Interactions Between Nucleopolyhedroviruses and Polydnaviruses in Larval Lepidoptera

The field dynamics of some insect populations are strongly influenced by two types of insect viruses: the nucleopolyhedroviruses (NPVs) and the polydnaviruses (PDVs). Although greatly different in origin and mode of infection, both viruses produce considerable mortality directly and indirectly in the field, and have evolved reproductive strategies that use the same life stage of the same host insects. The life histories of these host insects are such that there are many opportunities for coinfection and competition between baculoviruses and polydnaviruses, although these relationships remain largely theoretical and unexplored.


Introduction
The field dynamics of some insect populations are strongly influenced by two types of insect viruses: the nucleopolyhedroviruses (NPVs) and the polydnaviruses (PDVs). Although greatly different in origin and mode of infection, both viruses produce considerable mortality directly and indirectly in the field, and have evolved reproductive strategies that use the same life stage of the same host insects. The life histories of these host insects are such that there are many opportunities for coinfection and competition between baculoviruses and polydnaviruses, although these relationships remain largely theoretical and unexplored.
The hosts of both these viruses are found primarily in the insect order Lepidoptera, the moths and butterflies. The viruses are members of two large and diverse families: the Baculoviridae, pathogens of insects that fit the common conception of disease-causing organisms, and the Polydnaviridae, genome-integrated wasp symbionts with a unique natural history. Although the Baculoviridae are known from several insect orders, they occur in their greatest variety as the nucleopolyhedroviruses in the larval Lepidoptera. This is probably because the relatively long, plant-feeding caterpillar stages of moths and butterflies give ample opportunities for per os infection, as we will describe in detail below. Many nucleopolyhedroviruses known from the Hymenoptera typically occur in families feeding on leaves during the larval stage, as well, but they will not be included here. The ease of rearing lepidopteran larvae, and their place as pests of human crops, has also led to an accumulation of data on their viral diseases, and the many nucleopolyhedroviruses infecting Lepidoptera make this insect order a good starting point for exploring interactions with other viruses known to occur within them, polydnaviruses in particular. As symbionts of parasitoid wasps, polydnaviruses are also found in lepidopteran larvae. There they produce products that abrogate the larval immune system. Larval hosts immunosuppressed by a polydnavirus fail to muster the encapsulation response that would otherwise kill the parasitoid eggs and larvae; the wasp mutualist, or carrier, brings the polydnavirus to its host to ensure the survival of its own progeny. Our interest comes from past research in the field on population and disease dynamics of outbreaking lepidopteran forest pests, although we do not limit ourselves to these here.
NPVs initiate infection when a susceptible host ingests virus in the form of occlusion bodies (OBs) present on host plants (Fig. 1A). OBs are composed primarily of the protein polyhedrin, which forms a paracrystalline matrix into which occlusion derived virus (ODV) is embedded. The polyhedron provides the embedded ODV protection from environmental elements such UV light. NPVs produce another form of virus, termed the budded virus, that is produced in the early stages of viral replication (Rohrmann, 2011). Within the alkaline environment of the larval midgut, OBs dissolve, thereby releasing ODV. To initiate an infection, ODV must first traverse a physical structure termed a peritrophic membrane (Fig.  1B), which is composed of proteins, mucopolysaccharides, and chitin (Pritchett et al., 1984;Hegedus, et al., 2009). The peritrophic membrane provides a barrier to gut cells to bacteria, viruses, fungi, and physical damage from ingested plant material. The peritrophic membrane is in a constant state of regeneration from epithelial cells as larvae feed, and the movement of food material through the insect gut also causes loss of the peritrophic membrane.
After penetration of the peritrophic membrane, ODV gains entry into midgut cells by a type of fusion process (Fig. 1C), although this has resisted definitive characterization (Granados & Lawler, 1981). The type NPV, Autographa californica multiple NPV (AcNPV), initiates the infection cycle by infecting columnar epithelial cells within the midgut and regenerative epithelial cells in Trichoplusia ni (Keddie et al., 1989) or Spodoptera exigua larvae (Flipsen et al., 1995). Several factors are involved with the initial act of infection that includes ODV binding to midgut cells at cell receptors, and viral entry into the cells. All sequenced baculoviruses contain genes that code for per os infectivity factors (PIFs) that are associated with ODVs but not budded virus (Faulkner et al., 1997;Kikhno et al., 2002;Fang et al., 2006;Harrison et al., 2010;Fang et al., 2009). The pif genes include p74-pif, and pif genes 1-5, Ac119, Ac22, Ac115, Ac96, and Ac148, respectively. Deletion of any of the genes from a viral genome significantly decreases but does not eliminate per os infectivity (d'Alencon et al., 2004;Crouch et al., 2007). The PIFs, with the exception of PIF3, are thought to be involved in binding or interacting with the midgut cells that leads to infection (Ohkawa, et al., 2005;Li et al., 2007;Peng et al., 2010;Horton & Burand, 1993).
Upon entry into midgut cells the nucleocapsids are actively transported to nuclear pores in a process that uses actin polymerization (Ohkawa et al., 2010) (Fig. 1C). Viral DNA is then released into the cell nucleus and viral replication ensues (Rohrmann, 2011) (Fig. 1C).
During the early phase of viral replication, BV are produced that bud from midgut cells and infect tracheal epidermal cells, which penetrate the basal lamina (Volkman, 2007). Infection spreads via the tracheal system and haemocytes until many different cells are infected (Engelhard et al., 1994) (Fig. 1D). During the later phase of viral replication, ODV are produced and packaged within OBs. Upon the host's death, liquefaction occurs, releasing OBs into the environment to infect another host (Reardon, 1996;Riegel & Slavicek, 1997). Insect hosts have developed several mechanisms to thwart baculovirus infections including the peritrophic membrane physical barrier, developmental resistance (Engelhard & Volkman, 1995), melanization of infected cells, and apoptosis of infected cells. The Lymantria dispar MNPV (LdMNPV) exhibits a LD50 (74 OBs) in L. dispar (Hoover et al., 2002); however, by the middle of the fourth instar, the LD50 of LdMNPV is 18-fold higher than in newly molted larvae. The majority of cases of developmental resistance within an instar are midgut-based (Haas-Stapleton et al., 2003). As the virus infects midgut cells the cells are in a continual process of renewal and sloughing. Consequently, if BV are produced and escape the midgut cell before it sloughs off and is excreted, the virus can infect the tracheal cells and usually generate a systemic infection (Engelhard et al., 1994). However, loss of tracheal epidermal cell infections can occur (Haas-Stapleton et al., 2003), and consequently the host can escape viral infection. In addition, larval hosts are less susceptible to viral infection between and within larval instars, due to a process termed developmental resistance (Engelhard & Volkman, 1995).
Because the main way in which polydnaviruses might affect NPV pathogenesis is via immunosuppression, it is important to note that the immune response is a vital part of larval response to infection. Hoover et al., (2002) hypothesized that immune responses play a role in systemic resistance in L. dispar to LdMNPV. In non-or semi-susceptible hosts, cellular immune responses to AcNPV-infected tissues can occur (Rivkin et al., 2006;Trudeau et al., 2001;. Haemocytes of some hosts are refractory to AcNPV infection and replication, which slows or stops the spread of the virus within the haemocoel (Rivkin et al., 2006). Apoptosis of AcNPV-infected cells can also serve as an antiviral response (Clarke & Clem, 2003;da Silveira et al., 2005;Zhang et al., 2002). In a recent study using LdMNPV, McNeil et al. (2010a) found that encapsulation and apoptosis of infected tissues are mechanisms of host defense. Elimination of infected tissue through encapsulation and melanization can occur in lepidopterans infected with AcNPV, but has only been reported in non-susceptible hosts such as Helicoverpa zea and Manduca sexta (Trudeau et al., 2001; and did not differ in effectiveness at different times within the instar. Apoptosis of tracheal epidermal cells was hypothesized to contribute to AcNPV resistance in Spodoptera frugiperda (Haas-Stapleton et al., 2003) and it appears to be an effective defense against AcNPV infection in the fat body and epithelium of IH-injected S. frugiperda (Clarke & Clem, 2003) and S. exigua (Clem, 2005). Removal of midgut infections in L. dispar probably involves apoptosis of infected midgut epithelial cells (Dougherty et al., 2006) and is considered a conserved mechanism for midgut based resistance to NPVs in most lepidopterans (Volkman, 2007).
A recent study by McNeil et al. (2010b) found that mid-instar L. dispar larvae exhibited a higher degree of hemocyte immunoresponsiveness, a greater potential hemolymph phenoloxidase (PO) activity at the time the virus is escaping the midgut to enter the hemocoel, greater FAD-glucose dehydrogenase (GLD) activity, and more targeted melanization of infected tissue, which correlate with reduced viral success in the host. PO and GLD are components of the humoral immune response associated with melanized encapsulation (Lovallo & Cox-Foster, 1999;Nappi & Christensen, 2005). Phenoloxidase has potent anti-microbial properties, and its activation is tightly controlled within the hemolymph to prevent non-specific activation from harming the host (Jiravanichpaisal et al., 2006). PO activity in the hemolymph of Heliothis virescens larvae has been shown to have virucidal effects in vitro (Shelby & Popham, 2006) and previous research has supported the role of PO and cellular immunity in immune defenses against viruses (Stanley & Shapiro, 2007, 2009Shrestha & Kim, 2008). GLD is induced during encapsulation responses to pathogens (Cox-Foster & Stehr, 1994;Lee et al., 2005). GLD is also involved in the production of free radicals derived from quinones during melanization in insects, and has been hypothesized to strengthen melanized capsules (Cox-Foster & Stehr, 1994), which could also negatively impact viral success.

Polydnaviruses
Polydnaviruses are multipartite double-stranded DNA viruses originating in koinobiont endoparasitic Hymenoptera, commonly called parasitoid wasps. The polydnavirus genome is integrated into that of the "carrier" wasp species, and is transmitted through the germ line (Stoltz, 1990;Belle et al., 2002;Wyler & Lanzrein, 2003;Bezier et al., 2009 ). The origin, replication, and function of polydnaviruses, and their symbiotic relationship with their carriers, make this system unique among members of any taxon. What follows is a short review of the discovery of the polydnaviruses, and our current understanding of their reproduction and functions. We will use the term carrier to refer to wasps bearing an genomeintegrated polydnavirus. This avoids the connotations of host and possible confusion with the host of the wasp -here a lepidopteran larva, or caterpillar. Similarly, polydnavirus infection is used to refer to the polydnavirus in the larval host of the carrier wasp, not in the wasp itself.
The discovery of polydnaviruses began with the work of George Salt. Salt had a career that spanned more than a half-century, beginning in the 1920s. His research on the underlying mechanisms of insect immunity culminated in the treatise, The Cellular Defence Reactions of Insects (Salt, 1970). This synthesis included the results of earlier work, the most pertinent of which was an experiment involving placing washed eggs of the parasitoid Venturia (Nemeritis) canescens into living larvae of its host Ephestia kuehniella (Salt, 1965; see also Rotheram, 1973). Washed eggs were encapsulated by hemocytes within the larvae, while intact unwashed eggs were not. Salt used simple light microscopy to find the origin of this resistance; a "coating" on the eggs which originated in the calyx of the female parasitoid. This work certainly informed the subsequent seminal research of Stoltz, Vinson, and Mackinnon (Stoltz et al., 1976;Stoltz & Vinson, 1977. Surveys and research led by these and other researchers in the 1970s and 1980s led to a preliminary understanding of the origin and function of the baculovirus-like particles seen in the calyx fluid of some ichneumonids and braconids Cook & Stoltz 1983;Fleming & Summers 1986). These particles were injected by carrier wasps into larval hosts along with eggs and toxins, and expressed genes while in that host, but did not replicate in host cells. Parasitoid eggs from such wasps, when placed in a host without other calyx-derived products such as the virus-like particles and wasp toxin, were invariably recognized as foreign, encapsulated, and destroyed (Edson et al., 1981;. The particles were clearly necessary for successful reproduction of the carrier wasp, but the emergent question was how these baculovirus-like particles were acquired or produced. The advent of modern molecular analyses allowed researchers to answer this question definitively. The baculovirus-like particles discovered in braconid and ichneumonid wasps were determined to be a new form of virus, integrated into the genome of the carrier wasp as proviral genetic sequences (Stoltz, 1990;Bezier et al., 2009). Although these sequences were found in both male and female wasp genomes, they were excised to form free viral particles only in the ovary calyx cells of maturing female wasps. This is in stark contrast to a typical baculovirus life cycle (Cory & Myers, 2003). In the polydnavirus life cycle, replication and vertical transmission occur in the carrier wasp, while a form of horizontal transmission and deleterious infection occurs in the carrier wasp's larval host. This has also been previously defined as the two "arms" of the PDV life cycle (Turnbull & Webb, 2002). The larval host takes the brunt of the effects of the polydnavirus in the form of immunosuppression caused by polydnavirus DNA expression -but not through polydnavirus replication in and subsequent destruction of its cells.
Most of the known polydnaviruses are carried by wasps that parasitize larval Lepidoptera, in subfamilies of the Ichneumonidae (these polydnaviruses are termed ichnoviruses) and Braconidae (the bracoviruses) (Stoltz & Vinson, 1977;Fleming & Summers, 1991;Stoltz & Whitfield, 1992;Webb & Strand, 2005). At this time, known polydnavirus carrier wasps occur only in the braconid subfamilies Cardiochilinae, Cheloninae, Mendesellinae, Khoikhoiinae, Miricinae and Microgastrinae, and ichneumonid subfamilies Campopleginae and Banchinae (Table 1) LaPointe et al., 2007;Whitfield & O'Connor, 2012). Even if only currently named species are considered, these subfamilies contain a total of almost 30,000 wasp species, with many expected to harbor polydnaviruses. Although these braconid and ichneumonid PDVs are at least as diverse as the wasp families themselves and are apparently unrelated to each other, we may use the braconid lineage as an example of how this association originated (Whitfield, 1997; see the excellent review by Whitfield & O'Connor, 2012). Between 85-100 million years ago, an early braconid acquired an integrated nudivirus in a manner not fully understood (Banks & Whitfield, 2006;Murphy et al., 2008). This virus has evolved along with the carrier wasps to the present day, with greater or lesser diversity as a function of the diversification of the carrier wasp (Murphy et al., 2008;Bezier et al., 2009). Modern inexpensive sequencing techniques continue to provide ever-greater detail of these phylogenies.

Polydnavirus
Wasp A description of the life cycle and transmission process of a typical PDV can be started with the developing wasp pupa as well as at any other point in the cycle. As noted, PDVs have a transmission cycle unlike that of any other pathogen, and it has even been argued that they may not truly fit the definition of an insect virus (Whitfield & Asgari, 2003;Stoltz & Whitfield, 2009). They exist as proviral DNA integrated into the genome of the carrier wasp species (both male and female), and are "free-living" only during reproduction in female wasp ovarian calyx tissue, or after injection into the host of the carrier wasp. The injected PDV does not reproduce; it appears that some PDVs even lack the genes needed for selfassembly (Bézier et al., 2009). The PDV is at a virtual dead end, except that it allows for the reproduction of its genome in the developing wasp larva by blocking host immune responses that would otherwise encapsulate and kill it. By the same token, for most if not all carrier wasps, it is likely that successful parasitism is impossible if the polydnavirus is removed from the system: indeed this has been shown explicitly in a few systems (Edson et al., 1981;. The large, segmented, double-stranded PDV is excised from the wasp genome only in later pupal and adult stages, and only in female wasps (Gruber et al., 1996;Pasquier-Barre et al., 2002;. The exact process by which this occurs has only recently been elucidated for some polydnaviruses (Annaheim & Lanzrein, 2007;Bezier et al., 2009), although, as stated previously, it is believed that some PDVs, particularly the bracoviruses, rely on the wasp host for assembly. Virus particles are assembled in the nuclei of calyx cells and accumulate in the lumen of the oviduct.
When the female wasp has emerged and is sufficiently mature, she searches for a suitable host to parasitize ( Fig. 2A). During a sting event, the PDV is injected into the wasp's host along with additional proteins, toxin, and parasite eggs (Fig. 2B) (Pennachio & Strand, 2006). Once in the larval host, polydnavirus genes produce proteins that degrade the host's immune response and prevent encapsulation of the parasitoid egg or larva (Fig. 2C) (Stoltz et al. 1988). This is done in concert with other injected materials such as toxins. Cotesia melanoscela bracovirus (CmeBV), for example, acts in concert with wasp-produced venom on hemocytes in parasitized gypsy moth larvae (Guzo & Stoltz, 1985;Stoltz et al., 1988;Summers & Dib-Hajj 1995;Nalini et al., 2008). Among the best understood effects of PDVs on lepidopteran larvae are those occurring within the hemocytes of parasitized larvae. If PDVs were known only for their ability to prevent encapsulation of foreign bodies, they would rightfully be the subjects of intense research. However, the immunosuppressive activity of polydnaviruses has been shown to extend to effects on baculovirus pathogenesis. This was first shown explicitly by Washburn et al. (1996), who showed that parasite injection of the Campoletis sonorensis ichnovirus led to immunosuppression that resulted in a more rapid spread of the AcMNPV in H. zea larvae. This effect was attributed to PDVmediated abrogation of the hemocytic response as pertains to recognition of altered-self: H. zea is believed to clear some AcMNPV infections by encapsulating its own infected cells. This phenomenon, of special interest to the microbial ecologist, will be explored in greater detail below. So far as naturally-occurring interactions between the viruses are concerned, they are the most likely to be relevant.

Field interactions between polydnaviruses and nucleopolyhedroviruses
We have seen that parasitoids both produce and transmit polydnaviruses in order to suppress the immune system of their caterpillar host, and are thus able to successfully reproduce within it. In many cases baculoviruses and polydnaviruses require equivalent and possibly incompatible resources from the host, if they occur in the same host species (Caballero et al., 1991;Hochberg, 1991a). For example, LdMNPV and C. melanoscela are commonly encountered mortality agents of the gypsy moth in field populations in eastern United States. The timing of mortality caused by these agents overlaps completely (Woods et al., 1991;Crossman, 1922). Both enter a host during its larval stage and convert some or most of the host biomass into virus particles (baculoviruses), or aid in its conversion to a parasitoid (polydnaviruses). If a baculovirus kills or physiologically degrades a larva past a certain point, it will no longer sustain a larval parasitoid. If a parasitoid kills its host outright, as most do, baculovirus pathogenesis will be stopped at the point of host death. This would result in reduced or absent production and release of occluded virus, which is necessary to transmit the baculovirus to new hosts. Theoretically, then, a conflict arises whenever polydnaviruses and baculoviruses share a host. How may these organisms interact? Key to our exploration of this question is an analysis of the steps and barriers facing baculoviruses and polydnaviruses, and where these stages of infection overlap and are subject to interference or assistance from each other. To aid in this we will again refer to previous figures (Figs. 1, 2) mapping out both the baculovirus and polydnavirus infection processes, and dividing each process into distinct stages. At some of these stages the viruses are unlikely to influence the progression of one another, while there is great potential for them to do so at other stages. At each stage, the possibilities for interference from the other infection process will be discussed, if applicable. Where necessary for illustrative purposes we use a generalized MNPV, and a bracovirus such as those known from the braconid genus Cotesia. This will allow a reasonably cohesive discussion while we clearly recognize the diversity of both virus types.

Parasitoid selection of suitable host and injection of polydnavirus
Infection with a polydnavirus requires selection of a host by a wasp carrier and an attempt at parasitization of a larvae. Whether or not this selection process is affected by prior infection with a baculovirus is a question that has been explored a number of times in the past. Some studies have shown parasitoid avoidance of baculovirus infected hosts (e.g. Levin et al., 1983), while others found no effects on parasitoid preference (e.g. Sait et al., 1996). Raimo et al. (1977) showed that C. melanoscela readily parasitizes LdMNPV-infected larvae, at least in a laboratory setting. This particular braconid is of some importance in the literature, having also been used in some of the earliest studies of PDVs (Stoltz et al., 1976;Stoltz & Vinson, 1977;Guzo & Stoltz, 1985). In all cases, the length of time since infection began is an important component of parasitoid discrimination. For example, it is unlikely that C. melanoscela can detect infection before it produces overt behavioral or physiological changes, by which time the larvae in the field may be three or four days into the infection process. Indeed, in their work on the question of parasitoid discrimination using C. melanoscela, Versoi & Yendol (1982) felt it necessary to use infected larvae that were so moribund that they "could not right themselves when turned over".

Ingestion of polyhedra and release of virions
If full-blown infection by an NPV is likely to deter a parasitoid and thereby prevent PDV infection, a similar phenomenon can occur in the opposite direction. The first stage of infection of a lepidopteran host by an NPV is ingestion of virus-contaminated plant material. It may be argued that this process can be influenced by polydnavirus infection before it has a chance to occur. Larvae that have been stung by a carrier wasp suffer from a range of physiological effects caused by wasp-injected materials. One of these materials is the polydnavirus, and it has been shown that polydnavirus effects on hosts may occur in the presence or absence of a parasitoid egg (Strand & Dover, 1991;Beckage et al. 1994;Fathpour & Dahlman, 1995;Shelby & Webb, 1997). Most relevant of these effects is the arrested development and changes in eating habits seen in parasitized larvae; an unsurprising side effect of toxification, immunosuppression, and hormonal manipulation. The chance of baculovirus infection subsequent to parasitization will be reduced if food consumption decreases, in that the risk of infection for nucleopolyhedroviruses is directly related to the amount (leaf area) of foliage eaten (Dwyer & Elkinton, 1993;Dwyer et al., 2005), and the converse is also true. The amount of food present in the midgut could have an impact on the release of virions from the polyhedral occlusion body as well. Dissolution of the polyhedron is a pH-dependent event that occurs in the highly alkaline larval midgut. Polydnavirus infection is unlikely to directly influence midgut pH or the release of virions from polyhedra, but the presence or absence of plant material can do so (Keating et al., 1988;Rossiter et al., 1988). There may be no true competitive component to this aspect of polydnavirus interference with baculovirus transmission in evolutionary terms. However, it significantly influences subsequent infection risk for the parasitized larvae, and thus the parasitoid larvae as well. This is important when assessing the overall probability of both viruses occurring in the same host at the same time.

Might carrier wasps themselves increase the risk of baculovirus ingestion and infection?
Parasitoid wasps have been acknowledged as having a role in the field dissemination of baculoviruses. The limited research that exists on this topic has focused particularly on the baculoviruses that infect lepidopteran hosts (Kurstak & Vago, 1967;Hochberg 1991a;1991b;reviewed by Cossentine, 2009) for reasons of practicality and economic importance already discussed. These baculovirus infections in the field typically occur via ingestion of OBs on foliage, although Raimo et al. (1977) and others have noted that parasitoids can themslves act as a physical vectors of NPVs. These studies found that contamination of the ovipositor occurs when a wasp stings larvae, resulting in NPV infections in subsequently stung larvae. Early work (e.g. Raimo et al., 1977) did not always compare NPV-induced mortality between NPV + parasitization and NPV-only treatments, partly because such comparisons would have been of special interest only if polydnaviruses had been recognized as causing immunosuppression of the host at that time. When a nucleopolyhedrovirus is present in caterpillar larvae or on their food, parasitoid-mediated transmission of the NPVs could also occur in a number of other ways through contamination of and transfer from the parasitoid's body (Raimo et al., 1977;Young & Yearian, 1989;1990;Hochberg, 1991aHochberg, , 1991b). Yet another possibility was raised by Stoltz & Makkay (2003), who found strong circumstantial evidence that latent viruses in T. ni larvae could be activated after parasitism by the ichneumonid wasp Hyposoter exiguae, which is also known to harbor a polydnavirus (Stoltz & Makkay, 2000). None of these except for the last can be considered direct viral interactions, but they are nevertheless noteworthy.

Penetration of the peritrophic membrane
As described in Section 2, virions freed from the dissolved occlusion body must penetrate the peritrophic membrane in order to initiate infection of the insect by a baculovirus. Movement through the peritrophic membrane is thought to be facilitated by metalloproteinases termed enhancins (Slavicek, this volume). Enhancins are located within the ODV membrane in the LdMNPV (Slavicek & Popham, 2005) or are co-occluded in the granule of granuloviruses (Wang et al., 1994;Lepore et al., 1996). However, most baculoviruses lack enhancin genes; consequently other means that have yet to be elucidated are used to traverse the peritrophic membrane. Effects of ingestion-related changes to the midgut discussed in 4.2, such as changes in pH, could also influence the penetration of the membrane by virions. This hypothetical effect of gut contents or pH on the post-release penetration of the peritrophic membrane has not been explored and may be noteworthy; otherwise we feel that this stage of infection is not likely to be directly influenced by polydnavirus infection.

Spread of infection in the larval host
Once through the peritrophic membrane, NPVs infect midgut epithelial cells by binding to the cell membrane and releasing nucleocapsids into the cytoplasm. At this stage of infection, one of the strategies used by larvae to avoid infection is through apoptosis of infected midgut cells (Clarke & Clem, 2003;da Silveira et al., 2005). It has been shown that the PDV associated with Micropletis demolitor, the bracovirus MdBV, induces apoptosis in primary hemocytes (Strand & Pech, 1995). Recent research has shown that in some cases, however, PDV-derived proteins can inhibit baculovirus-induced apoptosis of insect cells , apparently via suppression of caspase activity. This is currently of interest to researchers wishing to explore methods of improving baculovirus efficacy via PDV genes. Inhibited apoptosis of midgut epithelial cells is not likely to be altering transmission probabilities in the field; however, expression of PDV has not been shown in this cell type. Budded virus leaving infected midgut cells infects other cells, and this stage is the most likely point at which the nucleopolyhedrovirus infection process could be be influenced by the action of a polydnavirus. At this point in the infection process the immune response is most likely to have a chance to clear an NPV infection, although by no means has this been shown to occur in all NPV-host systems studied.
Experiments done using injections of budded virus have produced different results. Three recent studies using budded virus injection (Rivkin et al., 2006;McNeil et al., 2010a;2010b) all found some evidence for increased virulence when larvae were previously infected with a polydnavirus. This can be interpreted as resulting from the inability of compromised hemocytes to encapsulate foci of infection. Failed encapsulation of virus-infected cells may explain why some normally refractory larvae exhibit different pathogenesis when infected with a PDV. However, Trudeau et al. (2001) reported that a more important mechanism of resistance in H. zea was the unsuitability of hemocytes for infection; budded virus (BV) entered hemocytes but was unable to replicate within them, so that they served as "sinks" for BV. A similar result was obtained by Rivkin et al. (2006), who showed that the hemocytes of S. littoralis larvae are particularly resistant to infection by AcMNPV, and that this is an important component of resistance in this insect.
One possibility explaining these results is that unimpaired larval hemocytes could effectively encapsulate a very small number of NPV-produced foci as would be produced in the study by McNeil et al. (2010a), which used only a few budded virus particles to initiate infection, but could not encapsulate the far greater number that occur in an infection initiated per os. While exact numbers of foci created in typical per os infections are not known, analysis of budded virus production has shown that Ld652Y (L. dispar) cells in culture produce from approximately 15 to 125 budded virus particles per infected cell, depending on the viral isolate used (Slavicek et al., 1996;Slavicek et al., 2001). This would quickly result in an enormous number of budded virus particles in the hemocoel. The McNeil et al. (2010a) study methodology may have an analog in the field; however, if a wasp transmits NPV on a contaminated ovipositor as suggested by Raimo et al. (1977), it is likely by transferring very small amounts of budded virus from previously parasitized larvae that were already infected.
Research has been also done in granulovirus-parasitoid systems (Matthews et al., 2004;Santiago-Alvarez & Caballero, 1990), and in studies that used NPVs that are normally not found in the hosts used . These are not considered here. In the experiments on normally permissive hosts (Beegle & Oatman, 1974;Raimo et al., 1977;Eller et al., 1988;Murray et al. 1995;Escribano et al., 2000; that explored polydnavirus interactions with baculovirus challenge, or can be inferred as having done so, none found clear positive effects of coinfection on baculovirus-caused mortality. All of those studies used a per os challenge, and are described in greater detail in Section 5 below.

PDV-NPV experiments using per os nucleopolyhedrovirus
Of primary interest to researchers of field interactions between PDVs and NPVs, are those few experiments that have used per os techniques to infect parasitized larvae with NPVs. This technique can be considered the most accurate representation of naturally-occurring exposure to both viruses, and we therefore have given more detailed information on these experiments below. All parasitoid names are followed by parentheses containing the name of the associated PDV.
In an early comparison of NPV mortality in parasitized and unparasitized larvae, Beegle & Oatman (1974 ) used the T. ni MNPV, the parasitoid Hyposoter exiguae (+HeIV), and T. ni caterpillar larvae. In this experiment, larvae were given a range of doses of TnMNPV per os immediately following parasitization. Ld 50 values for parasitized larvae were approximately twice as high as those for unparasitized larvae (3.16 × 10 3 vs 1.58 × 10 3 ), and LD 95 values five times higher (7.18 × 10 4 vs. 1.43 × 10 4 ). The format of this experiment can be considered a reasonable representation of field conditions: per os acquisition of NPV after infection with a PDV via a stinging event. Subsequent experiments will be shown to bear out such results for per os experiments; no increase in pathogenicity of NPVs coincident with PDVs, and typically some effect in the opposite direction. In another experiment conducted before PDVs were fully recognized, Raimo et al. (1977) explored the ability of a parasitoid wasp to physically vector an NPV. This experiment, using C. melanoscela (+CmBV), L. dispar, and LdMNPV, was not designed to look at the effects of parasitization on NPV mortality. However, in the course of the study a small amount of data was collected that allows for a comparison of virus-caused mortality of unparasitized and parasitized larvae. These larvae were reared from surface-sterilized field-collected eggs. Natural mortality caused by virus in larvae not exposed to parasites was 3.8%. Of larvae exposed to uncontaminated parasites for 2 hours and 24 hours, 9.5 % and 10.0%, respectively, died of virus. We note that this evidence is suspect, considering the possibility of contamination inherent in the use of fieldcollected material. We also note, however, the possibilities raised by Stoltz & Makkay (2003), of a PDV producing overt infections of a latent baculovirus. This phenomenon has been observed more often in field-collected insects than in laboratory colonies.
The study by Eller et al. (1988) compared mortality in M. croceipes (+McBV) parasitized and unparasitized H. zea larvae challenged per os with HzMNPV. A dose of 15,000 OBs of virus was delivered 0, 1, 2, and 4 d after parasitization (and 6, 7, 8, and 10 days after hatching). In all treatments, parasitization had no significant effects on the HzMNPV-caused mortality of larvae.
In a study that used Chelonus insularis (+CinsBV) to parasitize S. frugiperda larvae that were then challenged with SfMNPV, Escribano et al. (2000) compared SfMNPV mortality in parasitized larvae to that in virus-challenged unparasitized larvae. The LC 50 for all instars was higher for parasitized larvae; 2 nd instars (1.93 ×10 5 vs. 1.46 × 10 5 OBs/ml), 3 rd instars (1.15 ×10 6 vs. 6.03 × 10 5 OBs/ml), and 4 th instars (5.34 ×10 6 vs. 3.24 × 10 6 OBs/ml). The parasitized larvae in this case were exposed to C. insularis as eggs, which raises serious questions as to the continued presence of CinsBV in the 2 nd , 3 rd , and 4 th instars. In Escribano et al., (2001), using the same agents as in Escribano et al. (2000), parasitized and unparasitized larvae were given the previously determined LC 90 of SfMNPV. Although not designed to compare virus mortality, the experiment showed no significant differences in mortality between the groups (greater than 95% in both).
Three different parasitoid treatments were used in an experiment by Murray et al., (1995). In this experiment, M. demolitor (+MdBV), C. kazak (+CkBV) H. didymator (+HdIV), were used to parasitize larvae of Helicoverpa armigera. At 0 or 2 d post-parasitization these larvae were dosed with HzMNPV applied to the surface of artificial diet, in a 5-dose range from 2 to 6,000 OB/mm2 of diet. In all cases, LD 50 values were either the same (0 d treatment) or greater (2 d treatment) for parasitized larvae.
D' Amico et al., in review. In a small experiment designed to explore the effects of C. melanoscela (+CmBV) parasitization on LdMNPV mortality in gypsy moth larvae, we dosed larvae per os with LdMNPV 2 d pre-and 2 d post-parasitization with a range of doses. In all cases, no significant differences were found between larvae that were parasitized and those that were not. We consider it unlikely that changes in CmBV treatments to longer times preor post-per os LdMNPV challenge would change the amount of LdMNPV-caused mortality in this type of experiment. Both the LdMNPV and CmeBV infection processes have passed critical stages after 48 hours or sooner, if movement of budded LdMNPV out of midgut cells (McNeil et al., 2010a) or the presence of CmeBV-coded products (Guzo & Stoltz, 1985) are considered indicators of success for these viruses in this host. Changes in timing could, however, negatively affect the success of C. melanoscela parasitism as the quality of the gypsy moth larval host is compromised by LdMNPV infection.

Conclusions
For many larval Lepidoptera, parasitism by a polydnavirus-carrying wasp or infection by a nucleopolyhedrovirus are among the top mortality factors in the field. It is also clear that by necessity, parasitoid wasps and their PDVs compete with NPVs for the larval resource. It will require focused research in this area, however, to elucidate the role played by interactions between polydnaviruses and nucleopolyhedroviruses. In the few studies that have been done, there is no obvious indication that concurrent PDV and NPV-infection leads to greater NPV mortality in lepidopteran larvae challenged with NPVs with which they are normally associated. This is true despite the widely-accepted range of immunosuppressive effects of PDV on the host insect. There is some evidence for the opposite effect, which makes good competitive sense: if a parasitized host continues to be exposed to risk of NPV infection, it is in the interest of the parasitoid to reduce the chance of host death prior to successful maturation of the parasitoid progeny.