Vesicular Stomatitis Virus Evolution during Alternation between Persistent Infection in Insect Cells and Acute Infection in Mammalian Cells Is Dominated by the Persistence Phase

Abstract

Vesicular stomatitis virus has the potential for very rapid evolution in the laboratory, but like many other arboviruses, it evolves at a relatively slow rate in the natural environment. Previous work showed that alternating replication in different cell types does not promote stasis. In order to determine whether other factors promote stasis, we compared the fitness trajectories of populations evolving during acute infections in mammalian cells, populations evolving during persistent infections in insect cells, and populations evolving during alternating acute and persistent infection cycles. Populations evolving under constant conditions increased in fitness in the environment in which they replicated. An asymmetric trade-off was observed such that acute infection had no cost for persistence but persistent replication had a dramatic cost for acute infection in mammalian cells. After an initial period of increase, fitness remained approximately constant in all the populations that included persistent replication, but fitness continuously increased in populations evolving during acute infections. Determination of the consensus sequence of the genes encoding the N, P, M, and G proteins showed that the pattern of mutation accumulation was coherent with fitness changes during persistence so that once fitness reached a maximum, the rate of mutation accumulation dropped. Persistent replication dominated both the genetic and the phenotypic evolution of the populations that alternated between acute infection of mammalian cells and persistence in insect cells, and fitness loss was observed in the mammalian environment despite periodic replication in mammalian cells. These results show that stasis can be achieved without good levels of adaptation to both the mammalian and the insect environments.

The high mutation rates exhibited by RNA viruses during replication contribute to their ability to evolve rapidly and adapt to the changes in their environments (14, 15). The mutation rate has been estimated to be between 10⁻³ and 10⁻⁵ substitutions per nucleotide (nt) per round of replication (16). Considering the lengths of most RNA viral genomes, this translates into one mutation per genome copied, which gives the viral populations great diversity and plasticity.

As do other RNA viruses, vesicular stomatitis virus (VSV) exhibits high adaptability and the capacity to increase fitness very rapidly in laboratory experiments, making it an ad hoc model for studying viral evolution (28). VSV is an arbovirus and a member of the Vesiculovirus genus in the Rhabdoviridae family. A single molecule of negative-sense, single-stranded RNA composes its genome (41). The life cycle of VSV involves horizontal transmission from insects to mammals (9, 24, 25) and vertical transmission from infected female insects to their offspring (37). Despite the high evolvability of VSV, virus isolates from the same ecological zones have the same genotypes regardless of the year of isolation and viruses from different ecological zones have different genotypes. These results indicate that the evolution of VSV is the result of the selective pressure exerted on the virus by ecological factors that render genomes stable within a zone (34). This genetic stability is not a peculiarity of VSV; other RNA arthropod-borne viruses also appear to have lower rates of evolution than would be expected from their mutation rates (6, 20, 43). It has been proposed that the necessity for the virus to replicate in the two disparate environments provided by insects and mammals may constitute a strong purifying selection and mutations that are beneficial in one host but deleterious in the other would be lost from the population. While drift would not be not fully eliminated, this mechanism would minimize it and would result in genetically stable populations (36).

The hypothesis that the two-host life cycle may be responsible for the genetic stability observed in arboviruses has been tested in several studies. Weaver and collaborators found that alternating replication of eastern equine encephalitis virus (EEEV) between mosquito and mammalian cells increases the fitness of the virus in both hosts but that viruses passaged in either cell type alone show a loss of fitness in the other cell line. Genomic evolutionary rates are slower in cases of alternating environments than in cases of single-host adaptation, indicating a constraint in the mutation accumulation rate (but not in fitness gains) when the viruses are subjected to alternating host transmission cycles (42). EEEV populations adapting to either mosquito or avian cells are subjected to different selective pressures as well, but lineages grown in alternation between hosts express intermediate phenotypes consistent with dual adaptation to both environments (10). This suggests that alternating replication in different hosts renders viral populations well adapted to replicate in both cell types.

Previous work demonstrated that VSV adaptation to acute replication in insect and mammalian cells shows no limitations, either in terms of mutation accumulation rates or fitness increase, on the adaptation of the virus to either insect or mammalian cells (31). This result contrasts with the trade-off observed when VSV is allowed to replicate continuously during persistent infection of sand fly cells for 10 months, which results in two million-fold higher fitness in insect cells than in mammalian cells (29). Similar results were observed with VSV replicating in HeLa, MDCK, and BHK-21 cells. Single-host passages showed that adaptation to a particular host may or may not involve trade-off, but alternating replication in two cell types always leads to increased fitness in both hosts (39).

Since cell type by itself does not promote evolutionary stability of arboviral populations, we tested the effect of alternating persistent and acute infections on the evolution of VSV. Three different viral lineages were generated. For one, we alternated persistent viral replication in sand fly cells (LL-5) with acute infection in mammalian cells (BHK-21), and for the other two, we passaged viruses in persistently infected LL-5 cells or carried out repeated acute infections in BHK-21 cells. We determined the fitness trajectories of the populations recovered from these infections in both cell lines, and we found that the viruses replicating in alternating environments and the ones maintained in insect cells showed a marked increase in fitness in LL-5 cells and a decrease in fitness in BHK-21 cells. We also found that after 15 passages the fitness values for these evolved viruses remained approximately constant, in contrast with those for the viruses passaged in BHK-21 cells, whose fitness continuously increased in the mammalian environment as previously reported (30, 31). The populations undergoing persistent replication in insect cells evolved decreased virulence in mammalian cells. We determined the consensus sequence of one population of the alternating-replication lineages at passages 15 and 25, and we found a reduction in the evolutionary rate that approached genetic stasis. Overall, our results showed that the evolution of populations that alternate between acute replication in mammalian cells and persistent replication in insect cells was dominated by the latter and that the genetic stability observed in natural arboviral populations could be achieved in the absence of adaptation to the mammalian environment.

MATERIALS AND METHODS

Cells and viruses.

Monolayers of BHK-21 cells from John Holland's laboratory were grown at 37°C in Eagle's minimal essential medium supplemented with 7.5% bovine calf serum and 0.06% proteose peptone 3. LL-5 is a continuous sand fly cell line derived from Lutzomyia longipalpis (38), and LL-5 cells were grown at 28°C in Mitsuhashi and Maramorosch (MM) insect medium (Sigma) supplemented with 20% fetal bovine serum. I1 monoclonal antibody (MAb) hybridoma cells were obtained from Douglas Lyles (21) and grown for antibody production as previously described (17).

The viruses we employed were a wild-type VSV Indiana serotype (Mudd-Summers strain) and a MAb-resistant mutant, MARM U, that is neutral compared to the wild type in BHK-21 and LL-5 cells (19, 29). MARM U has a single amino acid substitution, Asp→Ala, at amino acid 259 of the G glycoprotein that confers resistance to I1 MAb (40). Because MARM U is neutral, we can use it as a wild-type surrogate in competition experiments.

VSV growth curve in LL-5 cells.

LL-5 cells were grown in a 75-cm² flask and infected with wild-type virus at a multiplicity of infection (MOI) of 0.1. The cells were incubated for 10 min at room temperature and 30 min at 28°C to allow virus adsorption and penetration. After this incubation, 30 ml of medium was added and the infection was allowed to proceed for 2 weeks. We took samples of the supernatant at different times, and fresh medium was added to keep the volume constant. The aliquots were stored at −80°C until all the samples were taken, and then they were titrated by plaque assay on BHK-21 cells.

Virus passages.

We carried out three different regimens of viral passages; four replicates were done for each regimen. The passages alternating between LL-5 and BHK-21 cells were done as follows: 2 × 10⁵ PFU of a wild-type viral stock was used to infect a monolayer of LL-5 cells grown in 25-cm² flasks. The cells were incubated for 10 min at room temperature and 30 additional minutes at 28°C to allow viral penetration, and then 5 ml of medium was added. At 3 days postinfection (pi), another 5 ml of medium was added. At 1 week pi, we replaced the medium with 5 ml of fresh medium. Finally, at 10 days pi, another 5 ml of medium was added. Eliminating the medium at 1 week pi minimized the sampling of virus selected for stability. After 2 weeks of incubation at 28°C, the virus was recovered from the LL-5 cell supernatant and 2 × 10⁵ PFU was used to infect a monolayer of BHK-21 cells. After 10 min of incubation at room temperature and 30 min at 37°C to allow virus penetration, we added 5 ml of medium and the infection was allowed to proceed for 24 h at 37°C. The virus was then recovered and appropriately diluted, and a new round of persistent replication was carried out in a fresh LL-5 monolayer. These strains were labeled AnA, AnB, AnC, and AnD, where n represents the passage number.

The serial passages in BHK-21 cells were done using 2 × 10⁵ PFU of the wild-type viral stock to infect a monolayer of BHK-21 cells (MOI = 0.1). The infection was allowed to proceed for 24 h at 37°C. The recovered virus was diluted, and 2 × 10⁵ PFU was used to infect a fresh monolayer of BHK-21 cells. These lineages were labeled KnA, KnB, KnC, and KnD, where n represents the passage number.

The serial passages in LL-5 cells also started with 2 × 10⁵ PFU of the wild-type viral stock added to a monolayer of LL-5 cells, which was incubated for 2 weeks at 28°C as described for the insect phase of the alternating passages. After 2 weeks, the cells were detached from the culture flask and diluted 1:20 to start a new culture. An aliquot of the supernatant was tested to determine the presence of infectious virus at each passage. These populations were labeled PnA, PnB, PnC, and PnD, where n represents the passage number.

Fitness assays.

The fitness assays were carried out by direct competition of test virus with a neutral wild-type surrogate (MARM U) as previously described (19). Briefly, mixtures of the test virus and MARM U (10⁶ PFU/ml) were used to infect either BHK-21 or LL-5 monolayers (MOI = 0.1). After incubation for 24 h at 37°C for the BHK-21 cells or 48 h at 28°C for the LL-5 cells, the viruses were diluted again to start a second competition passage. The ratio of test virus to MARM U in each passage was determined by triplicate plaque assays in the presence and absence of I1 MAb on BHK-21 cells. Titrations on LL-5 cells were not possible because plaques are not formed. The ratio of the test virus to MARM U was plotted against the competition passage number, and the relative fitness value was determined from the slope of the adjusted linear regression.

After repeated passages in persistently infected LL-5 cells, the viral titers dropped to approximately 10⁴ PFU/ml and were too low to carry out the fitness assay. In order to get a higher-titer virus for these assays, viruses from the persistent infections (passages 10, 15, 20, and 25) were amplified in LL-5 cells. A monolayer of LL-5 cells was infected with 200 μl of the low-titer virus. The infection was allowed to proceed for 48 h, and the recovered supernatant was titrated. After this single amplification passage, viral titers were at least 10⁶ PFU/ml and fitness assays were carried out.

Virulence tests.

Virulence, defined as the level of damage to the host cells we employed for this work, was tested by determination of cytopathic effect (CPE) in infected monolayers and by characterization of plaque size. Matched BHK-21 cells were infected with the wild type, P5B, and P25B at an MOI of 0.05, and the development of CPE was monitored under a microscope at different time points. To determine plaque size, matched BHK-21 monolayers were infected with appropriate dilutions of the same viral strains to produce isolated plaques. Infections were developed at 30 h pi, and the diameters of the plaques were measured.

Stability test.

The stability of viruses from persistently infected cells was determined by incubating an aliquot of P1A at 28°C for 5 days. Samples were taken from the aliquot before incubation, after 2 days, and after 5 days, and titers were determined by plaque assay on BHK-21 monolayers.

Nucleotide sequencing.

Viral RNA was extracted from 140 μl of the infection supernatant by using QIAamp viral RNA mini kit (QIAGEN) according to the manufacturer's recommendations. For the low-titer viruses, 3 ml of the amplification passage supernatant was concentrated 10-fold using a Centricon filter (Millipore) and 280 μl was used for RNA extraction. The purified RNA was eluted with 30 μl of water, and 8 μl of template was used in a reverse transcription reaction with Superscript II (Invitrogen) and random primers to obtain cDNA copies of the viral RNA. Viral cDNA was amplified by PCR by using primers that covered the coding regions of the N, L, P, and G protein genes as described previously (35). The PCRs were carried out using the following program: 3 min at 94°C; 40 cycles of 30 s at 94°C, 45 s at 48°C, and 2 min at 72°C; and finally 10 min at 72°C. All the PCR products were purified using QIAquick multiwell PCR purification kit (QIAGEN), and their concentrations were determined by spectrophotometry.

Automated fluorescent dye-terminator sequencing on an ABI Prism 3730xl sequencer was used to obtain the sequence of 100 ng of PCR product per reaction. The primers employed for amplification and sequencing are available upon request. AssemblyLIGN software (Oxford Molecular Group PLC) was used to assemble and analyze the sequences obtained.

RESULTS

Two phases of replication during infection of insect cells.

We plotted a growth curve for VSV in LL-5 for 2 weeks. As shown in Fig. 1, the virus titer increased at the beginning of the infection, reaching its maximum of 4.7 × 10⁷ PFU/ml after 48 h. After this point, the titer decreased and then remained approximately constant, at around 6 × 10⁵ PFU/ml, for the rest of the course of infection. Thus, infection of insect cells can be described as a two-phase process, with an initial phase of acute infection that lasts for about 2 days followed by a phase of persistent infection with low but continuous release of virus throughout the lives of the infected cells.

FIG. 1.

Two-phase growth curve for VSV in LL-5 cells. LL-5 cells were infected with wild-type VSV (MOI = 0.1). Viral titers were determined at different times pi and plotted against the time the sample was taken. An initial increase in viral titer up to 4.67 × 10⁷ PFU/ml was observed at 48 h pi. Then the viral titer dropped to 6 × 10⁵ PFU/ml and remained approximately constant for the rest of the course of infection.

In order to get an estimate of the level of replication taking place during the persistence phase, we tested the stability of virus from persistently infected cells during incubation at 28°C. The decay of log virus concentration through time was characterized by a linear regression (R = 0.98) with an average loss of 33% per day. Since titers remained approximately constant, this loss was compensated by new generation of about 10⁵ PFU/day.

Fitness trajectories in mammalian cells.

We determined the relative fitness in BHK-21 cells of the viruses passaged in BHK-21 or LL-5 cells alone and those passaged alternately in LL-5 and BHK-21 cells. As shown in Fig. 2A, the fitness of the viruses evolved in alternating hosts increased slightly during the first three passages but after the fifth passage the fitness decreased. However, overall fitness values beyond passage 20 remained constant (P > 0.5; unpaired t test). The viruses subjected to regimens of persistent replication also showed an overall decrease in fitness, but this decrease happened at an earlier passage and the fitness loss was larger. The fitness values for the persistent-replication viruses appeared to recover after passage 10 but were still lower than the fitness values for the wild-type progenitor. When we adjusted the data from passages 5 to 25 to a linear regression, we got a positive slope that was significantly different from zero (P < 0.0001; F test). Interestingly, the fitness values for the alternating-replication viruses at passage 25 were not significantly different from those measured for the viruses that were maintained in LL-5 cells during the entire experiment (P > 0.1; unpaired t test). Unlike viruses undergoing regimens including persistent replication, the viruses passaged only in BHK-21 cells showed continuous increase in fitness, as previously reported (30, 31).

FIG. 2.

Evolution of VSV lineages in BHK cells. (A) The fitness of the four replicates of the viral lineages undergoing the three regimens (A, alternating; P, persistent; K, acute) was determined at different times and plotted against the passage number. The horizontal line indicates neutrality. (B) VSV cytopathology in BHK-21 cells. The photos correspond to monolayers at 12 h pi with no virus (1), wild-type virus (2), P5B (3), or P25B (4) and were taken with a 10× lens objective.

Changes in virulence in BHK-21 cells.

Viruses from persistently infected insect cells not only displayed fitness losses in BHK-21 cells, they also evolved in virulence in these mammalian cells during this regimen. We need to stress that virulence in BHK-21 cells cannot be extrapolated to virulence in any animal host. We characterized the changes observed in one of the replicates in terms of plaque size and CPE. BHK-21 cells infected with the wild type showed extensive cell rounding at 12 h pi, similar to that observed in infections with virus obtained after five passages of persistent infection, P5B (Fig. 2B). In contrast, virus obtained after 25 passages of persistent infection, P25B, caused little cell rounding in infected monolayers (Fig. 2B). At 18 h pi, wild type-infected monolayers showed a significant number of syncytia while none of the viruses from persistently infected cells caused this effect at any time (data not shown); at 36 h pi, all the cells were killed by the three strains and detached from the flask surface.

Differences in plaque sizes correlated well with differences in levels of cytopathology. At 30 h pi, 85% of wild-type plaques had a diameter of 1 mm or more (50% were 2 mm or more in diameter). In contrast, 92% of P5B plaques were less than 0.5 mm in diameter (77% were pinpoint plaques). At 30 h pi, the plaques formed by P25B were not visible yet. This tendency to show less CPE and smaller plaque size on BHK-21 cells was also observed in other replicates from the persistent-infection regimens, although a detailed characterization was not done. These results showed that persistent replication selected for virus populations of increasingly lower virulence for the mammalian host cells.

The changes in fitness and virulence during persistence may be explained by accumulation of defective interfering particles (DIPs), which may interfere with replication. The potential presence of significant levels of DIPs was tested by infecting BHK-21 monolayers with MARM U diluted in (i) Mitsuhashi and Maramorosch medium, (ii) supernatant from P25B, and (iii) supernatant from P25B after amplification in LL-5 cells. The potential of formation of plaques by P25B virus was eliminated by addition of I1 MAb to the agar overlay and by development of the plaque assays at 20 h pi. These infections were done in triplicate. The numbers and the sizes of plaques formed under these three conditions were indistinguishable (data not shown), suggesting that if DIPs are present, their level is low and causes no major effects on the phenotypes observed.

Fitness trajectories in insect cells.

We initially carried out fitness assays in LL-5 cells by determining viral yields at 2 weeks pi, to be consistent with the way the adaptation passages were carried out. However, for the persistent and alternating strains, we could not detect the presence of MARM U (the neutral surrogate used in the competitions) at this time, even when competitions were started with a 50-fold excess of MARM U. These changes in ratio correspond to a fitness increase of over 3 orders of magnitude. We tested the fitness of one of the mammalian strains (KA) at passage 25 in competition with MARM U during persistent infection of LL-5 cells. The BHK-21-adapted virus showed a fitness value of 85 under these environmental conditions, indicating that replication in mammalian cells improved the fitness of the virus for LL-5 persistent infections (Fig. 3A). A second set of fitness assays were done for all the strains by recovering the viral progeny at 48 h pi, when viral titers reached their peak (Fig. 1). Figure 3A shows that the fitness of the viruses subjected to alternating persistent-cytolytic infection passages increased by 3 orders of magnitude from passage 1 to passage 10. However, the fitness seemed to reach a plateau at passage 10 and additional changes were not observed beyond that point. The viruses from the persistent infections also exhibited an increase in fitness (Fig. 3A) that was not significantly different from that observed for the viruses coming from the alternating-infection passages once the plateau was reached (P > 0.1; unpaired t test). Thus, viruses from these two different regimens, both including persistent replication, reached similar states of phenotypic stasis. Consistent with previous results (31), BHK-adapted virus increased its fitness in LL-5 cells. The extremely high fitness values obtained for viruses from persistent and alternating infections during competition in LL-5 cells suggest that the ability of these viruses to produce plaques on BHK-21 cells was not compromised, although plaque size was small (see below).

FIG. 3.

Evolution of fitness of VSV in LL-5 cells. (A) The fitness of viruses undergoing regimens of persistent (P), alternating (A), and acute (K) infections in LL-5 cells was determined and plotted against the passage number. The fitness of both lineages undergoing regimens that included persistent replication increased by 3 orders of magnitude and then remained approximately constant. The filled diamond represents the fitness of a mammalian virus (K25A) measured during persistent replication in LL-5 cells. For comparison purposes, we also included the fitness in LL-5 of a VSV population evolving during acute LL-5 infections (LL-VSV; i.e., passages with virus progeny harvested at maximum titers at 48 h pi; data were taken from Novella et al. [31]). (B) Effect of temperature on the fitness of the alternating-infection lineages at passage 15. The fitness values determined for BHK-21 cells at different temperatures were not significantly different (P > 0.5). Fitness values for BHK-21 cells at either temperature (28 or 37°C) were significantly different from fitness values for LL-5 cells (P < 0.001). The horizontal line indicates neutrality.

Effect of incubating temperature on fitness values.

One possible cause for the differences in the fitness values observed in BHK-21 and LL-5 cells was the incubation temperatures. While BHK-21 infections were carried out at 37°C, the infected LL-5 cells were incubated at 28°C. To explore this possibility, we measured the fitness of the viruses from the alternating passage 15, maintaining the BHK-21 cells at 28°C during competition. Figure 3B shows that the fitness values obtained in BHK-21 cells at both temperatures were not significantly different from each other (P > 0.5) but that the differences between the fitness values in LL-5 cells and in BHK-21 cells were highly significant (P < 0.001; analysis of variance; Tukey-Kramer test). This rules out the possibility that the differences in the fitness values determined for BHK-21 and LL-5 cells were due to the difference in the incubation temperatures.

Genetic evolution.

Next, we analyzed the accumulation of mutations in the evolved populations. We sequenced one randomly selected replicate from each experimental regimen at passage 25, targeting the regions encoding N, P, M, and G proteins as in previous work (31). In addition, we sequenced the alternating-infection lineage at passage 15 to see whether the fitness plateau observed correlated with genetic stasis. The mutations found and their positions are shown in Table 1.

TABLE 1.

Distribution of mutations within N, P, M, and G protein open reading frames^a

Nucleotide no.b	Corresponding proteins (amino acid position)	Nucleotide change and corresponding amino acid change in virus:
		K25	P25	A15	A25
163	N (33)			A→G, Lys→Glu	A→G, Lys→Glu
187	P (141)		C→U, Thr→Met	C→U, Thr→Met	C→U, Thr→Met
2257	M (3)		C→U, Ser→Phe	C→U, Ser→Phe	C→U, Ser→Phe
2937	M (248)	U→C + U, Stop→Arg
3421	G (115)		C→A, Thr→Lys	C→A, Thr→Lys	C→A, Thr→Lys
3978	G (301)		C→A, Gln→Lys	C→A + C, Gln→Lys	C→A + C, Gln→Lys
4122	G (349)		G→U, Val→Phe	G→U, Val→Phe	G→U, Val→Phe
4139	G (354)				A→U, Pro
4434	G (453)		U→G, Phe→Cys

^aMutations were analyzed after 25 passages in acutely infected BHK-21 cells (K25) and persistently infected LL-5 cells (P25) and 15 and 25 passages alternating between acute infection in BHK-21 cells and persistence in LL-5 cells (A15 and A25).

^bThe nucleotide changes correspond to the sequence of the positive strand.

The sequence of K25A showed one point mutation when compared with that of the parental strain. This mutation changed the stop codon in the M protein gene with an arginine residue and resulted in the addition of eight amino acids to the M protein. A15A revealed six mutations; all of them resulted in amino acid changes in the corresponding polypeptide (Table 1). A25A had one additional mutation, which was silent. Interestingly, the six nonsynonymous mutations found in the virus from the alternating passages were also present in P25A. An additional nonsynonymous mutation was found in P25A that was not present in the alternately passaged virus. The six mutations found in A15A corresponded to an evolutionary rate of 8.8 × 10⁻⁴ substitutions per nt per year. At passage 25, we found only one additional mutation, and the corresponding evolutionary rate was 2.12 × 10⁻⁴ substitutions per nt per year, in agreement with the fitness stasis observed in this population.

DISCUSSION

Previous work on VSV adaptation to alternating hosts showed that trade-off in a cell type does not ensue from replication in a different cell type (31, 39). Even if trade-off is observed, alternating replication in two hosts had always resulted in fitness increases in both hosts for several arboviruses, including VSV (31, 39), EEEV (10, 42), and Dengue virus (5). Bacteriophage ΦX174 replicating alternatively in Escherichia coli and Salmonella enterica shows an asymmetric trade-off. When the bacteriophage is adapted to grow in Salmonella, it has an impaired ability to grow in E. coli, but replication in E. coli does not affect the ability of the phage to grow in Salmonella (12). However, the bacteriophage recovers the ability to grow in E. coli after replication in this host. We also found an asymmetric trade-off, but in our case a single passage of the virus in BHK-21 cells was not enough for VSV to recover fitness in this cell line.

It is important to note that even though the viruses were maintained in LL-5 cells for 2 weeks, the viral production was still orders of magnitude lower than that of viruses replicating in BHK-21 cells for a single day. This implies that the dominance of the persistence phase in the phenotype is not due to a larger part of the replication's taking place in LL-5 cells and indicates that the selective pressure in the insect cells is stronger. It is worth mentioning that viral titer does not necessarily correlate with fitness (Table 2). The viruses obtained after 25 passages in persistently infected cells exhibited a 1,000-fold increase in fitness in LL-5 cells compared to the parental strain, but the titers dropped from 10⁷ PFU/ml at passage 1 to 10⁴ PFU/ml at passage 25. This finding may be due to smaller burst sizes of infected cells or to decreased numbers of infected cells. Virulence of the persistent-infection viruses in LL-5 cells did not seem to change, since cell damage was not observed at any point of the regime. However, persistence in LL-5 cells selected populations with low virulence for BHK-21 cells.

TABLE 2.

Summary of phenotypic properties of the wild type (wt) and virus with a history of persistent replication in LL-5 cells

Passage	Fitness in cells of line:		Virus productiona (PFU/ml) in cells of line:		Virulenceb in cells of line:
	LL-5	BHK-21	LL5	BHK-21	LL-5	BHK-21
0 (wt)	1	1	>106	1010	−	++++
5	2,000	0.06	105	ND	−	++
25	6,200	0.15	<104	ND	−	+

^aND, not determined.

^bMinus signs indicate no cell damage; different numbers of plus signs correspond to different levels of virulence, with + being the lowest.

The most remarkable observation is probably the nearly identical behavior obtained from lineages undergoing alternating-infection and purely persistent-infection regimens. Since the two sets of experiments were carried out separately, we can rule out cross contamination as an explanation for these results. The fact that six of eight mutations were present in the consensus sequence of both sets of viruses suggests that strong positive selection was driving the evolution of these populations. Not only did these two regimens produce similar fitness trajectories in LL-5 cells, but the genotypes of the corresponding viruses revealed that they had also fixated the same mutations. These results revealed highly deterministic behavior with complete dominance of the persistence phase and little or no contribution from the acute phase. Deterministic evolution has been often observed during experimental evolution in other viral systems (4, 13).

The fitness trajectories of persistent-infection viruses in BHK-21 cells showed two distinct phases of trade-off. After an initial period of very large fitness loss, there was a gradual recovery of fitness and reversal of the trade-off, even though neutrality was never reached. This behavior may be interpreted as the fixation of two types of mutations in these populations. During the initial period, mutations that were responsible for the trade-off were fixed, but at later times, additional mutations were fixed that not only had no cost but also were beneficial for acute replication. Consistent with this interpretation are the results obtained from the strains undergoing regimens of acute replication in mammalian and/or insect cells (31). These viruses show increased fitness in both cell types, and the determination of their genomic sequences reveals three mutations found in the strains from persistently infected cells (32). The C→U transition at nt 2257 (corresponding to a Ser-3→Thr change in the M protein) and the G→U transversion at nt 4122 (Val-349→Phe in the G protein) have been identified in viruses with a history of replication exclusively in BHK-21 cells. Mutations in nt 3421 (Thr-115→Lys in the G protein) have been found in viruses with a history of alternation between acute infections in LL-5 and BHK-21 cells. Despite variations during the initial passages of the experiment, the fitness of all the populations recovered from persistent infections (with or without a step of mammalian infection) converged towards a value of approximately 0.1 at the end of the experiment. Similar results were observed with phage φX174, in which convergent (parallel) evolution of several populations was reached but differences were observed in the trajectories followed by each replicate (44).

It is generally accepted that arboviral replication in both arthropod and vertebrate hosts is necessary in the life cycles of arboviruses. Sand fly females can transmit VSV vertically (transovarially) to offspring, as well as horizontally to mammals during blood meals (9). Even though transovarian transmission of VSV in insects can occur over several generations, eventually the virus is lost, indicating the necessity of mammalian infection. The loss in titers observed during the vertical transmission is also observed during persistent passages, suggesting that the cell culture regimen of persistent infection is an appropriate model for transovarial transmission during natural infections. The role of the mammalian host in the life cycle of VSV is not clear. While specific antibodies against VSV have been detected in different animals exposed to VSV, no viremia has been reported in mammalian hosts (7, 8). Moreover, even though nonviremic animals show high titers of antibodies against VSV, immune selection has not been observed, suggesting that the acute infection of mammals may have little effect on the evolution of the virus (33). Also pointing in this direction are the results of Mead and collaborators, who found that when infected and noninfected black flies are allowed to feed on the same host, the noninfected flies acquire the virus. Since no viremia is detected in the mammalian host, this result indicates that viral replication in this host is not necessary for the insects to become infected with VSV (26) and that transmission of VSV may take place mainly by cofeeding. The role of mammals in the virus life cycle may simply be to provide a physical support to allow sand fly-to-sand fly transmission.

The idea that virulence (defined as host mortality) is not necessarily related to the fitness of a pathogen and transmission to another host is not new (18). Our results are relevant only to the specific environments tested for this work: BHK-21 and LL-5 cells. The behavior of any of these laboratory strains in living swine, cows, horses, or sand flies is unpredictable from our in vitro results. However, high density of hosts should favor relatively high virulence (2, 3). Early transmission also favors the selection of more-virulent strains (11). Long-term replication in sand fly cells, either during acute infections (31) or during persistent passages, did not select for strains of higher virulence in the cell culture systems we have used, illustrating that other virus-host interactions need to be considered. However, persistent replication in insect cells selected for strains of lower virulence in the mammalian host cells. Our results showed lack of correlation between virulence and fitness (Table 2). Bergstrom and coworkers (1) proposed a model for virulence evolution based on the operation of a mechanism similar to Muller's ratchet, such that inefficient vertical transmission would promote sampling errors and the loss of virulent mutants. In our case, the fitness trajectories during persistence are consistent with the absence of significant bottleneck effects, suggesting that this mechanism is not operating in VSV populations.

Sequence analysis of natural VSV isolates clearly indicates the absence of molecular clocks and suggests that ecological factors rather than geographical clines determine the evolution of this virus (27, 34). A possible model for VSV evolution may be based on different requirements for the establishment of persistence in different species of insect hosts. The ecological characteristics of a particular area determine the species of resident insects that can serve as hosts for VSV. Unfortunately, many different insect species carry VSV during epizootic periods, and it is unclear which serve as vectors and which do not (22). However, a New Jersey serotype isolate of VSV from a sand fly was able to establish productive, long-lasting persistent infections in LL-5 cells but was relatively inefficient at sustaining persistence in mosquito C6/36 cells (23). More work towards identifying insect VSV vectors and their distribution in nature would help us to understand the forces that drive the natural evolution of this virus.