Markov Networks of Collateral Resistance: National Antimicrobial Resistance Monitoring System Surveillance Results from Escherichia coli Isolates, 2004-2012

Risk factors for fluoroquinolone- and macrolide-resistance among swine Campylobacter coli using multi-layered chain graphs

Campylobacter spp. resistant to fluoroquinolones and macrolides are serious public health threats. Studies aiming to identify risk factors for drug-resistant Campylobacter have narrowly focused on antimicrobial use at the farm level. Using chain graphs, we quantified risk factors for fluoroquinolones- and macrolide-resistance in Campylobacter coli isolated from two distinctive swine production systems, conventional and antibiotic-free (ABF). The chain graphs were learned using genotypic and phenotypic resistance data from 1082 isolates and host exposures obtained through surveys for 18 cohorts of pigs. The gyrA T86I point mutation alone explained at least 58 % of the variance in ciprofloxacin minimum inhibitory concentration (MIC) for ABF and 79 % in conventional farms. For macrolides, genotype and host exposures explained similar variance in azithromycin and erythromycin MIC. Among host exposures, heavy metal exposures were identified as risk factors in both conventional and ABF. Chain graph models can generate insights into the complex epidemiology of antimicrobial resistance by characterizing context-specific risk factors and facilitating causal discovery.

Global surveillance of antimicrobial resistance in food animals using priority drugs maps

Antimicrobial resistance (AMR) in food animals is a growing threat to animal health and potentially to human health. In resource-limited settings, allocating resources to address AMR can be guided with maps. Here, we mapped AMR prevalence in 7 antimicrobials in Escherichia coli and nontyphoidal Salmonella species across low- and middle-income countries (LIMCs), using 1088 point-prevalence surveys in combination with a geospatial model. Hotspots of AMR were predicted in China, India, Brazil, Chile, and part of central Asia and southeastern Africa. The highest resistance prevalence was for tetracycline (59% for E. coli and 54% for nontyphoidal Salmonella, average across LMICs) and lowest for cefotaxime (33% and 19%). We also identified the antimicrobial with the highest probability of resistance exceeding critical levels (50%) in the future (1.7-12.4 years) for each 10 × 10 km pixel on the map. In Africa and South America, 78% locations were associated with penicillins or tetracyclines crossing 50% resistance in the future. In contrast, in Asia, 77% locations were associated with penicillins or sulphonamides. Our maps highlight diverging geographic trends of AMR prevalence across antimicrobial classes, and can be used to target AMR surveillance in AMR hotspots for priority antimicrobial classes.

Early evaluation of the Food and Drug Administration (FDA) guidance on antimicrobial use in food animals on antimicrobial resistance trends reported by the National Antimicrobial Resistance Monitoring System (2012-2019)

Antimicrobial resistance (AMR) is one of the biggest challenges to global public health. To address this issue in the US, governmental agencies have implemented system-wide guidance frameworks and recommendations aimed at reducing antimicrobial use. In particular, the Food and Drug Administration (FDA) prohibited the extra-label use of cephalosporins in food animals in 2012 and issued the guidance for industry (GFI) #213 about establishing a framework to phase out the use of all medically relevant drugs for growth promotion in 2012. Also in 2015, the FDA implemented veterinary feed directive (VFD) drug regulations (GFI# 120) to control the use of certain antimicrobials. To assess the potential early effects of these FDA actions and other concurrent antimicrobial stewardship actions on AMR in the food chain, we compared the patterns of the phenotypic (minimum inhibitory concentration (MIC) and percentage of resistance) and genotypic resistances for selected antimicrobials before and after 2016 across different enteric pathogen species, as reported by the National Antimicrobial Resistance Monitoring System (NARMS). Most of the antimicrobials analyzed at the phenotypic level followed a downward trend in MIC after implementing the guidance. Although, most of those changes were less than one 1-fold dilution. On the other hand, compared to MIC results, the results based on phenotypic resistance prevalence evidenced higher differences in both directions between the pre- and post-guidance implementation period. Also, we did not find relevant differences in the presence of AMR genes between pre- and post-VFD drug regulations. We concluded that the FDA guidance on antimicrobial use has not led to substantial reductions in antimicrobial drug resistance yet.

Identifying patient-level risk factors associated with non-β-lactam resistance outcomes in invasive MRSA infections in the United States using chain graphs

Background

MRSA is one of the most common causes of hospital- and community-acquired infections. MRSA is resistant to many antibiotics, including β-lactam antibiotics, fluoroquinolones, lincosamides, macrolides, aminoglycosides, tetracyclines and chloramphenicol.

Objectives

To identify patient-level characteristics that may be associated with phenotype variations and that may help improve prescribing practice and antimicrobial stewardship.

Methods

Chain graphs for resistance phenotypes were learned from invasive MRSA surveillance data collected by the CDC as part of the Emerging Infections Program to identify patient level risk factors for individual resistance outcomes reported as MIC while accounting for the correlations among the resistance traits. These chain graphs are multilevel probabilistic graphical models (PGMs) that can be used to quantify and visualize the complex associations among multiple resistance outcomes and their explanatory variables.

Results

Some phenotypic resistances had low connectivity to other outcomes or predictors (e.g. tetracycline, vancomycin, doxycycline and rifampicin). Only levofloxacin susceptibility was associated with healthcare-associated infections. Blood culture was the most common predictor of MIC. Patients with positive blood culture had significantly increased MIC of chloramphenicol, erythromycin, gentamicin, lincomycin and mupirocin, and decreased daptomycin and rifampicin MICs. Some regional variations were also observed.

Conclusions

The differences in resistance phenotypes between patients with previous healthcare use or positive blood cultures, or from different states, may be useful to inform first-choice antibiotics to treat clinical MRSA cases. Additionally, we demonstrated multilevel PGMs are useful to quantify and visualize interactions among multiple resistance outcomes and their explanatory variables.

TGF-β Signaling Pathway-Based Model to Predict the Subtype and Prognosis of Head and Neck Squamous Cell Carcinoma

Background: Although immunotherapy with immune checkpoint therapy has been used to treat head and neck squamous cell carcinoma (HNSCC), response rates and treatment sensitivity remain limited. Recent studies have indicated that transforming growth factor-β (TGF-β) may be an important target for novel cancer immunotherapies.

Materials and methods: We collected genomic profile data from The Cancer Genome Atlas and Gene Expression Omnibus. The least absolute shrinkage and selection operator method and Cox regression were used to establish a prognostic model. Gene set enrichment analysis was applied to explore biological functions. Tracking of indels by decomposition and subclass mapping algorithms were adopted to evaluate immunotherapy efficiency.

Result: We established a seven TGF-β pathway-associated gene signature with good prediction efficiency. The high-risk score subgroup mainly showed enrichment in tumor-associated signaling such as hypoxia and epithelial-mesenchymal transition (EMT) pathways; This subgroup was also associated with tumor progression. The low-risk score subgroup was more sensitive to immunotherapy and the high-risk score subgroup to cisplatin, erlotinib, paclitaxel, and crizotinib.

Conclusion: The TGF-β pathway signature gene model provides a novel perspective for evaluating effectiveness pre-immunotherapy and may guide further studies of precision immuno-oncology.

Analysis of Multidrug Resistance in Staphylococcus aureus with a Machine Learning-Generated Antibiogram

Multidrug resistance (MDR) surveillance consists of reporting MDR prevalence and MDR phenotypes. Detailed knowledge of the specific associations underlying MDR patterns can allow antimicrobial stewardship programs to accurately identify clinically relevant resistance patterns. We applied machine learning and graphical networks to quantify and visualize associations between resistance traits in a set of 1,091 Staphylococcus aureus isolates collected from one New York hospital between 2008 and 2018. Antimicrobial susceptibility testing was performed using reference broth microdilution. The isolates were analyzed by year, methicillin susceptibility, and infection site. Association mining was used to identify resistance patterns that consisted of two or more individual antimicrobial resistance (AMR) traits and quantify the association among the individual resistance traits in each pattern. The resistance patterns captured the majority of the most common MDR phenotypes and reflected previously identified pairwise relationships between AMR traits in S. aureus Associations between β-lactams and other antimicrobial classes (macrolides, lincosamides, and fluoroquinolones) were common, although the strength of the association among these antimicrobial classes varied by infection site and by methicillin susceptibility. Association mining identified associations between clinically important AMR traits, which could be further investigated for evidence of resistance coselection. For example, in skin and skin structure infections, clindamycin and tetracycline resistance occurred together 1.5 times more often than would be expected if they were independent from one another. Association mining efficiently discovered and quantified associations among resistance traits, allowing these associations to be compared between relevant subsets of isolates to identify and track clinically relevant MDR.

Spatiotemporal persistence of multiple, diverse clades and toxins of Corynebacterium diphtheriae

Diphtheria is a respiratory disease caused by the bacterium Corynebacterium diphtheriae. Although the development of a toxin-based vaccine in the 1930s has allowed a high level of control over the disease, cases have increased in recent years. Here, we describe the genomic variation of 502 C. diphtheriae isolates across 16 countries and territories over 122 years. We generate a core gene phylogeny and determine the presence of antimicrobial resistance genes and variation within the tox gene of 291 tox+ isolates. Numerous, highly diverse clusters of C. diphtheriae are observed across the phylogeny, each containing isolates from multiple countries, regions and time of isolation. The number of antimicrobial resistance genes, as well as the breadth of antibiotic resistance, is substantially greater in the last decade than ever before. We identified and analysed 18 tox gene variants, with mutations estimated to be of medium to high structural impact.

Prevalence and trend analysis of antimicrobial resistance in clinical Escherichia coli isolates collected from diseased pigs in the USA between 2006 and 2016

Antimicrobial resistance (AMR) is an emerging threat to both human and animal health. Antimicrobial use and resistance in food animal production, including swine, has received increased scrutiny as a source of resistant foodborne pathogens. Continuous surveillance of AMR in bacterial isolates of swine origin can guide in conservation of antimicrobials used in both human and swine medicine. The objective of this study was to evaluate the prevalence and trends of the phenotypic AMR in Escherichia coli of swine origin isolated from clinical samples at the Minnesota Veterinary Diagnostic laboratory between 2006 and 2016. The prevalence of resistance to ampicillin, tetracyclines and sulphadimethoxine remained greater than 50% throughout the period. There was a drastic change in enrofloxacin resistance, increasing from less than 1% to more than 20% between 2006 and 2016 (annual relative increase of 57% between 2006 and 2013 and 16% between 2013 and 2016). The prevalence of resistance to other antimicrobials remained constant (ceftiofur, oxytetracycline and chlortetracycline) or changed significantly (annual relative changes of less than 10%) for at least some time-period between 2006 and 2016 (ampicillin, florfenicol, gentamicin, neomycin, sulphadimethoxine, trimethoprim-sulphamethoxazole and spectinomycin). Rarefaction analysis revealed an increase in the number of unique combinations of AMRs per year. Network analysis was performed by estimating and plotting partial correlations between minimum inhibitory concentrations (MICs) of various antimicrobials. An increase in strength of these networks was observed, particularly in networks created after 2010, which can be indicative of increased multiple AMR in these isolates. These results provide valuable insight into the trends in AMR in E. coli of swine origin in the USA and act as supplementary information to the existing active AMR surveillance systems.

Shared Multidrug Resistance Patterns in Chicken-Associated Escherichia coli Identified by Association Rule Mining

Using multiple antimicrobials in food animals may incubate genetically-linked multidrug-resistance (MDR) in enteric bacteria, which can contaminate meat at slaughter. The U.S. National Antimicrobial Resistance Monitoring System tested 14,418 chicken-associated Escherichia coli between 2004 and 2012 for resistance to 15 antimicrobials, resulting in >32,000 possible MDR patterns. We analyzed MDR patterns in this dataset with association rule mining, also called market-basket analysis. The association rules were pruned with four quality measures resulting in a <1% false-discovery rate. MDR rules were more stable across consecutive years than between slaughter and retail. Rules were decomposed into networks with antimicrobials as nodes and rules as edges. A strong subnetwork of beta-lactam resistance existed in each year and the beta-lactam resistances also had strong associations with sulfisoxazole, gentamicin, streptomycin and tetracycline resistances. The association rules concur with previously identified E. coli resistance patterns but provide significant flexibility for studying MDR in large datasets.

Estimation of multidrug resistance variability in the National Antimicrobial Monitoring System

Multidrug resistance is a serious problem raising the specter of infections for which there is no treatment. One of the most important tools in combating multidrug resistance is large scale monitoring programs, because they track resistance over large geographic areas and time scales. This large scope, however, can also introduce variability into the data. The primary monitoring program in the United States is the National Antimicrobial Resistance Monitoring System (NARMS). This study examines the variability of a previously identified resistance pattern in Escherichia coli among ampicillin, gentamicin, sulfisoxazole, and tetracycline using samples isolated from chicken during the years 2004 to 2006 and 2008 to 2012. 2007 is excluded because sulfisozaxole resistance was not measured at slaughter that year. To assess variability in this resistance pattern susceptibility/resistance contingency tables were constructed for each of the 15 combinations of the 4 drugs for each of the years. For each table, variability across the years was assessed at the full table multinomial level as a measure of general variability of the resistance pattern and at the level of the highest order interaction term in a log-linear model of the table as a measure of variability in that particular component of the resistance pattern. A power analysis using the traditional asymptotic normal approximation and one using a Dirichlet-multinomial simulation were carried out to determine the effect of variation on ability to detect nonzero highest order loglinear model terms and the validity of the normal approximation in carrying out such tests. All tables exhibit overdispersion at the multinomial level and in their highest order model parameters. The normal approximation performs well for large sample sizes, low levels of dispersion, and small log-linear model parameters. The approximation breaks down as dispersion or the log linear model parameter grows or sample size shrinks. Taken together these analyses indicate that the level of variability in the NARMS dataset makes it difficult to detect multidrug resistance patterns at the current level of sample collection. In order to better control this dispersion NARMS could collect more variables on each of the samples.

Characterization of Two New Multidrug-Resistant Strains of Mycobacterium smegmatis: Tools for Routine In Vitro Screening of Novel Anti-Mycobacterial Agents

Mycobacterium tuberculosis is a pathogen of global public health concern. This threat is exacerbated by the emergence of multidrug-resistant and extremely-drug-resistant strains of the pathogen. We have obtained two distinct clones of multidrug-resistant Mycobacterium smegmatis after gradual exposure of Mycobacterium smegmatis mc² 155 to increasing concentrations of erythromycin. The resulting resistant strains of Mycobacterium smegmatis exhibited robust viability in the presence of high concentrations of erythromycin and were found to be resistant to a wide range of other antimicrobials. They also displayed a unique growth phenotype in comparison to the parental drug-susceptible Mycobacterium smegmatis mc² 155, and a distinct colony morphology in the presence of cholesterol. We propose that these two multidrug-resistant clones of Mycobacterium smegmatis could be used as model organisms at the inceptive phase of routine in vitro screening of novel antimicrobial agents targeted against multidrug-resistant Mycobacterial tuberculosis.

Phenotypical resistance correlation networks for 10 non-typhoidal Salmonella subpopulations in an active antimicrobial surveillance programme

Antimicrobials play a critical role in treating cases of invasive non-typhoidal salmonellosis (iNTS) and other diseases, but efficacy is hindered by resistant pathogens. Selection for phenotypical resistance may occur via several mechanisms. The current study aims to identify correlations that would allow indirect selection of increased resistance to ceftriaxone, ciprofloxacin and azithromycin to improve antimicrobial stewardship. These are medically important antibiotics for treating iNTS, but these resistances persist in non-Typhi Salmonella serotypes even though they are not licensed for use in US food animals. A set of 2875 Salmonella enterica isolates collected from animal sources by the National Antimicrobial Resistance Monitoring System were stratified in to 10 subpopulations based on serotype and host species. Collateral resistances in each subpopulation were estimated as network models of minimum inhibitory concentration partial correlations. Ceftriaxone sensitivity was correlated with other β-lactam resistances, and less commonly resistances to tetracycline, trimethoprim-sulfamethoxazole or kanamycin. Azithromycin resistance was frequently correlated with chloramphenicol resistance. Indirect selection for ciprofloxacin resistance via collateral selection appears unlikely. Density of the ACSSuT subgraph resistance aligned well with the phenotypical frequency. The current study identifies several important resistances in iNTS serotypes and further research is needed to identify the causative genetic correlations.

Establishing Statistical Equivalence of Data from Different Sampling Approaches for Assessment of Bacterial Phenotypic Antimicrobial Resistance

To assess phenotypic bacterial antimicrobial resistance (AMR) in different strata (e.g., host populations, environmental areas, manure, or sewage effluents) for epidemiological purposes, isolates of target bacteria can be obtained from a stratum using various sample types. Also, different sample processing methods can be applied. The MIC of each target antimicrobial drug for each isolate is measured. Statistical equivalence testing of the MIC data for the isolates allows evaluation of whether different sample types or sample processing methods yield equivalent estimates of the bacterial antimicrobial susceptibility in the stratum. We demonstrate this approach on the antimicrobial susceptibility estimates for (i) nontyphoidal Salmonella spp. from ground or trimmed meat versus cecal content samples of cattle in processing plants in 2013-2014 and (ii) nontyphoidal Salmonella spp. from urine, fecal, and blood human samples in 2015 (U.S. National Antimicrobial Resistance Monitoring System data). We found that the sample types for cattle yielded nonequivalent susceptibility estimates for several antimicrobial drug classes and thus may gauge distinct subpopulations of salmonellae. The quinolone and fluoroquinolone susceptibility estimates for nontyphoidal salmonellae from human blood are nonequivalent to those from urine or feces, conjecturally due to the fluoroquinolone (ciprofloxacin) use to treat infections caused by nontyphoidal salmonellae. We also demonstrate statistical equivalence testing for comparing sample processing methods for fecal samples (culturing one versus multiple aliquots per sample) to assess AMR in fecal Escherichia coli These methods yield equivalent results, except for tetracyclines. Importantly, statistical equivalence testing provides the MIC difference at which the data from two sample types or sample processing methods differ statistically. Data users (e.g., microbiologists and epidemiologists) may then interpret practical relevance of the difference.IMPORTANCE Bacterial antimicrobial resistance (AMR) needs to be assessed in different populations or strata for the purposes of surveillance and determination of the efficacy of interventions to halt AMR dissemination. To assess phenotypic antimicrobial susceptibility, isolates of target bacteria can be obtained from a stratum using different sample types or employing different sample processing methods in the laboratory. The MIC of each target antimicrobial drug for each of the isolates is measured, yielding the MIC distribution across the isolates from each sample type or sample processing method. We describe statistical equivalence testing for the MIC data for evaluating whether two sample types or sample processing methods yield equivalent estimates of the bacterial phenotypic antimicrobial susceptibility in the stratum. This includes estimating the MIC difference at which the data from the two approaches differ statistically. Data users (e.g., microbiologists, epidemiologists, and public health professionals) can then interpret whether that present difference is practically relevant.

Inferring the interaction structure of resistance to antimicrobials

The growth of antimicrobial resistance presents a significant threat to human and animal health. Of particular concern is multi-drug resistance, as this increases the chances an infection will be untreatable by any antibiotic. In order to understand multi-drug resistance, it is essential to understand the association between drug resistances. Pairwise associations characterize the connectivity between resistances and are useful in making decisions about courses of treatment, or the design of drug cocktails. Higher-order associations, interactions, which tie together groups of drugs can suggest commonalities in resistance mechanism and lead to their identification. To capture interactions, we apply log-linear models of contingency tables to analyze publically available data on the resistance of Escheresia coli isolated from chicken and turkey meat by the National Antimicrobial Resistance Monitoring System. Standard large sample and conditional exact testing approaches for assessing significance of parameters in these models breakdown due to structured patterns inherent to antimicrobial resistance. To address this, we adopt a Bayesian approach which reveals that E. coli resistance associations can be broken into two subnetworks. The first subnetwork is characterized by a hierarchy of β-lactams which is consistent across the chicken and turkey datasets. Tier one in this hierarchy is a near equivalency between amoxicillin-clavulanic acid, ceftriaxone and cefoxitin. Susceptibility to tier one then implies susceptibility to ceftiofur. The second subnetwork is characterized by more complex interactions between a variety of drug classes that vary between the chicken and turkey datasets.

Monte Carlo Simulations Suggest Current Chlortetracycline Drug-Residue Based Withdrawal Periods Would Not Control Antimicrobial Resistance Dissemination from Feedlot to Slaughterhouse

Antimicrobial use in beef cattle can increase antimicrobial resistance prevalence in their enteric bacteria, including potential pathogens such as Escherichia coli. These bacteria can contaminate animal products at slaughterhouses and cause food-borne illness, which can be difficult to treat if it is due to antimicrobial resistant bacteria. One potential intervention to reduce the dissemination of resistant bacteria from feedlot to consumer is to impose a withdrawal period after antimicrobial use, similar to the current withdrawal period designed to prevent drug residues in edible animal meat. We investigated tetracycline resistance in generic E. coli in the bovine large intestine during and after antimicrobial treatment by building a mathematical model of oral chlortetracycline pharmacokinetics-pharmacodynamics and E. coli population dynamics. We tracked three E. coli subpopulations (susceptible, intermediate, and resistant) during and after treatment with each of three United States chlortetracycline indications (liver abscess reduction, disease control, disease treatment). We compared the proportion of resistant E. coli before antimicrobial use to that at several time points after treatment and found a greater proportion of resistant enteric E. coli after the current withdrawal periods than prior to treatment. In order for the proportion of resistant E. coli in the median beef steer to return to the pre-treatment level, withdrawal periods of 15 days after liver abscess reduction dosing (70 mg daily), 31 days after disease control dosing (350 mg daily), and 36 days after disease treatment dosing (22 mg/kg bodyweight for 5 days) are required in this model. These antimicrobial resistance withdrawal periods would be substantially longer than the current U.S. withdrawals of 0–2 days or Canadian withdrawals of 5–10 days. One published field study found similar time periods necessary to reduce the proportion of resistant E. coli following chlortetracycline disease treatment to those suggested by this model, but additional carefully designed field studies are necessary to confirm the model results. This model is limited to biological processes within the cattle and does not include resistance selection in the feedlot environment or co-selection of chlortetracycline resistance following other antimicrobial use.

A proposed analytic framework for determining the impact of an antimicrobial resistance intervention

Antimicrobial use (AMU) is increasingly threatened by antimicrobial resistance (AMR). The FDA is implementing risk mitigation measures promoting prudent AMU in food animals. Their evaluation is crucial: the AMU/AMR relationship is complex; a suitable framework to analyze interventions is unavailable. Systems science analysis, depicting variables and their associations, would help integrate mathematics/epidemiology to evaluate the relationship. This would identify informative data and models to evaluate interventions. This National Institute for Mathematical and Biological Synthesis AMR Working Group's report proposes a system framework to address the methodological gap linking livestock AMU and AMR in foodborne bacteria. It could evaluate how AMU (and interventions) impact AMR. We will evaluate pharmacokinetic/dynamic modeling techniques for projecting AMR selection pressure on enteric bacteria. We study two methods to model phenotypic AMR changes in bacteria in the food supply and evolutionary genotypic analyses determining molecular changes in phenotypic AMR. Systems science analysis integrates the methods, showing how resistance in the food supply is explained by AMU and concurrent factors influencing the whole system. This process is updated with data and techniques to improve prediction and inform improvements for AMU/AMR surveillance. Our proposed framework reflects both the AMR system's complexity, and desire for simple, reliable conclusions.