Carbapenem resistance has not, as yet, been reported in any Shigella species and is not currently common in other Gram-negative bacteria causing diarrhoea. However, given the propensity of Shigella to develop resistance to commonly used antimicrobials through the acquisition of AMR plasmids and individual gene cassettes, we additionally need to consider the requirement of a companion diagnostic to avoid treatment failures or overemployment of the drug, which my trigger resistance.
We considered that the emergence of resistance against tebipenem (and other carbapenems) may be mitigated through combining tebipenem with drugs with alternative other modes of action, including existing standard of care antimicrobials. Azithromycin, which reduced the MIC for Shigella isolates for tebipenem when tested in combination, offers an encouraging primary combination option, given the common use of azithromycin for the treatment of severe diarrhoea. Additionally, a second potential combination showing in vitro synergy was identi ed with an LpxC inhibitor, which again showed good activity when tested against the majority of enteric bacteria studied. Again, this observation is encouraging and adds further support to LpxC development programs and may provide a solution for limiting spontaneous resistance against this newer class of antimicrobials. Our data highlight the utility of tebipenem for treating a spectrum of enteric pathogens and the potential bene t of co-administration with recommended antimicrobials suggest that combination therapy with tebipenem is clearly worth exploring through clinical trials and/or through controlled human infection studies with Shigella. The existing safety pro le of these drugs opens the possibility of direct progression to a clinical proof of concept study as soon as possible.
Encouragingly, tebipenem pivoxil is approved and marketed in Japan as Orapenem (Meiji Seika) for the treatment of adult and paediatric respiratory tract infections 18 . In addition, Spero Therapeutics are working on a novel formulation with extended half-life for administration in adults which is under clinical evaluation for recurrent urinary tract infections 19,20 . The MICs of the panel of enteric bacteria against tebipenem were substantially below the clinical breakpoints reported for e cacious treatment of respiratory tract infections. Shigella infections have a pathophysiology which is distinct to that of respiratory tract infections, but our MALDI-TOF data suggests that tebipenem reaches a high concentration in the enterocytes lining the gastrointestinal surface. We hypothesise that the currently approved dose of Orapenem will reach a higher local concentration in the Shigella infection-relevant cells in the gut than in the respiratory tract. Therefore, the drug distribution behaviour of Orapenem is especially convenient for the treatment of enteric infections, as the active component is speci cally released at the site of infection. The release of free tebipenem intracellularly would clear intracellular bacteria (vital for Shigella and Salmonella) and active tebipenem in the intestinal lumen would reduce extracellular bacteria and limit cellular invasion. The animal model data implies this process is occurring, as the drug was highly potent after Shigella challenge and able to clear the gut of infecting organisms with 24 hours of administration in gnotobiotic piglets.
In conclusion, we performed a screening study to identify novel compounds with antibacterial activity against MDR/XDR Shigella. We found that tebipenem had excellent in vitro pro le against MDR/XDR Shigella and other Gram-negative bacteria; the oral bioavailability of tebipenem pivoxil makes it ideal for the treatment of gastrointestinal infections. Experimentation in animal models of shigellosis found the drug to be highly e cacious with activity comparable to cipro oxacin, when animals were challenged with cipro oxacin susceptible Shigella isolates. We are in desperate need of alternative antimicrobials for infections caused by MDR/XDR bacteria and our data outline the oral carbapenem tebipenem as an ideal repurposing candidate. The testing and controlled introduction of tebipenem pivoxil combinations for MDR Shigella infections may slow the rate of spread of organisms that exhibit resistance to current standard of care antimicrobials or provide an extended window of opportunity for the further development and introduction of Shigella vaccines.

Reagents and compounds
All commercially available chemicals were purchased from either Sigma Aldrich, UK, AK Scienti c, or Fischer Chemicals. The GSK 1.7 million compound screening set was sub-selected from the 8 million compound GSK collection using the principles described by Harper and co-workers 21 . Background of the structural lters or property constraints developed to create a "clean" set of compounds with regard to compound quality (purity/surety), no undesirable/reactive chemotypes and in the appropriate property space. This process has been described previously by Chakravorty and co-workers 22 . The set of 4,000 compounds with known antibacterial pedigree from the GSK antibacterial programs was selected from GSK compound collection using an MIC < 1µM against E. coli as cut-off value for selection.

Organisms and growth conditions
For the initial screening we used Shigella exneri 2457T (ATCC 700930). Organisms were grown at 37°C in Mueller Hinton medium overnight (Sigma Aldrich, UK). Hits from the initial screening were then pro led  24 . was used to determine the IC50 and the IC90 of potential inhibitors of Shigella exneri ATCC 2457T. The assay measures the effect of compounds on the bacteria using resazurin reduction (Thermo Fisher Scienti c, UK) as a surrogate of bacterial viability. Prior to day of experimentation, an overnight bacterial culture was inoculated 1/100 in Mueller Hinton medium and incubated for 24 hours; the bacterial culture was inoculated 1/5 into fresh Mueller Hinton medium.
Bacterial growth was measured at 600nm after three hours growth (expected OD 0.4-0.6) and referenced to the corresponding CFU/mL. The bacterial inoculum was diluted 1x10 6 CFU/mL to perform the assay.
Subsequently, 1,536-well polystyrene assay plates (HiBase, µClear, Greiner Bio-One) were prepared by dispensing 50 µl of compound from master plates at 1 mM in each well using a LabCyte Echo system.
The nal assay volume was 5 µl and nal compound concentration was 10 µM. The 11th and 12th columns were positive growth controls containing 50 nL of DMSO. The 34th and 35th columns contained 50 nL of Moxi oxacin at 2 mM as negative growth control. 5 µl of bacterial inoculum was dispensed into compound pre-dispensed plates using a Multidrop Combi dispenser (Thermo Scienti c). Plates were agitated for 10 seconds to ensure mixing and incubated at 37°C overnight (16-17 hours). Following overnight culture, 2 µl resazurin solution (Resazurin tablets for milk testing, BDH) was added to each well and incubated in at ambient temperature in the dark for 3 hours. Fluorescence intensity was measured using an Envision plate reader (PerkinElmer). For IC50 and IC90 determination, data were normalized by using DMSO only as positive control (100% bacterial growth) and 20 µM moxi oxacin as negative control (0% bacterial growth). Data were analysed using Base software and statistical cut-off analysis was performed for hit selection following GSK standard protocol. The percentage growth inhibition was calculated per well according to the plate controls. These values were aggregated to calculate the robust mean and standard deviation (SD) of the distribution. The robust mean was calculated starting from the median and the MAD of the distribution with an iterative process that aimed to remove outliers. Once the mean and SD were known, all compounds above mean + 3SD were selected as positive hits. After compound analysis and removal of structures with alerts, prioritized compounds were progressed to dose-response analysis. To perform the dose-response curves, compounds were diluted in DMSO at 1:3 serial dilutions in concentrations from 100 to 0.00169 µM.

Growth inhibition assays with clinical isolates
Compound screening and dose response assays have been described previously 25,26 . All chemical testing was performed at an initial concentration of 10 µM. The bacteria selected for screening were cultured on the Luria-Bertani agar overnight and their colonies would be separately mixed in sterile PBS to Before initiating the drug combination assays, the MIC values for both compounds were determined against ATCC organisms (as previously described) to design the checkerboard layout. The standard checkerboard titration method based on isobolograms was used to study the in vitro interaction between tebipenem and other compounds (WHO recommended treatments and an LpxC inhibitor). 20uM moxi oxacin was used as negative control (100% bacterial growth inhibition) in one column and DMSO as positive control (100% bacterial growth). 25µl of 1x10 6 CFU/mL bacterial culture was added to each well of 384-well plates (µClear; Greiner Bio-One). Plates were incubated overnight at 37ºC and 10µl of resazurin were added. Plates were incubated at ambient temperature for three hours. Bacterial growth inhibition was measured by uorescence in a Envision plate reader (PerkinElmer). Data were managed in Microsoft Excel and analysed in Gra t 7 (Erithacus) data analysis software. The frequency of spontaneous resistant mutants (FoR) was determined for tebipenem. Moxi oxacin, cipro oxacin, and ampicillin were used as internal controls; bacterial inoculum without drug was used a as positive control. The bacteria inoculum (1x10 5 CFU/mL) for the assay was prepared by diluting a log phase incubated at 37ºC for 3-4 hours. Mueller Hinton agar was prepared containing 4xMIC and 10xMIC and inoculated with different concentrations of bacteria (1x10 9 , 1x10 8 , and 1x10 7 CFU/mL). Plates were incubated overnight at 37ºC. The frequency of spontaneous resistant mutants was calculated by using the formula resistant CFUs/inoculum CFUs.

Extrapolation of e cacy doses from Orapenem approved treatment in paediatric population
Pharmacokinetic parameters of Orapenem in paediatric population at the approved doses (4-6 mg/kg, bid) were obtained from Sato et al. and pharmacokinetic parameters in rats were generated in house 17 .
Information regarding Orapenem oral bioavailability was extracted from Kijima et al. 17 . For the calculation of tebipenem pharmacokinetic behaviour in different animal species (i.e. mouse and piglet) conserved intrinsic clearance (CL int ) and distribution volume (Vss) across the different animal species have been assumed. A mono-compartmental PK model has been used for PK simulation using matlab simbiology (2019b) software. The equivalent doses in the different animal species were calculated using tebipenem free drug t/MIC as comparable parameter.

In vivo pharmacokinetic analysis
All mice experiments were performed in female C57 mice obtained from Charles River Laboratories, (Wilmington, MA) and housed in cages in groups of three animals with water and food ad libitum.
Animals were acclimated for seven days prior to infection. At the beginning of the experiment the mice weight was 20-30 mg. Tebipenem was dissolved in 20% Encapsine (Sigma Aldrich), 5% DMSO (Sigma Aldrich) in saline solution (Sigma Aldrich) at 0.5 mg/ml for intravenous administration and in 1% Methylcellulose (Sigma Aldrich) in water at 5 mg/ml for oral administration. Tebipenem pivoxil was dissolved in 1% Methylcellulose (Sigma Aldrich) in water at 5mg/ml for oral administration.
Drug distribution studies (intestine homogenate and MALDI analysis) To determine the distribution of tebipenem pivoxil and free base in the different parts of the intestine, tebipenem pivoxil was orally administrated at 5 mg/ml to C57 mice (n = 12, target dose 50 mg/kg) in four groups of three animals. Animals were euthanized at 0. 25 Corp.) at 40ºC with Acetonitrile (Sigma Aldrich) and 0.1% formic acid as eluents.
In vivo mouse e cacy studies The S. exneri lux1 strain was used to infect mice for the mouse e cacy experiments. S. exneri lux1 expresses a bacterial luciferase reporter as previously described and was created from S. exneri M90T Sm (ATCC BAA-2402) purchased from the American Type Culture Collection (ATCC, Manassas, VA). Broth cultures of S. exneri strain lux1 were grown at 37°C in Trypticase soy broth (TSB) with agitation at 220 rpm. For solid media cultures of S. exneri strain lux1, an agar form of Trypticase soy broth with the addition of 0.01% Congo Red. 30 µg/mL streptomycin was used for selection of the lux1 strain and 0.1% sodium deoxycholate was added to the broth to promote expression of proteins involved in Shigella cellular invasion 31 .
The in vivo e cacy of anti-Shigella drug candidates was evaluated using a previously published mouse model of shigellosis 31 . In brief, an overnight S. exneri lux1 broth culture was diluted 1:100 and grown to log-phase (3-4 hours). Log-phase bacteria (OD 600 ~ 0.4) were centrifuged, and the resulting pellet was washed in DPBS without calcium and magnesium (Thermo Fisher Scienti c). Bacteria were centrifuged again, and pellet was resuspended in DPBS to the target concentration of 5 x 10 7 S. exneri/200 µL (10 mL/kg). Each dose of bacteria was con rmed with luminescence using an EnVision Plate Reader (Perkin Elmer) for S. exneri lux1.
Female B6 mice (000664 Black 6, Jackson Laboratories) aged 9-10 weeks were weighed and infected IP with 5 x 10 7 CFU/mouse (total n = 7-10 per group     Notably, the drug concentration in all preparations remained above the estimated MIC of S. exneri 2457T (0.39µM). D) The e cacy of 50 mg/kg oral tebipenem pivoxil (dark grey) and 39 mg/kg oral tebipenem (mid grey) in reducing the CFU/g of faeces, 24 hours after infection with a 1 x 107 intraperitoneal dose of S. exneri 2457T compared to no drug (black) and 40 mg/kg cipro oxacin.