Low multiplicity of infection in vivo results in purifying selection against baculovirus deletion mutants

The passage of baculoviruses in cultured insect cells leads to the rapid accumulation of deletion mutants, some of which have been demonstrated to have defective interfering properties (Kool et al., 1991; Pijlman et al., 2001; Wickham et al., 1991). Populations with a high frequency of these deletion mutants can completely lose infectivity for insect larvae (Heldens et al., 1996). The presence of such deletion mutants is therefore likely to be deleterious to the fitness of a virus population in vivo, although deletion mutants that can increase fitness are found in some natural baculovirus isolates (López-Ferber et al., 2003). Dai et al. (2000) found that alternate passaging of baculoviruses in insect cells and larvae resulted in sustained infectivity for insect larvae, suggesting that purifying selection (Li, 1997) occurs in these animals. In other words, there may be stabilizing selection for a particular trait, in this case the ability to replicate autonomously. Consequently, individuals that do not possess this trait are removed from the population. However, the fate and dynamics of these cell culture-derived deletion mutants after reintroduction into insects have not been systematically addressed. Therefore, a clonal baculovirus was first serially passaged in insect cells to enrich for deletion mutants, and the resulting population was reintroduced into insect larvae. The effects of both in vitro and subsequent in vivo passaging were investigated by determining the virulence and occlusion-derived virion (ODV) content of polyhedra, and the frequency of occurrence of deletion mutants by quantitative real-time PCR (qPCR).

In order to generate a clonal population of Autographa californica multiple nucleopolyhedrovirus (AcMNPV), bacmid technology (Luckow et al., 1993) was employed. A bacmid with restored polyhedrin (polh) expression was generated using the pFastBac-DUAL/Polh construct (Zwart et al., 2008). This bacmid was used to transfect Spodoptera frugiperda Sf-AE-21 (Sf21) cells (Vaughn et al., 1977) according to the method of Pijlman et al. (2001). The virus resulting from transfection of Sf21 cells was then passaged serially with minimal dilution (1 : 4) in Sf21 cells for 20 passages (final population: P_cell20). RFLP analysis revealed that viruses with large genomic deletions appeared in the population starting at P_cell10, as shown previously (Kool et al., 1991).

Viruses of passages P_cell2, P_cell5, P_cell10 and P_cell20 were subsequently amplified in Sf21 cells. After 96 h, the cells were detached by agitation and sedimented by centrifugation (5 min at 2500 g). The supernatant was removed, stored at 4 °C and subsequently used as budded virus (BV) preparation (BV-P_cell20). The pellet was resuspended in 5 ml 1 % Triton X-100 and the cells were lysed by sonication. The resulting suspension was layered onto a 5 ml 30 % sucrose (in 0.2 % Triton X-100) cushion and polyhedra were pelleted by centrifugation (30 min at 15 000 g at 4 °C). The pellet was resuspended in 10 ml Milli-Q water, pelleted (5 min at 2500 g) and finally resuspended in 500 μl Milli-Q water and stored at 4 °C.

The median lethal concentration (LC₅₀) was determined by a droplet feeding bioassay (Hughes & Wood, 1981) for polyhedra of P_cell2, P_cell5, P_cell10 and P_cell20. Spodoptera exigua larvae were reared as described by Smits et al. (1986). Twenty-four newly molted S. exigua L4 larvae were challenged for each concentration and 24 larvae were taken as healthy controls. Tenfold dilutions of polyhedra (10⁶–10² polyhedra ml⁻¹ in sterile distilled water) were used and the bioassay was performed in triplicate. Probit analysis was used to determine LC₅₀. Statistical tests were performed with spss version 12.0 unless otherwise specified. Linear regression demonstrated that there was a significant increase in LC₅₀ over passages (Fig. 1a⇓; F_1,10=6.937, P=0.025). Unlike S. exigua MNPV (Heldens et al., 1996), passaged populations of AcMNPV retain infectivity – albeit reduced – through to passage 20. It was thus possible to use polyhedra of the P_cell20 population to reinfect insect larvae.

Polyhedra of P_cell2, P_cell10 and P_cell20 were also analysed by electron microscopy (EM) as described by Zwart et al. (2008). The number of ODVs per polyhedron cross-section was counted, for a minimum of 30 polyhedra per sample. There was a significant drop in the number of ODVs per cross-section with increasing passage number [Fig. 1b⇑; Jonckheere–Terpstra (JT) test, n=151, standardized JT statistic=−7.237, P<0.001]. By P_cell20, most polyhedra no longer contained ODVs.

DNA purification from polyhedra and qPCR were performed as described by Zwart et al. (2008) to determine the frequency of occurrence of deletion mutants. Primers used for amplification of the immediate-early gene ie-1 (Kovacs et al., 1992) (Fig. 2a⇓) have been described previously (Zwart et al., 2008). IE-1, an early transcriptional regulator, is essential for replication and the gene was expected to be relatively stable over passaging and therefore a suitable reference locus. Primers for the non-homologous region origin of DNA replication (non-HR ori) (forward, 5′-CCGAGACATACCACAAAGCCG-3′; reverse, 5′-GCACATAAACGACGCAGAATACAT-3′) were also used, as this locus has been found to become enriched during passaging (Pijlman et al., 2002). Primers for polh (forward, 5′-GGGTGGGCAGCAACAACGAGTA-3′; reverse, 5′-CCGATGTAAACGATGGGCTTGTAG-3′; Smith et al., 1983) were included to test for the presence of the gene. No selection for this gene was expected in vitro as BV was passaged, whilst in vivo there is weak selection for polh (Bull et al., 2001). Primers for the mini-F replicon in the bacmid insert (forward 5′-CTTTACGACGGCGACTCCCATC-3′, reverse 5′-GCTTACTGAGGACGCACTGGATG-3′; Luckow et al., 1993) were also used, as this locus is not expected to be under selection in either insect cells or larvae. Ratios of the individual loci to ie-1 (±SEM) were calculated as described by Zwart et al. (2008), rendering a locus frequency relative to ie-1 (referred to as the frequency).

qPCR analysis demonstrated that the frequency of the non-HR ori had increased to 1.68 by P_cell20 (Fig. 2b⇑). The frequency of polh decreased to 0.47. The mini-F replicon decreased in frequency to 0.03, probably also due in part to the instability of a bacmid insert (Pijlman et al., 2003). The P_cell20 virus population thus had an unbalanced gene content, as confirmed by qPCR, and had polyhedra with fewer ODVs and lower infectivity.

Amplified polyhedra from the P_cell20 population were used to perform two passages in S. exigua L4 larva (final population: P_insect2). Thus, insects were challenged with an LC₅₀ (2×10⁴ polyhedra ml⁻¹, final population: P_insect2-LC₅₀) and LC₉₉ (6×10⁷ polyhedra ml⁻¹, final population: P_insect2-LC₉₉), as calculated from the bioassay data for polyhedra from P_cell20. Two replicates were performed for each concentration, with polyhedra being pooled from ten larvae per replicate. Polyhedra of P_insect1-LC₉₉ and P_insect2-LC₉₉ (replicate 1) were analysed by EM as described above. The LC₅₀ of P_insect2-LC₉₉ (replicate 1) was also determined by bioassay as described above. For comparative purposes, P_cell2 was also included in the bioassay. DNA isolation and qPCR as described above were performed for all four P_insect2 populations.

After two passages in insects at high dose (P_insect2-LC₉₉, replicate 1), the infectivity of polyhedra was restored to control values (mean log LC₅₀±sem: P_cell2=3.754±0.064, P_insect2-LC₉₉=3.963±0.086; pairwise t-test: t₂=−2.392, P=0.139). The ODV content of polyhedra from this population (P_insect2-LC₉₉, replicate 1) was also restored to wild-type levels (Fig. 1b⇑). The observed number of virions per polyhedron cross-section was not significantly different from that in polyhedra from larvae transfected with the original bacmid (mean±sem: P_insect1-LC₉₉=5.74±0.58, bacmid in vivo=7.60±1.17; Mann–Whitney U-test, U=1568.5, n=123, P=0.168).

To analyse the qPCR data on the P_insect2 populations (Fig. 2b⇑), a two-way analysis of variance (Genstat 7.0) was performed to test whether there was an effect of dose (LC₅₀ or LC₉₉) on the frequency of the three loci. Neither dose (F_1,8=0.87, P=0.397) nor the dose–locus interaction (F_2,8=0.77, P=0.554) had a significant effect, whereas locus did (F_2,8=72.71, P<0.001). As there was not a significant effect of dose on the frequency of the three loci, locus frequencies for the four populations were compared with that in the P_cell20 population by means of a one-sample t-test. For the non-HR ori, there was a significant decrease in frequency (t₃=−10.491, P=0.002). Moreover, the mean P_insect2 frequency of the non-HR ori was not significantly different from the wild-type frequency (t₃=0.006, P=0.996). The mean increase in polh frequency (0.08) was not significant (t₃=2.717, P=0.073). The mini-F replicon also did not significantly increase in frequency (t₃=0.957, P=0.409).

The 3000-fold difference in dose between the LC₅₀ and LC₉₉ treatments did not have a significant effect on changes in locus frequencies. If the number of initially infecting viruses is proportional to the dose, then this could be an indication that the number of initially infecting viruses was very small, even at the high dose. In other words, despite the higher number of viruses initiating infection at high dose, the initial in vivo m.o.i. was still so low that it had effectively not changed. This is another indication that the number of founders of infection is small, as has previously been suggested (Smith & Crook, 1988).

Deletion mutants missing ie-1, or with multiple copies of the non-HR ori, were rapidly purged from the populations during passaging in vivo (Fig. 2b⇑). During continuous bioreactor experiments with baculoviruses, cyclic fluctuations in virus titres similar to those described by von Magnus (1954) have been observed (van Lier et al., 1992). The explanation proposed for this phenomenon is that, following a strong drop in virus titres due to defective interference, viruses capable of autonomous replication are enriched at low m.o.i. (de Gooijer et al., 1992). The fact that infection of larvae is initiated by a small number of viruses will similarly lead to low m.o.i. during initial infection. Thus, the purifying selection observed in vivo may come about mainly because of these low m.o.i. This is not to suggest that there is no selection for the presence of, for example, polh during passaging in insect larvae, but rather that the predominant selective condition resulting in the purifying selection observed is the ability to replicate autonomously.

In order to test whether this explanation has merit for the rescue effect observed by passaging in larvae, the titre of BV-P_cell20 was determined by an end-point dilution assay and one round of infection of Sf21 cells was performed at m.o.i. of 10, 1, 0.1 and 0.01. BV was purified, DNA isolated and qPCR performed as described above. The experiment was performed in duplicate. Linear-regression analysis was used to determine whether locus frequencies changed as the m.o.i. decreased. For the non-HR ori, there was a significant decrease in frequency (F_1,6=21.937, P=0.0034), whilst for polh (F_1,6=0.117, P=0.7445) and the mini-F replicon (F_1,6=0.1750, P=0.6903), no change in frequency was observed (Fig. 3a⇓). It was therefore possible to mimic in vitro the purifying selection observed in vivo simply by altering the m.o.i. This is in agreement with results obtained by van Lier et al. (1996) in bioreactor systems: it was empirically shown that low m.o.i. resulted in the slower accumulation of defective viruses. The BV-P_cell20 sample and a number of time points during passaging were analysed by qPCR to confirm that the ancestral virus (BV-P_cell0) had the same genomic AcMNPV profile as the bacmid and to determine when deletion mutants could first be detected (Fig. 3b⇓). Changes in locus frequencies were only visible around passage 12 (P12), in agreement with RFLP data (not shown).

In the experiments described here, polh did not increase significantly in frequency by P_insect2. Bull et al. (2001) found that a polh-negative virus persisted, during co-infection with a polh-positive virus, for six serial transfers in insect larvae. The polh-negative virus needs to co-infect a cell with a polh-positive virus in order to produce polyhedra and be carried over to the next round of infection. As most cells in the insect will be infected during the final round of infection, Bull et al. (2001) reasoned that the rate at which a polh-negative virus is lost from a population depends directly on m.o.i. in the final round of cellular infection. Thus, presumably there was not strong selection against the polh-negative virus because of a high in vivo m.o.i. at the end of infection. ODV of the P_cell20 population has an observed polh frequency (f_virus) of 0.463. If the frequency of polh is conserved during infection, then in the final wave of infection, the frequency that cells will be infected by at least one BV containing polh in its genome (f_cell) is described as:

. Here, m is the m.o.i. in vivo, estimated at 4.3 BV per cell (Bull et al., 2001, 2003). This results in a value of f_cell of 93 %. As most cells infected in the final wave of infection have been infected by a BV containing the polh gene, it can be understood that selection is weak.

In contrast to polh locus frequency, the ODV content of polyhedra was rescued to wild-type levels within a single passage (P_insect2-LC₉₉, replicate 1). One explanation for this discrepancy is that the presence of deletion mutants, for example those missing ie-1, is somehow interfering with the generation of polyhedra. The relative frequency of polh to non-HR ori – rather than ie-1 – gives an indication of the frequency of viruses capable of generating polyhedra to all viruses in the population, not only those containing ie-1. The frequency of polh compared with the non-HR ori rises from 0.28 (P_cell2) to 0.54 (mean P_insect2) during passaging in vivo. Another compatible explanation for this observation is that the purifying selection observed results in a critical increase in frequency of another gene essential for generation of normal polyhedra (e.g. 25K).

During baculovirus infection of insect larvae, the m.o.i. in vivo changes drastically. As the number of viruses initiating infection appears to be small (e.g. Smith & Crook, 1988), initially the m.o.i. is extremely low. Towards the end of infection, the m.o.i. is rather high (Bull et al., 2001, 2003). These different conditions will result in different selection pressures on the virus population during the course of infection. Genes involved in replication (ie-1) or transmission (polh) will also be under different selection pressures. There will be a very strong selection on genes involved in replication early in larval infection. Later in infection, selection is somewhat relaxed as m.o.i. become high – if the presence of a single gene copy is sufficient for cellular infection to proceed normally. Conversely, there will not be selection for genes involved in transmission early in infection. Later in infection there will be moderate selection for these genes, but not of the strong purifying type seen for replication early in infection. The fate of baculovirus deletion mutants derived from tissue culture and reintroduced into insect larvae illustrates a virological principle that may apply to many other virological systems.