The final countdown: presence of an invasive mosquito extends time to predation for a native mosquito

Larvae of the predatory mosquito Toxorhynchites rutilus consume arthropods within container habitats, including native Aedes triseriatus and invasive Aedes japonicus mosquitoes. Previous studies, which did not account for common habitat attributes such as habitat structure and predation cues, conflict on whether Ae. triseriatus and Ae. japonicus differ in their vulnerability to predation. We conducted two laboratory experiments to assess how habitat attributes modulate Tx. rutilus predation on Ae. triseriatus and Ae. japonicus. In experiment one, we added fine particulate organic matter (FPOM) and assessed vulnerability for each species separately. Experiment two contained the following treatments: presence/absence of predation cues, presence/absence of habitat structure (FPOM and leaves) and three species combinations: Ae. triseriatus or Ae. japonicus alone, and both species together. We added one Tx. rutilus to feed in each microcosm for 24 h (experiment one and two) and until all prey were consumed (experiment two only). When reared alone, Ae. triseriatus had higher survival compared to Ae. japonicus in experiment one (71% vs. 52%) but there were no significant differences at 24 h in experiment two. When we followed the cohort to total predation, Ae. triseriatus had a lower daily survival rate compared to Ae. japonicus (hazard ratio 1.165) when the species were kept separately. When the species were mixed, however, Ae. japonicus was more vulnerable than Ae. triseriatus (hazard ratio 1.763), prolonging Ae. triseriatus time to total cohort predation. Both species were less likely to be consumed in the presence of predation cues. We detected no effect of habitat structure. These results demonstrate vulnerability is context dependent and the presence of an invasive congener can relax predation pressure on a native prey species when they co-occur in the same habitat.


Introduction
Non-native species may become invasive if they undergo rapid population growth in the absence of their co-evolved predators and parasites (Colautti et al. 2004;Heger and Jeschke 2014). However, non-native species often encounter native predators in their new range (Pintor and Beyers 2015), and outcomes resulting from invasive prey-native predator encounters partially depend on how well the two are able to detect or avoid each other (Mennen and Laskowski 2018;Narimanov et al. 2021). Many prey species that have evolved with predators, whether in their native or introduced habitat, have developed strategies to detect them, or when predation has occurred, and modify their behavior to minimize their risk. The act and byproducts of predation produce chemical cues that prey respond to (Sih 1986;Lima and Dill 1990;Chivers et al. 2001;Juliano and Gravel 2002). However, the strength of the behavioral response to predation cues seems to be a product of an evolutionary relationship between native predators and native prey. As a result, non-native species may be more vulnerable to predation by novel (native) predators (Grill and Juliano 1996;Juliano and Gravel 2002;Costanzo et al. 2011;Twining et al. 2020) and may indirectly benefit native prey by relaxing predation pressure (Rodriguez 2006;Skein et al. 2018) or competition (Wagner et al. 2010).
Small, discrete, aquatic container ecosystems; either natural (bromeliads, pitcher plants, treeholes) or associated with humans (tires, buckets, trashcans), are dominated by mosquito larvae. Within these habitats, predation-risk cues are primarily solid wastes and partially eaten conspecifics (Kesavaraju and Juliano 2010). In response to detection of predationrisk cues, larval mosquitos exhibit anti-predatory behaviors such as reduced foraging and increased resting near the water surface (Juliano and Gravel 2002). These behaviors reduce predation risk, but not without fitness trade-offs such as smaller body-size, reduced fecundity, and increased development time (Juliano and Reminger 1992;Chandrasegaran and Juliano 2019;Costanzo et al. 2011). Rapid behavioral adaptation to changes in predation pressure have been observed in the Eastern treehole mosquito Aedes triseriatus in just two generations of selection under laboratory conditions (Juliano and Gravel 2002). Additionally, inter-population differences in response to predation risk by the predator Toxorhychites rutilus between two geographically isolated populations of Ae. triseriatus with low and high levels of cohabitation with Tx. rutilus, have been observed, highlighting the importance of recent evolutionary history (Juliano and Gravel 2002).
Within natural container habitats (treeholes) across the Eastern United States, the native Elephant Mosquito, Tx. rutilus is a common and dominant predator that can exert strong pressure on prey communities Griswold and Lounibos 2006). Wild larval Tx. rutilus are generalist predators, voraciously feeding on smaller mosquito larvae, fallen terrestrial arthropods, other dipterans (Psychodidae & Chironomidae), and microcrustaceans (copepods & ostracods) (Campos and Lounibos 2000). Tx. rutilus is an ambush predator that relies on mechanoreceptors (setae) all over its body to detect prey movement while hunting (Steffan and Evenhuis 1981). There are some cases, however, where Tx. rutilus and related species will exhibit searching behavior (Linley and Darling 1993;Linley 1995;Zuharah et al. 2015). Typically, established container habitats such as treeholes and buckets hold large amounts of benthic sediment (fine particulate organic matter, FPOM) and plant/insect detritus (Macia and Bradshaw 2000, Yee et al. 2007, Beckermann andWestby, personal observation). Resuspension of the sediment by prey activity can disclose prey location and reduce visibility, making predators harder to detect. Alternatively, sediment/detritus adds complexity to the larval habitat, potentially giving prey increased refuge from predation. Previous studies evaluating the effects of habitat complexity on vulnerability of native Ae. triseriatus and invasive Aedes albopictus to Tx. rutilus found the addition of artificial leaves as habitat structure did not alter predation on either species (Alto et al. 2005). However, natural leaf substrate can alter the feeding behavior of another non-native invasive mosquito, Aedes japonicus. Increased browsing activity has been observed for Ae. japonicus in the presence of senescent white oak leaves compared to a plastic strip in the larval environment (O'Donnell and Armbruster 2007). The increase in browsing activity due to the presence of phagostimulants from natural leaf substrate (vs. artificial substrates) could potentially leave prey larvae more vulnerable to predation by Tx. rutilus. Browsing activity is considered one of the more risky behaviors (Juliano and Gravel 2002;Kesavaraju and Juliano 2004). To our knowledge, the effects of FPOM on predation risk for mosquitoes has not been tested and the role of natural substrates, such as leaves, is under-explored.
Tx. rutilus commonly cohabitates and preys upon Ae. triseriatus within their respective native range in the United States (Bradshaw and Holzapfel 1983). The non-native, invasive mosquito Ae. japonicus, which established and began expanding its US range in the late 1990's, has substantial niche overlap with Ae. triseriatus, becoming a potential competitor and a new potential prey source for Tx. rutilus (Kaufman and Fonseca 2014). Aedes triseriatus and Ae. japonicus display differences in foraging behavior which 2509 The final countdown: presence of an invasive mosquito extends time to predation for a native… may make Ae. japonicus more vulnerable to predation by Tx. rutilus, and Ae. japonicus has been shown to actively forage more often than Ae. albopictus. Aedes triseriatus' larval feeding behavior does not differ substantially from Ae. albopictus (O'Donnell and Armbruster 2007;Skiff and Yee 2014), thus we expect Ae. japonicus to spend more time foraging than Ae. triseriatus. Although it has not been quantified to our knowledge, our personal observations of these two species in detritus rich larval habitats indicate that Ae. japonicus spends more time foraging near the bottom of their container in FPOM and leaf litter compared to Ae. triseriatus. Additionally, both species exhibit increased resting behavior when exposed to predation risk cues produced by Tx. rutilus (Kesavaraju et al. 2011(Kesavaraju et al. , 2007Kesavaraju and Juliano 2004). While one laboratory study found the differences in prey vulnerability to Tx. rutilus for the two species to be insignificant (Murrell and Juliano 2013), others suggest Ae. japonicus may be more vulnerable (Freed et al. 2014;Juliano et al. 2019) with larger population declines for Ae. japonicus compared to Ae. triseriatus in a manipulative field experiment . How predation cues and habitat structure change the relative vulnerability for the two species remains to be tested.
Our goals were to identify differences in prey vulnerability to Tx. rutilus between native Ae. triseriatus and invasive Ae. japonicus, when alone or together, and in the presence or absence of predation risk cues and realistic habitat complexity. No previous prey vulnerability studies between these two species, to our knowledge, have included multiple levels of natural habitat structure and predation risk cues that mimic established container habitats in the wild. Specifically, we predicted that (1) predation cues and habitat structure alter predation rates for both species; (2) the two species will differ in their vulnerability to predation; and 3) that vulnerability may change when the species are exposed to predation singly or together.

Methods
Experiment one: habitat structure, single prey species Prey larvae rearing Ae. triseriatus and Ae. japonicus eggs were collected from Tyson Research Center (TRC) in Eureka, MO in June of 2016 and stimulated to hatch by being submerged in a 0.35 g/L nutrient broth (Difco Laboratories, Detroit, MI) solution for 24 h. Once hatched, larvae were reared on 0.001 g of 1:1 lactalbumin and liver powder and 6 mL of deionized water per larva. Once larvae reached 2nd instar, they were identified and separated by species and larvae were maintained in an environmental chamber held at 25 ℃ and a 16:8 h (L:D) photophase.

Predatory larvae rearing
Wild Tx. rutilus larvae were collected in July of 2016 from 140 L barrels established in forest habitat at TRC in 2013 (Westby and Juliano 2017). Individuals were separated into 20 mL vials with 15 mL of deionized water and fed a diet of size-appropriate larval Aedes spp. (Ae. triseriatus and Ae. japonicus) and Culex restuans, Culex territans, Anopheles barberi, and Orthopodomyia signifera when sufficient numbers of Aedes spp. were not available, all collected from barrels at TRC. Each larva received 15 prey larvae daily until they reached 4th instar, wherein we reduced the prey to 10 larvae per day.

Experimental design
We established 40 laboratory microcosms in clear polypropylene Tri-Pour® cups with 350 mL of deionized water, half of which received our treatment of 40 mL of fine particulate organic matter (FPOM) collected from the bottom of buckets that had been in the field for one year. All large detritus was removed from the FPOM which was then boiled twice to sterilize microbes. We then poured off all the extra water and vigorously stirred the FPOM until it was homogenized before we measured out the 40 mL and added it to the treatment microcosms. This created a treatment with very turbid, dark water which resembled treehole conditions (Fig. 1). Each microcosm received 20 3rd or 4th instar prey larvae of a single species (Ae. japonicus or Ae. triseriatus), and one 4th instar Tx. rutilus larva, which had been starved for 24 h to standardize their level of hunger. The experiment was run for 24 h after which predator larvae were removed and the remaining live prey larvae were counted. The experiment and all mosquito husbandry were conducted in environmental chambers set at 25 °C with a 16:8 L:D photophase.
We analyzed the proportion of live larvae using a GLM (generalized linear model) with a quasi-binomial error distribution and two fixed effects: prey species (Ae. triseriatus or Ae. japonicus) and FPOM (present or absent). The analysis was performed in SAS 9.4 using the GLIMMIX procedure.
Experiment two: habitat structure, predation cues, single and mixed species Prey larvae rearing Aedes triseriatus and Ae. japonicus were hatched from field collected eggs obtained in June of 2017. The larvae were then reared via the same methods as in experiment one.

Predatory larvae rearing
Toxorhynchites rutilus larvae were field-collected or hatched from field-collected eggs from TRC and reared individually in 20 mL vials containing 15 mL of deionized water. Each predator larva was fed a diet of 15 size-appropriate field-collected prey larvae per day with a species composition of Ae. japonicus, Ae. triseriatus, Cx. restuans, Cx. territans, An. barberi, and Or. signifera. Once larvae reached 4th instar, they were fed 10 4th instar field-collected larvae per day with the same species composition in experiment one.

Experimental design
We held Ae. triseriatus and Ae. japonicus 3 rd and 4 th instar larvae in three different prey larval assemblages: 20 Ae. triseriatus larvae alone, 20 Ae. japonicus larvae alone, and 10:10 of both species together and in the presence or absence of predation cues and habitat structure. We created 6 replicates of each treatment combination for a total of 69 (72 minus 3 that were spilled or dropped) microcosms in 400 mL clear polypropylene Tri-Pour® containers. Each microcosm contained 340 mL of deionized water which was replenished as needed. We modified the habitat structure treatment from experiment one by reducing the amount of FPOM from 40 to 1.25 mL which increased visibility, making it easier to locate prey larvae daily, but remaining a sufficient amount to mimic realistic field conditions. Additionally, we added 1.25 g of senescent sycamore leaf detritus collected from the forest floor at TRC. Leaves are a common structural element in both treeholes and artificial container habitats; thus, we were able to test the effect of two common types of habitat structure between experiment one and two (Fig. 1). FPOM was collected from 3-year-old field mesocosms, homogenized, and boiled using the same protocol as in experiment one. To create our predation risk cues, we modified the methods presented by Costanzo et al. (2011). Predation water (cue) treatments were prepared over the course of five days by adding 3rd Fig. 1 Habitat structure treatments, from left to right. Experiment one: lots of fine particulate organic matter (FPOM) only, Experiment two: a reduced amount of FPOM and sycamore leaves, Control: water only or 4th instar prey larvae according to our three species treatments (e.g. 20 Ae. triseriatus for the Ae. triseriatus treatment group) with one 3rd or 4th instar predator larva. Control treatments were also prepared over the course of five days by holding prey larvae in the three species treatments in microcosms without predation. This was done as a control to account for chemical cues left by larvae. Prey larvae were not fed during this time. After the five preparation days, all prey and predator larvae were removed.
For the experiment, fresh prey and predators, that were not part of the creation of cue or control water, were used. Predator larvae were starved for 24 h before being added as 3rd or 4th instar to each microcosm. We recorded the number of prey alive daily. Pupated or dead Tx. rutilus were removed and replaced as needed. The experiment was allowed to run until all prey larvae were consumed by Tx. rutilus. Experimental microcosms, and all mosquito husbandry, were held in an environmental chamber at 25 ℃ with a 16:8 h (L:D) photophase.
To directly compare our results from experiment one to experiment two, we analyzed the proportion of larvae alive at the 24 h mark using two GLMs (generalized linear model) with a quasi-binomial error distribution, one for the when the prey species were held singly, and the second for when the species were held together. The models contained three fixed effects: prey species (Ae. triseriatus or Ae. japonicus) predation cues (present or absent), and habitat structure (present or absent). To assess the time to total cohort predation, we used two separate Cox Proportional Hazards models due to the different numbers of prey larvae of each species if they were in single species (20 Ae. triseriatus or 20 Ae. japonicus) or mixed species (10 Ae. triseriatus and 10 Ae. japonicus). Both analyses included the same three fixed effects as the above analysis. All analyses were performed in SAS 9.4 using either the GLIMMIX or PHREG procedures.

Discussion
Aedes triseriatus and Ae. japonicus are mosquito species of public health importance and known or potential vectors of several viruses (Westby et al. 2015;DeCarlo et al. 2020). Both species are vulnerable to predation by Tx. rutilus, but there has been ambiguity among prior studies about whether the native species Ae. triseriatus, which coevolved with Tx. rutilus, is more or less vulnerable than the invasive Ae. japonicus (Murrell and Juliano 2013;Freed et al. 2014). In this paper, we further explored this question and added the nuances of: (1) aquatic habitat structure, which has been shown to act as a refuge for prey in other systems and is abundant in mosquito habitat, and (2) the presence of predation cues. Additionally, we explored whether relative vulnerability changed when the prey species were available alone or together and on short and long time scales.
Interestingly, we observed no significant effects of habitat structure from either of our experiments. We predicted that the addition of structure that closely mimics conditions experienced in treeholes would impact predation rates for one or both of the prey species tested. Specifically, we predicted FPOM would have a relatively greater effect on Ae. japonicus as we have frequently observed this species spending more time burying themselves in sediment compared to Ae. triseriatus (Westby and Beckermann, personal observations). Our current study did not include an assessment of prey behavior, so we are unable to determine if habitat structure impacted behavior, or if the two species behaved differently. It appears from our results, however, that if our habitat structure did influence behavior it was not sufficient to alter predation rates. Our results are consistent with the conclusions of Alto et al. (2005) wherein Ae. albopictus did not alter its behavior with the addition of plastic artificial leaves as habitat structure; however, prey larvae are known to browse the bacteria on natural substrates (Kaufman et al. 2008) and to increase browsing behavior when offered natural leaves compared to plastic artificial leaves (O'Donnell and Armbruster 2007). Browsing has repeatedly been shown to be risky in behavioral studies of mosquito larvae   Fig. 2 A Proportion of prey alive after 24 h from experiment one (least squares means and standard errors). Ae. triseriatus had significantly higher mean survivorship in experiment one with no effect of habitat structure. B Proportion of prey alive at 24 h when the prey were offered as a single species and C proportion alive at 24 h when the two prey species were mixed together (least squares means and standard errors). No treatment effects were significant at 24 h in experiment two (Juliano and Reminger 1992;Grill and Juliano 1996), but results from the studies presented here and behavioral studies are not directly comparable. The addition of FPOM and real leaves make visual recording prohibitive.
One hypothesis as to why habitat structure did not affect predation rates is that Toxorhynchites spp. eyesight is poorly developed in larvae, potentially owing to the dark and turbid treehole habitat where this species evolved (Steffan and Evenhuis 1981). Thus, Toxorhynchites spp. rely less on visual cues and more on tactile cues such as water movement (Steffan and Evenhuis 1981). Aquatic ambush predators may be unaffected by the addition of habitat structure whereas active searchers are impeded (Saha et al. 2009;Deboom and Wahl 2013); however, whether structure increases, decreases, or has a neutral effect on predation, as observed here, is determined by a suite of biological, behavioral, and environmental factors (Michel and Adams 2009;Klecka and Boukal 2014).
We observed significant increases in the time to total cohort predation for both Ae. japonicus and Ae. triseriatus when held in water containing predation cues. These results support the conclusions reached in behavioral assays using video capture that invasive Ae. japonicus is sensitive to predation cues produced by Tx. rutilus, despite its recent evolutionary history with this predator (Kesavaraju et al. 2011). Ae. japonicus encounters predators in its native range (Soto 1996;Sunahara et al. 2002) and apparently maintained its ability to detect and respond to  predation cues during its establishment and range expansion in the US. It is also plausible that Ae. japonicus has had sufficient time to evolve an antipredatory response to Tx. rutilus specifically since its introduction in our study area at least ten years before we performed these experiments (Gallitano et al. 2006). Prey naivete is known to decrease since time of introduction (Anton et al. 2020) and rapid evolution in response to predation cues has been observed for Ae. triseriatus in the laboratory (Juliano and Gravel 2002). Predation by Toxorhynchites spp. in natural habitats, and these types of experiments, leaves behind small amounts of organic matter that support bacterial growth (Albeny-Simões et al. 2014). This can serve as a food source for filter-feeding prey larvae with surviving larvae emerging larger than those from non-predation or cue treatments (Costanzo et al. 2011). It is not known if the bacteria also serves as a food source for Toxorhynchites spp. which could indirectly reduce pressure on prey. While Toxorhynchites spp. can be reared in laboratory colonies on liver powder, or pieces of dead insects, they are not considered filter-feeders or browsers and the resulting adults are small (Steffan and Evenhuis 1981). In our experiment, there were few pieces of dead prey left by Tx. rutilus so any nutritional benefit would have to have been gained by filter-feeding in the presence of live prey. While future research on this subject is warranted, we believe that the increased survival in the cue treatments was due to prey behavioral changes resulting in prolonged time to predation (Grill and Juliano 1996).
Whether the native or non-native invasive mosquito species was more vulnerable to predation was condition-dependent: our experiments came to different conclusions depending on when the data was collected (24 h vs. time to total predation) and if the species were offered as prey separately or together. We found evidence that Ae. triseriatus was less vulnerable after 24 h (experiment one), no difference between the species at 24 h when held separately (experiment two), Ae. triseriatus was more vulnerable to total predation when reared alone (experiment two) and that Ae. japonicus was more vulnerable to total predation when offered with Ae. triseriatus at the same time (experiment two), which extended the time to total predation for Ae. triseriatus. Based on the conflicting results from our own studies it is not surprising that there was ambiguity among the other three published studies (Murell et al. 2013;Freed et al. 2014;Juliano et al. 2019). We suggest that differences in experimental design, such as the length of the trials and the number of prey offered, can alter the outcome of these experiments and that context dependence is an important driver of the differences in results (Juliano 2009).
In our second experiment, we found that when Ae. japonicus was offered as the single prey species, it experienced extended time to predation compared to native Ae. triseriatus. Since Ae. triseriatus exhibits such a strong anti-predatory behavioral response and shares a longer evolutionary history with Tx. rutilus, this result was not in line with our original predictions. Nor did this align with the results from our first experiment where Ae. triseriatus was seemingly less vulnerable to predation than Ae. japonicus. Moreover, Murrell and Juliano's (2013) single species prey vulnerability assays between Ae. japonicus and Ae. triseriatus showed predation rates of the two species by Tx. rutilus over a period of 4 days were statistically indistinguishable. However, our methods and objectives differed significantly from Murrell and Juliano (2013) which focused on attack rates over a 24 h period and replenished prey species after they were eaten. We did not replenish prey, instead allowing the predator to eat until local extinction of prey occurred. This could explain the differences in our results since Tx. rutilus is a Type II functional predator and changes its foraging strategies based on the proportion of available prey (Russo 1983). Our results indicate it is still unclear whether Ae. triseriatus or Ae. japonicus are less vulnerable to predation from Tx. rutilus comparatively when offered separately.
In our mixed species treatments, we found Ae. triseriatus was significantly less vulnerable to predation when held with Ae. japonicus in both cue and non-cue treatments, respectively. This shows a reversal in the results of the single species treatments where Ae. japonicus was less vulnerable to predation, and perhaps a more realistic picture of prey vulnerability in the wild since container habitats frequently contain multi-species assemblages. Most interestingly, we show that the presence of an invasive competitor may assist a native species through predator-mediated effects. When the presence of one species reduces vulnerability to a shared predator for another prey species, it is known as apparent mutualism (Holt and Lawton 1994). This phenomenon is important to consider when studying systems with polyphagous predators, such as Tx. rutilus, due to the possibility of the predator's prey preference changing when presented with multiple species. Testing the interaction of multiple prey species and their shared predator also gives a better picture of how predation of invasive species may occur. As we saw in the results from our two studies, when prey were offered as single species, or at 24 h, there was still ambiguity on whether Ae. japonicus was more vulnerable, but when held with Ae. triseriatus, Ae. japonicus were more vulnerable to predation. In our case, had we not offered the prey species together, or followed them over time, our conclusions from this study would be different, highlighting the need for more context-dependent studies of predator-prey interactions (Juliano 2009). Furthermore, the results from the Juliano et al. (2019) experimental field predator removal study reported strong evidence that Ae. japonicus were more negatively affected by Tx. rutilus predation, whereas there were no significant differences in pupal and larval abundances of Ae. triseriatus whether Tx. rutilus was present or removed from their habitat. There is now both field and laboratory evidence that Tx. rutilus predation more negatively affects Ae. japonicus than its native competitor, Ae. triseriatus when held together. Future work should explore the interactions between these three species as smaller instars. It is possible that overcompensation via the Hydra effect could occur if the predation occurs while the larvae are small (McIntire and Juliano 2018) which we did not test in the experiments presented here.
Our results show that Ae. japonicus' presence in container habitats can relax predation pressure on Ae. triseriatus ). This, coupled with its presence reducing gregarine parasite prevalence  suggests that Ae. japonicus may be more of an apparent mutualist for Ae. triseriatus than an antagonist. Furthermore, there is limited evidence of asymmetric competition between the two species (Alto 2011;Freed et al. 2014) or of displacement of Ae. triseriatus in containers in forested areas (Westby et al. , 2021.