The Orsay Virus as a model for population-wide viral infection dynamics

To this day, epidemics pose a considerable threat to mankind. Experimental models that simulate the spread of infectious diseases are thus crucial to the inception of effective control policies. Current models have had great success incorporating virulence and host immune response but do rarely take host genetics, behavior and host environment into account. Here, we present a full-scale imaging setup that utilizes the infection of the nematode C. elegans with a positive-stranded RNA virus (Orsay Virus) to probe key epidemiological parameters and simulate the spread of infection in a whole population. We demonstrate that our system is able to quantify infection levels and host behavior at a high sampling rate and show that different host genetic backgrounds can influence viral spread, while also highlighting the influence of infection on various host behaviors. Future work will allow the isolation of key behavioral and environmental factors that affect viral spread, potentially enabling novel policies to combat the spread of viral infections. Significance Statement In the ongoing COVID-19 pandemic, we struggle to find effective control policies that “stop the spread”. While current animal models of virus spread in populations are highly sophisticated, they rarely explore effects of host behavior and its environment. We developed an experimental animal model system that allows us to visualize virus transmission in whole populations of C. elegans while also measuring behaviors. We were able to demonstrate how C. elegans genetics influences the progression of viral infection in a population and how animals adjust their behavior when infected. In the future, we envision that animal model systems like ours are used to test the effects of viral control policies on viral spread before they are applied in real world scenarios.


Introduction
Throughout history, humankind has faced threats from infectious diseases, and human populations have been repeatedly shaped by pandemics. Yersinia pestis (the "black death") wiped out 30-60% of Europe's population in the 14th century, and more recently, 3-5% of the world's population died as the result of influenza virus infections in 1918 (1).
Today, the spread of viral and bacterial pathogens is ever more amplified due to the world's unprecedented interconnectedness as seen in the recent COVID-19 outbreak (2)(3)(4)(5). Hence, it becomes ever more important to predict and control such deadly outbreaks.
Epidemics are driven by the transmission of a pathogen from infected hosts or carriers to susceptible ones, leading to the pathogen's spread in the population. The dynamics of these transmissions are often difficult to predict, as they result from complex and interdependent interactions between biological processes in both the pathogen and the host, the environment, and the behavior of the host population (6)(7)(8)(9)(10)(11)(12)(13). Therefore, experimental animal models are frequently used to simulate pathogen dynamics and to facilitate the design of control policies. Laboratory mouse populations, for example, have been previously used for transmission experiments of a bacterial pathogen to directly measure the bacterial reproductive ratio through blood and fecal sampling (14,15). In other studies, infections that lead to host death were used to measure infection levels via the host mortality rate (16,17). A more direct and less invasive visualization of infection can be achieved by using transparent hosts. For example, microparasite infections in transparent frog embryos and fungal pathogens in Daphnia dentifera allowed pathogen detection in vivo by eye (18,19). While these models have greatly advanced our understanding of infectious diseases, they also have some limitations: viral infections are difficult to follow over time in non-transparent hosts, and genetic tools for organisms like Daphnia are often limited.
This manuscript introduces the infection of the nematode Caenorhabditis elegans by the Orsay virus (OrV) scored by a fluorescent reporter as a novel host-pathogen model ("epidemiology-in-a-dish"). C. elegans is small (~1 mm), feeds on bacteria such as Escherichia coli, and can be easily maintained on agar plates, thus allowing easy longterm imaging and tracking of individuals.
Discovered in immunocompromised wild C. elegans isolates in France in 2011, the OrV is the only natural viral pathogen of C. elegans found so far (20) (see review (21)). This positive-stranded RNA virus primarily infects intestinal cells (22). The infection is not lethal and results in a mild phenotype, inducing fusion of intestinal cells and slowed progeny production (20). In populations, the virus spreads only by horizontal transmissions, in a similar way to feces-to-oral transmissions: viral particles are released from the intestine of infected animals into their environment through the rectum and enter other individuals while they are feeding on the contaminated bacterial layer (20). Wild-type animals (N2) show no symptoms of infection, as they have two alternative pathways for viral defense, analogous to an innate and an acquired immune response. While N2 animals are asymptomatic and only carry a low viral load, some immunodeficient mutants have been identified to be susceptible to the virus. One family of mutants is deficient in the antiviral RNA interference (RNAi) response (20,23), which relies on factors of the classical RNAi pathways (24), but also specifically requires the Dicer-related helicase (DRH)-1 protein for recognition of the viral RNAs (23). A second family is mutated in the co-suppression defective (CDE)-1 protein, which uridylates viral RNAs, a signal that leads to their degradation (25). Finally, general transcriptional responses have also been characterized that, upon infection, regulate expression of innate immunity genes (23,26

Results
We previously developed a fluorescent infection reporter which enables a visual read-out of OrV infection at the individual level (25). The reporter comprises two parts. To detect and track individual worms, a fluorescent protein (mcherry) is constitutively expressed by a tissue-specific promoter in the nematode pharynx. A second fluorescent protein (GFP) is under the control of an immune-response-activated promoter (lys-3), leading to an infection-specific fluorescent signal in the intestine. (Fig. 1A, D, E). We validated the reporter by comparing the fraction of GFP positive nematodes to real time PCR measurements of viral RNA, collected in the same animals (Suppl. Fig. 1).
Our experimental setup utilized a customized upright fluorescence stereomicroscope that can be controlled by a laptop and to allow simultaneous imaging of four spatially separated nematode populations in a 4-well plate (filled with agarose) (Fig. 1B-C). To highlight the influence of host genetics -in particular its susceptibility to the virus -on the dynamics of infection in populations, we worked with two characterized mutations that result in immunocompromised animals: cde-1 (compromised in antiviral uridylation response) and drh-1 (compromised in antiviral RNAi response). We also investigated a double mutant that carries both mutations (drh-1;cde-1).
Viral filtrate was homogeneously mixed with a bacterial food layer (Escherichia coli HB101) before seeding as done previously (27). Hence, nematodes were infected from a homogeneous environment, and individuals were expected to rapidly ingest the virus while feeding. We seeded each 16mm wide well of the 4-well plates with an HB101-OrV mixture, and an isogenic population of 20-25 L4-stage nematodes of one of the four genetic backgrounds (wild-type (N2), drh-1, cde-1, and drh-1;cde-1). Nematode infection was subsequently recorded over 2-3 days with images (GFP/viral load & mcherry/pharynx channels) taken every 10-20 seconds. The resulting movies were post-processed to enhance the fluorescence signal (see Our setup enabled us to follow and quantify the infection of individuals over the course of several days at high temporal resolution (Fig. 2). Animals were tracked over 55 hours with simultaneous scoring of infection reporter expression. We quantified infection levels by dividing the number of infected animals by the total number of animals in each frame.
Different genetic backgrounds exhibited characteristic infection dynamics that were similar between two separate biological repeats (Fig. 2B, Table 1). OrV infection onset in both cases occurred earlier in the cde-1 mutant than in the drh-1 mutant and the infection spread rapidly and early on, reached a peak and then attenuated ( Table 1). In contrast, in the drh-1 population, the infection spread more slowly and remained active until the end of our imaging period (55h) while the cde-1 population was able to clear the virus ( Table 1). The double cde-1;drh-1 mutant exhibited an intermediate behavior with similar early dynamics as the cde-1 mutant (rapid onset of infection) and with comparable late dynamics to the drh-1 mutant (longer persistence) ( Table 1). In the wild-type population, some worms were GFP positive, i.e. infected, for short periods of time but there was no widespread epidemic (Fig. 2B).
Next we tested whether nematode behavior was affected by the Orsay virus by analyzing behavior pre-and post-infection for the four different genotypes, wild-type, drh-1, cde-1, and drh-1;cde-1 (Fig. 3). Infected and uninfected worms exhibited similar behavioral changes upon infection, e.g. speeds and inter worm distances, across different genotypes ( Fig. S2) allowing us to pool the results of the four genotypes together into one group of infected and one group of uninfected nematodes. Velocities of uninfected nematodes were significantly (quantified as effect size 95% confidence interval) higher than those of infected worms (Fig. 3A-B). Infected nematodes were by an average of 330 [300-365] µm/sec slower than uninfected ones in biological repeat 1 and 115 [92-138] (µm/sec) slower for biological repeat 2 ( Fig. 3A-B). [..] signifies the 95% confidence interval of the average effect size value. The average inter-animal distance is an approximation of the population density and the "contact network" (i.e. the number of average contacts a worm has). The average inter-animal spacing was 1.43 [1.28-1.57] mm (mean effect size [95% confidence interval]) larger for infected worms compared to uninfected ones for biological repeat 1 and 1.26 [1.08-1.44] mm larger for biological repeat 2 ( Fig. 3C-D), suggesting that uninfected worms avoid infected worms. We investigated this further by transferring 8 uninfected worms on virus plates on which worms had been infected previously.
Transferred uninfected worms avoided the food suggesting that infected worms secret an avoidance signal that is sensed by other worms (Fig. S4).
Our model system can be used to explore how behavior affects viral spread. Both biological repeats exhibited different infection dynamics with the onset and severity of population infection being higher in the 2 nd repeat ( Table 1). Uninfected immunocompromised worms exhibited lower velocities in repeat 2 compared to repeat 1 ( Fig. S3) suggesting that worm mobility is linked to infection dynamics.

Discussion
Here we introduced a new "epidemiology in a dish" model to quantify viral infection spread in C. elegans populations and started connecting information about the behavior of individual uninfected and infected worms with the dynamics of infection of the whole population. In the future, parameters that potentially contribute to viral spread, such as nematode behavior, environment, and genetic heterogeneity, can be selectively manipulated and their effect on viral transmission dynamics be tested.
We observed that cde-1 mutants exhibited an earlier onset of infection than drh-1 mutants ( Fig. 2B-C, Table 1). This suggests that the antiviral uridylation response, which is downregulated in cde-1 mutants, provides an earlier immunity response (also supported by (25)), while drh-1 mediated antiviral RNAi is important for later stages of infection. We speculate that the RNAi response is delayed due to effectors, i.e. the small viral RNAs, that need to be synthesized before DRH-1 can recognize them. The double mutant (drh1;cde-1) exhibited an early onset of infection combined with a longer persistence of infection than the cde-1 mutant background, indicating that cde-1 and drh-1 are indeed part of independent pathways (25).
In addition to host immunity and genetic background, host behavior is also known to influence the dynamics of infectious disease transmission. Host mobility and relative population density, in particular, shape the population's contact network, i.e. the frequency of contact between individuals, which the infection needs in order to spread.

9
The role of direct contacts in infection spread of the OrV is currently unknown, but colocalization of individuals may enhance transmission even without direct contact by increasing the likelihood of individuals coming into contact with shed virus. Nematodes slowed down upon infection while inter-animal distances increased (Fig. 3A). Both effects combined are likely to decrease the rate at which infected and uninfected individuals colocalize, potentially decreasing viral spread. We speculate that the behavioral response is intended to mitigate viral spread, as we have observed that uninfected worms can exhibit avoidance behavior towards food patches exposed to infected animals (Fig. S3).
Similar observations have been made in other organisms, where infection induced a characteristic "sickness behavior" (28). Our current algorithm is able to reliably track worms over an average of 1 hour.
We observed that infection progressed differently in populations with different behaviors (Fig. S3). In future work, nematode behavior, such as motility and worm aggregation (the frequency of contact between individuals), can be directly perturbed to measure its effects on viral spread. Worms can be slowed down, for example, by using strains with abnormal locomotion phenotypes, such as unc (29) or rol (30) mutants, as well as mutants for neurotransmitters that are known to affect locomotion rate by rendering worms unable to sense food or their movement more erratic (31). In particular, dopamine-defective mutants (32) could be used to study the observed slowing down upon infection, as the dopaminergic pathway is thought to induce the reduction in locomotion characteristic of this response (33). In addition, other mutants of C. elegans, such as npr-1 (a mutant with an aggregation phenotype (34)

Preparation of viral filtrates
A stock of viral filtrates was prepared using a liquid culture protocol from David Wang's lab.
Firstly, "thick" bacteria food was prepared as follows. HB101 bacteria were inoculated in In parallel, three young JU1580 adults were transferred on 50 mm NGM agar plates seeded with HB101. 20 µl of a previous viral filtrate were pipetted onto the bacterial layer in drops and the worms were incubated at 20°C for 3 to 4 days, to let them multiply. Just before getting starved, worms were chunked onto 90 mm plates and incubated at 20°C.
Just before starvation, the worms were washed off the plates with 5 ml of S-medium (prepared as described in (35)

Infection protocol
The Orsay virus filtrate was diluted within the HB101 bacterial food (1:10) to ensure homogeneous distribution of the virus on the plate. The mixture was then seeded on NGM agar plates. These plates, referred to as "virus plates", were left to dry at room temperature for 2-3 days and then stored at 4°C for 1 month maximum. Before use, plates were kept at room temperature for at least one hour. In parallel, mock plates were prepared by diluting M9 medium into the HB101 bacterial food (1:10).

RT-qPCR and viral load measurements
Worms were harvested using M9 medium supplemented with 1:100 worm lysis buffer (to avoid worms to stick within the tip) and let on ice to settle down. Supernatant was removed and the worm pellet was washed 3 times with M9 medium only. After the washes, most of the M9 medium was removed and the pellets were frozen in liquid nitrogen and stored at -80°C. The lysis was performed in PCR strips by transferring 5 µl of worm pellet into 45 µl of Lysis Solution supplemented with 1:100 DNase I (Ambion). The samples were then lysed by at least 10 freeze-thaw cycles using liquid nitrogen and a warm water bath, followed by 30 min of vortexing at room temperature. Lysis was stopped by adding 5 µl of Stop Solution (Ambion). Retro-transcription (RT) and qPCR reactions were then performed using the Power SYBR Green Cells-to-Ct kit (Ambion), according to the manufacturer's instructions. Either 5 or 10 µl of worm pellet were used in the RT reaction (depending on the pellet size) and the cDNA samples were stored at -80°C. 2 µl of a 1:10 dilution of the cDNA were used as template in the qPCR reactions, which were run on a StepOne Plus qPCR machine (Applied Biosystems) following this program: 95°C for 10 min, 40 cycles of 95°C for 15 sec and 60°C for 1 min. Orsay virus RNA1 was amplified using the primers GW194 and GW195 (20). Samples were normalized to gapdh mRNA, using primers developed by (36). Primers sequences are provided in Table 3.2.

Manual scoring
The The GFP is continually expressed in opening areas, such as the rectum, the vulva and the pharynx. As being continually exposed to environmental pathogens, these areas are thought to be under permanent immune activation. When an immunocompromised mutant is exposed to the virus, the GFP signal becomes activated in the anterior part of 14 the intestine. On the other hand, the RFP signal is constitutively expressed under any conditions in the pharynx, in both wild-type and mutant backgrounds.

Reliability
To assess the reliability of our reporter, we infected populations of 40 eggs of each genotype in 2 independent experiments (with n=3) for drh-1 and cde-1;drh-1, and 3 independent experiments (with n=3) for wt and cde-1. After 3 or 4 days, we manually

Long term imaging
Adult C. elegans from the 4 genotypes were transferred to 4 well plates with each well containing a food drop with HB101 bacteria that were mixed with OrV. 15h post plating the plates were transferred under a customized Zeiss Axio Zoom microscope which was able to take high resolution images of all 4 wells (Fig 1D-E). Both mcherry and GFP channels were than acquired with exposure time 1 and 5 seconds respectively every 10 -20 seconds for ~40h. Room temperature was kept at 20 degrees. After that time the next generation of larvae hatched, and the food tended to dry out.

Image Analysis
Movies were quantified as follows: 1. The ImageJ Rolling Ball Average plugin was applied individually to both the mcherry (ball size 10 pixels) and gfp (50 pixels) channels to reduce background noise. This helped to remove large bright areas in particular. with an adjusted version of the tracking script used in (37). This approach enabled us to track individual worms for approximately 1h. Worms were assumed to be infected if they exhibited an GFP positive signal for at least 50 frames.
The red circle represents the outline of the well the worm is imprisoned in. Quantifications shown in Table 1.  Tables   Table 1: Quantification of Infection dynamics in Figure 2B-C.

Supplemental Figures
Supplemental Figure 1