Foot-and-mouth disease (FMD) has a high economical impact, affecting domestic and wild cloven-hoofed animal species worldwide (reviewed in references 2
, and 46
). The etiological agent FMD virus (FMDV) of Picornaviridae
occurs as seven distinct serotypes and multiple subtypes, reflecting significant genetic and antigenic heterogeneity. In the field, this heterogeneity is reflected by the lack of cross-protection even between intraserotype variants (2
VP1 (1D), the highly variable FMDV capsid protein with roles in virus entry, immunity, and serotype specificity, has been the subject of extensive comparative sequence analysis (reviewed in reference 22
). These studies have shown cocirculation of FMDV genotypes in single outbreaks, with genotypes usually grouping into geographically and genetically distinct lineages (less than 15% nucleotide differences) known as topotypes (41
). With the expansion of FMDV genomic databases, however, evidence is accumulating for the inadequacy of VP1 analysis alone for epidemiological studies and for the importance of recombination in FMDV evolution (4
The selective forces at work during the emergence of FMDV populations in nature are likely to be influenced by specific epidemiological and immunological aspects of host-virus interaction as well as the quasispecies composition of the viral population. Many important questions, including those regarding the significance of high mutation rates in adaptive virus evolution, of Darwinian selection in diversification of viruses with short infection cycles, and of genetic drift as a mechanism for FMDV evolution, remain unanswered. Similarly, there is no knowledge of the limits within which a highly variable pathogen, such as FMDV, can accumulate genomic changes and still reproduce the disease in the natural host and spread in the natural environment. Very few studies have been published regarding FMDV-natural host interactions at the genetic level (5
). No studies have been conducted to examine FMDV evolution during replication in the natural host, and very few evolutionary analyses have examined genomic regions other than those corresponding to VP1 or its precursor, P1 (4
). Paradoxically, the few experimental studies conducted with natural isolates suggest extreme constrains for 1D variation (5
) and loss of fitness during passages in natural hosts (1
). In fact, enhanced mutagenesis experiments have shown infectivity loss for a number of RNA viruses, including FMDV, lymphocytic choriomeningitis virus, and Hantavirus (18
), suggesting that critical variability thresholds that may explain the restrictions for change observed in vivo exist. However, the characteristics and boundaries of those limits in genetic variation and phenotypic expression remain unknown.
Here, we analyzed genetic changes in full-length FMDV genomes during serial passages of O Taiwan 97 (O Tw97) virus in pigs and in BHK-21 cells. Originally isolated from pigs during an FMD outbreak, O Tw97 virus exhibits rapid spread and high virulence in pigs (13
). New FMDV genetic variants with altered pathogenicity in pigs and the rapid replacement of the original consensus sequence by new variant genotypes with acquired mutations, mostly outside the capsid coding region P1, were observed. The data indicate rapid accumulation of nucleotide substitutions and fitness loss, suggesting bottleneck transmission effects. Fixation of amino acid changes in nonstructural proteins (NSPs) likely resulted in deleterious effects for virus biology, leading to the establishment of a subclinical infection that resembles the carrier state described for cattle (2
). Furthermore, we found significant differences in evolution parameters between in vivo- and in vitro-passaged virus, reflecting differences in selective pressures operating on virus populations, expressed as differences between numbers of synonymous and nonsynonymous substitutions, frequencies of transitions and transversions, and levels of tolerance for changes in specific viral proteins.
MATERIALS AND METHODS
O Tw97 virus was obtained from the Animal and Plant Health Inspection Service, USDA, as a 10% swine epithelial tissue homogenate and was completely sequenced (GenBank no. AY593835). For stock virus production, two pigs (no. 48 and 50) were inoculated with the supernatant of the epithelial homogenate (0.5 ml; 105.7
50% tissue culture infectious doses [TCID50
]/ml) by the intradermal (i.d.) route. Within 24 h after inoculation, both animals developed high fevers and generalized lesions with large vesicles on the feet and snout, symptoms consistent with previous descriptions of O Tw97 virus infection (14
). Vesicular fluid was collected from independent lesions and pooled to create a viral stock, T00. The T00 stock was titrated and the FMDV genome completely sequenced and then used to infect a pig i.d. in pig passage T0 (Table 1
serial dilutions of vesicle fluid were used to inoculate BHK-21 cells. After 1 h of virus adsorption, cells were cultivated with 50 μl medium containing 2% fetal calf serum (FCS) at 37°C and 5% CO2
for 72 h. Cell monolayers were stained with neutral violet, plaques counted, and titers determined using the Reed and Muench method (37
) and expressed as numbers of TCID50
For the serial passage experiment, 4-week-old pigs were randomly paired and housed in containment rooms. One pig, T0 (meaning time zero of infection), was inoculated by the i.d. route with 100 μl of T00 (106.47 TCID50/ml) and housed with two recipient pigs (T1 [time 1]) in the same room. When the body temperatures of the T1 pigs reached 104°F or above, the T1 pigs were moved to a clean room with two noninfected pigs (T2 [time 2]). The period of time between the T1-T2 contact and the appearance of vesicular lesions on the feet and/or mouth of any of the T2 pigs is what we define as the “infectious round.” The number of cohabitation days for each infectious round differed between passages. This procedure was repeated for every infectious round up to T13, the last round of the transmission-of-infection chain. For each infectious round, when vesicles in donors became evident, vesicle fluid was collected, and the animals were kept in contact with the recipient animals until fever occurred (i.e., T3); then, the animals were culled. For each infected animal, we recorded daily body temperatures and the presence of clinical symptoms.
Vesicular fluids were individually collected with sterile syringes, placed on ice, and transported to the laboratory, where titrations were immediately performed. The remaining volume was stored at −70°C until used for RNA extraction and sequencing. Epithelial tissue from broken vesicles was collected using clean sterile scissors, immersed in cryotubes containing 500 μl of Dulbecco's modified Eagle's medium (DMEM), and immediately frozen at −70°C. Tonsil scrapings and nasal swabs were collected from animals that did not present signs of disease after 26 days in contact with donor animals. This material was used for both reverse transcription (RT)-PCR and virus isolation in BHK-21 cells.
BHK-21 cell culture infections.
BHK-21 cells were grown in T25 tissue culture flasks with DMEM containing 5% FCS. Cells were serially passaged 23 times at a concentration of 105 cells/ml. Infections were carried out when cells were approximately 95% confluent using a multiplicity of infection of 1 to 10 virus particles per cell from the previous viral passage and cultured in DMEM with 2% FCS. The first passage, P1, was carried out with 0.5 ml of vesicular fluid containing 106.47 TCDI50s/ml from the T00 stock virus, resulting in a multiplicity of infection of 1 to 10 virus particles per cell. In this case, we define the infectious round as the period of time between the culture inoculation and the detection of a complete cytopathic effect. When the cytopathic effect was complete, the culture was frozen and thawed three consecutive times; the supernatant was clarified by centrifugation at 3,000 rpm for 10 min and fractionated in 1-ml aliquots at −70°C. For the next infectious round, 1 ml of the supernatant was used for infecting BHK-21 tissue cultures in duplicate (P2a and P2b). By repetition of these steps, serial infections of separated lineages (A and B) were carried out up to passages 23A and 23B. Titrations of every infected cell culture supernatant were performed for each passage.
RT-PCR and sequencing.
Total RNA was directly extracted from 140 μl of the DMEM-vesicular fluid mixture or from infected cell culture supernatants. Full-length FMDV genome sequences were obtained by RT of the viral genomic RNA, followed by amplification and sequencing of overlapping cDNA fragments spanning the entire viral genome as previously described (4
Direct DNA sequencing of amplicons derived from a given FMDV isolate yielded a consensus sequence representing the most probable nucleotide for each position of the sequence. This approach prevented analysis of minor sequence variants, polymerase misincorporation errors, and sequencing ambiguities through multiple independent cDNA synthesis, PCR amplification, and direct sequencing events. Due to the quasispecies nature of FMDV populations, polymorphisms were detected in some nucleotide positions. Nevertheless, all positions could be unambiguously assigned to a single dominant nucleotide due to the high degree of redundancy generated by the sequencing strategy.
As described previously (4
), bases were called from chromatogram traces with the Phred program, which also produced a quality file containing a predicted error probability at each base position. Viral sequences were assembled with the Phrap and CAP3 assemblers. Gap closure was performed as described previously (4
). Multiple sequence alignments were performed with the ClustalW (version 1.7) computer program. Analyses of codons and synonymous/nonsynonymous substitution ratios were calculated using the programs SNAP, CodonW (http://www.molbiol.ox.ac.uk/cu/
), and codeml (PAML3.14 package), which was also used for statistical evaluation of heterogeneous selection pressures at amino acid sites. For protein analysis, the PRETTY program was used. Protein secondary-structure predictions were performed using the GOR and Pratt computer programs. The codeml program was used to analyze and compare predicted positively selected sites in the FMDV genome under in vitro and in vivo growth conditions. A Bayes Empirical Bayes (BEB) analysis-based codeml model giving the highest probability values (P
= 0.001) was chosen for the analysis.
Repeated contact transmission of the highly virulent O Tw97 isolate of FMD resulted in complete attenuation of pathogenicity in pigs, characterized by an asymptomatic infection resembling the FMD carrier state previously described for ruminants but not previously described as occurring in swine (39
). This work is the first to correlate decay of infection levels with fixation and accumulation of genetic mutations in in vivo viral quasispecies during serial infections, which could be interpreted as the result of bottleneck transmission and the genetic effect of Muller's ratchet (7
). Such a rapid substitution of the consensus sequence has been demonstrated to occur during in vitro clone-to-clone replication of FMDV (17
) and other viruses (7
), leading to a detrimental accumulation of mutations. In the case of plaque-to-plaque transfers; however, the mutations were never transitory but remained fixed in the genomes. We have shown that loss of transmission was not due to lower viral titers in vesicular fluids; however, as we mentioned above, significant decreases in numbers and sizes of lesions may have impacted total viral yields from one to the next infectious round. In any case, progressive loss of virulence upon pig passages is related to viral genetic factors. These results suggest that if the observed loss of pathogenicity and accumulation of genetic mutations is due to the so-called Muller's ratchet effect, our data may reflect features of FMDV infection of great importance for pathogenesis, virus transmission, and FMD epidemiology in nature (7
). In contrast to the observed resistance to extinction of FMDV subjected to plaque-to-plaque in vitro transfers (15
), animal-to-animal transmission in nature may lead to virus extinction. Under conditions of natural host transmission, a number of viral phenotypic functions are likely involved in virus-host interactions and an unknown number of bottlenecks lead to a continuous purification of the population, narrowing the mutant spectrum composition of the quasispecies in such a way that it is unable to successfully retain its fitness. This has recently been observed for another picornavirus, a poliovirus mutant displaying enhanced polymerase fidelity (35
For survival in nature, the viability, virulence, and transmission capability of FMDV must be maintained by as yet unknown mechanisms. In light of recent insights from bottlenecking effects observed during poliovirus infection in mice (36
), it is likely that bottlenecking is the result of organ tropism and tissue-specific amplification within the host, resulting in the generalization of the progeny from very few particles of the parental quasispecies. Two observations suggest that more than one point of selection may act as a bottleneck during FMDV infection of the host. Our present data (Table 2
) and previous reports have demonstrated with different FMDV isolates that 100 or more TCID50
of BHK-21 cell passaged infectious viral particles injected i.d. into pigs result in identical parental and progeny viral consensus sequences (5
). The rapid imposition of new genetic variants observed here indicates that the initial route of animal infection imposes a serious barrier and acts as a bottleneck for the initial viral population. Additionally, it has been previously shown that an FMDV variant isolated during the febrile phase of the disease from blood from a pig infected with a highly purified homologous population of C-S8c1, an FMDV variant isolated from pigs and plaque purified three times in BHK-21 tissue culture before being inoculated into pigs (5
), showed a consensus sequence different from that of the virus obtained from vesicles. This viremic variant was genetically stable upon cell culture and pig passages and showed phenotypic differences from the parental strain, which correlated with its origination from viremic blood (6
). Thus, different viral variants cocirculate during FMDV infection of the host, likely as a result of bottlenecks during spread and replication within the host, although unless the genetic mutation selected during the bottleneck is advantageous with respect to the parental virus, the epitheliotropic FMDV consensus sequence will be the major progeny population in vesicular fluid. Therefore, more than one bottlenecking event may occur during FMDV infection and this may affect subsequent transmission in natural hosts.
The RasMol 2.7.1 program (www.rasmol.org
) and published crystal structures of FMDV proteins were used to analyze predicted effects of amino acid changes detected in viral proteins during pig and cell culture passages. The E186/A substitution in Lpro
falls in a highly disordered, unresolved region of the protease. The nonconservative Q580/R change affecting position 76 in VP3 is in close contact with P132 of the same protein, and its replacement by R may have an effect on the folding of the protein since R76 seems to interrupt the long α-helix structure of the parental sequence to induce a β-sheet structure. Secondary-structure analysis (Chou-Fasman) of the P114S substitution in the 3C viral proteinase predicts no significant effect. Both residues are small and uncharged; examination of the A10 virus 3C crystal structure revealed that P114 is on the protein surface, and although distant from the active site, P114 may somehow affect the protease substrate specificity pocket (S. Curry, personal communication). Substitutions in the 3D region, E11/Q and K156/E, seem to affect the protein surface and may have possible consequences for functional interaction between 3D and other proteins in the replication complex (Table 5
The quantification of the FMDV genomic variability following the cell culture passage observed here is consistent with previous published reports (11
). Characterizations of FMDV genomic regions most extensively affected by mutations are difficult to reconcile with the many reports using VP1 as an indicator of variability to obtain phylogenetic information from field isolates (24
). Previous analyses of partial sequences of the VP1-coding region following a single passage in vivo (5
), along with recent full-length genome studies performed with UK2001 field isolates, support our present data indicating that a surprisingly low number of mutations are found in SPs in animals that have not been vaccinated (9
). These differences could be the result of early transmission events which precede development of antigenic variants due to host immune responses, while the field isolates compared in epidemiological studies come from animals with developing immune responses to previous infections and/or vaccination, which could act as driving forces for positive selection of antigenic variants (26
Pigs 5017 and 4822 did not exhibit any sign of disease after 26 days in contact with T14 donors. Nevertheless, infectious virus was isolated from tonsil scrapings and nasal swabs collected from both animals, with titers of 103.2
/ml and 103.9
/ml and 103.9
/ml and 102.4
/ml, respectively. Virus isolation from pig tonsils at day 26 postcontact confirmed that these animals had been infected without clinical symptoms of disease and that virus had persisted in them for 4 weeks. This case resembles what has been described as a carrier state of FMDV for cattle, sheep, and goats (2
) but not yet demonstrated for pigs. Bottleneck transmission may confer on the virus the ability to ratchet down fitness and virulence to ensure the immunization of at least a fraction of the population rather than kill or debilitate the entire susceptible host population. Indeed, most lineages would be destined to be self-limiting in subclinical and nonproductive infections. Interestingly, phylogenetically based epidemiological studies have indicated that FMDV topotypes appear to represent evolutionary cul-de-sacs (41
). Our results suggest strong selection against changes in capsid proteins and higher flexibility for changes in NSP 2C and 3D in vivo, while a strong selection for substitutions in the P1 region (Table 3
) is shown in vitro. These data confirm previous reports of spontaneous mutations in VP1 and the rise of antigenic variants occurring during FMDV replication in cell cultures in the absence of immunological selective pressure (11
). We do not understand this difference, since in both cases there is no immunological selective pressure. This observation may result from as yet unknown selective pressures involving viral receptor binding and/or particle internalization present in vivo. Finally, our results demonstrate that the effects of host adaptation can be objectively quantified and compared through the calculation of parameters of evolution and selective pressure, like those obtained with the CODELM analysis program (PALM). Although preliminary and limited, this is a novel and promising approach for analysis of FMDV genomic variability suggesting that the extension of our knowledge regarding viral evolution under experimental conditions in natural hosts will allow development of molecular epidemiology tools for improved identification of viral strains.