Establishing an inexpensive, space efficient colony of Bemisia tabaci MEAM1 utilizing modelling and feedback control principles

Abstract A stable, synchronized colony of whitefly (Bemisia tabaci MEAM1 Gennadius) was established in a single ~30 cu.ft. reach‐in incubator and supported on cabbage host plants which were grown in a 2 × 2′ mesh cage without the need for a greenhouse or dedicated growth rooms. The colony maintenance, including cage cleaning and rotation of plants, was reduced to less than 10 h per week and executed by minimally experienced researchers. In our hands, this method was approximately 10‐fold less expensive in personnel and materials than current typical implementations. A predator‐prey model of whitefly colony maintenance that included whitefly proliferation and host plant health was developed to better understand and avoid colony collapse. This quantitative model can be applied to inform decisions such as inoculum planning and is a mathematical framework to assess insect control strategies. Extensive measurements of colony input and output (such as image analysis of leaf area and whitefly population size) were performed to define basic ‘feedback control’ parameters to gain reproducibility of this inherently unstable scaled‐down whitefly colony. Quantitative transfer of ~100 whiteflies repeatedly produced more than 5000 adult whiteflies over a 6‐week, two‐generation period. Larger scale experimentation could be easily accommodated by transferring adult whiteflies from the maintenance colony with a low flow vacuum capture device. This approach to colony maintenance would be useful to programs that lack extensive plant growth room or greenhouse access and require a “clean” implementation that will not contaminate an axenic tissue culture laboratory.

| 649 THOMPSON eT al. 1991). Researching whiteflies and their transmitted viruses is an important aspect of developing a plan to mitigate crop loss.
One of the barriers to carrying out research on whiteflytransmitted pathogens is the required establishment and maintenance of a whitefly colony and the considerable cost of operation in personnel, materials and facilities. Typically studies of whitefly virus transmission require isolation of viruliferous and virus-free colonies using separate rooms or greenhouses, as well as the use of small greenhouses to produce host plants for the colonies (Lapidot et al., 2014;Polston & Capobianco, 2013;Schuster et al., 2009).
An additional obstacle to whitefly-vector research is that many of the current methods utilized are a potential point of contamination -a particularly prevalent concern is the routine introduction of host plants usually produced in a greenhouse. While some of these limitations can be overcome through collaborations, such multiinvestigator projects are limited due to the size of grant necessary to execute the research as well as problems with meeting phytosanitary regulations. More importantly, the fragile nature of whitefly as a research component makes it time-consuming and difficult to transport between performance locations. We designed, implemented and refined a scaled-down version of a whitefly colony that avoids the aforementioned obstacles by being readily implemented in an indoor laboratory at minimal cost with an accompanied model to predict the quantity of whiteflies available for experimentation.
Whiteflies have a roughly 3-week life cycle for feeding, egg-lay and nymph development at 28°C. At this temperature, whitefly females are capable of laying 50-110 eggs (Aregbesola et al., 2020;Butler et al., 1983;Powell & Bellows, 2009), resulting in an amplification ratio that can cause a rapid decline in host plants, which is particularly problematic for scaled-down colony maintenance (see Supplemental Figure S8-1). Given the haplodiploidy nature of whiteflies, unfertilized eggs will hatch as male; therefore, a colony initiated from small numbers of whiteflies can result in a male/female imbalance. The method developed here is our 'engineering' solution to this challenge of reducing cost and effort while still maintaining robust and high-quality whiteflies for research studies.

| Indoor facilities design
The colony was maintained in a single reach-in incubator (SP Scientific, model 317512, 33 cu.ft. Environmental Stability Chamber) that had sufficient inner dimensions (33″W × 27″D × 60″H) to accommodate four cages that would provide two whitefly life cycles. Cages were designed to accommodate front-loaded standard 1020 greenhouse flat trays (Agron #HGC726165) and whiteflies using whitefly-proof screening. The optimal design was 12″W × 24″L × 18″H with a 12″ × 12″ front panel door reach-in sock with a 6″ vinyl upper segment window for observation of watering and plant health (BioQUIP #1450NS85). The use of a remote-control watering pump cart made for extremely convenient and time-efficient watering that could be systematized based on the timing of the watering per pot (see Supplemental Info S1). Lighting was provided by (6)

| Primary cage setup
Five 5-week-old cabbage plants were weighed as an indicator of the soil moisture content entering the whitefly colony followed by addition of up to 30 ml of water if weight was under 300 g. A picture taken from above the plants provides a basis for assessing leaf surface area at this initial time point (see Figure S7-1). These inputs (plant size and soil moisture content) were chosen based on experience of over a year of colony maintenance as important measures of colony health. Cage initiation was implemented bi-weekly on Monday with a goal of providing procedural consistency for synchronization of adult emergence.

| Inoculation
Serial inoculation of whiteflies onto host plants underwent considerable changes in an effort to achieve reproducible colony performance. Two basic methods were found to provide an near zero-escape of whiteflies: (1) qualitative inoculation on a transfer plant exposed for a defined period of time and (2) quantitative inoculation with a population size determined from pictures on a single leaf. Both qualitative and quantitative methods were accomplished by exposure of an inoculum plant to the colony cage entering its 7th week of proliferation. This corresponds to the emergence of the second generation of whiteflies. Qualitative inoculation was accomplished using a randomly selected host plant in an enclosed transfer device (see Figure S4-2). Since the emphasis of this description is on quantitative colony monitoring, the details of the qualitative approach are provided in Supplemental Info S4 as that approach would be sufficient for most experimental work. For quantitative inoculation, the best approach was found to use a 5-week-old cabbage seedling removing all but one leaf that was fixed upright using a wooden applicator stick (see Figure 1d). The plant was illuminated obliquely with a spotlight to provide high contrast for pictures of both sides of the leaf for counting. Once the target quantity of adult whiteflies was reached, a 90 mm disposable Petri dish was used to pinch off the leaf inside the mesh cage for transfer. Wiping the outside of the Petri dish with an anti-static dryer sheet prior to use was found to prevent whiteflies sticking to the plastic due to static charges. Inoculation involved opening the Petri dish in the newly prepared primary cage setup. Combining experience with preliminary models of whitefly proliferation, a whitefly count of 80-110 whiteflies was chosen as a balance of numbers, plant health and subsequent proliferation of both the plant host and the whiteflies.

| Quality control parameters
The quality control (QC) procedures described below were an important aspect of establishing consistency between many different inexperienced researchers, and while not necessarily needed for routine maintenance, they are recommended for transition to new personnel during training.  tall table number holder (New Star TBH-12/23237) as seen in Figure 2. The sticky card is covered with a small plastic bag during placement in the middle of the cage (to prevent premature whitefly capture). After allowing 15 min for the colony cage to settle after this disturbance, the sticky is uncovered for 15 min. The manual whitefly count on the yellow sticky card was conducted with a stereo microscope.

| Emerged 4th instar exoskeleton count
Assessments of adult whiteflies based on sticky card count was found to be quite variable. Counting of eggs was undertaken for many months but was found to be tedious and not amenable to image analysis. The various larval instars can be difficult to distinguish; however, once the adults emerge from the 4th instar, they generate a relatively large, distinct high-contrast white exoskeleton ( Figure 2). Initially a circular punch was used to sample a consistent leaf surface area; a refinement to further improve exoskeleton count is to mark the punch area without excision, then recursively image concentric circles of increasing the leaf surface area until a minimum of 25 exoskeletons is measured as a basis of a normalized exoskeleton count per surface area.

| Total whitefly count via image analysis
Counting the whiteflies at 7 weeks (47 days) was performed after the cage was kept at 4°C for at least 48 h to ensure that the whiteflies were immobilized for imaging. The leaves of each plant were excised (for spatial and weight analysis), and the whitefly adults were gently brushed onto a flat dark surface with a very soft 2″ paint brush (Artist'sLoft™ necessities). As shown in Figure 3c, the whiteflies were gently distributed to analyse their number using ImageJ (64-bit Java 1.8.0_172) ( Figure 3d). An independent manual count of a small quadrant of the same image was used to validate and calibrate the automated procedure to within a few percent error ( Figure 3a,b). Spraying the whiteflies with water after distribution on the surface was found to dissolve whitefly honeydew and improve image analysis.

| Discrete-time model for whitefly colony proliferation
A discrete-time model was developed to predict the number of whitefly adults and the health evolution of the plants on a weekly basis in the colony given an initial inoculum amount and plant size.
The model follows the predator-prey model structure. The full list of variables and parameters are presented in Table 1. The number of eggs at any given week is based on a modified predator-prey model as presented in Equation 5. At low counts of whiteflies, there is plenty of space to lay eggs and therefore they exhibit unrestrained proliferation. At high counts of whiteflies, space on the plants becomes constrained for both egg laying and feeding, which imposes saturation kinetics on proliferation. The egg-laying rate (r) was determined based on an average total number of eggs laid during a female's lifetime being 109 eggs at 28°C (Aregbesola , 2020). Based on the assumption that adults live for 2 weeks, we assume they lay half of their eggs (~50) each week. Finally, this number was divided by two to account for half male whiteflies for the fit to the experimental data. The egg death rate (µ E ) was defined in a similar fashion to the nymphs and adults and finally fit to the experimental data within the bound region E ∈ [0.20, 0.40].
Note that although a model that accounts for female and male populations could be created, our experimental observations lack the needed resolution for such a finer description. Additionally, increased numbers of whiteflies deteriorate the health of the plants.
Host plant deterioration negatively affects the health of the colony, which is captured in a predator-prey model by the bilinear term in Equation 5 by the consequence of fewer eggs being laid. While the model captures general characteristics of whitefly and plant growth that could be adapted to more diverse conditions, its utility is intended to provide insight into the system behaviour and specific prediction for colony maintenance.

| Quality control measures
The goal of achieving a compact whitefly colony that occupies roughly 20 square feet of laboratory space and relatively inexpensive to maintain was accomplished and could consistently provide hundreds to thousands of whiteflies for experimentation on a biweekly basis. However, despite the focus on consistency of inputs, and a qualitatively healthy colony productivity, the quantitative outputs reflected considerable instability. Numbers of exoskeletons and adult whiteflies at the 7th week of colony proliferation reflected a consistent correlation in the variation of these productivity indices.

| Model results
The model parameters fit to the experimental data of total harvested whiteflies ( Figure 5, iterations #8-13) and plant surface area are presented in Table 1 and calculated predictions presented for an inoculation of 100 whiteflies in Figure 6. The delay period in total adult whiteflies from week zero to two reflects the hatching of eggs to nymphs. The subsequent increase in the viable adult count was dramatic due to the 1:80 amplification ratio for the adult females. This high amplification ratio also results in the vast majority of whiteflies being viable as indicated by the dotted red line of Figure 6a being comparable to the total. Viable whiteflies decrease from weeks 0-2 due to adult whiteflies dying while the eggs/nymphs To assess the sensitivity of the colony performance, a deviation analysis was conducted on the three main variables (input leaf surface area, input whitefly count and proliferation time) and is presented in Figure 7. The change in the final total whitefly count due to an increase of 60% in plant leaf surface area (i.e. plant size) is minimal compared to whitefly inoculation count or proliferation duration (no change = 1). The 60% deviation percentage in initial whitefly count is within the range of possibility for a bias of predominantly male or female inoculum whiteflies. The strong impact of proliferation time that results from logarithmic whitefly proliferation is apparent, as a comparable deviation in total whitefly output to variations in whitefly number inoculation is achieved for only an increase or decrease of one week (15% deviation) in proliferation time.

| DISCUSS ION
Preliminary efforts at establishing a whitefly colony at Penn State in a walk-in growth chamber with plants initiated at a greenhouse had extensive problems including humidity control and the associated mildew contamination, resulting in poor colony health.
In trying to replicate a comparable whitefly colony capability at the new performance site, the original project costs for technician time, heavily subsidized greenhouse and walk-in incubator space were in excess of $100K U.S. per year. This led to a complete 'reengineering' of the approach to colony maintenance to improve whitefly productivity while reducing cost with the goal of a whitefly colony that could be maintained within a typical laboratory space that did not have dedicated plant growth infrastructure or technical maintenance personnel. Reducing whitefly colony initiation to a bi-weekly basis (after more than a year of weekly maintenance) was a simple substantial reduction in space and materials that had minimal impact on whitefly availability for experimentation and balanced efforts for predominantly plant and predominantly insect manipulation as a convenience for colony workflow.
The scaled-down implementation described here represents a high-performance whitefly colony (~5000 whiteflies bi-weekly) that costs an order of magnitude less, with capital costs (incuba-  (Tahara et al., 2018). It should be kept in mind, that classic biological models of growth for suspensions of cells, a 1 ml inoculum of optical density of 1 will contain on the order of 10-million cells, which typically grow by simple division and allow for nearly continuous time point measurements (Myers et al., 2013). Refinements in fecundity and death rates are far more difficult to assess; nonethe-less, obtaining more refined models and experimental platforms could be invaluable for assessing very different strategies of insect control such as reducing fecundity, attenuating lifespan, male sterility etc.
Since the goal in this work was to dramatically scale down the colony (in both size and effort), the SOP was iteratively refined as we sought improvements. Feedback for changes were informed by qualitative observations and quantitative quality control measurements. Monitoring the health of the colony through sticky-paper counts, exoskeleton counts and total whitefly counts allowed us to troubleshoot and implement a qualitative feedback control methodology. Some discussion on this evolution of methods is informative towards future improvements. A problem encountered week 4 was shown to be dramatically lower than previous weeks.
Looking through the QC logs, this correlated with a particularly low inoculation number (back-calculated from the model corresponds to between 20 and 30 whiteflies). With this observation of a strong sensitivity to whitefly inoculum counts, a quantitative inoculation based on known whitefly numbers was developed to avoid acquisition based on time. This was accomplished while avoiding highly accurate but tedious methods for whitefly manipulation, for example aspirator counting (Polston & Capobianco, 2013).
This experience with colony maintenance and model predictions converged to a target inoculum count of between 80 and 110 whiteflies. An inherent assumption of our inoculation method is that the procedure acquires random males and females; however, this may be biased towards a greater number of males due to being more mobile relative to the larger feeding and egg-laying females.
It is noteworthy that for extended acquisition times, when fewer whiteflies were available in the inoculum cage, there was a clear pairing of males and females (see Supplemental Figure S8-4) that is not apparent when higher whitefly loads in the inoculum source cage resulted in more rapid acquisition. Such a female/male biasing provided by this inoculation method could be expected to help stabilize colony performance. While a refinement based on sexing of whiteflies based on size from image analysis might provide more accurate modelling, we wanted to retain the simplicity of total whitefly inoculum of ~100. More whiteflies than this overloads the host plants and rapidly deteriorates their health. To confirm this, an inoculation with 200 whiteflies was undertaken. As seen in Figure 5, iteration 13 corresponds to this high inoculum and it resulted in a final count of 22,740 whiteflies and dramatically lowered plant health that was clearly evident by 4 weeks for this proliferation cycle (see Supplemental Figure S8-3).
While the focus of whitefly colony maintenance is to provide in 2 , more than half of the leaf surface area is not directly exposed to the PAR light levels of ~110 µE m −2 s −1 measured inside the cage at mid-height between pot and cage top. Given the improvements F I G U R E 7 Whitefly production model deviation analysis for the predicted total whitefly population relative to the standard condition: initial leaf area of 25 sq.in. per plant, 100 whitefly inoculum and 7 week harvest duration. The assessment is the deviation ratio in predicted total harvested whitefly in response to 60% changes in the plant leaf surface area, whitefly inoculation and 15% duration (±1 week) in the harvest duration. Consistent with qualitative observations over several years, initial whiteflies and the duration of colony amplification have a far greater impact on the final whitefly count than the plant size [Colour figure can be viewed at wileyonlinelibrary.com] F I G U R E 8 Timeline of quality control measures and colony implementation changes. The sticky count indicator of whitefly capture rate was the initial focus for feedback changes to colony implementation. During that period (I < 170) large variations in plant health and whitefly count was observedincluding near complete colony collapse. Additional quality control measures were incorporated as the process was improved and refined (I > 170). The graph illustrates the difficulty of achieving stability in this small-scale system and ultimately the effectiveness of quantitative inoculation [Colour figure can be viewed at wileyonlinelibrary.com] in plant growth and the associated observation of light limitation, the use of only four cabbage plants is sufficient for the inoculum of roughly 100 whiteflies -thereby further reducing time and materials for maintenance while still resulting in a production of around 5000 whiteflies.
Overcoming problems of scaled-down plant and whitefly performance was approached by moving from qualitative to quantitative monitoring a variety of inputs ranging from pre-colony plant growth to soil volume and moisture as they related to total whitefly performance indices after nearly 7 weeks of proliferation. We were able to create a cost-effective high-performance approach to a whitefly colony maintenance that can reduce the burden and investment required for researchers to execute whitefly research and thus address a critically important aspect of crop protection and viral disease transmission research.

S U PP O RTI N G I N FO R M ATI O N
Additional supporting information may be found in the online version of the article at the publisher's website.
How to cite this article: Thompson, N., Waterton, N