David A. Kennedy

Comparative genomic studies of Marek’s disease virus (MDV) conducted in the past two decades have proven powerful for identifying genomic loci driving historical shifts in virulence. However, past efforts were limited to comparing a few strains at a time, lacked access to certain regions of the MDV genome, and were restricted in their ability to characterize viral phenotypes. To address these limitations, we performed whole-genome sequencing on a collection of 65 MDV strains previously characterized using a standardized phenotyping assay. In addition to phylogenetic and recombination analyses, this approach enabled us to perform a genome-wide association study of virulence for this pathogen. In total, we identified 10 genome-wide significant loci associated with virulence, including a tandem repeat variant resulting from two alternative versions of a 132-bp repeating motif. Our findings support prior descriptions of virulence as a complex trait in MDV, and highlight the potential contribution of repeat-based and intergenic variants to the phenotypic diversity of herpesviruses.

Host jumps—defined as the process by which a pathogen establishes sustained transmission in novel hosts—are threats to human and animal welfare, but anticipating which pathogen will be the next to successfully host jump remains elusive. A spillover event must precede a host jump, and so spillover rate is thought to be related to risk. However, nonendemic pathogens that spill over frequently have demonstrated a poor ability to host jump from any given spillover. So which is riskier, pathogens that spill over rarely or commonly? Applying a Bayesian framework to a general model of host jump risk, we show that 1) the riskiest pathogens can be those that spill over at low, intermediate, or high rates, and 2) as the rate of spillover gets large, the information gained from past spillovers is exactly counterbalanced by the increased number of future spillovers. Taken together, this means that spillover rate has little to no value in explaining host jump risk. Rather, we show that novel pathogens (i.e., pathogens with a relatively short history of spilling over in their current form) are substantially more likely to result in host jumps than pathogens that have had long-associated opportunities for spillover into the novel host. Notably, a pathogen might be thought of as novel if spillover only recently became possible, or if it recently underwent substantial evolutionary change. We therefore propose that the length of historical association, but not spillover rate, will be an important predictor of host jump risk.

Monoclonal antibody therapies are being developed to treat measles in response to its recent resurgence. These therapies risk driving measles virus evolution in ways that might undermine the protection offered by vaccination, outweighing potential benefits.

Spillover of viruses into novel host species occurs frequently. Often, spillover results in dead-end infections in novel hosts, sometimes, in stuttering transmission chains that die out, and rarely, in large epidemics with sustained transmission. If we could identify early which outcome will occur following a spillover event, we could more appropriately invest in efforts to surveil, respond to, or prevent disease emergence. Our goal was to identify early epidemiological characteristics that correlate with these outcomes, including those predictive of population-level virus persistence in novel hosts. To identify these characteristics, we experimentally induced spillover in the Caenorhabditis nematode-Orsay virus system and measured infection prevalence in exposed populations and virus shedding and infection intensity from infected hosts in replicate populations of eight strains belonging to seven non-native host species. We then passaged 20 adult nematodes from exposed populations to virus-free plates where they reproduced, initiating new populations to which they had the potential to transmit virus. We used quantitative PCR to track virus presence in passaged host populations for 10 passages or until virus was undetectable, indicating its loss. We then used a correlative modeling and a mechanistic modeling approach to understand which epidemiological characteristics were associated with population-level viral persistence. In our correlative models, we found that the number of passages until virus loss was associated with early epidemiological characteristics in the spillover host populations, including infection prevalence in the initially exposed population, the ability of hosts to detectably shed the virus, and the relative susceptibility of the host species, but not infection intensity. When all these characteristics were included simultaneously in a correlative model, only infection prevalence and shedding were significantly associated with virus maintenance, and the model explained over half of the variation in the data. We then developed a mechanistic model that attempts to explain virus passage success by using our epidemiological characteristics data to calculate the probability that at least one worm infectious enough to infect a conspecific is transferred during passage. This mechanistic model explained 38% of the variation in the data on its own. With the goal of understanding how our mechanistic model falls short, we used model selection to test a suite of larger models that included or excluded each epidemiological characteristic and included random effects of strain, experimental line, passage number, and block while the mechanistic prediction was included as an offset. We found that 66% of the variation in our data could be explained by a model that included our mechanistic prediction in addition to infection prevalence, infection intensity, and random effects. Altogether, our study demonstrates that early epidemiological characteristics can play a substantial role in explaining the ultimate outcome of a spillover event.

While it is well established that host heterogeneity in transmission and host heterogeneity in susceptibility each individually impact disease dynamics in characteristic ways, it is generally unknown how disease dynamics are impacted when both types of heterogeneity are simultaneously present. Here we explore this question. We first conducted a systematic review of published studies from which we determined that the effects of correlations have been drastically understudied. We then filled in the knowledge gaps by developing and analyzing a stochastic, individual-based SIR model that includes both heterogeneity in transmission and susceptibility and flexibly allows for positive or negative correlations between transmissibility and susceptibility. We found that in comparison to the uncorrelated case, positive correlations result in major epidemics that are larger, faster, and more likely, whereas negative correlations result in major epidemics that are smaller and less likely. We additionally found that, counter to the conventional wisdom that heterogeneity in susceptibility always reduces outbreak size, heterogeneity in susceptibility can lead to major epidemics that are larger and more likely than the homogeneous case when correlations between transmissibility and susceptibility are positive, but this effect only arises at small to moderate R ₀ . Moreover, positive correlations can frequently lead to major epidemics even with subcritical R ₀ . To illustrate the potential importance of heterogeneity and correlations, we developed an SEIR model to describe mpox disease dynamics in New York City, demonstrating that the dynamics of a 2022 outbreak can be reasonably well explained by the presence of positive correlations between susceptibility and transmissibility. Ultimately, we show that correlations between transmissibility and susceptibility profoundly impact disease dynamics. Highlights: Systematic review finds that effects of correlations on epidemics are understudied Positive correlations lead to larger, faster, more likely epidemics Negative correlations lead to smaller, less likely epidemics Positive correlations allow for major epidemics even with subcritical R ₀ Mpox outbreak dynamics in New York City were consistent with positive correlations

The presence of heterogeneity in susceptibility, differences between hosts in their likelihood of becoming infected, can fundamentally alter disease dynamics and public health responses, for example, by changing the final epidemic size, the duration of an epidemic, and even the vaccination threshold required to achieve herd immunity. Yet, heterogeneity in susceptibility is notoriously difficult to detect and measure, especially early in an epidemic. Here we develop a method that can be used to detect and estimate heterogeneity in susceptibility given contact by using contact tracing data, which are typically collected early in the course of an outbreak. This approach provides the capability, given sufficient data, to estimate and account for the effects of this heterogeneity before they become apparent during an epidemic. It additionally provides the capability to analyze the wealth of contact tracing data available for previous epidemics and estimate heterogeneity in susceptibility for disease systems in which it has never been estimated previously. The premise of our approach is that highly susceptible individuals become infected more often than less susceptible individuals, and so individuals not infected after appearing in contact networks should be less susceptible than average. This change in susceptibility can be detected and quantified when individuals show up in a second contact network after not being infected in the first. To develop our method, we simulated contact tracing data from artificial populations with known levels of heterogeneity in susceptibility according to underlying discrete or continuous distributions of susceptibilities. We analyzed these data to determine the parameter space under which we are able to detect heterogeneity and the accuracy with which we are able to estimate it. We found that our power to detect heterogeneity increases with larger sample sizes, greater heterogeneity, and intermediate fractions of contacts becoming infected in the discrete case or greater fractions of contacts becoming infected in the continuous case. We also found that we are able to reliably estimate heterogeneity and disease dynamics. Ultimately, this means that contact tracing data alone are sufficient to detect and quantify heterogeneity in susceptibility.

Current strategies to understand the molecular basis of Marek’s disease virus (MDV) virulence primarily consist of cataloging divergent nucleotides between strains with different phenotypes. However, most comparative genomic studies of MDV rely on previously published consensus genomes despite the confirmed existence of MDV strains as mixed viral populations. To assess the reliability of interstrain genomic comparisons relying on published consensus genomes of MDV, we obtained two additional consensus genomes of vaccine strain CVI988 (Rispens) and two additional consensus genomes of the very virulent strain Md5 by sequencing viral stocks and cultured field isolates. In conjunction with the published genomes of CVI988 and Md5, this allowed us to perform three-way comparisons between multiple consensus genomes of the same strain. We found that consensus genomes of CVI988 can vary in as many as 236 positions involving 13 open reading frames (ORFs). By contrast, we found that Md5 genomes varied only in 11 positions involving a single ORF. Notably, we were able to identify 3 single-nucleotide polymorphisms (SNPs) in the unique long region and 16 SNPs in the unique short (US) region of CVI988_GenBank.BAC that were not present in either CVI988_{Pirbright.lab} or CVI988_{USDA.PA.field}. Recombination analyses of field strains previously described as natural recombinants of CVI988 yielded no evidence of crossover events in the US region when either CVI988_{Pirbright.lab} or CVI988_{USDA.PA.field} were used to represent CVI988 instead of CVI988_GenBank.BAC. We were also able to confirm that both CVI988 and Md5 populations were mixed, exhibiting a total of 29 and 27 high-confidence minor variant positions, respectively. However, we did not find any evidence of minor variants in the positions corresponding to the 19 SNPs in the unique regions of CVI988_GenBank.BAC. Taken together, our findings suggest that continued reliance on the same published consensus genome of CVI988 may have led to an overestimation of genomic divergence between CVI988 and virulent strains and that multiple consensus genomes per strain may be necessary to ensure the accuracy of interstrain genomic comparisons.

The use of antibiotics to treat bacterial infections often imposes strong selection for antibiotic resistance. However, the prevalence of antibiotic resistance varies greatly across different combinations of pathogens and drugs. What underlies this variation? Systematic reviews, meta-analyses, and literature surveys capable of integrating data across many studies have tried to answer this question, but the vast majority of these studies have focused only on cases where resistance is common or problematic. Yet much could presumably be learned from the cases where resistance is infrequent or absent. Here we conducted a literature survey and a systematic review to study the evolution of antibiotic resistance across a wide range of pathogen-by-drug combinations (57 pathogens and 53 antibiotics from 15 drug classes). Using Akaike information criterion-based model selection and model-averaged parameter estimation we explored 14 different factors posited to be associated with resistance evolution. We find that the most robust predictors of high resistance are nosocomial transmission (i.e., hospital-acquired pathogens) and indirect transmission (e.g., vector-, water-, air-, or vehicle-borne pathogens). While the former was to be expected based on prior studies, the positive correlation between high resistance frequencies and indirect transmission is, to our knowledge, a novel insight. The most robust predictor of low resistance is zoonosis from wild animal reservoirs. We also found partial support that resistance was associated with pathogen type, horizontal gene transfer, commensalism, and human-to-human transmission. We did not find support for correlations between resistance and environmental reservoirs, mechanisms of drug action, and global drug use. This work explores the relative explanatory power of various pathogen and drug factors on resistance evolution, which is necessary to identify priority targets of stewardship efforts to slow the spread of drug-resistant pathogens.

The presence of heterogeneity in susceptibility, differences between hosts in their likelihood of becoming infected, can fundamentally alter disease dynamics and public health responses, for example, by changing the final epidemic size, the duration of an epidemic, and even the vaccination threshold required to achieve herd immunity. Yet, heterogeneity in susceptibility is notoriously difficult to detect and measure, especially early in an epidemic. Here we develop a method that can be used to detect and estimate heterogeneity in susceptibility given contact by using contact tracing data, which is typically collected early in the course of an outbreak. This approach provides the capability, given sufficient data, to estimate and account for the effects of this heterogeneity before they become apparent during an epidemic. It additionally provides the capability to analyze the wealth of contact tracing data available for previous epidemics and estimate heterogeneity in susceptibility for disease systems in which it has never been estimated previously. The premise of our approach is that highly susceptible individuals become infected more often than less susceptible individuals, and so individuals not infected after appearing in contact networks should be less susceptible than average. This change in susceptibility can be detected and quantified when individuals show up in a second contact network after not being infected in the first. To develop our method, we simulated contact tracing data from artificial populations with known levels of heterogeneity in susceptibility according to underlying discrete or continuous distributions of susceptibilities. We analyzed this data to determine the parameter space under which we are able to detect heterogeneity and the accuracy with which we are able to estimate it. We found that our power to detect heterogeneity increases with larger sample sizes, greater heterogeneity, and intermediate fractions of contacts becoming infected in the discrete case or greater fractions of contacts becoming infected in the continuous case. We also found that we are able to reliably estimate heterogeneity and disease dynamics. Ultimately, this means that contact tracing data alone is sufficient to detect and quantify heterogeneity in susceptibility.