Densoviruses are arthropod parvoviruses (classified as Densovirinae) characterized by a nonenveloped capsid protecting a short single-stranded DNA genome. Their pathogenicity for insects at larval stages suggests that they could be interesting agents for the control of insect populations, provided that their potential unwanted side effects on nontargeted populations are understood.
display a great diversity in structure, sequence, and host range. There are four known genera; the genus Densovirus
(DNV) infects exclusively Lepidoptera
). According to the International Committee on Taxonomy of Viruses (ICTV), two viruses have been classified so far in this genus: the type species Junonia coenia
DNV) and the greater wax moth Galleria mellonella
DNV). They are characterized by ambisense genomes displaying 88% sequence identity. They infect similar tissues in their hosts, both excluding the midgut, but differ in their host range. Jc
DNV infects several lepidopteran species but not G. mellonella
, while Gm
DNV appears restricted to this host.
DNVs are orally transmitted. Once ingested by the larvae, viral particles must cross the intestinal barrier of their host to reach and replicate in the underlying target tissues (12
). The host digestive tract, and particularly the entry in the epithelial cells of the midgut, is the first step that determines specificity.
Parvovirus capsids are icosahedral with a T=1 symmetry, and the surface is shaped by 60 copies of a common core of structural proteins (VP) (4
). Substitutions on the surfaces of vertebrate parvovirus capsids have been associated with changes in tissue tropism and host shift (7
). Four proteins sharing a common C terminus and extended N termini compose the DNV capsid (VP1 to -4), displaying 88% sequence identity. They are translated at different ratios by a leaky scanning mechanism; VP4, the most abundant VP, shapes the surface of the capsid. The VP4s of Gm
- and Jc
DNV are each made of 437 amino acids (aa) and display 96% sequence identity.
The molecular anatomy of several parvoviruses has been solved by X-ray crystallography and cryo-electron microscopy (cryo-EM), which elucidate the early mechanisms of specificity (3
). The molecular mechanisms driving DNV's specificity are currently unknown, and no cellular receptors have been characterized so far. The structure of Gm-
DNV has been solved by X-ray crystallography and cryo-EM, respectively (2
). They share with vertebrate parvoviruses a highly conserved beta barrel core, although their surface is smoother, with only two small reliefs in the 5-fold axis and none of the prominent spike on the 3-fold axis shown to be involved in the parvovirus attachment to the receptor (8
By comparing the VP4s of Gm-
DNV, Bruemmer et al. identified 22 differences in amino acids, of which 8 aa are exposed at the surface and may act as potential determinants of host specificity (2
). Five of these differences are located on the two unique reliefs on the 5-fold axis: two at the tip and the base of the smaller spike (at positions 177 and 174, respectively), and three around the larger spike (positions 167, 276, and 196) (). The last 3 aa changes were located close to the 5-fold axis (position 123), in the 3-fold axis (position 297), and in a dimple on the 2-fold axis (position 305) (2
). By analogy with parvovirus data, the two small reliefs on the 5-fold axis in DNV capsid have been considered the strongest candidates for host range or tropism (2
Fig 1 Close-up view along the 5-fold axis of (a) JcDNV and (b) GmDNV. An example of the smaller spike is marked with a circle, and an example of the second spike is marked with an oval in each panel. In panel b, an asterisk marks the small protuberance near (more ...)
Based on these structural data, we produced eight mutants of JcDNV by substituting at these sites the GmDNV amino acids. We then characterized in vivo the role of these residues on virulence. The results presented here provide functional insight in insecta on densovirus determinants of virulence and tissue specificity.
A JcDNV mutant harboring GmDNV amino acids has a reduced virulence in the Spodoptera frugiperda host.
To construct Jc
DNV mutant viruses, site-directed mutagenesis was carried out to replace VP4 candidate residues with those of Gm
DNV. Substitutions were carried out in an infectious molecular clone of the Jc
DNV wild-type (wt-Jc
DNV) genome, pBRJH (9
). In total, eight Jc
DNV mutants (m-Jc
DNVs) were constructed: six mutants carried a single substitution (A123V, S167T, A196T, A276V, I297T, or S305N), one mutant combined two close substitutions in the strongest candidate relief (I174V/T177D), and one mutant carried all the substitutions (8m) (). Transfecting mutants and wt pBRJH in S. frugiperda
larvae generated all viral stocks, which were subsequently purified and quantified by quantitative PCR (qPCR). The proper assembly of the viral particles was analyzed by electron microscopy, and the presence of mutations confirmed by sequencing the viral genomes before and after their amplification in larvae.
Candidate amino acid positions and localization on VP4 capsid surface proteinsa
We first estimated the virulence by assessing the time to death and the mortality of larvae induced by these viruses. Cohorts of 30-s instar S. frugiperda larvae were infected with wt-JcDNV and all m-JcDNVs at two viral doses, 1E+06 and 1E+08 viral equivalent genomes (veg) per larva. These doses were previously determined by qPCR to correspond, respectively, to wt-JcDNV doses killing 50% and 95% of a larval cohort (LD50 and LD95, respectively). Virulence was assessed until death or nymphosis (15 days). Infections were similarly performed with cohorts of G. mellonella larvae.
Caterpillars infected with A123V, I174V/T177T, and 8m mutants displayed a higher survival rate at both doses (P < 0.001; Dunnett test) compared to those infected with wt-JcDNV, suggesting a lower virulence of these mutants (A). Virulence was intermediate for the S167T mutant, i.e., similar at LD95 but lower at LD50 (P values of 0.29 and <0.01, respectively; Dunnett test), while no difference in virulence was observed for the A196T, A276V, I297T, and S305N mutants. The time to death was delayed in S. frugiperda, the strongest effect was observed for infections with mutant 8m. At LD50, almost all 8m-infected larvae went through nymphosis within 2 weeks (B). Increasing the dose to LD95 delayed the death of 8m-infected larvae by 6 to 7 days compared to controls (data not shown).
Fig 2 Specific amino acid substitutions on the JcDNV capsid surface affect virulence on S. frugiperda larvae. (A) Mean larval survival rates following infections with m- or wt-JcDNV at the LD50 and the LD95 of wt-JcDNV (1E+06 and 1E+08 veg/larva, respectively). (more ...)
No mortality of G. mellonella larvae was observed, even at an over-lethal dose (1E+11 veg/larva), showing that these residues are not sufficient to exchange virus specificity.
We next sought to determine whether the stability of the capsid was affected by mutations. In the route of infection, viral particles cross a wide range of pHs, from highly basic (pH 9 to 10) in the gut of phytophagous caterpillars to acidic (pH 4 to 5) in endosomal compartments. Such environments could negatively affect the stability of potentially weakened mutant capsids. To test this hypothesis, we challenged 8m and wt particles at pH 9 for 1 h prior to oral infections of S. frugiperda larvae and compared the resulting virulence by determining the LD50s. The pH 9 challenge did not significantly affect the LD50s.
By analogy with parvoviruses, mutations in or near the 5-fold axis might be detrimental for infectivity, impairing for example the externalization of VP1 during the traffic toward the nucleus (1
). To assess this point, we determined the infectivity of all viruses in vitro
using the Ld652 permissive cell line (9
). Infectious titers on these cells were all similar, further suggesting that viral structures were not compromised by the mutations (11
). No loss in virulence was observed either by challenging viral particles to a high temperature (up to 50°C for 1 h) before ingestion.
The loss in virulence is thus unlikely to result from an altered stability of the mutated capsid in the course of infection.
To quantify the loss in virulence, we next determined the LD50 and LD95 for the four mutants by infecting 30 larvae with serial viral dilutions. The A196T mutant, referred to as a “neutral” mutant, and wt-JcDNV were included as controls. As shown on C, a 10- to 1,000-fold increase in LD50 and LD95 was obtained depending on the virus, and the strongest effect was again observed for the 8m mutant carrying all aa substitutions. No difference was observed between the two controls.
The results showed that four of the eight predicted residues located on or near the 5-fold axis protrusions are important for JcDNV virulence, but not stability, in S. frugiperda. We next enquired which step in the infection process was affected.
Substitutions on JcDNV capsid do not affect replication within target cells.
We first analyzed the tissue tropism. Larvae were infected with mutants and wt-Jc
DNV at their respective LD95
s (C); tissue tropism was assessed for dying larvae by immune-labeling as previously described (12
). Viruses were antigenically cross-reactive, and a similar tissue tropism was observed for all the mutants tested (data not shown). The low virulence of the 8m mutant was thus not due to a change in tissue tropism.
We next quantified with time the replication of m- and wt-Jc
DNV in the larvae. Previous work carefully dissected the Jc
DNV pathogenesis in S. frugiperda
). These studies revealed that viral replication is not observed in midgut cells and is restricted to underlying target tissues, e.g., visceral muscles and trachea, where a viral clearance and a phase of viral eclipse follow entry (12
). Further, at day 2 postinfection (p.i.), the input virus is entirely eliminated and replication is barely detected in target tissues. By this time, infectious viruses have crossed the midgut.
To compare viral loads between viruses, cohorts of 50 larvae were infected with 1E+09 and 1E+11 veg/larva (respectively, the LD50
rounded values of the less-virulent mutant, 8m). Next, four larvae per virus and per dose were sacrificed every day. Viral loads per larva were estimated by qPCR as previously described (12
). At day 2 p.i. and at both doses, viral loads were significantly lower in larvae infected with m-Jc
DNVs than those in controls (P
< 0.05; Dunnett test), suggesting a decreased midgut crossing. Similar loads were recovered with time for mutants A123V, S167T, and I174V/T177T and controls (P
> 0.85; Dunnett test) (). Larvae infected with the 8m mutant were the only ones still displaying lower viral loads at 10 days p.i. (P
< 0.05), although few dead larvae were observed. Similar values were calculated for the slope of viral accumulation curves (data not shown) suggesting that the abilities of all viruses to replicate in target tissues are similar. Infection for mutant viruses might then be delayed as a matter of dose at the early step of viral entry. Mutations might affect the capacity of the virus to reach the target tissues by limiting its capacity to cross the midgut.
Fig 3 The A123V, S167T, I174V/T177D, and 8m mutant viruses efficiently replicate in S. frugiperda larvae. Larvae were infected at the LD50 of the mutant with the lowest virulence, the 8m mutant (1E+09 veg/larva). Results are presented as log10(copy number of (more ...) Four aa on JcDNV capsid define the efficiency of midgut crossing.
If mutations affect the midgut crossing, then virulence should be similar for infections bypassing the midgut, i.e., by direct virus injections in the larval haemocoel. To test this hypothesis, we first determined the LD50 and LD95 by injecting 4th instar S. frugiperda larvae with serial dilutions of m- and wt-JcDNV. The LDs of all m-JcDNVs were similar to those of the controls, wt-JcDNV and mutant A196T (P > 0.41) (A). As hypothesized, no loss in virulence was observed when the midgut was bypassed, strongly suggesting that capsid substitutions affect the midgut tropism. No mortality was observed following the injection of mutants in G. mellonella.
Fig 4 The A123V, S167T, I174V/T177D, and 8m mutant viruses have a decreased capacity to cross the midgut of S. frugiperda larvae compared to that of wt-JcDNV. (A) Larval survival rates 15 days following injections with m- or wt-JcDNV at the LD50 and the LD (more ...)
To underpin this hypothesis, we next measured the ability of viruses to cross the midgut of S. frugiperda ex vivo
, using Ussing chambers. This device allowed to isolate the luminal (L) from the hemolymphatic (H) sides of an epithelium and study the trafficking of macromolecular complexes or ions (15
). The Ussing chambers have been successfully adapted to large lepidopteran midgut epithelium, including S. frugiperda
but not G. mellonella
, which is too small to be mounted (6
). In this setting, tissue integrity is retained for 2 to 3 h as measured by the transepithelial resistance assessing the permeability of the tissues. Midguts of S. frugiperda
6th instar larvae were dissected and opened to produce a single layer, which was mounted in between the two chambers as previously described (6
). After a 30-min equilibration, m- or wt-Jc
DNV (1E+13 veg) were added to the L compartment and left for 30 min. The total volume of both compartments (500 μl each) was then recovered separately to extract DNA and estimate the amount of viral genomes by qPCR. At least six repeats on different midguts were performed for each virus. The total amount of DNA recovered from both compartments (L and H) was similar between experiments and between viruses (P
> 0.76). The ratio of the number of wt viral genomes that crossed the midgut, over the total amount of input viral genomes was about 1 in 1E+04 during the 30 min that the experiment lasts (B). The amounts of m-Jc
DNV crossing the midgut were significantly lower for the A123V, I174V/T177T, and 8m viruses compared to those for the controls (P
< 0.05; Dunnett test). These results further confirmed that substitutions on residues located near or at the 5-fold axis protrusion affected the capacity of Jc
DNV to reach the haemocoel. Results obtained for the S167T mutant were always statistically less significant than those with the A123V and I174V/T177T mutants, suggesting a more minor role of this residue when alone and/or a potentiating role when combined with the three other residues. Altogether, our results provide the first functional data (both in vivo
and ex vivo
) on densovirus virulence and midgut tropism determinants on the capsid. Four of the eight residues structurally predicted to be involved in host specificity significantly affect virulence upon ingestion by S. frugiperda
larvae. Their substitutions markedly decrease the number of viral particles crossing the midgut but do not appear to affect replication once this barrier is overcome, clearly indicating that these four residues are involved in midgut tropism. The mutated capsids may have a lower affinity for midgut-specific molecules involved in recognition and/or entry of the viral particles; we are currently investigating the mechanism of midgut crossing. These results suggest that two different mechanisms control tissue specificity, one for midgut entry and one for target tissues entry and replication.
This study highlights the great potential of the densovirus/Lepidoptera interaction model and the suitability of virus chimeras for in vivo mechanistic studies on virulence and specificity.