Acute Diarrhea in West African Children: Diverse Enteric Viruses and a Novel Parvovirus Genus

Abstract

Parvoviruses cause a variety of mild to severe symptoms or asymptomatic infections in humans and animals. During a viral metagenomic analysis of feces from children with acute diarrhea in Burkina Faso, we identified in decreasing prevalence nucleic acids from anelloviruses, dependoviruses, sapoviruses, enteroviruses, bocaviruses, noroviruses, adenoviruses, parechoviruses, rotaviruses, cosavirus, astroviruses, and hepatitis B virus. Sequences from a highly divergent parvovirus, provisionally called bufavirus, were also detected whose NS1 and VP1 proteins showed <39% and <31% identities to those of previously known parvoviruses. Four percent of the fecal samples were PCR positive for this new parvovirus, including a related bufavirus species showing only 72% identity in VP1. The high degree of genetic divergence of these related genomes from those of other parvoviruses indicates the presence of a proposed new Parvoviridae genus containing at least two species. Studies of the tropism and pathogenicity of these novel parvoviruses will be facilitated by the availability of their genome sequences.

INTRODUCTION

Metagenomic analyses of human fecal samples have shown frequent coinfections with known enteric pathogens (26, 70) and the presence of numerous novel viral species, particularly members of the Picornaviridae (20, 31, 33, 37, 40, 51), Astroviridae (26, 27, 41) Parvoviridae (39, 45), and Circoviridae (49, 58) families of small enteric viruses. Metagenomics studies of animal feces have similarly shown frequent coinfections and yielded highly divergent viral species (10, 16, 25, 30, 43, 44, 50, 52, 53, 57, 60, 61, 64).

Parvoviruses are small nonenveloped icosahedral viruses with linear single-stranded DNA genomes of 4.5 to 5.5 kb, infect a variety of mammals, and are associated with a wide spectrum of acute and chronic diseases (14, 32, 55). The family Parvoviridae is divided into two subfamilies based on their host range, Parvovirinae and Densovirinae, which infect vertebrate and nonvertebrate animals, respectively (46). Under the International Committee on Taxonomy of Viruses (ICTV), the mammal- and bird-infecting family Parvovirinae is currently divided into five genera, Bocavirus, Erythrovirus, Dependovirus, Amdovirus, and Parvovirus (46). Other, still unclassified, parvoviruses have also been reported (46). At least four different groups of parvoviruses have been reported to infect human (14). Adeno-associated viruses (AAV) in the Dependovirus genus, initially isolated in 1967 (9), were the first parvoviruses identified in humans and are not known to cause any symptoms. AAV are dependent on coinfection with a helper adenovirus or herpesvirus for efficient replication. The second identified human parvovirus, B19V, was identified in 1975 (22) and sequenced in 1984 (23). B19V consists of three genotypes and is the prototype of the Erythrovirus genus, which includes related viruses infecting nonhuman primates and chipmunks (46). B19V infection in utero can cause fetal hydrops and developmental abnormalities (4, 15), while infection of children typically leads to the minor rash erythema infectiosum (5, 6, 59). In children and adults with sickle cell anemia, B19V can cause arrest of erythropoiesis and chronic anemia in AIDS patients (28). In 2005, a human bocavirus was characterized in respiratory secretions of Swedish children (2) and has been associated with lower respiratory tract infections worldwide, often in combination with other viral infections (1, 21, 35, 38, 62). Related parvoviruses named human bocaviruses 2 to 4 have also been detected in the feces of children with acute flaccid paralysis (39, 45) and diarrhea (7). Members of the Bocavirus genus also infect bats (GenBank accession no. JQ814850), cows (17), dogs (42, 63), pigs (11, 19, 24, 48, 64), chimpanzees (67), and gorillas (43). Human parvovirus 4 (PARV4) was initially described in 2007 in the plasma of a febrile injection drug user (36). Because of the high prevalence of anti-PARV4 antibodies in hemophiliacs using plasma pool-derived coagulation factors and in hepatitis C virus (HCV)- and HIV-positive injection drug users, the transmission of PARV4 in developed countries is thought to be mainly parenteral (29, 54, 66, 72, 73). The detection of PARV4 DNA in the blood of sub-Saharan African children indicates that other routes of PARV4 transmission may exist (56, 68). Related viruses have been detected in chimpanzees (67), baboons (67), bats (16), sheep (69), pigs (47), and cows (47), and a Parvoviridae genus named Partetravirus has been proposed (69). PARV4 has been reported in patients with different symptoms, but the full extent of its pathogenicity remains uncertain (8, 18, 36, 65).

Here we report on a viral metagenomic analysis of feces of children from Burkina Faso with acute diarrhea (12) allowing the characterization of their enteric viruses, including the genetic characterization of novel parvoviruses from a previously unrecognized genus (12, 13).

MATERIALS AND METHODS

Biological samples.

A prior study analyzing feces from children <5 years of age with acute diarrhea reported that 34% of them were positive for rotavirus by an immunochromatographic assay (SD Bioline Rota/Adeno; Standard Diagnostics, Inc., Kyonggi-Do, South Korea) (12, 13). Samples were collected between November 2008 and February 2010 from the capital city of Ouagadougou in Burkina Faso (12). Rotavirus antigen-negative samples were used in this study approved by the University of California at San Francisco committee on human research.

Viral metagenomics.

Viral particles were first enriched by filtration and nuclease treatment to digest non-particle-protected nucleic acids (3). Nonspecific RNA and DNA amplification was then performed by random reverse transcription (RT)-PCR using primers with randomized 3′ ends (70). Forty-nine different primers with distinct 5′ regions were used to molecularly label 49 random RT-PCRs. Amplicons were then pooled, and 454 libraries were generated and pyrosequenced using the 454 Titanium FLX+ sequencer (70). Each of 49 molecular tags was used on two fecal samples, for a total of 98 diarrhea samples. Because fecal samples were analyzed in pairs, the viruses identified are reported per sample pair rather than for individual samples. The 454 singlet reads and assembled contigs of >100 bp were compared to the GenBank protein databases using BLASTx.

Bufavirus PCR.

Primers BF.F1 (5′-TCAACAATCACTCAGGCAAATGG-3′) and BR.R1 (5′-AGTTTGCCTGGATGTTCTTTGA-3′) were used for the first round of PCR, and primers BF.F2 (5′-CTAACACTGGTACTTGCTATGGAC-3′) and BF.R2 (5′-TTCTCTGGTGATGATTCTTTTGTC-3′) were used for the nested PCR round, resulting in an expected amplicon of ∼440 bp. The PCR protocol for the first and second rounds was as follows: denaturation at 95°C for 5 min and 35 cycles of 95°C for 30s, 48°C for 30s, and 72°C for 1 min.

RESULTS

Viral particles in fecal samples from children with acute diarrhea were first enriched by filtration. Non-particle-protected nucleic acids were digested with nucleases to reduce the nonviral background, and the remaining nucleic acids were extracted, amplified by random RT-PCR, tagged, and then pyrosequenced, generating over half a million reads (see Materials and Methods). Using a BLASTx E score cutoff of 10⁻⁵, we identified 24,626 reads to known human viruses whose translated sequences were >90% identical to those of viral sequences already in GenBank. The distributions of these viral hits among the 49 pairs of samples analyzed are shown in Table 1. Because diarrhea samples were pooled in pairs prior to random RT-PCR, viruses detected could only be assigned to pairs of samples. The most common human viral sequences amplified were those of members of the Anelloviridae family (9,093 reads), sapoviruses (6,586 reads), dependovirus (aka AAV) (2,798 reads), bocaviruses (2,773 reads), enteroviruses (2,349 reads), noroviruses (682 reads), human astroviruses (194 reads), parechovirus (55 reads), rotaviruses (41 reads), hepatitis B virus (HBV; 30 reads), adenoviruses (23 reads), cosaviruses (5 reads), and human astrovirus HMO-A/VA2 (2 reads) (Table 1). The most prevalent enteric viruses detected were Anelloviridae (found in 34 sample pairs), dependoviruses (12 sample pairs), sapoviruses (10 sample pairs), enteroviruses (9 sample pairs), bocaviruses (8 sample pairs), noroviruses (7 sample pairs), adenoviruses (5 sample pairs), parechoviruses (3 sample pairs), rotaviruses (2 sample pairs), cosavirus (2 sample pairs), astroviruses (1 sample pair), human astrovirus HMO-A/VA2 (1 sample pair), and HBV (1 sample pair) (Table 1). From 0 to 6 eukaryotic viruses (average, 1.9) were detected in fecal sample pairs. When only pathogens known to cause diarrhea were included (counting sapovirus, enterovirus, norovirus, human astrovirus, parechovirus, rotavirus, and adenovirus but excluding Anelloviridae, dependoviruses, bocaviruses, cosaviruses, human astrovirus HMO-A/VA2, and HBV) from 0 to 3 pathogens (average, 0.77) were detected per sample pair at this depth of sequencing (Table 1).

Table 1.

Distribution of sequence reads to different viral species and total of human viruses in 49 sample pairs

Sample paira	No. of human viruses detected	No. of known human pathogens	Anelloviridae (n = 34)	Dependovirus (n = 12)	Sapovirus (n = 10)	Enterovirus (n = 9)	Bocavirus (n = 8)	Norovirus (n = 7)	Adenovirus (n = 5)	Parechovirus (n = 3)	Rotavirus (n = 2)	Astrovirus (n = 2)	Cosavirus (n = 2)	HBV (n = 1)
27	6	3	220	1,025		11	1		1	19
37	4	1	794	1					1					30
47	4	2	6	1		3					38
96	4	2	2					1			3		1
20	3	1	210					1					15
43	3	2	128			1,639						194b
28	3	1	32	1	38
30	3	2	31			7		4
22	3	1	27	1		663
7	3	2	17		1					35
13	3	0	16	1,309								2c
49	3	1	1		11		6
26	3	3			1	1				1
41	3	2					24	16	1
44	2	1	2,379		1
11	2	1	1,446						1
4	2	1	1,193						19
38	2	0	134	2
15	2	1	112		1
10	2	0	93	1
3	2	1	21			1
33	2	1	17		1
19	2	1	13			13
42	2	2	13					647
17	2	0	6				95
35	2	0	5	1
34	2	1	3		1
32	2	2				11		6
8	2	1			43		1,227
23	2	0		2			1,375
25	2	0		2			9
31	1	0	1,799
18	1	0	182
29	1	0	98
21	1	0	35
12	1	0	24
1	1	0	16
2	1	0	12
5	1	0	4
6	1	0	4
9	1	1			6,488
86	1	0					36
45	1	1						7
40	1	0		452
14	0	0
16	0	0
24	0	0
39	0	0
48	0	0
Total no. of reads	95	38	9,093	2,798	6,586	2,349	2,773	682	23	55	41	196	16	30

^aThe sample pair underlined contained the initial bufavirus sequence reads. The samples shaded in gray were positive for bufavirus by nested PCR.

^bHAstV.

^cHMO-A/VA2.

Ten sequences out of a total of 1,057 contigs plus singlets from one specimen pair (Table 1, pair 96) showed significant similarities (E values, 1 × 10⁻² to 4 × 10⁻²⁸) to parvovirus proteins. Gaps between pyrosequences were filled by PCR, and 5′ and 3′ regions were amplified by rapid amplification of cDNA ends. The nearly complete genome (4,921 bp) of bufavirus 1-BF96 (for Burkina Faso parvovirus species 1 strain 96), including a partial 5′ untranslated region (UTR) (18 bp), the complete NS1 sequence (671 aa), the complete VP1 (707 aa) and VP2 (569 aa) sequences, and a partial 3′ UTR (400 bp), was acquired (GenBank accession no. JQ918261). The genome contained two major open reading frames (ORFs), with the left and right ORFs encoding NS1 and VP1, respectively (Fig. 1A). A middle ORF (130 aa) was also found that did not show any similarity to other parvovirus genomes by BLAST (Fig. 1A). Sequence alignment showed a high degree of sequence divergence, with <50% overall nucleotide identity to other parvoviruses (Fig. 1B). The putative bufavirus NS1 start codon was located in a strong Kozak sequence, CACCATGG. The ATP- or GTP-binding Walker loop motif (GXXXXGK[T/S]) was found in NS1 (⁴⁰²GPASTGKS⁴⁰⁹) (71). In addition, NS1 also contained two conserved replication initiator motifs, GLHIHVLVC and IANYFLIKKP (where conserved amino acids are in boldface type) (34). A sequence identity matrix was then generated (Table 2). NS1 showed less than 39% identity with other parvoviral NS1 sequences, including its closest relatives in the Parvovirus genus.

Fig 1.

Bufavirus genome and phylogeny. (A) Organization of the bufavirus genome. The PLA₂ similarity region, including the calcium-binding region and catalytic residues, is shown. Theoretical splicing for expression of VP1 is shown. nt, nucleotides. (B) Pairwise sliding window of percent nucleotide identity of bufavirus aligned with representatives of other parvovirus genera and bufavirus 2. (C) Phylogenetic analyses of the NS1, VP1, and VP2 proteins of bufavirus and related parvoviruses. Genetic distances were calculated by Kimura's two-parameter method (PHYLIP), and a phylogenetic tree with 100 bootstrap resampling of the alignment data sets was generated using the neighbor-joining method. Each scale indicates the number of amino acid substitutions per position. Proposed and official Parvoviridae genera are labeled. The names and accession numbers of the taxa used are listed in Table S1 in the supplemental material.

Table 2.

Pairwise percent amino acid sequence identities among NS1, VP1, and VP2 regions of bufavirus 1-BF96 and the prototypes of different parvovirus genera

Protein and virus	% Amino acid sequence identity
	Bufavirus	Bocavirus	Erythrovirus	Dependovirus	Amdovirus	Parvovirus
NS1
Bufavirus	100
Bocavirus	14	100
Erythrovirus	13	15	100
Dependovirus	18	15	20	100
Amdovirus	18	15	11	14	100
Parvovirus	38	16	13	18	18	100
VP1
Bufavirus	100
Bocavirus	17	100
Erythrovirus	10	13	100
Dependovirus	15	24	19	100
Amdovirus	26	10	10	10	100
Parvovirus	31	19	10	17	28	100
VP2
Bufavirus	100
Bocavirus	14	100
Erythrovirus	10	15	100
Dependovirus	14	21	20	100
Amdovirus	32	13	10	13	100
Parvovirus	29	14	12	14	32	100

Two potential splice sites were detected, a potential donor site (AG↓GT) at nucleotide 2027 and an acceptor site (AG↓G) at nucleotide 2396. The putative VP1 sequence started at the first ATG of the middle ORF at nucleotide 2003 upstream of a splice donor site at nucleotide 2027 (Fig. 1A). The phospholipase A₂ (PLA₂) motif (Fig. 1A), with its highly conserved calcium-binding site (YLGPF), was found in the main ORF of VP1. The phospholipase catalytic residues (HD and D) were present at amino acid positions 40 to 41 and 62 (Fig. 1A). Bufavirus VP1 showed <31% amino acid identity to other genera of the family Parvovirinae, including its closest relative in the Parvovirus genus (Table 2). The N terminus of the bufavirus VP2 protein contained a glycine-rich sequence (GGGGGGGGSGVG) also present in other parvoviral VP2 proteins. The VP2 protein was most closely related to those of amdoviruses (Table 2). Amdoviruses lack the PLA₂ motif found in the N terminus of VP1 in bufavirus and other parvoviruses. Three tandem repeats of the sequence TAGTTGATAAGT were seen in the 400-base-long 3′ UTR. Two other repeats, TAGTTTATAAGT and TAGTTTATAAAT (where mutated bases are in boldface italics), were also found in the 3′ UTR region. Such repeated sequences were not reported in other parvoviruses.

The three major proteins of bufavirus were aligned with those of other parvoviruses and phylogenetically analyzed (Fig. 1C). Bufavirus clustered with the members of the Parvovirus genus in NS1 but was basal to both the Parvovirus and Amdovirus genera in its capsid region. The VP2 protein identity among the five ICTV-approved genera of the subfamily Parvovirinae ranged from 10% to 28%. Bufavirus showed a range of 10% to 31% identity to the VP2 proteins of these five genera and may therefore qualify as the reference genome for a novel genus.

To investigate the prevalence of this novel parvovirus, nested PCR primers targeting the NS1 region were designed. Bufavirus DNA was detected in an additional 3 out of 98 fecal samples from children with acute gastroenteritis in Burkina Faso, with a total frequency of virus detection of 4% (4/98). PCR amplicons were directly sequenced to confirm their identification. The complete protein-coding regions of these three related viral genomes were then determined. The coding regions of two strains, BF7 (GenBank accession no. JX027295) and BF86 (GenBank accession no. JX027296), showed a high level of nucleotide identity (99%) to prototype bufavirus 1-BF96, but the coding regions of the bufavirus BF39 genome (GenBank accession no. JX027297) showed an overall nucleotide identity of only 87%. The VP2 proteins of the three closely related strains BF96, BF7, and BF86 showed only 72% amino acid identity to that of strain BF39, indicating that it might represent a second species in the proposed genus. The NS1 of BF39 was >95% identical to those of the other bufaviruses. We therefore provisionally named this genome bufavirus 2-BF39 to reflect the presence of a second tentative species highly divergent in its VP2 protein sequence. The high degree of identity between the two species in their NS genes relative to their more divergent capsid genes may indicate that the VP region was acquired by recombination from a still uncharacterized viral genome or that the structural region diverged at a much higher rate than the nonstructural region. An alignment of NS1 and VP1 from both bufavirus species with the most closely related parvoviruses is shown in Fig. S1 in the supplemental material.

In order to investigate the wider distribution of bufavirus, we tested fecal specimens collected from children with diarrhea in Chile (n = 100) and with nonpolio acute flaccid paralysis in Tunisia (n = 63) by nPCR. One bufavirus 1 was amplified from the feces of one Tunisian child, indicating that these viruses are not geographically restricted.

DISCUSSION

A metagenomic analysis showed the presence of a diverse enteric virome in diarrheic children <5 years old in Burkina Faso. A prior study of this population showed 34% of the diarrhea cases to involve rotaviruses (12). The detection of two samples containing rotavirus sequences despite the exclusion of rotavirus-positive samples likely reflects viral loads below the assay's antigen detection level. The most common infections detected were of anelloviruses and dependoviruses, neither of which are recognized pathogens, likely reflecting common and/or chronic commensal intestinal infections in these children. The hierarchy of the next most common viruses detected was sapoviruses > enteroviruses > bocaviruses > noroviruses > adenoviruses > parechoviruses > astroviruses > cosaviruses, most of which have been associated with diarrhea, while a few of the more recently characterized viruses are of unknown pathogenicity (bocaviruses, astrovirus HMO-A/VA2, cosaviruses). The detection of blood-borne HBV DNA sequences may be the result of bleeding into the intestinal tract. The metagenomic detection of bufavirus sequences in only one sample while three more samples were positive by nested PCR reflects the lower sensitivity of pyrosequencing at the depth used here. The rate of infection with other viruses measured here by metagenomics is therefore likely also an underestimate of their actual prevalence in feces.

A potential new genus in the Parvoviridae family was genetically characterized, and a PCR survey showed a prevalence of 4% among the rotavirus antigen-negative cases of childhood diarrhea. The genetic characterization of a second Bufavirus species with a highly divergent capsid gene indicates that wider geographic sampling for related viruses will likely reveal other related species. The genetic diversity within this proposed genus may also indicate a range of phenotypes upon their hosts.

Wider geographic sampling of human and animal fecal samples will provide a better description of the genetic diversity of this proposed Parvoviridae genus. Serological assays will help determine whether bufaviruses infect humans or are simply passing through the gut from a dietary source. Case-control studies comparing the presence and viral loads of bufavirus DNA in feces or the prevalence of anti-bufavirus IgM will also help determine whether members of this viral clade are associated with diarrhea or other symptoms.