Live In Vivo Imaging of Plasmodium Invasion of the Mosquito Midgut

Malaria is one of the most devastating parasitic diseases in humans and is transmitted by anopheline mosquitoes. The mosquito midgut is a critical barrier that Plasmodium parasites must overcome to complete their developmental cycle and be transmitted to a new host. Here, we developed a new strategy to visualize Plasmodium ookinetes as they traverse the mosquito midgut and to follow the response of damaged epithelial cells by imaging live mosquitoes. Understanding the spatial and temporal aspects of these interactions is critical when developing novel strategies to disrupt disease transmission.

invaded cells (5). A coordinated response of midgut epithelial cells, hemocytes, and the complement-like system is necessary for mosquitoes to mount an effective antiplasmodial response. Our current understanding of the cell biology of ookinete invasion is based mostly on immunofluorescent studies performed with fixed materials at preselected time points; such studies do not provide precise information on the timing of key cellular events (1). For example, it is clear that ookinetes must traverse invaded epithelial cells before the damaged cells are extruded into the midgut lumen, but the time frame of these events in vivo has not been well established.
The first live imaging studies of ookinete midgut invasion were done using differential interference contrast (DIC) microscopy of in vitro-cultured Plasmodium gallinaceum (bird malaria) ookinetes placed on top of fragments of dissected Aedes aegypti midguts (6). This method involves extensive manipulation of the mosquito midgut that can result in tissue injury. The samples were imaged for a maximum of 3 h due to the limited viability of dissected midguts. Many ookinetes were observed gliding on the luminal surface of epithelial cells, and the images documented that ookinetes are flexible and can suffer temporary constriction as they move through the microvillusassociated network. DIC imaging allowed rapid temporal resolution at 5-to-15-s intervals, and the images documented that it takes an average of 1.3 min from the onset of cell penetration for ookinetes to complete internalization. They also showed that ookinete-invaded cells underwent caspase activation and apoptosis (6). However, spatial resolution in the "Z" axis was limited because ookinetes could not be visualized once they had invaded the cell.
A second imaging study dissected Anopheles gambiae and Anopheles stephensi midguts infected with green fluorescent protein (GFP)-labeled Plasmodium berghei parasites. Midguts could be imaged for a maximum of only 30 min using confocal microscopy due to loss of viability following midgut dissection (7). The gut epithelium was fluorescently labeled by incubation of tissues ex vivo with a lipophilic dye that stained only the basal membrane of epithelial cells. Examination of the results confirmed that ookinetes predominantly use an intracellular route to traverse the midgut epithelium and that ookinetes can serially traverse the cytoplasm of several midgut cells before egressing to the basolateral intercellular space to reach the basal lamina. In addition, ookinetes were observed gliding on the membrane foldings of the basolateral labyrinth, and wound repair following epithelial extrusion into the gut lumen involved extensive lamellipodia crawling from adjacent cells. The ookinetes were subjected to three-dimensional (3D) visualization and exhibited different modes of motility, including stationary rotation, translocational spiraling, and straight᎑segment motility (7). However, because the midgut luminal membrane was not labeled, it was not possible to capture the initial invasion events.
In this study, we labeled the midgut of Anopheles gambiae G3 females with specific fluorescent markers in vivo to capture the full invasion process. New imaging methods were developed to minimalize the processing of mosquitoes, as well as to increase the viability of tissues, allowing longer time frames of live-tissue imaging. This facilitated the capture of both ookinete midgut traversal and epithelial cell responses to parasite invasion.

RESULTS
Sample preparation for live in vivo imaging. Two major obstacles had to be overcome to image Plasmodium ookinete midgut invasion in real time. The first obstacle entailed developing a new strategy to mount blood-fed infected mosquito midguts while maintaining tissue viability for several hours. Our initial attempts to mount dissected midguts ex vivo were unsuccessful due to the loss of tissue integrity. Dissected midguts could not be imaged for more than 20 to 30 min, in agreement with previous reports (7). Levels of ookinete invasion peak at 24-h post-blood meal, a time when the mosquito midgut is under physical stress because of mechanical stretching to accommodate the blood meal and when it can also be damaged by the extensive secretion of digestive proteases. To solve this problem, whole-blood-fed mosquitoes, infected by feeding on mice infected with P. berghei-mCherry parasites (Fig. 1A), were mounted after decapitation and removal of legs and wings to immobilize them (Fig. 1B, top panel). Previous studies have shown that decapitated mosquitoes can survive for long periods of time, even for several days (8). Imaging whole mosquitoes allowed us to maintain an adequate oxygen supply to the gut through the tracheal respiratory system. To minimize any possible injury, mosquitoes were gently knocked down using CO 2 (instead of using air suction followed by ice cooling) and were placed sideways to enable imaging of the midgut through the thinnest and lightest areas of the abdominal cuticle. Mosquitoes were immobilized using malleable adhesive craft putty as a spacer, to confine them to the narrow space between the glass slide and a coverslip (Fig. 1C). Using this mounting setup, we were able to image the blood-fed midgut through the cuticle for more than 4 h using fluorescent confocal microscopy ( Fig. 1D and E).
The second obstacle was the difficulty of identifying appropriate fluorescent markers to label midgut epithelial cells in P. berghei-infected mosquitoes. These markers must retain strong fluorescence for more than 24 h (time of peak ookinete invasion after blood feeding) at low concentrations, to avoid toxicity. Blood-fed mosquitoes refrain from subsequent feeding until digestion is complete; therefore, dyes were fed in saline solution using a membrane feeder before mosquitoes were fed on P. bergheiinfected mice. An alternative method for labeling mosquitoes by systemic injection was also tested for nine potentially suitable fluorescent live-cell tracer dyes (see Table 1). Membrane feeding of CFDA-SE (carboxyfluorescein diacetate succinimidyl ester) in a saline solution worked very well to label all midgut epithelial cells. This is a cellpermeative dye with low toxicity that is commonly used for long-term cell tracing (9). Epithelial caspase-3 activity was detected by feeding NucView to mosquitoes, a substrate that, when cleaved by caspase-3, releases a high-affinity DNA binding dye which produces nuclear fluorescent staining (10). Epithelial F-actin was visualized in vivo by  feeding mosquitoes SiR-actin, and this made it possible to follow the reorganization of epithelial actin after ookinete invasion. Systemic injection of the lipophilic membrane dye DiD labeled hemocytes very efficiently, as previously reported (5). These dyes were often used in combination by feeding them together. Labeling multiple components in the same mosquito made it possible to obtain a more comprehensive picture of mosquito-parasite dynamics in vivo.
Ookinetes have different fates after they develop in the midgut lumen. Knowing that P. berghei ookinete midgut invasion peaks around 24 h postfeeding (PF), we imaged infected midguts from 22 to 26 h PF to maximize the probability of capturing ookinete dynamics before, during, and after midgut invasion. Z-stack sections were obtained at 1-m intervals, every 30 s, to capture quick ookinete movements and obtain cross sections that encompassed the full thickness of the mosquito epithelium, from the midgut lumen up to the basal side of epithelial cells. Images were imported into Imaris, a visualization software program, to perform tracking analysis of the ookinetes (see Fig. S1A in the supplemental material).
We tracked 143 individual ookinetes from four mounted midguts and found that the parasites had three possible fates. Although 50% (71/143) of them (see Movie S1 in the supplemental material) invaded the midgut, others (30%; 43/143) came in contact with the midgut surface (Movie S2) but did not invade the cell and 20% (29/143) were motile but remained within the gut lumen (Movie S3) and never made contact with the midgut surface ( Fig. 2A to E). Any spherical parasite that lacked the characteristic "banana shape" of ookinetes was presumed to be a zygote and was not included in this analysis. A successful cell invasion event does not guarantee that the parasite will reach the basal lamina and develop into an oocyst. Of the 71 ookinetes that invaded the midgut, 76% (54/71) had already reached the basal lamina and remained there for the duration of the imaging (4 h), 11% (8/71) egressed back into the lumen of the gut and we were able to track 9 ookinetes (13%) as they actively traversed the epithelial cells. We also estimated the time required for an ookinete to traverse the midgut. Because we scanned the tissues every 30 s, we were able to determine that it took these 9 parasites between 60 and 120 s to traverse epithelial cells, with a median of 90 s and an average of 83 s (see Table S1A in the supplemental material).
We then compared the motility of the 54 ookinetes that reached the basal lamina with that of parasites that remained in the gut lumen (n ϭ 29) or that only came in contact with the microvillar surface but did not invade the cell (n ϭ 43). A circular spinning motion was often observed in ookinetes gliding on the gut surface (Movie S1), and those parasites displaced significantly longer distances of traversal (ANOVA, P Ͻ 0.0001) (Fig. 2F, left) and moved faster (ANOVA, P Ͻ 0.0001) (Fig. 2F, right) than the ookinetes after they invaded the midgut or than those that moved within the blood bolus but did not contact the gut surface.
Caspase activation in invaded midgut cells. We also investigated the kinetics of midgut epithelial cell apoptosis in response to ookinete invasion by tracking caspase activity in cells invaded by ookinetes. NucView was fed to mosquitoes in a saline solution prior to Plasmodium infection. Caspase signal was not detected in midguts of control mosquitoes fed on a healthy mouse, indicating that apoptosis of midgut cells is rare under these conditions, while it was readily observed following ookinete invasion (Fig. S1B).
Caspase activity was detected in a total of 16 single cells or cell clusters from four P. berghei-infected midguts that were imaged for 4 h. Interestingly, the NucView signal was cytoplasmic and not nuclear, indicative of cell damage and DNA release into the cell cytoplasm of invaded cells. A relative level of signal intensity was established for each midgut, where the strongest signal was 100% and the weakest 0%. A minimum threshold was applied, according to which caspase activity with signal intensity that was at least 20% of the highest in the sample was considered to represent a positive result (Fig. 3A and B; see also Table S1B) (Movies S4 and S5). Most of the 16 cells (13/16 ϭ 81%) were already positive for caspase activity when the imaging started ( Fig. 3B; see also Table S1B), and 8 cells were still positive when the imaging ended after 240 min ( Fig. 3A; see also Table S1B). In 8 cells, caspase activity was detected for close to 4 h (225 to 240 min) ( Fig. 3C; see also Table S1B).
Actin reorganization in response to ookinete invasion. Previous confocal studies with fixed tissues showed that the actin cytoskeleton undergoes dramatic rearrangements as cells damaged by Plasmodium invasion are extruded from the midgut epithelial layer, by forming an actin ring at the base of damaged cells that is gradually constricted (1). Because traditional actin markers, such as phalloidin, are highly toxic (11) and impermeable to live cells, we decided to use the F-actin-specific SiR-actin fluorophore (12).
Midgut cells invaded by Plasmodium ookinetes exhibited strong F-actin staining. These damaged cells could be seen protruding into the lumen of the midgut as they were extruded from the epithelial sheet ( Fig. 4A; see also Movie S6). Five midguts were imaged for 4 h starting at 24 h PF. A total of 34 single cells or cell clusters with F-actin signal were observed, and a 20% minimal signal threshold was also applied. The median duration of positive F-actin signal was 217 min for 26 cells that were imaged for a total of 240 min ( Fig. 4B; see also Table S1C). We also established how long it took for a cell to achieve maximum F-actin levels and for it to decrease to background levels. We captured the appearance of F-actin in 12 invasion events, and F-actin levels increased over a median period of 135 min ( Fig. 4C; see also Table S1D). Once maximum levels were reached, it took about 20 min longer for F-actin to decrease to background levels, with a median of 155 min (n ϭ 17) (Fig. 4D; see also Table S1E). We also found that some ookinetes that remained intracellular when F-actin levels increased were still able to egress from the cell (Movie S7A), while others are completely surrounded by F-actin and trapped inside the cell. Parasites that failed to egress were eliminated when the cell budded from the midgut epithelium (Movie S7B).
Hemocytes patrol the basal surface of the midgut. We previously showed that the mosquito midgut releases prostaglandin E2 (PGE2) when epithelial cells come in contact with bacteria from the gut microbiota or with the immune elicitors that they release (13). Ookinete invasion triggers PGE2 synthesis, and this is a strong hemocyte chemoattractant (13). We investigated hemocyte dynamics following ookinete invasion in live mosquitoes. Hemocytes were fluorescently labeled by systemic injection of DiD, a lipophilic dye that specifically labels this cell population (5). The next day, mosquitoes were infected by feeding on a P. berghei-mCherry-infected mouse. We found that hemocytes were actively patrolling the basal surface of infected midguts (Movie S8A) and that a few hemocytes appeared to be tracking specific ookinetes (Movie S8B), but the specificity and functional relevance of these transient interactions remain to be determined.

DISCUSSION
Three key components had to be integrated for successful live in vivo imaging of ookinetes as they traversed the mosquito midgut: development of a novel mounting protocol, identification and delivery of appropriate fluorescent markers to label mosquito cells, and establishment of a confocal microscope imaging strategy. Mounting live blood-fed mosquitoes made it possible to capture the invasion process and the midgut epithelial responses to parasite invasion by imaging each midgut continuously for several hours, without loss of tissue viability, overcoming a major limitation in previous studies. Confocal point scanning microscopy made it possible to capture the fluorescent signal from the labeled mosquito cells and the transgenic parasites, while avoiding the background signal from the mosquito cuticle, a tissue with strong autofluorescence. A 2-photon laser system was tested, but it failed because the mosquito cuticle absorbed the energy of the laser, and this damaged and sometimes even burned the sample.
Ookinetes in the lumen that contacted the midgut surface traveled the longest distances and moved faster than ookinetes that did not contact the midgut (Fig. 2F). Ookinetes were often observed spinning in circles along the luminal surface of the midgut (see Movie S1 in the supplemental material). This "spiraling" motion has been previously reported (7). Once ookinetes enter the epithelial cell, they slow down and travel a shorter distance than when they glide on the surface (Fig. 2F). This is likely due to the interaction with the actin cytoskeleton of the epithelial cell, a barrier that the parasite must overcome to traverse the midgut. A preliminary study using live in vivo imaging of dissected midguts also showed similar kinetics of P. berghei ookinetes in A. gambiae and A. stephensi, with slower motility as parasites crossed the epithelium and faster displacement when they glided in the midgut lumen (14). We were able to observe the active process of invasion of nine parasites and found that it took them an average 83 s to traverse epithelial cells (see Table S1A in the supplemental material).
Ookinete invasion triggers a caspase-mediated apoptotic response (1), and a previous ex vivo study also reported higher caspase activity in invaded cells using PhiPhiLux-G 1 D 2 , a transient dye to detect caspase-3 activity (6). In the current study, NucView, a stable caspase marker, was used. The use of NucView allowed extended continuous imaging and revealed that epithelial cells damaged by the parasite induced caspase activity and were able to remain associated with the epithelium for several hours, in some cases for the 4-h duration of the imaging session ( Fig. 3C; see also Table S1B).
A strong F-actin signal was also observed in ookinete-invaded cells, indicative of reorganization of the actin cytoskeleton, but was absent in uninfected midguts. In some instances, we observed cells with strong F-actin signal without a parasite in direct contact with the cell. However, ookinetes were always present in neighboring cells, suggesting that parasites can sequentially invade multiple cells, as previously reported (1) and confirmed in a recent preprint report (14). During this dynamic rearrangement of the actin cytoskeleton, a few parasites were surrounded by F-actin and trapped inside the cell (Movie S7B). Similarly to caspase activity, F-actin activity also can be detected for several hours, with a median of 217 min (Fig. 4B), and can sometimes persist for 4 h. This indicates that, under these physiological conditions (22°C), the extrusion of midgut epithelial cells takes several hours, while ookinetes move much faster, as it takes them an average of 83 s to traverse an epithelial cell. We have previously shown that caspase activation and epithelial nitration following ookinete invasion are critical for mosquitoes to activate the mosquito complement system and mount an effective antiplasmodial response (4).
Circulating hemocytes are also key players in this defense response. Ookinete invasion triggers PGE2 release by the mosquito midgut, which attracts hemocytes to the basal surface of the midgut (13). If hemocytes come in contact with a nitrated surface, they release microvesicles, promoting activation of the mosquito complement system (5). Epithelial responses are not fast enough to effectively kill ookinetes during cell traversal, but by attracting hemocytes that move to the invasion site and subsequently release microvesicles, effective complement-mediated elimination of most ookinetes can be achieved after they egress from the epithelial cell. We found instances where patrolling hemocytes appeared to track nearby ookinetes, suggesting that they may have the ability detect and target ookinetes (Movie S8A and B). However, further studies with extensive imaging would be necessary to confirm if this is indeed the case.
We developed a new protocol for live, in vivo imaging of Plasmodium-infected mosquitoes that allows direct observation of ookinetes and their interactions with the mosquito midgut as they traverse the epithelium. This method preserves the tracheal and circulatory systems, allowing oxygen and nutrient delivery to intact mosquito organs, making it possible to acquire images for several hours without loss of organ viability. The results from our time-lapse imaging and tracking analysis are in agreement with the original timebomb model, based on confocal analysis of fixed midgut samples. They provide information on the temporal dynamics of ookinete motility and midgut traversal, as well as actin rearrangements, apoptosis, and extrusion of invaded epithelial cells. This in vivo imaging method is versatile and can be readily adapted to study later stages of Plasmodium development in the mosquito, such as the oocyst and sporozoite stages, as well as other vector-pathogen combinations.

MATERIALS AND METHODS
Sample preparation. A. gambiae CDC G-3 mosquitoes were reared according to standard protocols at 28°C in 80% humidity with a 12-h light/dark cycle and a 10% Karo syrup cotton ball. Three-day-old adult females were given water supplemented with penicillin (Sigma-Aldrich) (100 U/ml) and streptomycin (Sigma-Aldrich) (100 g/ml) antibiotics overnight and were then fed the fluorescent dye of choice the next day using an artificial feeder at 38°C. Fluorescent probes were delivered in a mixture of 0.15 M NaCl and freshly prepared 0.01 M NaHCO 3 , adjusted to pH 7.3 using HCl (15). Added fluorescent dyes included CFDA-SE (Invitrogen) (1:100) to label the midgut, SiR-actin (Cytoskeleton) (1:300) for F-actin activity, and NucView488 (Biotium) (1:50) for caspase-3 activity. Only fully engorged females were kept for subsequent infection. If hemocytes were to be labeled, those mosquitoes were additionally injected with freshly diluted (1:20 in water) Vybrant DiD (Thermo Fisher). Antibiotic treatment was kept in a 10% sucrose solution after dye feeding until infection day. The following day, mosquitoes were starved in the morning and then fed in a BALB/c mouse in the afternoon before being transferred to 19°C until mounting. Mice were infected with P. berghei-mCherry at a parasitemia level of 6% to 8%.
Sample preparation started the next day at approximately 1 h before the imaging start time to allow careful mounting. To capture ookinete invasion, imaging began at 22 h PF. For experiments observing the later events of actin and caspase dynamics, imaging began at 24 h PF. Mosquitoes were placed in a CO 2 chamber containing dry ice for knockdown and were then gently transferred to a dish on ice. Engorged females were placed in a glass well containing enough Schneider's insect medium (S0146, Sigma-Aldrich) to keep them wet but not floating. The head, wings, and legs were removed using a blade and no. 5 forceps. For mounting, a thin strip of adhesive craft putty tape was placed along the edge of a #1.5 square coverslip. Up to 15 mosquitoes were lined up directly on the cover slip, alternating anterior and posterior ends for a compact fit and reduction in twitching. Roughly 5 l of Schneider's insect medium was added to coat all mosquitoes and prevent desiccation. A glass slide was then gently pressed evenly on top of the mosquitoes-putty-coverslip until the midgut could be seen to lay flat against the slide. Samples were immediately imaged, keeping them at room temperature (22°C).