Although, resource allocation by heterotrophs has been studied extensively, the extent to which their behavioral responses is manipulated by pathogens/parasites reflects the fine-tuned evolutionary adaptation in nature (Roy and Raguso, 1997; Ebbert and Nault, 2001; Eigenbrode et al., 2002; Hurd, 2003; Maris et al., 2004; Belliure et al., 2005; Lacroix et al., 2005; Lefèvre et al., 2006; Mauck et al., 2010; Bosque-Pérez and Eigenbrode, 2011). The choice of plant hosts for oviposition by insects remains an important research area in entomology and ecology (Bernays and Chapman, 1994; Schoonhoven et al., 2005; Tilmon, 2008). For insects, the preference-performance hypothesis (PPH) suggests that females will maximize their fitness by laying their eggs on plant types on which their offspring will perform best (Gripenberg et al., 2010). This hypothesis, also known as the optimal, oviposition theory (Jaenike, 1978), the naive adaptionist theory (Courtney and Kibota, 1990), or the “mother knows best” principle (Valladares and Lawton, 1991), was originally proposed by Jaenike (1978). Some studies testing female oviposition preferences and larval performance have established positive correlations between females and their offsprings (Craig et al., 1989; Heisswolf et al., 2005; Staley et al., 2009). Other studies, however, showed that mothers could make poor oviposition decisions, leading to a weak preference–performance relationship between female adults and offsprings (Rausher, 1979; Scheirs et al., 2004; Digweed, 2006; Gripenberg et al., 2007). These “bad motherhood” cases have caused extensive debates (Mayhew, 2001; Gripenberg et al., 2010). Alternative hypotheses include optimal foraging (Scheirs et al., 2000), enemy-free space (Denno et al., 1990), and insect neural capacity (Bernays, 2001).

The current study concerns the effects of virus infection of host plants on herbivorous insects. The influences of virus-infected host plants on herbivorous insects can be neutral (Matsuura and Hoshino, 2009), beneficial (Maris et al., 2004; Belliure et al., 2005), or deleterious (Rubinstein and Czosnek, 1997) depending on the specific combination of herbivore, virus, and plant. In addition, viruses can affect plant-derived volatiles such that herbivorous insects are repelled (van Molken et al., 2012) or attracted (Srinivasan et al., 2006; Ngumbi et al., 2007). These manipulations can strongly influence rates of virus transmission and thus have important implications for ecology, human health, and agriculture (Hurd, 2003; Lefèvre et al., 2006, 2009). Although, host choice by insects has been continually studied, the influences of virus infection of host plants on host choice by vector and non-vector insects are not well-understood.

Datura stramonium L. (Solanaceae), one of the hosts included in the current study, is a summer weed that originated in the Americas and that has a worldwide distribution (Holm et al., 1997). D. stramonium is also a host for many plant viruses, including Tomato yellow leaf curl virus (TYLCV) (Cohen and Nitzany, 1966) and Tomato spotted wilt virus (TSWV) (Moyer, 1999). TSWV (genus Tospovirus, family Bunyaviridae) infects and causes disease in hundreds of plant species, including many ornamentals and vegetables (Moyer, 1999). TSWV is transmitted in a propagative and circulative manner by thrips (Thysanoptera: Thripidae; Ullman et al., 1992).

The sweetpotato whitefly, Bemesia tabaci (Gennadius) (Hemiptera: Aleyrodidae), is a vector of Begomoviruses but not for TSWV. Although, there has been much debate about whether B. tabaci is “a species complex or complex of species” (Perring, 2001; Brown, 2010; Dinsdale et al., 2010; De Barro et al., 2011), the most current molecular evidence reveals that B. tabaci actually includes at least 24 genetically distinct but morphologically indistinguishable cryptic species (De Barro et al., 2011). The most damaging and widespread cryptic species are the Mediterranean genetic group (biotype Q) and the East-Minor Asia 1 genetic group (biotype B). In a field survey conducted in 2012, we observed B. tabaci Q outbreaks on TSWV-infected plants in several provinces of China, although B. tabaci Q is not a vector of TSWV. Previous study has demonstrated that TSWV infection can decrease B. tabaci Q performance on pepper (Pan et al., 2013). TYLCV infection, in contrast, increases B. tabaci Q (vector) attraction and performance on its host, D. stramonium (Chen et al., 2013).

Previous research examining virus effects on plant-insect interactions has focused primarily on vectors, whereas the effects of virus-infected plants on non-vectors have been largely overlooked (Su et al., 2016). In this study, we investigated the effects of TSWV infection on a non-vector B. tabaci Q. The outbreaks of B. tabaci Q were occurred on TSWV-infected pepper even though the performance of this non-vector was compromised. In addition, we test the volatiles released by TYLCV- and TSWV-infested D. stramonium. We hypothesized that both viruses induce similar changes in volatiles emitted by D. stramonium to make B. tabaci Q select virus-infested D. stramonium.

We found that TSWV infection of D. stramonium increased the settling and oviposition of the non-vector B. tabaci Q on D. stramonium even though some of its life history traits were reduced. These results coupled with a previous study (Chen et al., 2013) indicate that virus-infection may, in some cases, cause females to select hosts that do not necessarily maximize their performances. The host selection may depend on the by-products of host infection by the virus. We further tested the hypothesis using peppers, and again found that the non-vector B. tabaci Q elected to settle and oviposit on virus-infected pepper. Similarily, B. tabaci Q preferred TSWV-infected over healthy peppers, even though TSWV decreased Q performance (Pan et al., 2013). Our results also demonstrated that four volatiles released from TYLCV-infected and TSWV-infected plants (o-xylene, α-pinene, phenol, and 2-ethyl-1-hexanol) can repel or attract B. tabaci Q in ways that are consistent with the effects of virus infection on host behavior (Figures 4, 5). In other words, virus-infected plants tended to release greater quantities of volatiles to attract B. tabaci Q adults, while healthy plants tended to release greater quantities of volatiles to fend off B. tabaci Q adults.

From an evolutionary viewpoint, we expect females to maximize their fitness by laying eggs on plants that maximize offspring's fitness (Gripenberg et al., 2010). However, these expectations may or may not be realized on virus-infected plants. In our previous study, B. tabaci Q preferred to settle and oviposit on TYLCV-infected rather than on healthy D. stramonium and had greater fitness when feeding on TYLCV-infected D. stramonium, suggesting an indirect mutualism between B. tabaci Q and TYLCV (Chen et al., 2013). Similar phenomenon has been documented in other insect-virus systems (Montllor and Gildow, 1986; Castle and Berger, 1993; Jiménez-Martínez et al., 2004a; Srinivasan et al., 2006; Ngumbi et al., 2007). In our case, B. tabaci Q preferred virus-infected D. stramonium, however, the overall performance of whiteflies on TSWV-infected plants was reduced. Although, we did not examine other traits, including female cost for oviposition or offspring protection from natural enemies, which also contribute to insect fitness, our results demonstrate a deviation from the mother-knows-best hypothesis. Under the ecological and evolutionary context, we speculate that the apparent “loss” for B. tabaci, a non-vector of TSWV, represents a “gain” for TSWV. By reducing the fitness/performance of B. tabaci, the insect vector for many other plant viruses, the transmission of competing pathogens will be compromised, which means a limited or no sharing of host plants for TSWV.

Mauck et al. (2012) suggested that vectors prefer to settle on plants infected with persistently transmitted (PT) viruses, while non-persistently transmitted (NPT) viruses generally have no effect on vector settling and feeding or cause infected plants to be less attractive than healthy plants. Mauck and his colleagues believe that PT viruses typically enhance host plant quality for vector herbivores, resulting in enhanced vector herbivore fitness. NPT viruses, in contrast, often decrease or have no effect on host plant quality (Mauck et al., 2012). For example, pepper plants infected with TSWV increased the survival and oviposition of the spider mite Tetranychus urticae (Belliure et al., 2010). The growth of Spodoptera exigua caterpillars was accelerated on tomato plants infected with Tobacco mosaic virus (Thaler et al., 2010). Our previous study showed the enhanced B. tabaci performance on TYLCV-infected D. stramonium (Chen et al., 2013). In this study, however, we demonstrated that quality of the virus-infected plants might not be the deciding factor for B. tabaci to choose the host plants.

To test whether our results applied to other hosts, we determined whether TSWV infection affected B. tabaci Q settling (Figure 3A) and oviposition (Figure 3B) on pepper. The results with pepper were the same as those with D. stramonium in terms of settling and oviposition preferences. Consistent with our finding, several previous studies indicated that Cucumber mosaic virus (CMV) significantly decreases host quality (the aphid Myzus persicae and Aphis gossypii performed poorly on infected host plants) but enhances the attractiveness of infected host plants to aphids (Mauck et al., 2010). We suggest that regardless of how virus infection influences insect fitness, both vector and non-vector insects may prefer to settle and oviposit on virus-infected plants rather than on healthy plants, i.e., both vectors and non-vectors may be “manipulated” by viruses. Although, the ecological relevance of such manipulation of non-vectors is unknown, we speculate that these non-vector insects might be victimized by the co-evolution between viruses and their vectors due to the shared chemical signatures emitted by the virus-infect plants.

Our results also suggest that the preference of B. tabaci Q for virus-infected plants may be mediated by virus-induced changes in plant volatile emissions. Virus-induced effects on plant odor cues are the induction of specific chemical emissions by plant pathogens (Anderson and May, 1991; Jiménez-Martínez et al., 2004b). For example, enhanced preference of sandflies to hamsters who infected with Leishmania was attributed to changes in host-derived odor cues (O'Shea et al., 2002), and it is considered that volatiles also might explain a study result that the children from western Kenyan harboring the gametocytes (the stage transmissible to mosquitoes) of the Plasmodium falciparum attracted strongly more Anopheles gambiae than uninfected children or those harboring the asexual (non-infective) stage of the parasite (Lacroix et al., 2005). The pathogen causing Dutch elm disease has been revealed to up-regulate volatiles that attract vectors Hylurgopinus rufipes to infected host trees (McLeod et al., 2005). van Molken and his colleagues found that virus infection reduces the attractiveness of host plants for a non-vector insect (van Molken et al., 2012), suggesting the ecological benefits of virus infection for host plants by inducing insect repellence. Our results, however, showed that virus infection attracted a non-vector insect to the host plant. Additional experiments are warranted to resolve this discrepancy.

In summary, this study provides empirical evidences showing that virus-infection increases insect settling and oviposition but reduces insect performance; the long-term evolutionary consequences of these effects are unclear. Specifically, the preference for virus-infected plants in B. tabaci is correlated with an increase in phenol and 2-ethyl-1-hexanol and a decrease in o-xylene, α-pinene, within the volatile profile emitted by D. stramonium plants. As a non-vector of TSWV, however, insect performance seems to be impaired since adult body size, longevity, survival, and fecundity diminished in TSWV infected D stramonium plants.

Nevertheless, other factors may affect host selection as well. Trichome plays an important role in the host selection process, ovipositional behavior, and larval performance (Shanower et al., 1999; Mayhew, 2001; Handley et al., 2005). Trichomes and their exudates can reduce the efficiency of natural enemies and thus may have negative or positive influences on insect herbivory insect populations (Blaney and Simmonds, 1990). Besides there are four major groups of defensive compounds in plants that impact insect performance (Schoonhoven et al., 1998), including phenolics (e.g., polyacetates, tannins, and lignins), terpenoids (e.g., cardenolides and terpenes), cyanogenic glycosides and glucosinolates, and nitrogen-containing compounds (e.g., non-protein amino acids and alkaloids). These compounds can have both direct and indirect influences on the performance of herbivorous insects (Awmack and Leather, 2002). TSWV infection may induce changes in trichome traits or the release of defensive compounds in host plants, which warrants further investigation.

GC contributed to the chemical ecological laboratory work, performed the major part of entomological manipulations, participated in data analysis, participated in the conception and design of the study and drafted the manuscript; QS and XS carried out ecological manipulations; XL, ZP, and HZ contributed to viral and microbiological manipulations; WX and BX contributed to insect and plant husbandry and data collection; SW and QW carried out the statistical analyses; YZ and XZ contributed to the conception and design of the study, coordinated the study and edited the manuscript. All authors gave final approval for publication.

This work was supported by the State Key Program of National Natural Science Foundation of China (31420103919), the 973 Program (2013CB127602), the National Natural Science Foundation of China (31572014), China Agriculture Research System (CARS-26-10), Beijing Training Project For The Leading Talents in S and T (LJRC201412), Graduate Research and Innovation Project in Hunan Province (CX2015B248) and the Beijing Key Laboratory for Pest Control and Sustainable Cultivation of Vegetables. The granting agencies had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.