|Home | About | Journals | Submit | Contact Us | Français|
The family Asfarviridae contains only a single virus species, African swine fever virus (ASFV). ASFV is a viral agent with significant economic impact due to its devastating effects on populations of domesticated pigs during outbreaks but has not been reported to infect humans. We report here the discovery of novel viral sequences in human serum and sewage which are clearly related to the asfarvirus family but highly divergent from ASFV. Detection of these sequences suggests that greater genetic diversity may exist among asfarviruses than previously thought and raises the possibility that human infection by asfarviruses may occur.
The family Asfarviridae contains a single double-stranded DNA (dsDNA) virus called African swine fever virus (ASFV), which is thought to have evolved from an ancestral virus common to the nucleocytoplasmic large DNA viruses including poxviruses, iridoviruses, and phycodnaviruses (10, 11). ASFV infects ticks and swine but has not been reported to infect humans. ASFV infection of wild swine typically causes persistent infection with few symptoms (9, 17, 24, 25), but domesticated pigs can develop severe disease including acute hemorrhagic fever with nearly 100% mortality. As there is no vaccine and disease is contained by animal quarantine and slaughter, ASFV outbreaks can decimate pig populations and have significant economic impact. A 2007 outbreak in the former Soviet republic of Georgia resulted in the death and slaughter of over 80,000 pigs (20).
ASFV is endemic in sub-Saharan Africa but has also been introduced to countries in Europe, South America, and the Caribbean (26). Characterization of various ASFV isolates has led to the identification of 22 genotypes based on sequence variation in the portion of the B646L gene encoding the C terminus of the major capsid protein (2, 4, 14). Within this segment of B646L, approximately 14% of the nucleotide sites are variable among the ASFV isolates studied (2, 4). The ASFV genome has been completely sequenced for the Vero cell-adapted BA71V strain (29) and several wild isolates (5). Like many other large dsDNA viruses, ASFV contains open reading frames with homology to cellular genes involved in DNA replication, transcription, repair, and protein modification (29). ASFV also has an array of open reading frames with potential function in modulating host cell function or immune response (7).
We report here the discovery of novel viral sequences in human serum from the Middle East and in sewage from Spain that have clear similarity to ASFV genes. Of the 36 sequences identified, 29 did not have significant overlap or nucleotide identity with any of the other sequences. These 29 sequences are similar to 18 different ASFV genes, with some sequences matching to different regions within the same ASFV genes. Sequence and phylogenetic analyses indicate that the novel viral sequences are most closely related to the asfarvirus family but are highly divergent from known ASFV strains. We therefore hypothesize that these viral sequences are derived from at least one novel virus in the Asfarviridae family, which we refer to herein as ASFV-like virus (ASFLV).
We analyzed total nucleic acid extracted from human serum samples by 454 sequencing as an approach to identifying potential novel human pathogens. Serum samples were collected from patients with acute febrile illness (AFI) and from healthy volunteers (normal serum [NS]) in the Middle East between 2002 and 2005 and stripped of identifying information before analysis to protect patient confidentiality. Total nucleic acid was extracted from 199 AFI and 200 NS samples and reverse transcribed to enable detection of both RNA and DNA viruses. Each sample was then amplified by sequence-independent PCR using a primer that incorporates a 6-nucleotide barcode unique to that sample. Amplicons from multiple samples were pooled and subjected to 454 pyrosequencing. Additional details on serum sample processing and sequence data analysis are provided in the “Supplemental methods” section of the supplemental material. From one AFI sample and three NS samples, we identified six novel viral sequences with no significant nucleotide similarity to known viruses but whose translated sequences had detectable sequence identity to several ASFV proteins (Table (Table1),1), as determined using tBLASTx (1).
ASFV-related sequences were also found in sewage collected from an urban wastewater treatment plant in Barcelona, Spain. Briefly, viral particles from the sewage samples were concentrated and separated into fractions by cesium chloride equilibrium gradient centrifugation. Fractions with high concentrations of virus, as determined by quantitative PCR for human adenovirus, were treated first with DNase I to degrade nonviral DNA not protected by a viral capsid and then with a lysis buffer to disrupt viral capsids and release viral nucleic acids. Total nucleic acid was extracted from these fractions and was reverse transcribed and PCR amplified prior to 454 sequencing as described above. Additional details on the processing of sewage samples are provided in the “Supplemental methods” section of the supplemental material. Fifteen ASFV-related sequences were identified in five sewage fractions (SF).
We also identified an additional 15 ASFV-related sequences from our sequencing runs which could not be assigned to specific samples due to lack of a perfect sequence match with the barcode sequence. All of these unassignable sequences were identified in sequencing runs containing multiple samples including the sewage fractions which were found to contain ASFV-related sequences. We did not detect ASFV-related sequences in the other samples in these sequencing runs and therefore speculate that these unassignable sequences are derived from the sewage samples within these runs that contained ASFV-like sequences. However, as we cannot verify the specific sample source of these unassignable sequences, they are reported here as unassigned (UA) sequences.
A total of 36 novel ASFV-related sequences were identified, all of whose best-scoring tBLASTx matches were to the respective ASFV genes from various ASFV isolates (not shown). These sequences are listed in Table Table11 and were positionally mapped to the complete genome for the ASFV BA71V strain (Fig. (Fig.1)1) (29). These novel viral sequences were similar not only to ASFV genes such as the DNA polymerase and RNA polymerase genes, which are also conserved in other large dsDNA virus families (10), but also to multiple ASFV genes including EP364R and M448R, which do not have significant similarity to genes in other viral families and are therefore, to date, specific to ASFV.
The low amino acid identity between the novel viral sequences and the corresponding ASFV proteins (Table (Table1)1) suggested that these sequences may belong to a genetically distinct virus rather than to a new isolate of ASFV. We therefore performed multiple sequence alignments to compare the DNA polymerase-like sequences from our samples to the corresponding sequences from ASFV isolates. Sequences SF-3, SF-6, SF-8.7, and UA.10 all mapped to the same region of ASFV DNA polymerase (Fig. (Fig.1)1) and were nearly identical in nucleotide sequence where they overlapped (see Fig. S1 in the supplemental material). These sequences formed a 575-nucleotide consensus sequence, whose translated sequence had 51% amino acid identity with ASFV DNA polymerase when compared by tBLASTx (E value of 2e−26). Alignment of the regions of similarity between our novel DNA polymerase sequence and the corresponding ASFV sequences showed that the ASFV sequences were highly conserved with each other, but the sequence from our samples was divergent except for small blocks of amino acids which were conserved with ASFV sequences (Fig. (Fig.2A).2A). The same result was obtained using the NS-1.2 and NS-2.2 sequences, which are similar to ASFV RNA polymerase and topoisomerase II, respectively (see Fig. S2A and B in the supplemental material). We also performed phylogenetic analysis using sequence SF-8.3, with similarity to the EP364R gene, which appears to be ASFV specific. Sequences from ASFV isolates were again found to be highly conserved with each other, whereas the novel viral sequence was divergent (Fig. (Fig.2B).2B). The same result was obtained using the UA.15 sequence, with similarity to another ASFV-specific gene, M448R (see Fig. S2C in the supplemental material).
As viruses in multiple dsDNA viral families encode homologs of DNA polymerase, RNA polymerase, and topoisomerase II, we compared those sequences from our samples to corresponding ones from ASFV isolates as well as other dsDNA viruses from families including Poxviridae, Iridoviridae, Phycodnaviridae, Herpesviridae, and Ascoviridae. We also included in our analyses several sequences that had the most significant E values aside from those for ASFV strains when the novel DNA and RNA polymerase sequences were queried against the nucleotide database by tBLASTx. The novel topoisomerase II sequence did not have any hits with significant E values other than those for ASFV strains.
We found that our translated DNA polymerase consensus sequence formed a separate branch which was most closely related to the cluster containing the ASFV isolates (Fig. (Fig.3A).3A). The same result was observed for the RNA polymerase (Fig. (Fig.3B)3B) (NS-1.2) and topoisomerase II (Fig. (Fig.3C)3C) (NS-2.2) sequences. For all three proteins, grouping of the various dsDNA viruses was consistent with their classifications and with previously published studies using other sets of conserved genes (10, 18, 22).
As several additional novel sequences from our samples are similar to other ASFV genes that have also been reported to have homologs in other dsDNA virus families (10, 11), we performed similar phylogenetic analyses on these sequences. We found that 10 of the 11 sequences with similarity to ASFV gene B354L, C962R, D250R, EP1242L, B646L, or B962L were most closely related to, but distinct from, ASFV sequences (not shown), consistent with our results for the DNA polymerase, RNA polymerase, and topoisomerase II sequences. The exception was sequence SF-9.1, which is similar to ASFV C962R but did not group with sequences from any of the dsDNA families in the phylogenetic analysis. SF-9.1 is 99% identical in nucleotide sequence over a 238-nucleotide overlap with sequence UA.6 (not shown), which is also similar to C962R. Two of the three nucleotide changes in SF-9.1 are insertions or deletions resulting in frameshifts that significantly altered portions of the SF-9.1 translated sequence relative to that of UA.6. It is therefore likely that the results of phylogenetic analysis for SF-9.1 differed from those for UA.6 because the frameshifts decreased the overall similarity of the SF-9.1 translated sequence to ASFV.
The detection in this study of multiple viral sequences that are clearly related to, but phylogenetically distinct from, ASFV suggests that at least one additional member of the family Asfarviridae exists. The facts that these sequences were identified in multiple 454 sequencing runs over a period of approximately 7 months and that the first sequences identified, NS-1.2 and NS-2.2, were not detected in subsequent sequencing runs of different samples strongly argue against the possibility that the detection of ASFV-like sequences in these samples is due to cross-contamination.
Since the ASFV-like sequence fragments were identified from two types of samples from different geographic regions, it is possible that these sequences are derived from more than one virus in this family. In support of this possibility, the translated SF-10, UA.3, and UA.4 sequences aligned to the same region in C475L of ASFV but were significantly different from each other in nucleotide sequence. An alignment of the translated SF-10, UA.3, and UA.4 sequences with the corresponding segment of ASFV C475L showed only 28% identity between UA.3 and SF-10 and 71% identity between UA.3 and UA.4 in their respective regions of overlap (see Fig. S3 in the supplemental material). This suggests the possibility that our samples contained more than one asfarvirus-related virus, although this cannot be conclusively determined since the regions of overlap between these sequences are short.
Although ASFV is not known to infect humans even where the virus is endemic, identification of ASFV-like sequences in serum from multiple human patients suggests that human infection may occur. Further studies are under way to prospectively screen patient samples by PCR for the presence of ASFV-like sequences using primers to the sequences reported here in order to assess prevalence and geographic distribution. These studies, in combination with serological analyses, will be required to determine whether the ASFV-like virus is, in fact, a human virus and whether it is associated with human disease.
The finding of ASFV-like sequences in sewage from Spain indicates that they are not geographically limited to the Middle East, where the human patient specimens were obtained. Although it is unclear whether the source of the ASFV-like sequences in sewage is human or animal, this also suggests the virus may be fecally shed and that screening of stool in addition to serum may be informative. Identification of additional samples containing ASFV-like sequences will be important for the key future goals of obtaining more sequence from the viral genome, determining if the virus can be cultured, and establishing a small-animal model of infection.
The sequences obtained in this study have been assigned GenBank accession numbers FJ957903 to FJ957938.
We thank Guillermo Pimentel and Erica Dueger for supplying critical reagents, contributing to discussions of this work, and providing feedback on the manuscript.
This work was supported in part by National Institutes of Health grant U54 AI057160 to the Midwest Regional Center of Excellence for Biodefense and Emerging Infectious Diseases Research.
Published ahead of print on 7 October 2009.
†Supplemental material for this article may be found at http://jvi.asm.org/.