SemOpenAlex |

SemOpenAlex

Matches in SemOpenAlex for { <https://semopenalex.org/work/W4320806969> ?p ?o ?g. }

Showing items 1 to 53 of 53 with 100 items per page.

W4320806969 abstract "Article Figures and data Abstract Editor's evaluation Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract Antibiotic consumption and its abuses have been historically and repeatedly pointed out as the major driver of antibiotic resistance emergence and propagation. However, several examples show that resistance may persist despite substantial reductions in antibiotic use, and that other factors are at stake. Here, we study the temporal, spatial, and ecological distribution patterns of aminoglycoside resistance, by screening more than 160,000 publicly available genomes for 27 clusters of genes encoding aminoglycoside-modifying enzymes (AME genes). We find that AME genes display a very ubiquitous pattern: about 25% of sequenced bacteria carry AME genes. These bacteria were sequenced from all the continents (except Antarctica) and terrestrial biomes, and belong to a wide number of phyla. By focusing on European countries between 1997 and 2018, we show that aminoglycoside consumption has little impact on the prevalence of AME-gene-carrying bacteria, whereas most variation in prevalence is observed among biomes. We further analyze the resemblance of resistome compositions across biomes: soil, wildlife, and human samples appear to be central to understand the exchanges of AME genes between different ecological contexts. Together, these results support the idea that interventional strategies based on reducing antibiotic use should be complemented by a stronger control of exchanges, especially between ecosystems. Editor's evaluation This work presents important findings on the role of ecological niches in shaping the distribution of antibiotic resistance genes in the environment and human populations. The evidence supporting the conclusion is compelling, presenting a rigorous analysis of a large dataset with many possible confounding variables. The work should be of broad interest to microbiologists, public health workers, and policymakers. https://doi.org/10.7554/eLife.77015.sa0 Decision letter Reviews on Sciety eLife's review process Introduction Antibiotic resistance (AR) is a persistent global public health problem that has increased over the last decades, with resistances spreading faster and faster after antibiotic introduction in clinical use (Witzany et al., 2020)⁠. Although first concerns about infections by antibiotic resistant bacteria (AR bacteria) were formulated in the 1940s, the discovery and development of new antibiotics allowed for treatment substitutions (Podolsky, 2018)⁠ during the first decades of antibiotic use. The discovery that AR was frequently acquired by horizontal gene transfer (HGT, Watanabe, 1963⁠) and the subsequent emergence of multi-resistant strains led the international health agencies to start raising the issue, at the end of the 1970s, that antibiotic resistance propagation was threatening to end the antibiotic golden area and jeopardize the huge progress made in the treatment of bacterial infectious diseases. In parallel, the discovery and design of new antibiotics had become more and more difficult (Livermore et al., 2011)⁠. This trend persisted despite the bulk of information provided by recent advances in genomics (Ribeiro da Cunha et al., 2019), thus decreasing the hope for potential treatment substitutions. A recent review (Antimicrobial Resistance Collaborators, 2022) evaluated that in 2019, around 1.25 million deaths were directly attributable to bacterial antimicrobial resistance. Antibiotic consumption and its abuses have been historically and repeatedly pointed out as the major cause of AR genes propagation (O’Neill, 2014; Podolsky, 2018; Schrijver et al., 2018): the frequency of AR increases in bacteria communities under the selective pressure of antibiotics. To fight this threat, most health agencies are thus focusing their policies on sanitation and mostly on a more reasonable use of antibiotics (World Health Organization, 2015)⁠. This perception of the factors driving AR spread and its associated policies still prevail nowadays: for example, a meta-analysis over 243 studies found a positive correlation between antibiotic consumption and presence of AR (Bell et al., 2014). However, even if antibiotic consumption decreases in most developed countries, AR does not always follow the same path: for example, a drastic reduction in sulfonamide consumption in the United Kingdom during the 1990s was not followed by a decrease in prevalence of sulfonamide-resistant Escherichia coli (Enne et al., 2004)⁠. Antibiotic consumption therefore does not seem to be the only factor maintaining AR in pathogenic bacteria communities. Indeed, although the impact of human activities on the circulation and spread of AR genes is now well documented through the accumulation of specific examples, integrative studies identifying large-scale trends are lacking and this absence of general view has been pinpointed as a gap in our knowledge of drivers of antimicrobial resistance (Holmes et al., 2016). Another strongly overlooked factor is ecology, and a growing number of studies has called for a more comprehensive analysis of AR outside of farms and hospitals (see e.g Bengtsson-Palme et al., 2018)⁠. Though data in natural ecosystems remain scarce, AR genes have probably always been natural members of the gene pools of environmental microbial communities (D’Costa et al., 2011)⁠. Natural ecosystems may contribute to the spread of AR, both as sources and as vectors of propagation (Bengtsson-Palme et al., 2018; Berendonk et al., 2015; Marti et al., 2014)⁠. This realization, combined with an increasing access to genomic data, led to bioinformatic studies where the goal was to extend our understanding of AR in ecological contexts that are often overlooked: for example β-lactam resistance in dairy industry (Pitta et al., 2016)⁠, in slaughterhouses (Lavilla Lerma et al., 2014)⁠, in wastewater treatment plants (Karkman et al., 2016)⁠, or in natural fresh water (Czekalski et al., 2015)⁠. However, even if these studies complement our knowledge on the presence, the frequency, the nature, and the circulation of AR in poorly documented environments (see e.g. Cuadrat et al., 2020; Zhang et al., 2020), descriptions of global distribution patterns and analyses of factors underlying them remain scarce. In this study, we investigate the relative importance of environmental and genomic factors in shaping the routes of antibiotic spread on a large scale, focusing on aminoglycoside resistance. Aminoglycosides are antibiotics that bind to the 30 S subunit of prokaryotic ribosomes and thus inhibit protein synthesis (Mingeot-Leclercq et al., 1999)⁠. This antibiotic class was first introduced in 1944 with the successful use of streptomycin against tuberculosis, and the first antibiotic able to fight Gram-negative bacteria. Several years later, other drugs produced by Streptomyces spp. were discovered (kanamycin, spectinomycin, tobramycin, neomycin, apramycin) and introduced in clinics. They were followed in the 1970s by a series of new isolates or derivatives synthesized compounds (amikacin, netilmicin, isepamicin, dibekacin, arbekacin, see van Hoek et al., 2011)⁠. However, the emergence of resistant strains during the following years, combined with the requirement of administration by injection and possible nephro- and ototoxicity has reduced the use of aminoglycosides in therapies (Murray and Murray, 1991)⁠. Nowadays, these drugs are only used in humans as a second-line or last-resort treatment for Gram-negative bacteria (Garneau-Tsodikova and Labby, 2016)⁠. However, they remain frequently used in agriculture and veterinary medicine (European Medicines Agency, 2019)⁠. As an example, in 2015, aminoglycosides still represented 3.5% of the total sales of antimicrobials for farm animals, and they are still important drugs for some pathologies, for example post-weaning diarrhoea in pigs (van Duijkeren et al., 2019). Thus, aminoglycoside resistance is still highly prevalent among farm animals: in Denmark, 85% of Salmonella spp. serovar Kentucky isolated in turkeys are not susceptible to aminoglycosides, and in Italy, up to 67% of E. coli samples are resistant to at least one aminoglycoside drug (van Duijkeren et al., 2019). Thus, although these drugs are no longer widely prescribed for patients, aminoglycoside resistance is still a major threat, at least toward food production and the treatment of patients infected by multi-resistant bacteria. Moreover, since most aminoglycosides originate from soil-dwelling bacteria, there are at least three ecological contexts (hospitals, farms, and soil) in which aminoglycoside resistance can evolve and spread to cause public health issues. The three main mechanisms of aminoglycoside resistance are (Garneau-Tsodikova and Labby, 2016)⁠: (i) decrease in drug uptake (through modifications of membrane permeability or of the peri-membrane ion gradient) and/or increase in drug efflux (through efflux pump activation); (ii) drug inactivating enzymes; and (iii) modification of the drug target, for example by point mutations in the genes coding for ribosomal small subunit (Finken et al., 1993; Sander et al., 1995; Toivonen et al., 1999)⁠. Aminoglycoside-modifying enzymes (AMEs) are a class of inactivating enzymes that catalyze the transfer of chemical groups on specific residues of the aminoglycoside molecules, leading to a modified drug which poorly binds to its target (Jana and Deb, 2006)⁠. AMEs represent the most common mechanism of aminoglycoside resistance in clinical isolates and are well characterized biochemically (Ramirez and Tolmasky, 2010)⁠. The classical nomenclature of AMEs is based on the group they transfer (i.e. acetyltransferases, AACs; nucleotidyltransferases, ANTs; and phosphotransferases, APHs), on the residue modified, and on the resistance profile they confer (Ramirez and Tolmasky, 2010)⁠. However, AMEs emerged several times during evolution (see e.g. Salipante and Hall, 2003 for AAC(6’) enzymes), so biochemical nomenclatures do not reflect the evolutionary history of any class of AMEs. Finally, many AME genes are carried by mobile genetic elements, which give them the potential to be transmitted both vertically and horizontally (Davies, 1983)⁠. They represent today a major threat for the treatment of multidrug-resistant bacteria, notably Mycobacterium tuberculosis (Labby and Garneau-Tsodikova, 2013)⁠. Through a computational approach, more than 160,000 publicly available genomes were screened to identify the presence of AME-encoding genes (AME genes) across the phylogeny of Eubacteria. The present study intended (i) to describe the genomic, geographical and ecological distribution of AME genes; and based on these data, (ii) to quantify the relative contribution of several key factors (geography, ecology, genomic context, human activities) potentially driving the spread of AME genes. Results Aminoglycoside resistance is widespread across geography, ecology, and phylogeny 160,987 publicly available Eubacteria genomes were screened for the presence of AME genes. The list of genomes and their metadata are listed in Source data 1. Published sequences of AMEs and genes coding for AMEs had been previously grouped in 27 clusters of homologous genes (CHGs), each containing sequences of genes and proteins very likely to share a common ancestor (see Materials and methods). A total of 46,053 AME genes were detected in 38,523 genomes (i.e. about one quarter of the genomes screened). Their distribution in 27 CHGs is listed in Table 1, all the gene coordinates are listed in Source data 2, and their distribution over time, phyla, geography and ecology in Source data 3. Our dataset included 54 phyla across the phylogeny of Eubacteria but 89.6% of genomes belong to the three most represented phyla: Proteobacteria, Firmicutes, and Actinobacteriota. In the same way, though we analyzed genomes from 13,879 species, only 10 species made up 43.6% of the dataset. This imbalance was even stronger regarding the repartition of resistance genes: we found AME genes in 23 phyla, 97.2% of them were detected in the three most represented phyla. Moreover, whereas 23.9% of all the genomes in our dataset carried at least one AME gene, this proportion decreases to 16.3% across the least sampled species making up 50% of the dataset, and even to 10.9% across the species making up 10% of the dataset. The frequency of the resistance-carrying genomes presented very contrasting patterns across CHGs and phyla (Figure 1). For each CHG, the phylogenetic diversity of the species in which it was detected was evaluated by calculating Faith’s distances (Faith, 1992)⁠. This reveals that the number of species carrying these genes ranges from 2 to 468 (respectively for ANTc with Faith’s distance dFaith = 1.4 and ANTa with dFaith = 78.4), but also that for CHGs present in numerous species, these species can belong to a small number of phyla (e.g. AACc, dFaith = 24.5 for 406 species) or be largely spread across the bacteria phylogenetic tree (e.g. AACf1, dFaith = 66.7 for 418 species). Regarding the phyla, some present a high diversity of CHGs (e.g. Proteobacteria, Actinobacteria, Firmicutes I) whereas others only contain a very small number of different CHGs. Figure 1 Download asset Open asset Prevalence of AME-gene-carrying bacteria across the phylogeny of Eubacteria. The phylogenetic tree corresponds to an aggregation of bac120 phylogeny (Parks et al., 2018) to the phylum level. In the heatmap, blank boxes correspond to the observed absence of a CHG in a phylum. For the other colors, blue to red boxes stand for CHG frequencies from near-zero to one. Gray bars in the top part correspond to the Faith’s distance (the sum of the lengths of all the branches on the bac120 tree) for the species in which each CHG was found. Gray bars on the right correspond to the prevalence of AME bacteria for each phylum, i.e. the proportion of genomes in which at least one AME gene was found. Table 1 List of CHGs by biochemical function. ClusterBiochemical functionNumber of genes identifiedAACaN-acetyltransferases619AACbN-acetyltransferases711AACcN-acetyltransferases4636AACdN-acetyltransferases1044AACe1N-acetyltransferases5230AACe2N-acetyltransferases13AACf1N-acetyltransferases11227AACf3N-acetyltransferases10AACgN-acetyltransferases4016AAChN-acetyltransferases3124AACiN-acetyltransferases871AACjN-acetyltransferases2246AACmN-acetyltransferases50ANTanucleotidyltransferases4508ANTbnucleotidyltransferases2990ANTcnucleotidyltransferases21ANTdnucleotidyltransferases982ANTenucleotidyltransferases59APHaphosphotransferases675APHbphosphotransferases351APHcphosphotransferases49APHd1phosphotransferases518APHd2phosphotransferases67APHephosphotransferases20APHfphosphotransferases1970APHgphosphotransferases23APHhphosphotransferases23 The location and biome metadata could be recovered for 45,574 genomes and from these data, it was established that AME genes were present in samples coming from all the continents, with the exception of Antarctica. In most regions, the prevalence of AME-gene-carrying bacteria (AME bacteria) was between 20% and 40%, but it ranged from 10% in Japan, Eastern Europe, and Eastern Africa to 50% in Indonesia, Mexico, and Turkey (Figure 2A). A quite high spatial heterogeneity in terms of the proportion of each CHG among the AME bacteria was also revealed (Figure 2B): AACf1 is over-represented in the Southern hemisphere (Africa, South-East Asia, Oceania, Brazil), APHf is over-represented in Canada and Mexico and at very low proportion elsewhere, whereas Western Europe, the United States, and Japan have a rather balanced representation of all CHGs. Moreover, this heterogeneity does not appear to be structured by geographical distance itself: there are large differences when making comparisons e.g. between Canadian and the United States’ resistomes, or between Western European and Central European resistomes. AME bacteria were identified in samples coming from all the biomes investigated. The vast majority of them come from clinical samples (55.3%), human samples (22.1%), and farm samples (12.3%). Despite this large bias in ecological distribution, the prevalence of AME bacteria per biome varies in a relatively narrow range from 9% of bacteria sampled in domestic animals, to nearly 40% of bacteria sampled in humans (Figure 3). The ecological spread of CHGs is strongly correlated to their phylogenetic spread (Pearson correlation between the number of ecosystems and dFaith, ρ=0.759, p=6.8.10–6). Thus, the CHGs with the largest dFaith are ecologically ubiquitous (AACc, AACe1, AACf1, AACg, ANTa and APHf), while 8 other CHGs with limited phylogenetic diversity were found in at most three biomes (AACe2, AACf3, ANTc, APHc, APHd2, AACm, ANTe, and APHg). Interestingly, clinical samples were found to carry resistance genes of all CHG except two and these two CHGs were specific to agrosystems only (APHg) and agrosystems and farms (AACf3). Moreover, there are significant negative correlations between the date of first sampling and the phylogenetic and ecological spreads (respectively ρ=0.712, p=4.4.10–5 and ρ=0.718, p=3.6.10–5): all ecologically ubiquitous CHGs were sampled before 1955, while the CHGs that were never sampled before 2000 could be found in at most four biomes. Figure 2 Download asset Open asset Distribution of aminoglycoside-resistant bacteria and AME-encoding genes over the world. (A) Distribution of sampled aminoglycoside-resistant bacteria in the world. The frequencies of resistant and sensitive bacteria are displayed in red and blue, and the size of pies represents the number of genomes sampled in a given region. (B) Distribution of sampled AME genes in the world. Pie size is irrelevant to the number of samples. Figure 3 Download asset Open asset Prevalence of AME-gene-carrying bacteria across land biomes. In the heatmap, blank boxes correspond to the observed absence of a CHG. For the other colors, blue to red boxes stand for CHG frequencies from near-zero to 0.3. Gray bars on the right correspond to the prevalence of AME bacteria for each biome, i.e. the prevalence of genomes in which at least one AME gene was found. APHe is excluded from the heatmap because it was sampled in seawater only. Our dataset includes bacteria sampled between 1885 and 2019. However, the vast majority of them were sampled recently: 96.3% of genomes were sampled after 1990, and 58.9% after 2010. In this dataset, the first occurrence of an AME bacterium dates from 1905 (Figure 4A), i.e. before the human discovery of streptomycin in 1943. Moreover, the prevalence of AME bacteria remains constant (about 30%) between 1990 and 2019 (Figure 4B), a period during which no new aminoglycoside drug was released for commercial use. However, the frequencies of individual CHGs varied much more over time in this same time frame (Figure 4C), with some CHGs having a regularly increasing (e.g. AACe1, AACh) or decreasing (e.g. AACf1) frequency over the whole period or showing one-time frequency peak (e.g. ANTb in 2003 and 2004). The worldwide dynamics shows more a coexistence across time of a diversity of CHG than sequential substitution of one CHG by another. However, though we also observe coexistence across time at continental levels (Figure 4—figure supplement 1), local time trends suggest that some CHGs might be progressively replaced: for example the frequency of AACc and AACe1 has increased a lot at the expense of AACf1 in Europe, Southeastern Asia and Oceania since the 1990s. This moreover shows that, despite geographical differences between continental resistomes (see Figure 2B), these differences are not consistent over time. Figure 4 with 2 supplements see all Download asset Open asset Global time trends for aminoglycoside resistance. (A) First sampling occurrence for each CHG in the analyzed dataset. First sampling occurrence of APHe is unknown. (B) Evolution of the worldwide and European prevalences of AME bacteria between 1943 and 2019. The dots and error bars represent the measured prevalence (± standard error) of AME bacteria each year. The dotted line represents a binomial regression of the prevalence of AME bacteria over time, fitted separately before 1995 and after 1995. The blue curve represents the number of bacteria genomes sampled each year (in logarithmic scale). Orange stars represent the dates at which the following aminoglycoside drugs were released for commercial use: streptomycin, neomycin, hygromycin, kanamycin, paromomycin, capreomycin, gentamicin, ribostamycin, dibekacin, tobramycin, amikacin, sisomicin, netilmicin, micronomicin, isepamycin, plazomicin. The red rectangle displays the period of time analyzed in panel: (C) Evolution of worldwide CHG frequencies among sampled AME genes between 1990 and 2019. 17 CHGs for which less than 400 sequences were sampled are grouped and displayed in gray. The European distribution of aminoglycoside resistance is driven by ecology and human exchanges We investigated the potential role of different factors on the distribution of AME bacteria in Europe: ecology (structuration in biomes), human exchanges (potential AME bacteria importations through immigration and merchandise imports), and aminoglycoside consumption. This analysis was performed on the timeframe 1997–2018 for which antibiotic consumption data by country and by year was available. It is important to note here that during this period, aminoglycoside consumption was quite constant over time, but varied strongly between countries. We included in the dataset 16 CHGs for which occurrences were detected in at least 30 genomes sampled in Europe during this timeframe. Most CHGs had very distinct distributions over time and space, so they were analyzed separately. Analysis of the whole dataset Within Europe, a Matérn spatial autocorrelation structure was kept in all the models: CHGs tend to cluster between close countries and the autocorrelation structure allowed to control for this. A time autoregressive structure was also kept in all the models, which showed positive autocorrelation for half of CHGs (the others show negative time autocorrelation). On average, the inputs of the dataset reduced deviance by R2adj=32.2% (ranging from 10.1% to 67.6%). Ecology, human exchanges, and antibiotics consumption are kept as explanatory variables for respectively 16 CHGs, 13 CHGs, and 10 CHGs out of 16. The importance of each explanatory variable varies across CHGs, but ecology overall appears to be the most important one: it increased adjusted R2adj by 24.4% on average (4.6% on average for human exchanges, and 2.9% on average for antibiotics consumption). Interactions between ecology and human exchanges increased R2adj by 6.9% (for 6 CHGs), and interactions between ecology and antibiotics consumption increased R2adj by 2.6% (for 6 CHGs) (Figure 5). The effects of all variables and their interactions for individual CHGs are given in Supplementary file 1. Figure 5 with 1 supplement see all Download asset Open asset Relative importance of several factors to explain the distribution of aminoglycoside-resistant bacteria in Europe between 1997 and 2018. Logistic regressions with a spatial Matérn correlation structure were computed to explain the frequency of 16 CHGs. This figure represents the contribution of each variable in each selected model, as the fraction of adjusted McFadden’s pseudo-R2 explained by adding this variable. The effect of AG consumption is not unidirectional across all the CHGs: when kept in the selected model, its effect is significantly positive for only three CHGs (AACa, AACc, and AACe1), but negative for three others (AACf1, APHd1, and APHf) and nonsignificant for the others. Human exchanges represented in the analysis by our proxy for potential AME bacteria influx due to trade, likewise has a positive significant effect on the AME bacteria prevalence for few CHGs (AACi, AACj, ANTb, and APHa), but a negative significant effect for AACc and APHf Our proxy for potential bacteria influx due to migration has a negative significant effect on the probability to sample AME bacteria for AACd, AACf1, and APHf, and a positive effect for AACb and AACc. No significant effect of trade and migration were found for other CHGs. Despite the explanatory importance of ecology and the numerous interactions identified, for most CHGs, the probability of sampling AME bacteria does not significantly differ between most biomes. However, the probability of sampling AME bacteria in clinical samples is significantly higher for AACa, AACc, AACe1, and AACh (respectively significantly lower for AACb and ANTa) than in other biomes. In the same way, AACb, ANTa, and APHf are significantly less likely to be sampled in farms, while AACc is significantly more likely and AACb less likely to be sampled in soil. Regarding interaction effects, aminoglycoside consumption can have a positive significant effect on the probability to sample AME bacteria in clinical samples (for AACb, AACf1, and APHf), soil (for AACf1), and farms samples (for AACb). Human exchanges also have ecology-specific impacts for four CHGs only. For AACf1, trade has a negative effect on the probability to sample AME bacteria in clinical, farms, human, and soil samples, whereas migration has positive effects in clinical, farms, and human samples and a negative effect in freshwaterAME bacteria. For AACg, migration has a positive effect in clinical samples, whereas for ANTa, trade has a negative effect in soil samples. For APHf, migration has a negative effect in clinical, farms, and human samples, while trade has a positive effect in farms and a negative effect in clinical samples. To sum up, differences in aminoglycoside use are less likely to explain differences in aminoglycoside resistance prevalence, than sampling across different biomes. We proceeded to the same analyses for worldwide data on the same timeframe, analyzing only the effects of ecology and human exchanges (see Supplementary file 2). Antibiotic consumption data could not be accessed at this spatial scale, but the same trend was found for the other factors: ecology is always the most important explanatory factor, and human exchanges are conserved as an explanatory factor for most CHGs (see Figure 5—figure supplement 1). What is the effect of biased sampling? However, since our dataset consists of genomes sequenced for multiple research projects, it was necessary to test the effect of a sampling bias on these results: indeed, it can be easily assumed that sampling effort in different habitats will be focused on different bacterial taxa, for example those that have the greatest impact in these habitats. To account for this bias, the dataset was split into two subsets, each making up for approximately 50% of sampled genomes in each biome: one consisting of the 99 most sampled species (e.g. Clostridioides difficile in sludge and waste), and the other of the 805 least sampled species (e.g. Lactobacillus kalixensis in the clinical environment). This implies that a given species can be represented among either the most sampled or the least sampled species, depending on the biome in which it was sampled. A new model selection was computed on each of these datasets for the 7 CHGs for which 30 AME bacteria were available in both datasets (i.e. AACc, AACe1, AACf1, AACg, AACh, ANTa, and APHf). The results of this model selection are detailed in Supplementary file 3. The inputs of the models have a better explanatory power for the most sampled species dataset (on average R2adj = 52.7%, ranging from 19.7% to 86.6%) than for the least sampled species dataset (on average R2adj = 18.3%, ranging from 7.1% to 25.6%). Ecology is still the most important variable to explain variations in AME bacteria prevalence, being kept as an explanatory factor for all 7 CHGs in the case of highly sampled species, and 6 of them for the least sampled species. However, its importance varied greatly between the datasets: ecology increased R2adj by 45.6% on average for most sampled species, but only by 14.8% for least sampled species. Human exchanges still accounted for variation in prevalence of 6 CHGs in the case of most sampled species (with an average increase of R2adj of 3.4%), whereas for least sampled species, it had a higher importance (average increase of R2adj of 4.5%) but for 4 CHGs only. Similarly, although aminoglycoside consumption contributed to the variance of 3 and 5 CHGs respectively for most and least sampled species, it had a higher importance for least sampled species (with an average increase of R2adj of 6.3%) than for most sampled species (respectively 1.9%). Finally, interactions between ecology and human exchanges increased R2adj by 3.2% for most sampled species (for 5 CHGs) and by 5.5% for least sampled species (for 3 CHGs), whereas interactions between ecology and aminoglycoside consumption increased R2adj by 2.9% for most sampled species (for one CHG only) and by 3.3% for least sampled species (for 4 CHGs). Thus, although ecology is always the most important variable to explain the distribution of AME bacteria, especially compared to aminoglycoside consumption, its importance is overestimated by an unbalanced sampling of taxa across ecosystems. The effect of human exchanges For the 13 CHGs sampled in Europe for which we kept human exchanges as explanatory variables in models of section 2.2.1, we computed new models based on the decomposition of these effects: we replaced the two effects beneath human exchanges (i.e. international trade and immigratio" @default.
W4320806969 created "2023-02-15" @default.
W4320806969 creator A5011931659 @default.
W4320806969 creator A5024965499 @default.
W4320806969 date "2022-09-27" @default.
W4320806969 modified "2023-10-16" @default.
W4320806969 title "Author response: Ecology, more than antibiotics consumption, is the major predictor for the global distribution of aminoglycoside-modifying enzymes" @default.
W4320806969 doi "https://doi.org/10.7554/elife.77015.sa2" @default.
W4320806969 hasPublicationYear "2022" @default.
W4320806969 type Work @default.
W4320806969 citedByCount "0" @default.
W4320806969 crossrefType "peer-review" @default.
W4320806969 hasAuthorship W4320806969A5011931659 @default.
W4320806969 hasAuthorship W4320806969A5024965499 @default.
W4320806969 hasBestOaLocation W43208069691 @default.
W4320806969 hasConcept C110121322 @default.
W4320806969 hasConcept C134306372 @default.
W4320806969 hasConcept C144024400 @default.
W4320806969 hasConcept C18903297 @default.
W4320806969 hasConcept C2777787772 @default.
W4320806969 hasConcept C30772137 @default.
W4320806969 hasConcept C33923547 @default.
W4320806969 hasConcept C36289849 @default.
W4320806969 hasConcept C501593827 @default.
W4320806969 hasConcept C86803240 @default.
W4320806969 hasConcept C89423630 @default.
W4320806969 hasConceptScore W4320806969C110121322 @default.
W4320806969 hasConceptScore W4320806969C134306372 @default.
W4320806969 hasConceptScore W4320806969C144024400 @default.
W4320806969 hasConceptScore W4320806969C18903297 @default.
W4320806969 hasConceptScore W4320806969C2777787772 @default.
W4320806969 hasConceptScore W4320806969C30772137 @default.
W4320806969 hasConceptScore W4320806969C33923547 @default.
W4320806969 hasConceptScore W4320806969C36289849 @default.
W4320806969 hasConceptScore W4320806969C501593827 @default.
W4320806969 hasConceptScore W4320806969C86803240 @default.
W4320806969 hasConceptScore W4320806969C89423630 @default.
W4320806969 hasLocation W43208069691 @default.
W4320806969 hasOpenAccess W4320806969 @default.
W4320806969 hasPrimaryLocation W43208069691 @default.
W4320806969 hasRelatedWork W176919298 @default.
W4320806969 hasRelatedWork W2022652693 @default.
W4320806969 hasRelatedWork W2080225608 @default.
W4320806969 hasRelatedWork W2084000051 @default.
W4320806969 hasRelatedWork W2354663347 @default.
W4320806969 hasRelatedWork W2384059439 @default.
W4320806969 hasRelatedWork W2415457881 @default.
W4320806969 hasRelatedWork W4240928414 @default.
W4320806969 hasRelatedWork W4309216359 @default.
W4320806969 hasRelatedWork W4361190065 @default.
W4320806969 isParatext "false" @default.
W4320806969 isRetracted "false" @default.
W4320806969 workType "peer-review" @default.