INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Yellow fever (YF) virus causes a viral hemorrhagic fever in humans. The case fatality rate of YF can exceed 50% (17). YF remains a major public health concern in sub-Saharan Africa and tropical South America despite the availability of a safe and effective vaccine. For example, in the period between 1987 and 1991, a total of 18,735 YF cases and 4,522 deaths were reported to the World Health Organization (15). These figures represent the greatest YF activity since 1948 (15). YF is transmitted by the bite of an infected female mosquito, usually Aedes species in Africa and Haemagogus or Sabethes species in South America. YF virus is the prototype member of the genus Flavivirus, family Flaviviridae. Flaviviruses have a small positive-sense, single-stranded RNA genome. The prototype strain of YF virus is Asibi, and the genome consists of 10,862 nucleotides (6). The genome is arranged into a short 5' noncoding region, a single open reading frame consisting of 10,233 nucleotides that encodes the structural genes C, prM, and E and nonstructural genes NS1, NS2A, NS2B, NS3, NS4A, 2K, NS4B, and NS5, and a 3' noncoding region (1a).

Genetic relationships among wild YF virus strains in Africa are not well understood, primarily because of the limited number of studies on the subject. Although six studies (2, 3, 4, 10, 19, 20) have specifically analyzed genetic and phylogenetic relationships among wild YF strains from Africa, these studies utilized relatively few YF strains, from 3 (20) to 21 (10). This is underrepresentative considering the size of the zone where YF is endemic in Africa and the frequency of YF outbreaks in this region. Most of these studies (2, 3, 10, 19) showed clear genetic and phylogenetic distinction between east/central and west African YF strains. However, the samples were biased towards west Africa, which is probably attributed to the availability of isolates due to more YF activity in this region than in east and central Africa. For example, Lepiniec et al. (10) analyzed 21 wild strains of YF virus, but only 6 (28.5%) were from east and central Africa.

Previous studies (2, 3, 4, 10, 19, 20) showed that east and central African wild YF strains were closely related genetically and belong to the same genotype. The same studies showed that strains from west Africa were genetically distinct from those in east/central Africa and represented two distinct genotypes. Deubel et al. (4) used oligonucleotide fingerprinting and identified three genetically distinct topotypes of YF virus in Africa, two in west Africa and one in east/central Africa. Subsequently, Lepiniec et al. (10) and Chang et al. (2) analyzed nucleotide sequence variation of the envelope (E) protein gene and described two genotypes, one in east/central Africa and the other in west Africa. In the present study, we examined a panel of 38 spatially and temporally diverse wild YF virus isolates from diverse regions of Africa to elucidate the precise genetic relatedness of wild strains of YF virus in Africa.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Viruses. Thirty-eight low-passage virus strains isolated between 1927 and 1993 from 13 African countries were used in this study (Table 1). The virus strains were lyophilized stocks obtained from the World Arbovirus Reference Collection at the University of Texas Medical Branch, Galveston, Tex. and Centers for Diseases Control and Prevention, Fort Collins, Colo., except strain 85-82H (13). Twenty-eight (73.7%) were epidemic strains, and 10 (26.3%) were enzootic. Similarly, 30 (78.9%) strains were from human cases, 7 (18.4%) were from mosquitoes, and 1 (2.6%) strain from Uganda (Uganda 72; strain number Z 19039) was from a monkey. The majority of strains (23 [60.5%]) were from west Africa. Nine strains were from eastern Africa, five from central Africa, and one (Angola 71; strain 14 FA) was from southern Africa. Each reconstituted virus preparation was passaged once in Vero cell cultures to produce seed virus. The seed virus was used to prepare working stocks by an additional passage in Vero cell culture.

Nucleotide sequencing studies. Methods used to grow virus and to extract viral RNA have been described in detail elsewhere (18). Reverse transcription (RT)-PCR was performed on purified viral RNA using the procedures described by Wang et al. (19). One set of studies involved amplification of a 670-bp DNA fragment for all 38 YF strains using the CAG (CTGTCCCAATCTCAGTCC) and the YF7 (AATGCTTCCTTTCCCAAAT) primers. This fragment included the 3' 108 nucleotides of the premembrane (prM) protein gene, the entire 225 nucleotides of the membrane (M) protein gene, and the 5' 337 nucleotides of the envelope (E) protein-coding gene. We have previously shown that this region is a representative sample of the YF genome (19). The second set of studies amplified the structural protein genes C, prM/M, and E for the Angola strain of YF virus. PCR products were screened using agarose gels and ethidium bromide staining. The PCR products were then extracted from the gels and either cloned into pGEM (Easy) vectors (Promega) or directly sequenced, depending on the quantity of cDNA recovered. For cloned cDNAs, three clones were sequenced to provide a representative consensus sequence for the strain. Sequencing was done using an ABI automatic sequencer at the University of Texas Medical Branch protein chemistry core facility.

Sequence data analysis. Nucleotide sequences for the YF strains were imported directly and aligned using Vector NTI sequence analysis program (Informax). Percentage similarities and differences were calculated using the MegAlign program (DNASTAR, Lasergene). Phylogenetic analysis of the aligned sequences was performed using PAUP (16) and NEIGHBOR, a neighbor-joining program, in the PHYLIP package (5). Parsimony analysis was implemented using the heuristic algorithm. The one-parameter formula was used to generate the distance matrix for neighbor-joining analysis (9). Bootstrap analysis with 1,000 resamplings was used to determine confidence values for groupings within the phylogenetic tree. The tree was rooted using a homologous sequence of dengue-1 virus (GenBank accession no. NC001477). We estimated nucleotide substitution rates by identifying sister sequences that were closely related and isolated at least 5 years apart. The differences in changes depicted in branch lengths separating each sister sequence from the predicted common ancestor were divided by the number of years between the sequences and the number of nucleotides in the sample sequence. Several estimates were compared to provide an estimate mean and standard deviation.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Nucleotide sequence variation among strains of YF virus from Africa. We selected a 670-nucleotide fragment that includes coding regions for three proteins, the premembrane (prM), membrane (M), and envelope (E), which has been shown to be representative of the entire genome of wild-type strains of YF virus (19). Nucleotide sequences of this region were determined for 31 strains of YF virus amplified in this study plus 7 previously published sequences and used to generate a phylogenetic tree (Fig. 1). The phylogenetic tree generated indicated that wild-type YF virus strains in Africa could be divided into two major lineages, strains in west Africa and those from east/central Africa (Fig. 1). The two lineages were further divided into five clades, two in west Africa and three in east/central Africa, based on phylogenetic relationships (Fig. 1) and variation in nucleotide sequence (Table 2 and Fig. 2). Genotypes were defined as distinct lineages that differed by greater than 9% at the nucleotide sequence level (Table 2). A similar criterion was used by Lepineic et al. (10), Wang et al. (18), and Chang et al. (2) to define YF virus genotypes in Africa and South America.

View larger version (32K): [in this window] [in a new window]   FIG. 1.   Neighbor-joining phylogram derived from the nucleotide sequence of the prM/E region of 37 wild YF virus strains from Africa.

View larger version (37K): [in this window] [in a new window]   FIG. 2.   Nucleotide sequence alignment for the distal 108 nucleotides of the premembrane protein gene of 37 YF viruses from tropical Africa. This alignment highlights nucleotide variations among different genotypes and similarities within genotypes. Dots indicate identity with the consensus sequence.

Nucleotide variation differs between genotypes. Nucleotide variation in west African genotype II was only 2.8% despite the fact that the strains were from four different countries (Senegal, Burkina Faso, Guinea-Bissau, and Ghana) and were isolated up to 65 years apart (1927 to 1992). Similarly, a clade within the east/central African genotype (Fig. 1) consisted of eight strains (CAR80, Ethiopia61a and -61b, Uganda64, Uganda72, Sudan40a and -40b, and Zaire58) from five countries (Central African Republic, Ethiopia, Uganda, Sudan, and Zaire) that were isolated over a period of 40 years. Again, nucleotide variation within this clade was only 2.4%. In contrast, nucleotide variation in west African genotype I (predominantly strains from Nigeria) was up to 6.8% for strains isolated over a 45-year period and was more than twice that observed for the two clades above. Also, four strains isolated in the Central African Republic over an 8-year period (1977 to 1985) and all within the east/central African genotype varied by up to 7.6% (Table 6).

High degree of amino acid sequence homology between strains of YF virus from Africa. In comparison to the extensive nucleotide variation, the deduced amino acid sequences for all 38 YF virus strains showed a high degree of sequence homology (91.9 to 100%). The variable amino acid positions did not show genotype-specific differences similar to those defined by the nucleotide analyses above. However, by comparison to a consensus YF sequence of all 38 strains used in this study, amino acid variations at eight amino acid positions segregated all the strains into two geographically distinct groups, west Africa and east/central Africa (Fig. 3). These two major groups were characterized by four amino acid substitutions (amino acids [aa] 96, 99, 100, and 104) towards the C terminus of the prM protein, one (aa 55) in the M protein, and three (aa 46, 62, and 87) in the N terminus of the E protein (Table 3).

View larger version (65K): [in this window] [in a new window]   FIG. 3.   Amino acid sequence alignment for the prM/E region of 37 YF viruses from Africa. Group-specific amino acid residues are highlighted for each gene. Dots indicate identity with the consensus sequence.

View larger version (41K): [in this window] [in a new window]   FIG. 4.   Amino acid sequence alignment for the structural protein region of four wild-type YF virus strains representing different genotypes.

Codon usage differs between genotypes of YF virus. Because of observed genotypic differences in transition/transversion ratios, we investigated codon usage by different genotypes of YF virus in Africa. We compared the number of times a particular codon was used to the expected value within and among YF strains by using the ² test. Expected values for the ² test were calculated assuming that all codons were used at the same frequency. Five YF strains, CAR77b, Nigeria70a, Angola71, Ghana27, and Kenya93, representing the five YF virus genotypes in Africa were selected for this study. Significant bias in codon usage was detected for four amino acids, Leu, Ile, Ala, and Lys (Table 7). However, codon usage for two amino acids (Lys and Ile) separated east/central African strains from west African strains, while codon usage for Leu separated Angola71 from all other strains (Table 7).

Rates of evolution of different genotypes are very similar. To investigate the nucleotide sequence variation in more detail, we determined the rates of evolution for three genotypes, west Africa I, west Africa II, and east/central Africa, and found that there was no statistically significant difference (4.58 × 10⁴ ± 7.36 × 10⁴, 2.344 × 10⁴ ± 1.35 × 10⁴, and 7.9 × 10⁵ ± 6.03 × 10⁴ nucleotides per site per year, respectively). We estimated rates of nucleotide substitutions using the times of isolation for the various YF virus strains. For each YF virus genotype, we selected several sister pair sequences isolated less than 7 years apart, with less than 10 nucleotide changes. Small nucleotide changes are better estimators for rates of evolution because multiple substitutions increase with the number of nucleotide changes. The mean for several sister pairs in each genotype was computed and used as the rate of evolution for each genotype.

Contribution of quasi-species to different genotypes. It is well recognized that RNA viruses are continually evolving and consist of quasi-species. To investigate the potential contribution of quasi-species to genetic variation observed within genotypes, an example of each of four genotypes was examined in detail. The 17D vaccine strain was included as a reference. A PCR product for each virus was cloned, and 20 clones of each PCR product were sequenced. The 20 clones were each compared to the consensus sequence obtained by direct sequencing of the PCR product, and nucleotide substitutions were identified. The vast majority of the clones were identical in sequence to the consensus sequence, giving confidence in the nucleotide sequence data generated (Table 8). The 17D vaccine and Nigeria69 viruses gave identical results, with 17 clones (85%) identical to the consensus sequence and 3 clones each with a single nucleotide difference. Guinea-Bissau65 was very similar, with only one clone different from the consensus sequence. Thus, the greater nucleotide variation within west Africa genotype I compared to west Africa genotype II could not be explained by quasispecies. The east African viruses Kenya93 and Ethiopia61a revealed greater nucleotide variation within the virus population, with both viruses containing 25 to 40% clones with up to three nucleotide differences from the consensus sequence. Thus, there was greater nucleotide variation within east African viruses compared to west African viruses.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The results of the present study support and extend the already established concept that wild-type strains of YF virus in Africa are genetically heterogeneous. We present evidence which suggests that wild-type strains of YF virus in Africa are genetically more diverse than originally reported by Lepiniec et al. (10), Chang et al. (2), and Wang et al. (19). These previous studies used a smaller panel of strains to define two major genetically distinct lineages of YF viruses in Africa, one in east/central Africa and the other in west Africa. The group in west Africa was further divided into two distinct subgroups (10), but until now it has been thought that only one genotype of YF virus circulated in east and central Africa. In the present study we have described five genotypes, two in west Africa and three in east and central Africa. Although we were able to demonstrate up to 25.8% nucleotide variation between strains, there was a maximum of 9.1% amino acid variation between strains, and there were few amino acid differences between most strains (Fig. 3).

Significantly, nucleotide variation within a given genotype varied greatly among genotypes. Members of west African genotype II and a clade within the east/central African genotype were highly homogeneous (2.8% nucleotide variation), whereas west African genotype I was more heterogeneous (up to 6.8% nucleotide variation). The strains in the former two clades are found in sylvatic transmission cycles in enzootic foci that exist year-round in the equatorial rain forest (enzootic zone) and are transmitted predominantly from monkey to monkey by Aedes africanus (17). Viruses within these two genotypes have been associated occasionally with epidemics (11) that, presumably, are associated with ecological conditions favoring human transmission. In contrast, there was up to 6.8% nucleotide variation in west Africa genotype I. Strains within this clade were predominantly from moist savannas (zone of emergence), dry savannas, and urban areas (epidemic zone), where transmission is seasonal, involving monkeys and Aedes furcifer, Aedes luteocephalus, and Aedes vittatus (17). Survival and continuation of epizootics are ensured by vertical transmission in the mosquitoes. The increased nucleotide variation within this genotype probably reflects adaptation to more diverse conditions in the savannah transmission cycles compared to the stability in the sylvatic environment. Interestingly, the ratio of transitions to transversions was different for the two west African genotypes, providing indirect evidence in support of the two genotypes, existing in different transmission cycles.

The viruses found in the clade including Sudan, Ethiopia, Uganda, Central African Republic, and Zaire have been associated with large epidemics separated by long periods with few human cases. It has been speculated that this phenomenon is due to introduction of viruses from other areas. The phylogenetic analysis supports the hypothesis of continual enzootic activity in these countries with occasional epidemics that are presumably associated with favorable ecological conditions for human transmission. The results also indicate that the recent outbreak in Kenya in 1993 involved viruses genetically related to those isolated in Uganda in 1948 rather than viruses found in the east/central African genotype. Since Angola71 is also an independent lineage, we speculate that multiple independently evolving lineages of YF virus exist in east/central Africa in sylvatic transmission cycles and these viruses only rarely come into contact with humans. In comparison, viruses within west African genotype I include strains often associated with epidemics, most notably from Nigeria. The frequent human epidemics appear to be associated with extensive genetic variation within this genotype.

Strain14 FA, isolated in Angola in 1971, was well differentiated from all the other strains identified to date at the nucleotide sequence level and was clearly a different lineage phylogenetically. However, the amino acid sequence of Angola71 was remarkably similar to those of other strains in east and central Africa, which suggests that the Angola71 strain probably evolved from a progenitor east/central African virus. Virological studies following the isolation of Angola71 showed that it was antigenically very similar to the Asibi (Ghana27) strain (12), suggesting a close relationship between west African YF virus strains and Angola71. This is consistent with the amino acid identity described in this paper. Moreover, the YF epidemic in Angola in 1971 was the first in Angola for almost 100 years and followed extensive YF activity in several west African countries (Cameroon, Equatorial Guinea, Ghana, Nigeria, and Togo) in 1970. Pinto and Filipe (12) speculated that the Angola outbreak was due either to YF virus moving south from west Africa or to a sylvatic strain. Our results clearly establish a strong genetic relationship between Angola71 and east/central Africa YF virus strains and suggest that the Angolan outbreak in 1971 was due to a sylvatic strain originating in east or central Africa. However, the extensive nucleotide differences between Angola71 and east/central African strains (15.7 to 18.5%) indicate that the progenitor to Angola71 diverged from central/east African strains many years ago. Unfortunately, we have been unable to locate any additional Angolan strains for inclusion in this study.

Earlier work by Deubel et al. (4) and Lepiniec et al. (10) described two distinct YF virus genotypes in west Africa. They observed that what we term west African genotype II was circulating in the area between western Ivory Coast, Mali, and Senegal, while what we term west African genotype I was circulating in the area from eastern Ivory Coast north to Burkina Faso and east to Nigeria and Cameroon (10). Our results confirm the same pattern of distribution of these genotypes, providing more support for the already reported genetic relationships within this region. However, one strain, Rendu, isolated in Senegal in 1953, was genetically distinct from other strains from Senegal in particular and other strains in west African genotype II in general. Studies by Wang et al. (20) showed that the same strain, Rendu, differed significantly from the other strains from Senegal antigenically and at the nucleotide level. They suggested that Rendu belonged to a different genotype. Our data confirm the observations of Wang et al. (20) and show that Rendu belongs to west African genotype I (the Nigeria and Ivory Coast genotype).

We evaluated nucleotide sequence variation among strains from Central Africa Republic, Senegal, Nigeria, and Uganda (Table 6). These countries were selected because we had data for four or more strains of YF virus from each country. Our data show that nucleotide variation of the strains in Nigeria and Central African Republic was below 9% (i.e., below the level of variation between genotypes). This indicated that the same YF virus genotype was circulating in Nigeria throughout the 45 years between isolation of the earliest and most recently examined strains. Likewise, one genotype was circulating in the Central African Republic. In Senegal, variation of up to 11.2% indicated that more than one genotype was present. However, Senegal53 (Rendu) is considered an imported strain (20). Thus, when Senegal53 is excluded from the analysis, nucleotide variation range was 0.1 to 2.8%, showing that one genotype circulates in Senegal (Table 6). Similarly, nucleotide variation of up to 10.7% in Uganda indicates more than one genotype. We have shown that the strains isolated in 1948 were genetically distinct from those isolated in 1964 and 1972, indicating that at least two YF virus genotypes circulate in Uganda. The above provides evidence that the geographical distribution of strains of YF virus is restricted in Africa and that it evolves very slowly.

Our results show significant bias in codon usage by different genotypes of YF virus in Africa. Differences in codon usage between east/central African genotypes and west African genotypes suggest variations in the enzootic-endemic cycles between these regions. Codon choice has been associated with AT or GC content (7), but the GC content of all the strains that we analyzed in this study was close to 50%, suggesting that it was not a significant determinant of the observed codon bias. Hooper and Berg (7) suggested that there is positive selection on codons that are translated more efficiently, either faster or more accurately. Since YF virus uses host cell macromolecular machinery for replication and protein synthesis, the observed codon usage bias may be attributed to the host. Differences in codon usage bias among the YF genotypes may then be attributed to regional differences among vector species and especially in regions where human YF epidemics are frequent. In such cases, the YF viruses adapt to peridomestic and urban vector mosquito species that may have varying codon usage strategies. We generally assume that the jungle cycle is more stable because it usually involves a single mosquito vector, A. africanus, and several species of Old World monkeys. However, genetic variations among regional populations of the jungle vector and vertebrate hosts may also result in the observed codon usage bias among the YF virus genotypes.

Overall, the genetic relationships between strains of YF virus in Africa were more complex than previously described. We have demonstrated that multiple genotypes of YF virus exist in Africa. The results are consistent with the hypothesis that YF virus is undergoing independent evolution in different areas in Africa. Although we have identified clear genotypic differences between strains of YF virus, the phenotypic differences of these viruses remain to be elucidated.