Ngari Virus Is a Bunyamwera Virus Reassortant That Can Be Associated with Large Outbreaks of Hemorrhagic Fever in Africa

Abstract

Two isolates of a virus of the genus Orthobunyavirus (family Bunyaviridae) were obtained from hemorrhagic fever cases during a large disease outbreak in East Africa in 1997 and 1998. Sequence analysis of regions of the three genomic RNA segments of the virus (provisionally referred to as Garissa virus) suggested that it was a genetic reassortant virus with S and L segments derived from Bunyamwera virus but an M segment from an unidentified virus of the genus Orthobunyavirus. While high genetic diversity (52%) was revealed by analysis of virus M segment nucleotide sequences obtained from 21 members of the genus Orthobunyavirus, the Garissa and Ngari virus M segments were almost identical. Surprisingly, the Ngari virus L and S segments showed high sequence identity with those of Bunyamwera virus, showing that Garissa virus is an isolate of Ngari virus, which in turn is a Bunyamwera virus reassortant. Ngari virus should be considered when investigating hemorrhagic fever outbreaks throughout sub-Saharan Africa.

Evolution of segmented negative-strand RNA viruses occurs by antigenic drift, antigenic shift (genetic reassortment), and genetic recombination (1, 4, 9, 18, 31, 37). However, the contribution of each of these mechanisms to the biogenesis of viruses within the family Bunyaviridae is largely unknown owing to the lack of genetic data for most members of this family. Viruses of the family Bunyaviridae have a genome composed of three predominantly negative-sense RNA segments designated S, M, and L that minimally encode the nucleocapsid (N), envelope glycoproteins (Gn and Gc), and RNA-dependent RNA polymerase (L), respectively (36). The family is one of the largest, containing more than 330 viruses, including several viruses of public health and agricultural importance such as Rift Valley fever (RVF) virus and Crimean-Congo hemorrhagic fever (HF) virus (12, 27). There is enormous potential for the emergence of both reassortant and recombinant bunyaviruses with altered pathogenic potential or host range, as many of the known members of this family use the same arthropod and vertebrate hosts.

Garissa virus isolates from an HF outbreak identified as a Bunyamwera virus reassortant.

A large HF outbreak occurred in 1997 to 1998 in Kenya, Tanzania, and Somalia that was thought to be predominantly associated with RVF virus infections (39). However, during the investigation of this outbreak, a previously unidentified member of the Orthobunyavirus genus (family Bunyaviridae) was isolated from two HF cases (one in Kenya and the other Somalia). In addition, evidence of acute infection with this virus was detected in 27% of the HF cases tested (6). Genetic analysis of fragments of the virus S, M, and L genome RNA segments revealed the virus S and L segments to be nearly identical to those of Bunyamwera virus, whereas the M segment was unlike that of any of the genetically characterized members of the Orthobunyavirus genus, suggesting that the virus (provisionally named Garissa virus) was the result of a reassortment event (6). Prior to the 1997 to 1998 outbreak, known diseases caused by viruses of this genus ranged from uncomplicated febrile illness to fatal encephalitis (16, 21, 27), leading to the suggestion that the unique nature of the M genomic segment may have contributed to the novel pathogenesis properties of Garissa virus.

Genetic reassortment between members of the family Bunyaviridae has been demonstrated to occur in coinfected tissue culture cells and arthropod hosts and also by analysis of naturally occurring virus isolates (3-6, 10, 17, 24, 31-35). While genetic recombination is considered to be relatively rare within the negative-strand RNA viruses (18), examples have been recently reported within the family Bunyaviridae (37). These findings raise the question of the origin of the Garissa virus M segment, as the earlier study suggesting that it was a reassortant virus was based on analysis of a relatively small portion of the segment (6). To confirm that the origin of the unique Garissa virus M RNA segment was by reassortment rather than recombination, the complete nucleotide sequence of the Garissa virus M segment was obtained through direct sequencing of overlapping reverse transcription (RT)-PCR products. RNA was extracted, and RT-PCR analysis was carried out essentially as described previously (6). The entire Garissa virus M segment sequence (GenBank accession no. AY593729) was found to differ significantly from that of Bunyamwera virus and that of previously genetically characterized members of the genus Orthobunyavirus (Table 1) (23, 29). These data confirm that the Garissa virus M segment does not represent a genetic recombination event but rather segment reassortment with an as yet to be identified virus of the Bunyamwera virus serogroup. The Garissa virus M RNA segment contains one long open reading frame predicted to encode the Gn and Gc envelope glycoproteins and the NSm nonstructural protein. No unique features that might relate to the association of Garissa virus with HF (i.e., novel glycosylation sites or protease recognition sequences) were noted. Comparison with other genetically characterized members of the genus Orthobunyavirus demonstrates that the Garissa virus M segment falls within the Bunyamwera virus serogroup (Table 1).

TABLE 1.

Amino acid and nucleotide sequence identity comparison for the entire M segment of the Garissa (Kenya) virus isolate and selected viruses of the Bunyamwera virus serogroup^a

Virus isolate	% Identity with virus
	Garissa (Kenya)	Cache Valley	Bunyamwera	Germiston
Garissa (Kenya)		75.0	64.1	54.7
Cache Valley	70.2		64.6	55.3
Bunyamwera	63.9	63.3		60.3
Germiston	60.2	58.7	63.3

^aThe Garissa (Kenya) virus M segment is unique across its entire length and is most closely related to the M segment of Cache Valley virus. Amino acid and nucleotide sequence identity values are shown above and below the diagonal, respectively. The RT-PCR primers used to generate the Garissa (Kenya) virus M segment sequence are available on request.

The noncoding regions on either end of the genomic segments are thought to contain promoter and encapsidation sequences recognized by the virus-encoded L polymerase (required for both virus RNA transcription and replication) and the N nucleocapsid proteins (2, 15, 28). Comparison of the Garissa and Bunyamwera virus M RNA segment noncoding regions showed a surprising amount of divergence, especially on the 5′ end (viral sense) of the segment. Although the M segment terminal 20 nucleotides were not determined for Garissa virus, 76 nucleotides of the noncoding region were and no sequence similarities were found within the corresponding region of the Bunyamwera virus M segment (data not shown). The fact that the Garissa virus M segment is efficiently transcribed and replicated by the L and N proteins of Bunyamwera virus suggests that the encapsidation and promoter sequence motifs must reside in the approximately 20 nucleotides at the highly conserved terminal regions or that considerable flexibility is allowed in these sequence signals. Furthermore, these data suggest that there are no sequence-specific signals within these 76 nucleotides, which begs the question of the function of this region of the genomic segment.

High genetic diversity among viruses of the genus Orthobunyavirus and genetic match of the Garissa virus M segment with that of Ngari virus.

The Orthobunyavirus genus is an extremely diverse collection of mosquito-borne viruses. With few exceptions, viruses of this genus are mainly identified by antigenic and serologic means, with genetic characterization limited to only a few genus members. In an attempt to identify the donor of the Garissa virus M segment, as well as to better characterize at the molecular level the viruses of the Orthobunyavirus genus, we amplified and sequenced an approximately 600-nucleotide region of the M segment of 21 viruses within the genus. The resulting sequences were aligned, and a phylogenetic tree was constructed by the maximum-likelihood method (Fig. 1). The tree was based on nucleotide sequence differences and agrees well with previously defined serologic groupings, with serogroups forming defined clusters (Fig. 1).

FIG. 1.

Phylogenetic relationships of Bunyamwera virus serogroup viruses. Virus RNA was extracted and purified, and RT-PCRs were carried out with primers M14C and M619R as described previously (6). In most cases, the RT-PCR product was sequenced directly; however, some products could not be sequenced directly because of mispriming. In those cases, the RT-PCR product was cloned into TOPO-TA (Invitrogen) and a minimum of three individual clones were sequenced with M13 forward and reverse primers (Invitrogen). Nucleotide sequences were aligned with PILEUP of the Wisconsin Package, version 10.2 (Accelerys, Inc.). Maximum-likelihood analysis of the 654-nucleotide aligned sequences of a region of the virus M segments was performed with PAUP4.0b10 (Sinauer Associates Inc., Sunderland, Mass.). The HKY85 substitution model was used and included empirical nucleotide frequencies, estimations of transition/transversion ratios, the proportion of invariable sites, and the gamma distribution shape parameter. The analysis took 2,148 h to run on a G4 Macintosh computer and resulted in a −Ln likelihood of 14848.26. Nucleotide sequences (accession numbers are in parentheses) from GenBank included those for Jatobal virus (AF312380), Oropouche virus (AF312381), Vinces virus (AF499012), Oriboca virus (AF499013), Cache Valley virus (AF186241, AF186242, AF186243, AF082576), Garissa virus (AF398345, AF398346), Bunyamwera virus (M11852), Germiston virus (M21951), Tahyna virus (AF23484, AF123485), Lumbo virus (AF229129), California encephalitis virus (AF123483), San Angelo virus (AF123486), La Crosse virus (U70205, U70206, U70207, U70208, U18979, U18980, D10370), snowshoe hare virus (K02539), Jerry Slough virus (AF123487), Jamestown Canyon virus (U88058), Inkoo virus (U88059, U88060), Jamestown Canyon-like virus (AF468197), South River virus (AF123488), Melao virus (U88057), Serra do Navio virus (AF123490), Keystone virus (AF123489), and Trivittatus virus (AF123491). The viruses analyzed in this study included the Anhembi virus (SpAr2984; GenBank accession no. AY593745), BeAr328208 virus (BeAr328208; AY593743), Birao virus (YMP-1; AY593748), Bozo virus (DakArB7343; AY593739), Fort Sherman virus (86-MSP18; AY593734), Iaco virus (BeAr314206; AY593746), Ilesha virus (RML2; AY593730), Lokern virus (FMS4332; AY593736), Macaua virus (BeAr306329; AY593742), Maguari virus (CBAAR426; AY593735), Mboke virus (DakArY357/6e; AY593731), Ngari virus (DakArD28542/4e; AY593747), Playas virus (75V3066; AY593733), Potosi virus (89-3380; AY593750), Shokwe virus (SAAr4042; AY593749), Sororoca virus (BeAr32149; AY593744), Taiassui virus (Ar671; AY593740), Tensaw virus (B479-490; AY593737), Tlacotalpan virus (61-D-240; AY593732), Wyeomyia virus (prototype; AY593741), and Xingu virus (BeH388464; AY593738).

Both genetic and serologic data suggest that there is considerable inter- and transcontinental movement of viruses. Within the Bunyamwera virus serogroup, for instance, there are viruses endemic to both the Americas and Africa. In fact, on comparison of entire M segment sequences, the closest relative to Ngari virus is Cache Valley virus (7), a virus endemic to North America. Also, the M segments of the African Ilesha, Mboke, and Shokwe viruses are closely related to that of the Brazilian Xingu virus (Fig. 1). This feature is not unique to the Bunyamwera virus serogroup, as similar examples of virus genetic mixing among Africa, Europe, and the Americas are seen within the California encephalitis virus serogroup. For instance, there are close genetic relationships among the Tahyna, Lumbo, and California encephalitis viruses, which are from Europe, Africa, and North America, respectively, and between the Jamestown Canyon and Inkoo viruses from North America and Europe (Fig. 1). These data suggest that transcontinental movement of these viruses has occurred with some frequency over the long evolutionary history of these viruses. All of the characterized members of the Orthobunyavirus genus are arthropod-borne viruses, with vertebrate species involved as amplifying hosts. The transcontinental movement of the viruses could involve migration of vertebrate hosts, particularly birds or mammals, and virus introductions could occur in a manner similar to that recently observed with West Nile virus, another arthropod-borne virus (8).

Unexpectedly, our analysis revealed several instances of viruses with nearly identical M segment sequences. The M segments of the Wyeomyia and Taiassui viruses, the Ilesha and Mboke viruses, and the Ngari and Garissa viruses are as close to each other genetically, as are different isolates of Cache Valley virus (Fig. 1). The finding of a high degree of identity between the M RNA segments of the Wyeomyia and Taiassui viruses was not unexpected, as these viruses are closely related members of the Wyeomyia virus complex of the Bunyamwera virus serogroup (20). Somewhat more surprising was the high genetic identity of the Ilesha and Mboke viruses. Genetic analysis of the remainder of the virus genome is necessary to determine if Mboke virus (originally isolated in 1970) should be considered an isolate of Ilesha virus (originally isolated in 1957) (20). The S segments of the Mboke (GenBank accession no. AY593727) and Ilesha (R. M. Elliott, personal communication) viruses also have a high degree of nucleotide sequence identity, consistent with the Mboke virus representing an isolate of Ilesha virus. As the prototype Ngari and Garissa viruses had nearly identical M segments, it appeared that we had identified the parental virus that provided the M segment in the reassortment event that generated the Garissa virus (S, Bunyamwera virus; M, Ngari virus; L, Bunyamwera virus).

Garissa virus isolates are Ngari virus, which is a reassortant Bunyamwera virus.

To provide a more complete analysis of Ngari virus, regions of the virus S and L genomic segments were amplified and sequenced by RT-PCR with primers BUNYA1 and BUNYA2 and primers M13CBUNL1C and BUNL605R as described previously (6). To our surprise, sequencing of an approximately 900-nucleotide region of the S segment and a 600-nucleotide region of the L segment of the prototype Ngari virus revealed that they too were nearly identical to those of the Garissa virus isolate and Bunyamwera virus (Table 2). Thus, Garissa virus is not a unique member of the Orthobunyavirus genus but rather an isolate of Ngari virus that also has the genotype S, Bunyamwera virus; M, Ngari virus; L, Bunyamwera virus. Ngari virus and Bunyamwera virus (and presumably the donor of the Ngari virus M segment) have similar geographic distributions across a broad region of sub-Saharan Africa (41; http://www.pasteur.fr/recherche/banques/CRORA/fixes/depart.htm), and both viruses have been isolated from the same species of Aedes mosquito (http://www.pasteur.fr/recherche/banques/CRORA/fixes/depart.htm) (Fig. 2).

TABLE 2.

Identity matrices for S, M, and L segments of Garissa, Ngari, and Bunyamwera virus isolates^a

Genome fragment and virus source	% Identity with virus
	Garissa (Kenya)	Garissa (Somalia)	Ngari	Bunyamwera
S segment
Garissa (Kenya)	100	99.8	99.6	93.6
Garissa (Somalia)		100	99.8	93.4
Ngari			100	93.6
Bunyamwera				100
M segment
Garissa (Kenya)	100	99.6	98.7	67.5
Garissa (Somalia)		100	99.1	67.9
Ngari			100	68.1
Bunyamwera				100
L segment
Garissa (Kenya)	100	99.8	98.5	97.4
Garissa (Somalia)		100	98.6	97.6
Ngari			100	98.3
Bunyamwera				100

^aGarissa virus is an isolate of Ngari virus. Ngari virus is a reassortant virus bearing the S and L segments from Bunyamwera virus and an M segment from an as yet unknown Orthobunyavirus. Percent identities over 524-, 542-, and 581-nucleotide fragments of the S, M, and L segments, respectively, are shown.

FIG. 2.

Geographic distribution of Ngari and Bunyamwera viruses. Ngari virus has been isolated from animals and mosquitoes across a broad region of sub-Saharan Africa, including Senegal, Burkina Faso, Central African Republic, Madagascar, Mauritania, Kenya, and Somalia (black or hatched areas). This distribution overlaps that of Bunyamwera virus, which has been isolated from animals and mosquitoes in Senegal, Central African Republic, Madagascar, Guinea, Côte d'Ivoire, Uganda, Nigeria, Kenya South Africa, and Cameroon (gray or hatched areas). The countries in which both Bunyamwera virus and Ngari virus have been isolated are indicated by hatch marks. Only countries in which actual virus isolations were made are indicated. Antibody surveys suggest that the range of Bunyamwera virus is much of sub-Saharan Africa.

Prior to the 1997 to 1998 outbreak in East Africa, Ngari virus had not been reported to be associated with hemorrhagic illness. This situation is reminiscent of the history of the other major cause of hemorrhagic illness during this outbreak, RVF virus (39). RVF virus was first identified in 1930 as the causative agent of febrile disease in humans and fatal disease in domestic ruminants (11). However, it was not until the 1975 and 1977 outbreaks in South Africa and Egypt, respectively, that the encephalitic and hemorrhagic forms of this disease in human patients were recognized (19, 22, 25, 26). Hemorrhagic and encephalitic diseases occur in only 1 to 3% of the cases, and it is likely that surveillance, as well as the shear magnitude of the epizootic (>500,000 human cases) in Egypt, is what uncovered the full range of RVF disease manifestations. The suggested scope of the Ngari virus epidemic in East Africa has brought the hemorrhagic form of Ngari virus disease to light. Whether Ngari virus is unique among the viruses of the Orthobunyavirus genus with respect to its association with HF remains to be determined.

The important role that genetic reassortment plays in the evolution and emergence of viruses with altered disease potential or host range is well documented for a number of segmented RNA viruses such as influenza virus and rotavirus (13, 40). However, its importance within the family Bunyaviridae is largely unknown because of the lack of genetic characterization of most of the more than 330 members of this family. In this specific instance, Ngari virus reacts with antiserum raised against Bunyamwera virus in a fluorescent-antibody test (data not shown) and without more detailed serologic or genetic characterization, isolates of such a reassortant virus could easily be misidentified. Recent data for bunyaviruses, such as the hantaviruses and RVF viruses, suggest that naturally occurring segment reassortment between different virus strains may not be uncommon (17, 24, 35). In addition, evidence is accumulating for RNA segment reassortment between different viruses within the family Bunyaviridae. For instance, Jatobal virus has been genetically shown to be a reassortant Oropouche virus (Simbu virus serogroup, genus Orthobunyavirus) (34), and similar to Ngari virus, the S and L segments are from Oropouche virus and the M is segment is from an unknown member of the Simbu virus serogroup.

While the majority of the viruses in the family Bunyaviridae have not been characterized genetically, most have been extensively compared by various antigenic and serologic tests (20). Neutralization and hemagglutination inhibition assays are thought to predominantly reflect the properties of the virus glycoproteins (M segment encoded), whereas complement fixation assays predominantly reflect the properties of the virus nucleocapsid proteins (S segment encoded). On the basis of these assays, several discrepancies exist in the grouping of different viruses of the family Bunyaviridae that could most easily be explained by genetic reassortment of the virus RNA segments. This includes examples within the genus Orthobunyavirus (group C, Gamboa and Patois groups) and the genus Phlebovirus (14, 30, 38). The combination of such serologic data and more recent molecular analysis strongly suggests that the role of reassortment in the evolution and emergence of bunyaviruses with unique biological properties and disease potential is underestimated.