Map 1: What research evidence is there about ARB exposure and transmission to humans from the environment?
Description of review process
For Map 1, the search yielded 27,186 hits which equated to 11,016 unique titles and abstracts for screening. 380 of these were considered eligible for full text screening of which 350 full texts were able to be retrieved. 40 were eligible for inclusion in the map (Fig. 2).
A total of 238 articles were excluded at full text (Additional file 6) and a further 72 articles highlighted in Additional file 2 for their relevance to the search question (although they do not fit the inclusion criteria—see details below). The full Map 1 database for all included articles, with all metadata, can be found in Additional file 7.
The geographical location of included studies can be viewed in Fig. 3. To view an interactive version of this map, please download the html file provided as an additional file (“Map 1 geographical interactive—Additional file 8”).
Articles were not evenly distributed around the globe, with clusters found in Europe (n = 15), South East Asia (n = 8) and North America (n = 9). There were also a small number of studies undertaken in Africa. There were no studies investigating direct transmission of AMR from the environment to humans in South America and Oceania. In general, research study sites tended to be located where convenient and close to researchers’ location (based on author affiliation information). Therefore, the uneven geographical distribution of study sites is likely caused by where researchers undertaking this type of research are based.
A number of studies specified sampling in multiple countries in the United Kingdom. To avoid confusion these studies have been placed at the coordinates (latitude and longitude): 50.15852, − 1.258472. Studies with no information on where the sample sites were, were placed at the coordinates 38.49942, − 38.8872.
Publications over time
Figure 4 shows a trend of increasing number of publications over time with fluctuations throughout the time period, with a peak of 11 in 2018. The lack of a clear increase of studies over time is likely a reflection of this being an under researched topic area.
The number of different study types associated with the 40 included publications can be seen in Fig. 5. The most commonly used study type were risk assessments (resulting in an estimated or measured exposure risk outcome) (n = 16), with systematic reviews being the least commonly used study type (n = 2).
The bar graph in Fig. 6 shows the number of studies exploring different exposure routes. Most studies reported a specific exposure route. However, for nine exposure routes (n = 7 from studies with one exposure route investigated and n = 2 from a study investigating two exposure routes), the exposure route was not explicit but could be inferred from other information in the paper. For one study, it was not possible to infer the exposure route from the information provided which has resulted in “No information” on the base graph . Information on which studies the exposure route was inferred can be found in the systematic map database for Map 1 (see Additional file 7). Consumption/ingestion was the most common (n = 30), followed by direct contact (n = 9) and inhalation (n = 7). Of the 40 studies included in the map, 7 of these studies investigated two exposure routes, with remainder (n = 33) exploring only one type of exposure route.
Figure 7 shows the number of different outcomes reported in the included studies. Colonisation (n = 11), estimated/measured exposure risk (n = 17) and infection (n = 16) are reported a similar number of times with mortality being reported significantly fewer times (2 studies). Out of the 40 studies identified, 6 of these reported two different outcomes with 34 measuring only one outcome.
The different bacterial species investigated in the included studies are reported in Fig. 8. E. coli is the most highly studied species having been reported in 16 publications with the majority of other species studied only investigated by one publication.
Exposure (main categories) by health outcome
Figure 9 shows a heatmap of the main categories of environmental exposure sources by human health outcome. Water environments were the most researched natural environment, followed by eligible food sources. For water environments, colonisation (n = 16) was the most frequently researched outcome, followed by estimated/measured exposure risk (n = 12) and infection (n = 11) and only one study investigated exposure to water and mortality. In contrast, the most frequently studied outcome for exposure via food was infection (n = 10), whereas colonisation from food was only reported in three studies. Out of the 40 studies identified in Map 1, 6 of these reported two outcomes. In addition, 13 studies investigated two or more exposure sources per study.
To identify the corresponding publications that are presented in the heatmap, Fig. 9, below [and subsequent heatmaps for Map 1 (Figs. 10 and 11)], we have provided a column in the Additional file 7 database titled “Corresponding heatmap figure” (column AC). This column shows the figure number(s) in this study that each publication is displayed in. In addition, to further interrogate this data, columns D, F, H, J, L, N and P display the exposure type(s) in each publication and columns U and V display the outcome which align with the X and Y axis of Fig. 9, respectively. For Figs. 10 and 11, the exposure sub categories can be viewed in columns E, G, I, K, M, O and Q.
Exposure (sub-categories) by health outcome
Two further heatmaps were produced to investigate exposure sub-categories and capture the diversity of research into different water environments (Fig. 10), and ‘other’ categories (Fig. 11). There are 19 different eligible water sources reported in this map with the most commonly reported being coastal water (n = 10). The most frequently studied health outcome with exposure to coastal water was estimated or measured exposure risk (n = 6). Across all aquatic exposure environments the most frequently reported health outcome was colonisation (n = 16). Only one study was identified that investigated mortality due to antibiotic resistance and exposure to water (tap water) .
Figure 11 shows all other exposure source subcategories by outcome. The most studied exposure source here was plants consumed raw (n = 11), followed by exposure to wild meat (n = 7). Both of these showed infection as the most frequently reported outcome. Considering all of the exposure sources, infection was the most commonly reported outcome overall (n = 18). Again, very few studies (n = 2) investigated mortality associated with exposure to antibiotic resistance in these environmental sub-categories.
Map 2: What research evidence is there measuring the prevalence of ARB in the environment in the UK?
Description of review process
For Map 2, the search yield 12,939 hits and 6874 unique titles and abstract for screening once duplicates have been removed. 167 of these studies were selected for full text screening, and 62 were included in the map (Fig. 12). The full Map 2 database, with extracted metadata, can be found in Additional file 7.
The sampling locations of the 62 included articles can be seen in Fig. 13. To view an interactive version of this map, please download the html file provided as an additional file (“Map 2 geographical interactive—Additional file 9”).
In contrast to Map 1, study sites were more evenly distributed around the UK for Map 2. However, there is still a trend that sampling is undertaken relatively close to where researchers are based according to author affiliation information.
For studies that do not specify the sampling site and instead use terms such as UK, England, Wales, Scotland or N. Ireland we have chosen a set of nominal longitude/latitude points to represent these studies. Nominal coordinates were placed in the sea to avoid confusion with studies with specified sampling sites. The nominal latitudes and longitudes were:
England: 53.96486, 1.615305.
N. Ireland: 55.51571, − 9.35855.
Scotland: 59.61634, − 4.27592.
Wales: N/A (no studies specified just Welsh study sites without specifying the exact site).
United Kingdom/Multiple countries sampled (e.g. sample sites in England and Wales): 50.15852, − 1.25847.
Publications over time
Like Map 1, there is a weak increasing trend of published articles over time for Map 2 (Fig. 14). The greatest number of articles were published in 2016 (12 studies).
Figure 15 shows the different methodologies used when researching antibiotic resistance in the environment. Phenotypic methods (such as disk diffusion assays and plating data) are the most commonly used methods with 37 studies reporting using this method. The other three most popular methodologies are molecular methods: polymerase chain reaction (PCR) (15 studies); quantitative PCR (qPCR) (14 studies) and metagenomic sequencing (9 studies). All other methodologies used are only found in a small number of studies and include other (non-metagenomic) sequencing technologies and systematic reviews.
Data were extracted from included studies on environmental bacteria species that were resistant to antibiotics. Most studies investigated mixed communities of bacteria (n = 32) (Fig. 16). Bacterial species commonly found in faeces, and/or transmitted by the faecal–oral route are also common in these studies (E. coli, Enterobacteriaceae, coliforms, E. faecalis, E. faecium, Enterococcus spp., Campylobacter spp. (n = 27)).
Exposure (main categories) vs outcome
Figure 17 is a heatmap showing the main exposure categories by outcome. As can be seen in the heatmap, water is the natural environment that is most studied of all natural environments in the UK with ARGs being the most reported resistance metric for water environments. Conversely, when looking at all environmental compartments, AMR bacteria are quantified the most frequently (in all Map 2 heatmaps and the database, AMR refers to phenotypic resistance and ARGs refers to where molecular methods were used to identify specific genes). Of the 62 full text articles studies 26 investigated more than one outcome and 26 studies investigated more than one exposure.
To identify the corresponding publications that are presented in the heatmap below [and subsequent heatmaps for Map 2 (Figs. 18, 19, 20 and 21)], we have provided a column in the Additional file 7 database titled “Corresponding heatmap figure” (column Y). This column shows the figure number(s) in this study that each publication is displayed in. In addition, to further interrogate this data, columns D, F, H, J and L display the exposure type(s) in each publication and columns N and O display the outcome which align with the X and Y axis of Fig. 17, respectively. For Figs. 18, 19, 20 and 21, the exposure sub categories can be viewed in columns E, G, I, K and M.
Exposure (sub-categories) vs outcome
As seen in Fig. 17, water was the most highly studied environmental setting in terms of total numbers of articles. However, as can be seen in Fig. 18, it is also an extremely diverse research area with many different types of water environments being studied. This is particularly evident when looking at the other main exposure sources in the following sections (see Figs. 19, 20 and 21 below for soil, sediment and other, respectively). Even within just wastewater treatment plants, there are many different types of wastewater being researched. This is related to both influent and effluent and different sources of wastewater (e.g. domestic, hospital, production and agriculture/aquaculture). The most commonly studied outcome in a water environment was ARGs in domestic wastewater treatment plant effluent (n = 10). Of the 62 full text articles, 18 investigated multiple water exposures and 15 investigated more than one outcome.
As can be seen in Fig. 19, unspecified soil is the most common environmental substrate researched, with 12 different soil environments being tested for the relevant resistance outcomes, followed by agricultural soil.
At total of 14 sediment sites were tested for either intI1/class 1 integrons, ARGs or ARB with river sediment being the most commonly studied sediment environment (8 articles) and reedbed sediment being the least studied (1 article) (Fig. 20).
Finally, Fig. 21 shows a heatmap of remaining relevant exposure sources (classed as “other”). This heatmap shows a variety of different exposure sources including different types of food (both wild meat and plants consumed raw), wild animal sources (both faeces and lesions on the animal), agricultural animal faeces (which normally wouldn’t be included as a relevant exposure source as agricultural animals are often treated with antibiotics, however the study included here states that no antibiotics were used in the rearing of these animals ) and plankton. Wild animal faeces was by far the most studied environmental matrix (n = 9), followed by wild meat food sources (n = 4).
Knowledge gaps and clusters
Generally, for both maps most work has investigated water environments but within this there is a diverse range of sample types. For Map 1, research on water environments (e.g. coastal or drinking water) has focused on those with which humans interact so may pose a greater risk of transmission. Conversely, in Map 2 there is a focus towards sample sites and types of sample matrices with a high density of bacteria, such as wastewater treatment plant influent and effluent. This could be as a result of bias of sampling by researchers towards samples that are likely to have a high density of bacteria which could increase the likelihood of ARB/ARG detection.
Air environments are under researched in both maps. Presumably, even though a large number of the population is in daily contact with outdoor air, the density of bacteria in this environment is low and there may, therefore, be more methodological challenges in obtaining accurate data.
Whilst it could be argued that air environments are not worth investigating because the low abundance of ARB/ARGs, humans are in constant contact with air environments. Whilst concentrations may be low, over a lifetime exposure to ARB/ARGs in air could be high through constant exposure. It is, therefore, deemed important to overcome methodological challenges fill knowledge gaps such as these.
Both maps also illustrate clustered sampling sites, presumably close to where the researchers are based. This has resulted in knowledge gaps for environments in certain countries (Map 1) and certain regions of the UK (Map 2).
There is a clear lack of global empirical evidence for the transmission of AMR from the natural environment to humans with only 40 relevant articles being collated for Map 1 with a lack of data coming from certain areas of the globe. However, there were 72 supplementary articles that compares ARB/ARGs in both clinical samples and environmental samples but do not investigate an exposure route which shows a growing interest in this research area. More effort must occur, therefore, to establish transmission routes from environmental to clinical setting.
As previously stated, water was the most highly studied environment (n = 40) with food (excluding animals/fish that are reared on high levels of antibiotics and crops that are always consumed cooked) being the next most studied matrix (n = 18). For estimated exposure risk assessments, these environments are easier to quantify than say exposure and transmission from soil and wild animals as there are defined quantifiable volumes of water and food ingested or consumed. For example, Leonard et al. 2015 identified and collated a number of publications reporting volume of water ingested during water sport sessions. Using densities of third generation cephalosporin resistant E. coli in bathing waters, exposure risk was calculated based on volume of water ingested and concentration of resistant E. coli . Similarly, O’Flaherty et al. 2019 quantified the number of antibiotic resistance E. coli found on lettuce and created a model to estimate exposure risk to humans based on the consumption of a nominal amount of lettuce .
When undertaking other types of research other than estimated exposure risk studies (such as cohort studies), food (n = 12) and water (n = 28) exposure sources are also easier to quantify contact with and have a comparator group, compared to air (n = 1), animal (n = 3) and soil (n = 0) where two groups of “exposed” and “not exposed” are harder to define. Similarly, from all study types, ingestion/consumption (n = 30) is the most studied exposure route whereas direct contact (n = 9) and inhalation (n = 7) are studied less frequently as a result of being able to easily identify a comparator group.
In regards to outcomes, “estimated exposure risk,” “colonisation” and “infection” are reported to a similar degree, whereas there are significantly fewer studies investigating “mortality,” presumably because it is difficult to trace these human health outcomes back to the natural environment. One of the two articles reporting this were as a result of two types of accidents resulting in environmental exposure [67, 68]. Because infections do not necessarily occur at the time of exposure, epidemiological tracing of outbreaks is extremely challenging as individuals may be colonised for weeks, months or years before infection. In addition, infections may occur via transmission from exposed individuals to more vulnerable people.
Finally, the majority of research articles investigate E. coli which are used as a faecal indicator species that indicate if the environment has been impacted by anthropogenic pollution . Although a number of strains of E. coli are human gut commensals, others are important opportunistic pathogens . Commensal E. coli not associated with infections may colonise the gut with no adverse human health outcomes but may be able to transfer ARGs it harbours, via horizontal gene transfer, to pathogenic organisms in the gut making the infection more difficult to treat. Molecular epidemiology approaches focusing on ARG or mobile genetic elements may help in attributing an environmental origin of AMR. However, this still poses many challenges due to the complexity of gene transfer events within microbial populations over time.
There are more studies included in Map 2 (UK based) than in Map 1 (global database). This shows that the quantification of ARB, ARGs, intI1 and point mutations is significantly better researched than empirical evidence of transmission from the environment to humans. Despite the topic area being better researched, there is still a bias towards quantifying AMR in water environments and more specifically different types of wastewater environments.
In terms of what is being measured, ARGs (n = 78) and ARB (n = 74) are both frequently studied in various natural environments, with intI1 (n = 31) being targeted less often. This is unsurprising as, although qPCR targeting intI1 has used as a proxy for environmental resistance [53,54,55,56,57], other methodologies (such as metagenomics) targeting multiple ARGs at once and phenotypic methodologies (such as plating and antibiotic susceptibility testing) are significantly cheaper to undertake. In addition, qPCR can also be used for specific ARGs. Point mutations are rarely characterised in the publications identified in this study (n = 1 study, n = 2 environments). Whilst point mutations are an important mechanism deployed by bacteria to resist antibiotics, they are not transferable as in the case of ARGs associated with mobile genetic elements and cannot, therefore, undergo horizontal gene transfer. Although mutation based resistance is extremely important, acquisition of ARGs through horizontal gene transfer is of greater concern . Horizontal gene transfer represents the mode of ARG acquisition by many bacterial pathogens, including ESKAPE pathogens  and Gram-negative opportunists such as epidemic E.coli strains .
In regards to species targeted in Map 2, mixed communities are targeted most frequently targeted by studies (32 studies). By taking a whole community approach and investigating mixed populations, the resistome of a particular environment can be explored. E. coli are the second most frequently investigated species (13 studies) as they are faecal indicator organism and are often used by surveillance studies of the natural environment to investigate anthropogenic pollution .
Limitations of the map
Limitations of this review are:
Searches were undertaken in English and articles were excluded if they were not published in English. This is unlikely to affect Map 2 which was collating publications undertaken in natural environments in the UK. Map 1 was, however, a global map and, therefore, excluding articles not written in English may result in a bias in articles retrieved.
Despite best efforts to obtain full texts of all publications designated for full text screening, there were a handful of articles that were not able to be retrieved. There were 30 (7.9% of total full text articles screened) and 6 (3.6% total of full text articles screened) articles for Map 1 and Map 2, respectively. A list of these files can be found in Additional file 7.
For Map 1, as discussed previously, we excluded articles which met the inclusion criteria fully except there is no exposure route. We have termed these “sample comparison studies” as they do not show direct evidence of transmission from the environment to humans but show similarities between AMR in environmental niches and humans. This could be as a result of showing direct transmission has occurred is difficult to quantify. As the number of included studies is relatively low for Map 1, we have included a list of the “sample comparison studies” in Additional file 2 as we expect they may be useful for policy makers and researchers alike and should be investigated to provide a better picture on research investigating transmission even though transmission route has not been investigate or only hypothesised.
Finally, for both maps, only 10% of references were cross reviewed as a result of the number of search hits retrieved and limited resources available. However, there was 91.8% (Map 1) and 96.1% (Map 2) agreement between reviewers and where disagreements occurred, discussions occurred between the project team to review these.
Limitations of evidence base
As this was a research mapping exercise, and no critical appraisal was undertaken of gathered research papers. The research gathered and presented here is not necessarily of good quality but does summarise the current research landscape. In addition, these maps could provide a starting point for future evidence synthesis work that includes critical appraisal and meta-analyses.
As a result of the limited number of studies found globally for Map 1, the evidence base is not distributed evenly across the globe. It is generally, therefore, clustered where researchers are based identified through author affiliations in articles. This may create a bias as to the choice of the surrounding natural environments and socioeconomic status of those affected.
As a result of limited resources and in response to the funder focus, we limited Map 2 to UK specific studies, with the aim of this being useful for UK policy makers. This means that the results of Map 2 will not be representative of the global research as a whole. Whilst the UK has overwhelmingly studied water environments, this may not be the case elsewhere where a broader coverage of environments may have been studied.