The ability of microbes to resist the effect of antimicrobials meant to inhibit their growth or kill them currently threatens humans with a dark age where common minor infections could be potentially deadly. The fallout of bacterial antimicrobial resistance (AMR) impacts heavily on global mortality, morbidity and economy, particularly in low- and middle-income countries (LMICs) where uncontrolled access to potentially life-saving drugs is expanding1. In 2019, bacterial antimicrobial resistance was associated with an estimated 4.95 million deaths globally1. Annual treatment costs arising from antimicrobial resistance are estimated at US$4.6 billion in the United States alone2. As a recent study1 on the global burden of bacterial antimicrobial resistance in 2019 showed, the AMR burden in that year was highest in sub-Saharan Africa and higher in low- and middle-income countries (LMICs) than in East Asia, Australasia and Western Europe, with species of One Health importance in their transmission accounting for the highest attributable mortality. Rising selective pressure from judicious and injudicious use, as well as from disposal of antimicrobial waste, underpins much of the AMR crisis but is only one evolutionary driver. Bacteria carrying resistance genes on mobile genetic elements, which are capable of transposition within and between bacterial species, can be selected by other forces and are difficult to contain when their transmission is assured. The contribution of transmission to the evolution of resistance has received scant attention as the focus is often on the very visible impact of AMR in human clinical settings. However, outside health facilities, resistant organisms rarely move directly from one sick person to another. Instead, they circulate undetected among healthy and sick humans, domesticated and wild animal populations, and the environment3,4,5,6. In all these niches, some of which are hotspots of varied selective pressures due to the accumulation of antimicrobial waste, new resistant clones evolve and resistance genes and elements are disseminated among co-habiting species7. In this review, we examine AMR across the human-animal-environment continuum in LMIC settings. We overview the scope of the problem, highlight instances where resistance transmission has and has not been evidenced, and make a case for interrogating transmission in LMICs using genomic approaches. We also outline interventions that might be especially valuable for containing AMR spread. ## Transmission of resistant bacteria among humans is rife and worsening Low-income settings with poor access to antimicrobials and dense transmission networks have recorded AMR prevalence rates greater than 50%8. Rapid population growth and urbanization in LMICs can promote resistance by increasing bacterial transmission and access to antimicrobials9. In these informal and unplanned urban settings, subsistence strategies that maximize resource extraction, as well as shortfalls in regulatory oversight for water and sanitation10 converge to amplify the transmission risk among humans, animals and their environment (Fig. 1). Eliminating resistance reservoirs and breaking transmission chains will stem the spread of pathogens, thereby reducing the need for antimicrobials11. Resistant bacteria and mobile elements selected in people, livestock, pets or pests are transferred among individuals or voided into sewage systems, underground soak-aways, pit latrines, drains or open spaces from where they can be acquired by wildlife or seep into irrigation, household and even drinking water, thereby creating hotspots for the emergence of antimicrobial resistance. Few studies examine these One Health niches or their connections, but worryingly high resistance levels have been recorded in all of them12,13,14. For example, only three Medline-indexed studies – all sampling ≤100 latrines – published before December 2021 reported antimicrobial resistance in pit latrines, a sanitation system predominant in many LMIC settings. In all three studies, >75% of the Escherichia coli isolates were resistant to at least one antimicrobial, while >45% were resistant to three or more of the tested antimicrobials; the two studies that searched for isolates carrying extended-spectrum beta-lactamase genes found them15,16,17. ## AMR in LMIC food-animal-related bacteria While there are several published LMIC meta-analyses on resistant bacteria from humans18,19,20, there is less synthesized information from non-human niches. Perhaps the most comprehensive systematic review is the study of Van Boeckel et al.21 and its accompanying database (https://resistancebank.org/), which focuses on food animals. Food animals are hosts to resistant bacteria from other animals, their human handlers and the environment, and can potentially disseminate these bacteria to human consumers22,23. Van Boeckel and colleagues21 used point surveillance data from 901 LMIC-based studies from 2000 to 2018 to report global trends of AMR in animals in LMICs for common indicator pathogens such as E. coli, Campylobacter spp., Salmonella spp. and S. aureus. These species are important not only in AMR in animal production but also in human infections24. The study revealed a 173% increase in the proportion of antimicrobials with resistance rates greater than 50% (P50) in chickens (0.15 to 0.41), including a 161.5% and 91.7% increase in P50 in pigs (0.13 to 0.34) and cattle (0.12 and 0.23). Hotspots of AMR (P50 > 0.4) were reported in parts of Asia (India, China, Pakistan, Iran, Turkey, Vietnam), South America (Brazil, Mexico) and Africa (Egypt, South Africa), whereas AMR emergence was reported in Kenya, Morocco, Uruguay, southern Brazil, central India and southern China. Commonly used antimicrobials in animal rearing, such as penicillins, sulfonamides and tetracyclines, recorded the highest resistance rates. For instance, the study reported quinolone and gentamicin resistance to be between 20–60% among E. coli, and 5–38% among Salmonella spp. The highest resistance rates to antimicrobials in Campylobacter were reported for tetracyclines and quinolones (both 60%). In S. aureus, the highest resistance rates were associated with penicillins (40–80%), whereas it was between 20–60% for tetracyclines, oxacillins and erythromycin antimicrobials. These findings emphasize the rapid increase in antimicrobial resistance in the food-animal sector in LMICs and highlight it as a sector deserving closer attention due to its high importance and sensitive position as a critical crossover point for resistance transmission. ## Animal–human routes of AMR spread An estimated 60% of all human pathogens and 75% of emerging diseases affecting humans are zoonotic (https://iris.wpro.who.int/bitstream/handle/10665.1/13654/9789290618171-eng.pdf). Human–animal interactions are increasing, raising the risk for zoonotic infections25,26 and the subsequent emergence of resistant pathogens. In many LMICs, antimicrobials, which are poorly stewarded and easily obtained over the counter, are introduced as food additives for disease prophylaxis, metaphylaxis and growth promotion in aquaculture, poultry and livestock27,28,29. The quantity of antimicrobials used in animal production is predicted to increase by 11.5% between 2017 and 203030. However, there are few or no reports on the sale of veterinary antimicrobials in African countries30 despite indiscriminate and high usage of antimicrobials in animals in LMICs31. Since initial discovery in animal and human isolates in southern China in 2015, and coincident with intense colistin use in agriculture, ten mobile colistin resistance (mcr) gene variants have been described in over 40 countries across five continents32,33,34. Also, the recently reported existence of the mobile tet(X3) and tet(X4) genes among isolates from food animals and humans in China and elsewhere further limits the efficacy of last-resort antibiotics and underscores the imminent global health threat35,36. Existing reports predominantly highlight the dissemination of resistance from animals to humans, but humans also represent reservoirs for the transmission of AMR to animals26. For instance, Lowder et al.37 described a single human-to-avian jump of S. aureus CC5, which acquired mutations adaptive to the avian niche and was subsequently disseminated around the world in poultry. Following the first report of the human mcr-1 gene in Algeria, the gene was soon detected in Barbary macaques38. Similarly, the human epidemic ST22 methicillin-resistant S. aureus SCCmec IV clone is concurrently occurring in swine and primates39. The exposure of animals to human waste and the poor sanitation and hygiene associated with food-animal handlers in farms, retail shops and slaughterhouses are important contributors to the spread of antibiotic-resistant pathogens among animals40. In animal production operations in LMICs, sanitation, antimicrobial usage, animal overcrowding, lack of human protective gear and other farm management practices are important risk factors29,41,42. Moser et al.43 reported that 429 (69.3%) birds from small-scale farms in Ecuador yielded multidrug-resistant isolates, compared with 58 (15.3%) birds in non-production villages. In that study, the increased prevalence of phenotypic antibiotic resistance and mobile genetic element markers in small-scale and household poultry in Ecuador were associated with confined spacing of birds and lacing of animal feed with antibiotics. Cleaning and wiping down farm equipment and sanitary disposal of dead birds were protective against Salmonella in farms, whereas a high bird density increased the risk of Salmonella detection in two Colombian state departments for poultry farming42. Disposal of human and animal waste close to human residences and farms is common and, indeed, inevitable in resource-limited settings with inadequate or unsafe sanitation41. This practice allows bacteria, including resistant ones, to filter into groundwater or onto produce. It also attracts insects such as houseflies and cockroaches, which are efficient carriers of multidrug-resistant bacteria44,45. In addition, without intervention, antimicrobials used in agricultural practices invariably run off into other soil and water ecosystems, eventually reaching human populations and scavenging wildlife46,47,48. Thus, even among wildlife, AMR is usually reflective of human anthropogenic activities49,50. Close interaction among the animal-human-environment tripod facilitates varied heterogeneous routes of AMR spread and can foster the amplification of novel AMR hotspots. ## Genomic approaches for mapping transmission across the One Health continuum There are multiple retrospective records of cross-species or cross-niche bacterial lineages37,51,52. However, recent studies that have set out to prospectively compare lineages from animals with those from humans have found surprisingly few common clones, engendering the suspicion that concerns about transmission of resistance among humans, animals and the environment may be overblown3,51,53,54,55. Why is this? Successful bacterial transfer across species could be uncommon overall but become significant and visible in the face of selective advantage or when there are numerous crossover opportunities52. Alternatively, carriage by some species disseminating resistant organisms may be common but transient, but not without consequence, and therefore difficult to detect in the mainly cross-sectional surveys unless further adaptive evolution follows56. Many studies focus on hypotheses defining the movement of resistant bacteria towards humans and assume short paths of transfer, and therefore bias study design accordingly. However, more open multidirectional or circular models of transmission involving one or more intermediate hosts may be in play. Moreover, the transfer of bacterial lineages may be less significant than the movement of mobile elements and genes, which is harder to detect and confirm57. The probability of any of these scenarios occurring in LMIC settings is probably higher than elsewhere, but fewer investigations have been conducted outside Europe and North America, and LMIC studies need to use methods more likely to detect transmission when it is uncommon. Methods based on whole-genome sequencing provide the necessary granularity to rule in or out the inter-host transmission or other niche-entry events52,53. With falling sequencing costs and improved methods for designing and interpreting metagenomic studies, approaches that could identify crossover points and determine their frequency may now be within reach. However, One Health genomic studies are few and far between in LMICs; the often large and diverse specimen collections needed for this type of study preclude the helicopter and postal research approaches that accounted for most of the LMIC-derived bacterial genomes at the turn of the century58. Today, the higher cost of sequencing overall in these resource-limited settings continues to present challenges to implementing a study sensitive enough to determine whether, when and how resistant organisms, elements or genes cross from one niche to another. However, as summarized in Box 1, when transmission is intensively occurring, the potential for capturing informative snapshots is high. ## LMIC-specific features of the livestock industry exacerbating resistance transmission In LMICs, the variety of livestock production systems generate much greater contact between humans, animals and the natural environment than in high-income countries (HICs). Agricultural employment represents less than 5% of total employment in HICs but reached 23% in the Philippines, 30% in Ecuador, 54% in Kenya, and up to 86% in Burundi in 2019 (https://www.theglobaleconomy.com/rankings/employment_in_agriculture/). Urban subsistence farming is rare in HICs but common in LMIC settings, including informal settlements. Biosecurity measures that mitigate sanitary risks in high-income settings require capital expenditure and higher technical skills, and are often not implemented in LMICs, particularly in smallholder/subsistence farming systems, which represent up to 80% of the production in Asia and sub-Saharan Africa (http://www.fao.org/fileadmin/templates/nr/sustainability_pathways/docs/Factsheet_SMALLHOLDERS.pdf). Larger contacts between livestock and humans in LMICs than in HICs can be evidenced along the entire food system value chain, as illustrated by the central place that live and wet markets play in the commercialization of livestock and the access to food in Asia and Africa59 (Fig. 1). Informal livestock trade in LMICs exacerbates the risk of resistance transmission. Informal trade cannot, by definition, be easily and precisely assessed; however, some evidence-based evaluations suggested that the proportion of informal livestock trade in LMICs is significant. For example, informal trade in live cattle from India to Bangladesh was estimated at US$620 to 660 million per year in 2018, and the year before that, 70% of frozen chicken trade in Nigeria was assessed as informal60. Similarly, an evaluation by the Colombian Federation of Cattle producers (FEDEGAN) found that 62% of the national cattle production in 2018 was commercialized through informal channels (https://www.fedegan.org.co/noticias/ganaderia-colombiana-hoja-de-ruta-2018-2022). On one hand, informal livestock trade in LMICs offers small-to-medium benefits such as improved food security, reduced price instability, employment opportunities and avoidance of large, multiregional foodborne outbreaks. On the other hand, it undermines the possibility of sanitary inspections and the overall effective governance of value chains11.

The third critical LMIC pattern that exacerbates the risks of resistance transmission from livestock is limited access to veterinary services. In a recent survey in five African countries (Ghana, Kenya, Tanzania, Zambia, Zimbabwe)61, the veterinarian-to-livestock ratio was shown to be on average 20 times lower than in HICs such as Denmark, France, Spain and the United States. The World Organization for Animal Health, formerly the Office International des Epizooties (OIE), has launched a global initiative to assess and improve the level of performance of national veterinary services (https://www.oie.int/app/uploads/2021/03/2019-pvs-tool-final.pdf). Poor animal health services stemming from inadequate access to veterinary resources are precursors to the misuse of antimicrobials for prophylaxis or as feed additives, including the poor institution of proper farm biosecurity and other animal health infrastructures in animal rearing62 (https://www.oie.int/app/uploads/2021/03/en-oie-amrstrategy.pdf). On the other hand, even when veterinary services are available, the high reliance on cultural and historical perspectives in animal rearing practices by animal handlers, as with the Maasai pastoralists in Tanzania, may mean that veterinary professionals are not consulted or are consulted only as a last resort63.

The importance of country-owned and developed national action plans to combat AMR using appropriately tailored interventions cannot be overstated. Indeed, the typical practice of modelling LMIC action plans on HIC templates may account for the understatement, or in many cases complete absence of key transmission-blocking interventions, such as water sanitation and hygiene (WASH)64. In our opinion, many other National Action Plan (NAP) pillars under-exploit opportunities for impact because they prioritize top-down interventions aimed at combatting resistance in high-income settings that may not always be suitable for or effective in other parts of the world (Table 1). In HICs, the top-down implementation of One Health AMR interventions has the benefits of speed, leverage and scale within specific socio-economic and regulatory environments65. In the complex systems of livestock sectors in LMICs, a top-down approach is less likely to succeed. The multitude of contacts between animals and humans, the large informal livestock trade and the deficiency of veterinary services require integrated top-down and bottom-up national action planning and implementation of interventions that may be less prone to failure where regulation is not strong61.

AMR is a complex problem, and its development has the characteristics of nonlinearity, emergence, positive and negative feedback, and adaptation that are particularly prominent in LMICs. The stakes for resistance are higher and the value chain is uncertain in LMICs. These, combined with the high uncertainty from limited surveillance and the urgent need to avoid catastrophe from resistance, necessitate a post-normal science approach66 for addressing resistance in a One Health context.

## Technological tools with promise for addressing resistance transmission

There is considerable academic interest but a lack of general appreciation and clear translational paths for promising technologies for preventing resistance transmission. Given that the development of new antibiotics has stalled globally in the last few decades67 and new drugs are typically only available in LMIC settings after a lag, anti-transmission tools and approaches deserve serious consideration. Technological tools with the potential to contain the spread of resistance, such as bottom-up approaches to directly prevent transmission, may have greater potential in LMIC settings.

Infection control and prevention measures present arguably the most feasible and cost-effective approaches to limiting the evolution and spread of antibiotic-resistant microbes in any setting68. With reduced prevalence of infections, antimicrobial use can be limited, consequently reducing the selection pressure and evolution of resistant organisms. Vaccines do exactly this; they are typically administered prophylactically to prevent microbial infections in humans and animals. Although vaccines also exert selective pressure on microorganisms, there is a reduced risk of evolution of resistance to vaccines compared with antibiotics69. Additionally, when vaccine resistance does occur, public health benefits such as herd immunity and reduced prevalence of infectious organisms are often sustained. In addition to lowering infectious burden and the need for antimicrobials, vaccines can also directly address resistance and, in particular, the contribution that transmission makes to resistance spread, by preferentially targeting resistant lineages of bacteria (as with polyvalent pneumococcal vaccines, which are the textbook example of how vaccines can temper resistance) or by specifically targeting antigens associated with AMR70.

Vaccine development has seen a recent resurgence, with more and newer technologies applied71. A bioconjugate vaccine (ExPEC4V) against extraintestinal pathogenic E. coli (ExPEC), which are among the leading causes of life-threatening invasive drug-resistant infections in humans72, is currently under development by Janssen Pharmaceuticals. This vaccine has been shown in clinical trials to be well-tolerated, safe and capable of eliciting robust immunological responses against all tested ExPEC serotypes, many of which are antimicrobial resistant73,74. ExPEC are disseminated faeco-orally and are also reservoirs for resistance genes that can be transmitted to other pathogens in the guts of humans and animals, as well as in the environment. Therefore, ExPEC vaccines in development may have beneficial effects on the prevalence of resistance, although yet to be demonstrated, and existing vaccines targeting organisms that colonize animate and inanimate niches could offer similar benefits (Box 2).

Antimicrobial misuse is driven by ineffective diagnostic stewardship75. Lack of access to essential diagnostics, shortage of skilled laboratory personnel, and high cost and long turn-around times exacerbate diagnostic insufficiency in LMICs76,77. The common practice for the management of infections in low-resource settings involves symptom-based and endemicity-guided diagnosis, and the subsequent prescription of broad-spectrum antimicrobials for unconfirmed infections76,78. A recent study in a tertiary health facility in Nigeria found that under 25% of the patients who received prescribed antibiotics had confirmed bacterial infections79. The proportion of patients for whom antimicrobial susceptibility testing (AST) is conducted is often lower. Culture-based AST, the cornerstone of bacterial infection and resistance detection, works well in many settings but its turn-around time is long and some LMIC institutions caring for infected patients lack the requisite personnel and infrastructure. However, culture-based AST is critical to identifying novel mechanisms of resistance, which are periodically reported from LMICs with better surveillance. Cases in point are the discovery of mobile elements conferring resistance against reserve antimicrobials, noteworthily including mcr-1 and tetX alleles in China32,33,35, and NDM-1 in India80, all of which have subsequently been shown to be globally disseminated. The development of rapid point-of-care pathogen detection and AST platforms to increase access and affordability, lower the skill barrier for diagnostic stewardship and decrease turn-around times would greatly limit the empirical administration of broad-spectrum antibiotics and also reduce transmission of resistant bacteria from undiagnosed patients81. New clustered regularly interspaced short palindromic repeats (CRISPR)-Cas-based diagnostic technology appears to be promising for the development of rapid point-of-care pathogen detection82.

Non-antibacterial approaches to eliminating pathogenic and non-pathogenic resistant bacteria are needed and could be used in humans, animals and even the environment. Therapeutic monoclonal antibodies have been explored, albeit with modest results83. Promising lower-cost therapeutic approaches include the use of bacteriophages, probiotics, synthetic peptides and nanostructured polymeric materials. Metal nanoparticles are being explored for use as self-sterilizing coatings for medical devices, as well as for controlled antibiotic and antibody release for treatment and prevention of biofilm- and medical device-associated infections84. Phage therapy has the advantages of being highly specific to the target bacteria, non-toxic to the host and can be administered in small doses. This may be desirable in LMICs where the use of unprescribed, non-targeted antimicrobials is common, thus reducing transmission and evolution of resistance. Although phage therapies are also amenable to bacterial phage resistance71,85 and require diagnostic proficiency to sub-type level, the approach still offers the potential for propagative effects outside the sick host that could help stem transmission, something that probiotics might also offer.

Alternative growth promoters and feed additives such as probiotics, prebiotics and synbiotics86, lactic acid bacteria and bacteriocins87, organic acids88, vitamins and minerals89 have been recommended to greatly improve food-animal health and production. For example, probiotics and prebiotics stimulate commensal flora growth and improve animal health and immune status86,90. Probiotics aid the neutralization of enterotoxins and necrotizing enterocolitis86. Varying levels of success have been reported for these feed additives and more research is required to support their use as alternatives to antibiotics. These could augment sanitary interventions in animal production facilities in LMICs to reduce the underappreciated impact of animal faeces as an AMR vehicle and reduce threats to human health not covered by conventional WASH programmes91. The potential that probiotics could offer for human health is under investigation, but their impact on resistance does not appear to have been intensively studied. It is also worth hypothesizing, albeit as yet untested, that probiotic or other non-antimicrobial tools could offer an alternative to mass drug administration for child survival92, indirectly ameliorating the potential abuse of antimicrobials.

Antimicrobials proffer strong selective pressure by killing the bacteria they attack or at least inhibiting their growth. Compounds that inhibit virulence or colonization factors would presumably have a lower selection for resistance and a few are in development93. As with many anti-transmission technologies, questions remain about how they could be used therapeutically. Potentially useful tools against resistant bacteria might not be able to scale a non-inferiority clinical trial against an antimicrobial, posing a regulator barrier to their use67. It has been proposed that anti-virulence agents be trialled in combination with existing antibiotics. However, anti-colonization factors, in particular, could have the added benefit of dislodging resistance reservoirs, and the design of studies that include resistance reduction endpoints is worthy of consideration. Also, the use of anti-transmission approaches outside clinical medicine will not require clinical trials. Other technological tools that could impact the transmission of resistance and therefore have potential applications in the settings we have described include agents that inhibit horizontal gene transfer. These include but are not limited to anti-conjugation factors. CRISPR technology has been used to target resistance genes in enterococci94. Similarly, there is experimental work in progress aimed at blocking the transmission of conjugative elements95. The translational pathway for such agents is even more unclear. As the specific application of tools with the potential to stem the transmission of resistance in HIC clinical medicine is somewhat elusive, development and deployment for proof of concept might be applicable in a One Health transmission context, particularly in a setting where inter-individual contact is high and drug use and transmission controls are low.

## Conclusion

Resistant bacterial variants evolve continuously but are selected when antimicrobials are applied and can be transmitted among humans, animals and their environments. Surveillance of resistance is improving and increasingly incorporating genomic approaches, but many intervention points for stemming the flow of resistant organisms and genes remain under-exploited or unknown. In LMIC settings where humans and animals converge at high density with insufficient sanitation and biosecurity infrastructure, the risks of selection of resistant bacterial variants and their transmission among humans, animals and the environment are high, and patterns of transmission are likely to be different. The relative importance of the numerous risk factors needs to be further investigated in specific LMIC settings to develop the most appropriate responses to the development of antibiotic resistance.

Significant modifications of One Health approaches optimized for HICs should combine top-down approaches with appropriate bottom-up actions for LMIC settings to compensate for institutional weaknesses. There may also be a role for technical disruption of transmission. The promise of recent innovations and technologies that could accomplish this may vary with landscape and thus require independent evaluations by LMIC scientists and actors who may be better served by engaging directly in early-stage innovation rather than merely adapting their uses in HIC to LMIC conditions.