Item No. 1 of 1

INVESTIGATOR: Mihaljevic, J.

PERFORMING INSTITUTION:
UNIVERSITY OF CHICAGO
5801 SOUTH ELLIS AVE.
CHICAGO, ILLINOIS 60637

USING EPIDEMIOLOGICAL MODELING TO IMPROVE MICROBIAL CONTROL OF THE DOUGLAS-FIR TUSSOCK MOTH

NON-TECHNICAL SUMMARY: The Douglas-fir tussock moth (DFTM), Orgyia pseudotsugata, causes severe defoliation and economic loss to multiple fir species across western North America during cyclical outbreaks. These outbreaks are typically terminated by epizootics of the naturally circulating baculovirus, OpNPV. Due to the host-specificity and high virulence of OpNPV, the USDA Forest Service uses the virus as a microbial pesticide, TM Biocontrol-1 (TMB). Recent field trials with TMB have shown that, although spraying TMB sometimes leads to reductions in defoliation, overall larval mortality rates often do not differ from unsprayed, control plots that become infected with naturally circulating OpNPV. This result is perplexing because in plots infested by naturally circulating OpNPV, initial infection rates are much lower than in TMB-sprayed plots.As my primary objective, and in the hopes of improving future management efforts, I will use epidemiological modeling, complemented by field and laboratory experiments, to determine if there are ecological differences between TMB and naturally circulating OpNPV strains that could account for previous management outcomes. As a secondary, broader project objective, I will seek to understand how future climate warming may influence the efficacy of TMB compared to naturally circulating OpNPV strains. I will conduct laboratory experiments to determine how TMB and two wild-type strains of OpNPV perform at different temperatures. I will then use my models to make predictions about how a changing climate may affect epizootic dynamics and microbial control in the DFTM system.

OBJECTIVES: The Douglas-fir tussock moth (DFTM), Orgyia pseudotsugata, causes severe defoliation and economic loss to multiple fir species across western North America during cyclical outbreaks. These outbreaks are typically terminated by epizootics of the naturally circulating baculovirus, OpNPV. Due to the host-specificity and high virulence of OpNPV, the USDA Forest Service uses the virus as a microbial pesticide, TM Biocontrol-1 (TMB). Recent field trials with TMB have shown that, although spraying TMB sometimes leads to reductions in defoliation, overall larval mortality rates often do not differ from unsprayed, control plots that become infected with naturally circulating OpNPV. This result is perplexing because in plots infested by naturally circulating OpNPV, initial infection rates are much lower than in TMB-sprayed plots. The over-arching goal of this project is to use rigorous quantitative methods to differentiate between mechanisms that could cause these surprising management results.Objective 1: As my primary objective, and in the hopes of improving future management efforts, I will use epidemiological modeling, complemented by field and laboratory experiments, to determine if there are ecological differences between TMB and naturally circulating OpNPV strains that could account for previous management outcomes.Objective 2: As a secondary, broader project objective, I will seek to understand how future climate warming may influence the efficacy of TMB compared to naturally circulating OpNPV strains. I will conduct laboratory experiments to determine how TMB and two wild-type strains of OpNPV perform at different temperatures. I will then use my models to make predictions about how a changing climate may affect epizootic dynamics and microbial control in the DFTM system.

APPROACH: Methods for Objective 1: The first project objective is to determine if the results of previous management efforts with TMB were due simply to epizootic dynamics that differ in their initial conditions or were due instead to true ecological differences between TMB and the naturally circulating OpNPV strain. To do this I will first fit a mechanistic disease transmission model to epizootic data collected during initial field trials, including data from both TMB-induced epizootics and naturally circulating OpNPV-induced epizootics.Using mechanistic models to analyze epizootic data is an increasingly common approach in disease ecology, because mechanistic models allow for deeper insights into transmission mechanisms than do more conventional statistical models.Modeling Framework: The biology of DFTM-OpNPV interactions can be represented by susceptible-exposed-infected-recovered (SEIR) epidemiological models, which have been used to accurately represent epizootics in other baculovirus-insect systems. In the OpNPV system, neonate larvae become infected through environmental contamination by accidentally ingesting infectious propagules (occlusion bodies) present on egg masses. Occlusion bodies are released within 1-3 weeks post-exposure from the resulting cadavers, which can lead to devastating epizootics. The simplicity of these infection dynamics allows them to be characterized by relatively simple models.Fitting the Model: I will fit the model to currently available infection rate data from TMB and naturally circulating OpNPV-induced epizootics. To do this, I will estimate model parameters using Bayesian MCMC techniques.One of the main benefits of using Bayesian probability is that it makes it easy to incorporate multiple data sources to estimate model fit and parameter uncertainty. This is important in our case because previous work in my primary mentor's lab has shown that epizootic data from the field are often insufficient to make robust statistical inferences about the mechanisms underlying baculovirus transmission. In particular, to infer potential differences between naturally circulating OpNPV and TMB from the available data, I may need to estimate some model parameters from experimental data. In the modeling literature, this is often done simply by using point estimates of the parameters in the model, and then qualitatively comparing the resulting model prediction to the infection rate data. By using Bayesian statistics, I can instead carry out this approach in a more robust fashion, by allowing for uncertainty in both the model parameters and in the fit of the model to the epizootic data.As a first attempt at model fitting, I will set uninformative priors for all parameters. This will allow the epizootic data to dominate the posterior so that there is no a priori assumption of differences between virus types. Based on the parameters' posterior distributions derived from this model fitting routine, I will determine whether any of the four key parameter values differ significantly between TMB and naturally circulating OpNPV epizootics.Evaluation: If the resulting parameter estimates are clearly different, I will conclude that there are clear ecological differences in virus types. However, given the extensive prior experience in my mentor's lab, it is very likely that the model fitting routine will have difficulty converging on precise estimates for all model parameters based solely on epizootic data. I am thus planning to use experiments to construct informative priors on key model parameters and then I will again carry out the model fitting routine.I will conduct three experiments to complement the model fitting procedure described above. First, I will conduct a field-transmission experiment. This experiment will help determine whether viral transmission from infected cadavers to susceptible larvae differs between the three virus types and how this transmission depends on the initial density of virus present. I will then conduct another field-transmission experiment to measure the decay rate of each viral strain. Finally, I will conduct a dose-response experiment to estimate the kill speed parameters, and the viral output of the three strains.These three experiments will allow me to better estimate parameters in the mathematical model. Having these better parameter estimates will, in turn, allow me to differentiate between hypotheses that could explain previous management outcomes. Methods for Objective 2: In the second objective, I will seek to understand how future climate warming may influence the efficacy of future DFTM microbial control programs. Specifically, OpNPV exhibits a latitudinal cline in western North America, such that OpSNPV is not observed much farther north than Washington state, while OpMNPV is not nearly as prevalent south of British Columbia. The two strains may therefore have different climatic tolerances, with OpMNPV less tolerant of warmer climates. I will test this hypothesis in a laboratory dose-response experiment, nested within the previously described experiment.This dose-response experiment will be conducted at the same time as, and nested within, the experiment described above. Specifically, I will extend the experiment to include two warmer temperatures at which I will incubate infected larvae. I will thus incubate larvae at three different temperatures in total: 18C, 22C, and 26C. The first temperature will be used for the model fitting routines in Objective 1, while the latter temperatures will be used to explore potential effects of climate warming.I will use the data from this experiment in my models to predict the effects of climate warming under various initial conditions. Experimentally estimating the effect of temperature on all model parameters is beyond what I can accomplish in the time available, but these data will be a first step towards understanding the potential effects of climate change on epizootics in the DFTM system. Additionally, because TMB was derived from an OpMNPV strain, it is imperative to know if future warming might impede the efficacy of the pesticide, and this experiment will be a first step to such an understanding. If there are strong differences between the two strains in their responses to temperature in terms of speed of kill, I will carry out a transmission experiment in the greenhouse at different temperatures, to test for effects on transmission. This will allow me to further extend the model as necessary. I will then predict the effects of climate change by combining my parameterized model with the down-scaled predictions of existing climate change models.

PROGRESS: 2014/09 TO 2016/12
Target Audience:This research is assisting in the microbial control program of the Douglas-fir tussock moth, which benefits research personell at the US Forest Service's Pacific Northwest Research Station in Wenatchee, WA. Additionally, we plan on submitting a substantial manuscript, imminently, which deals with epidemiological theory that will be of broad interest to academic ecologists. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?Working with my two advisors on this project has opened many doors professionally. First, I am currently a Postdoctoral Researcher at the University of California, Santa Barbara, where I am adapting the computational tools developed for my NIFA project to a system of amphibian fungal disease, which is a serious conservation concern. Second, I have been offered and have accepted a tenure-track, Assistant Professor position at Northern Arizona University in the School of Infromatics, Computing, and Cyber Systems. My training as a NIFA fellow was vital for securing this position, particularly because of the skills I learned in applying epidemiological theory to interesting biological questions. How have the results been disseminated to communities of interest?We are currently drafting a manuscript of our project, which will be published in a broadly-read scientific journal. This will have impacts for both applied and academic ecologists. I presented this work as a poster at the Ecological Society of America and at the NIFA fellowship conference with other fellows. I have also discussed this project in seminars at Northern Illinois University, University of Illinois - Urbana/Champaign, and at Northern Arizona University. What do you plan to do during the next reporting period to accomplish the goals? Nothing Reported

IMPACT: 2014/09 TO 2016/12
What was accomplished under these goals? My advisors and I have made substantial progress towards meeting our outlined objectives, and we will continue to collaborate on this project as I move forward in my academic career. In year 1, we completed a large field experiment, which allowed us to estimate the average transmission rates (and the variance in these rates) of 4 unique viral isolates. We used TEM to identify these isolates and discovered 2 were OpMNPV, and 2 were OpSNPV. Since then we have collected more isolates. In year 2, after collecting more OpNPV isolates, we conducted a lab experiment to estimate each isolate's average rate of kill and within-host viral production. These latter data are still be analyzed. In years 1 and 2, we also developed a computationally advanced algorithm to fit a stochastic SEIR (epidemiological) model to historical data of OpNPV epizootics in DFTM populations. These data are from 12 populations, approximately half of which represent epizootics derived from "natural" OpNPV isolates, while the other half derived from experimental spray treatments with TMB. Comparing these two population types allowed us to assess the differences in epizootic dynamics between natural and sprayed virus. We have used our model in two ways that advance our understanding of DFTM-OpNPV dynamics and our integration of epidemiological theory and computation generally. First, our preliminary results suggest that, while average transmission rates between OpSNPV and OpMNPV may be measurably different in small field experiments, at the scale of forest-wide epizootics, our fits to available data do not show enough evidence to suggest these differences strongly alter epizootic dynamics. In other words, the differences between TMB and naturally occuring strains are not strong enough to impact forest-wide epizootic dynamics. Instead, we find that the initial DFTM density and the intial fraction of the DFTM population that is infected are strongly predictive of epizootic dynamics. Second, we use this interesting biological question to show how experimental data and observational data can be combined in a Bayesian framework to fit complex, dynamical models to data. By first fitting the model to observational data naively, and then by fitting the model to the observational data, using the experimental data to construct informative priors for some parameters, we can assess the impact of conducting focused experiments on our ability to model epizootic dynamics across the forest. This also allows us to comment on how well experiments, which are conducted at relatively small spatial scales, can capture the dynamics that impact large-scale phenomena (e.g. epizootics across forests). Preliminary results suggest that our experiments do capture these dynamics well and that conducting these types of experiment improves model performance.

PUBLICATIONS (not previously reported): 2014/09 TO 2016/12
No publications reported this period.

PROGRESS: 2016/09/01 TO 2016/12/31
Target Audience:This research is assisting in the microbial control program of the Douglas-fir tussock moth, which benefits research personell at the US Forest Service's Pacific Northwest Research Station in Wenatchee, WA. Additionally, we plan on submitting manuscripts dealing with epidemiological theory that will be of broad interest to academic ecologists. Finally, we trained one undergraduate student on this project, and I wrote a blog about my experiences in the field during the first reporting period. Changes/Problems:We were not able to conduct the planned experiment that alters temperature, in order to understand the potential effects of climate change. While we tried to rear enough insects in the lab by conducting lab matings, most of the egg masses did not produce offspring. We therefore scaled down our experiment to conduct the dose-response experiment described in this report. What opportunities for training and professional development has the project provided?I have continued to hone my quantitative skills during this project period, learning more advanced epidemiological theory and computational techniques. I am also currently applying for academic positions with much advice from both of my advisors. In addition, I traveled to Washington DC to present my work at the NIFA Postdoctoral Fellow's conference, where I networked with other fellows. Finally, this project gave me the opportunity to expand my mentoring experiences by supervising an undergraduate technician, who conducted the dose-response experiment. How have the results been disseminated to communities of interest?I presented my work in poster format at the Ecological Society of America's annual meeting in Fort Lauderdale, and I also presented a poster at the NIFA PI conferece. I was also invited to speak about my past and current research at the University of Illinois (Urbana/Champaign) and Northern Illinois University. In these talks, I highlighted the history of Douglas-fir tussock moth biological control, and I described how our epidemiological models might help inform current and future control measures. I also (informally) blogged about my field work related to this project, which was posted to NumbatMedia at this address: http://www.numbatmedia.com/stories/2016/4/5/mihaljevic?platform=hootsuite. What do you plan to do during the next reporting period to accomplish the goals?Over the last few months of the project, I will wrap up the modeling and simulation tasks, and prepare this work for publication in a high-caliber ecological journal.

IMPACT: 2016/09/01 TO 2016/12/31
What was accomplished under these goals? During this reporting period, we used historic and recent data on natural and spray-induced epizootics in order to parameterize our mechanistic model. Our results suggest that no strong ecological differences exist between naturally circulating strains and TMB-1 (e.g. in terms of transmission rates). We are now using this parameterized model in a spatially explicit simulation that will allow us to understand how different intial conditions (initial larval densities and initial viral densities) affect tree defoiation and mortality across a Douglas fir forest. During this period, we also conducted a dose-response experiment with eight viral isolates, four of the MNPV type and four of the SNPV type. In this experiment we measured the time to death of individuals given a specific dose, the proportion of individuals infected with each dose, and we are also counting the viral output in each infected individual. These data will allow us to further understand ecological differences between the two morphotypes, and we will use epidemiological models to understand how these differences might affect biological control. This project has mainly been conducted by a University of Chicago undergraduate student, under my close supervision.

PUBLICATIONS: 2016/09/01 TO 2016/12/31
No publications reported this period.

PROGRESS: 2015/09/01 TO 2016/08/31
Target Audience:This research is assisting in the microbial control program of the Douglas-fir tussock moth, which benefits research personell at the US Forest Service's Pacific Northwest Research Station in Wenatchee, WA. Additionally, we plan on submitting manuscripts dealing with epidemiological theory that will be of broad interest to academic ecologists. Changes/Problems:We were unable to collect as many feral egg masses from our colleagues across the Forest Service as we would have liked this year. Specifically, in the first year, we had collected DFTM egg masses from an outbreak occuring in Colorado. However, during the following summer, anepizootic of OpNPV killed most of the naturally occuring larvae in that area of Colorado, leaving us with few egg masses at the end of the season to use for this year's planned experiments. At the end of the first year, we were able to conduct matings of DFTM adults reared in the laboratory; however, it is well known that these lab matings have low success rates, and larvae that hatch from lab-mated egg masses tend to have developmental problems. Thus, we are potentially limited in our abilities to conduct as large of experiments as we planned. We are working on accomplishing everything we can with the larvae we have available. What opportunities for training and professional development has the project provided?I have continued to hone my quantitative skills during this project period, learning more advanced epidemiological theory and computational techniques. I have sat in on my advisor, Dr. Dwyer's, course at University of Chicago on the Ecology and Evolution of Infectious Diseases in order to observe his teaching style and the content of his course. Also, from Dr. Dwyer and Dr. Polivka, I have been learning about applying to academic positions (e.g. tenure-track faculty) and government jobs. I was invited for an on-campus interview for a tenure-track positionat Emory University, which was a great learning experience, although I was not offered the position. How have the results been disseminated to communities of interest?I gave an invited guest lecture to undergraduate students and faculty in the Biology Department at Northeastern Illinois University in which I detailed the goals and preliminary outcomes of this project. I am also still closely working with Dr. Polivka and his colleagues at the Forest Service. What do you plan to do during the next reporting period to accomplish the goals?We have requested a no-cost extension of this project until January 2017. This is because I did not originally start the project until January 2015, after I received my PhD diploma in December 2014. From now until January 2017, I plan to write and submit at least 1 publication pertaining to the work I have accomplished during my tenure with this fellowship. I will be attempting an additional field experiment over the next two months. Additionally, I will be presenting my progress at the NIFA Fellows conference in August of this year, and I will be presenting my results to a broader audience at the Ecological Society of America's annual meeting in Florida, also in August of this year.

IMPACT: 2015/09/01 TO 2016/08/31
What was accomplished under these goals? We have almost completed the entirety of Objective 1, and we have begun writing a publication in this regard. Additionally, we gathered preliminary evidence that the two morpho-types of virus, SNPV and MNPV, may have different transmission dynamics. We also determined via transmission electron microscopythat, from our sample of 8 virus isolates, exactly 4 are SNPV and 4 are MNPV. Based on this finding, we are imminently planning another field experiment to estimate the transmission rates and transmission heterogenities (two important epidemiological parameters) from all 8 virus isolates. I will be leaving in a matter of days to Wentchee, WA to begin this work. Finally, I have begun working on a web application that would allow users to simulate possible outcomes of spraying TMB-1 in terms of the size of the epizootic. This simulation would allow the user to specify different initial conditions, including initial larval densities and the amount of virus to be sprayed. This could help with making the decision whether or not to spray.

PUBLICATIONS: 2015/09/01 TO 2016/08/31
No publications reported this period.

PROGRESS: 2014/09/01 TO 2015/08/31
Target Audience:This research is assisting in the microbial control program of the Douglas-fir tussock moth, which benefits research personell at the US Forest Service's Pacific Northwest Research Station in Wenatchee, WA. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?During the year, I have gained a much deeper knowledge of epidemiological modeling techniques, including fitting models to data. I have also gained hands-on experience in the Douglas-fir tussock moth-NPV system. I conducted a complex field experiment, which involved rearing insects and working with multiple NPV isolates. I am currently running follow-up statistical and laboratory analyses that will also broaden my expertise. How have the results been disseminated to communities of interest?I am working closely with my co-advisor, Dr. Karl Polivka, at the USFS Pacific Northwest Research Station in Wenatchee, WA and with Forest Health Protection Program personell. We are in regular conversations so that the data and models are used in a way that directly assists the Forest Service with their tussock moth microbial control program. What do you plan to do during the next reporting period to accomplish the goals?During the next reporting period, I will finish fitting the epidemiological model to field data from the USFS microbial control field trials. I will then answer the following questions related to the microbial control program using these fitted models. First, can we better predict the occurrence of natural epizootics, given data on initial moth population sizes and infection rates? Knowing whether natural epizootics will crash the moth population without the need for the microbial pesticides will allow us to deploy the microbial agent in a more economically efficient way. Second, do ecological differences between natural OpNPV and the microbial pesticide, TMB, account for differences in epizootic patterns in sprayed and control plots? Answering this question will help us understand whether TMB is still the best possible OpNPV isolate to use in the microbial control program. We will also conduct a lab experiment in February 2016 to accomplish our second objective. This lab experiment will help us understand how climate change may impact the performance of the microbial pesticide TMB as well as naturally occurring OpNPV isolates. We have started a lab colony of moths to supply us with the specimen needed for this experiment.

IMPACT: 2014/09/01 TO 2015/08/31
What was accomplished under these goals? I began working on this project on January 1, 2015, shortly after graduating from my PhD program. My collaborators and I successfully conducted our proposed field experiment from May-August 2015, in which we measured the transmission and decay rates of three naturally occurring OpNPV and the microbial pesticide TMB. Recently I analyzed these data, which show that the viruses have statistically different mean and variation in transmission rates, as well as some differences in decay rates. I have preliminarily constructed prior distributions from these data in order to be used in our epidemiological model-fitting algorithm. The algorithm is running now, which will eventually complete Objective 1. Objective 2 will also be completed in the second year of the project.

PUBLICATIONS: 2014/09/01 TO 2015/08/31
No publications reported this period.