Improved Molecular Typing Assay for Rhinovirus Species A, B, and C

Abstract

Human rhinoviruses (RVs), comprising three species (A, B, and C) of the genus Enterovirus, are responsible for the majority of upper respiratory tract infections and are associated with severe lower respiratory tract illnesses such as pneumonia and asthma exacerbations. High genetic diversity and continuous identification of new types necessitate regular updating of the diagnostic assays for the accurate and comprehensive detection of circulating RVs. Methods for molecular typing based on phylogenetic comparisons of a variable fragment in the 5′ untranslated region were improved to increase assay sensitivity and to eliminate nonspecific amplification of human sequences, which are observed occasionally in clinical samples. A modified set of primers based on new sequence information and improved buffers and enzymes for seminested PCR assays provided higher specificity and sensitivity for virus detection. In addition, new diagnostic primers were designed for unequivocal species and type assignments for RV-C isolates, based on phylogenetic analysis of partial VP4/VP2 coding sequences. The improved assay was evaluated by typing RVs in >3,800 clinical samples. RVs were successfully detected and typed in 99% of the samples that were RV positive in multiplex diagnostic assays.

INTRODUCTION

Rhinoviruses (RVs) are members of the genus Enterovirus of the family Picornaviridae and are currently classified into three species (A, B, and C). More than 160 RV types have been identified to date, including 99 classic serotypes of RV-A and RV-B (1) and 6, 5, and 51 novel genotypes of RV-A, RV-B, and RV-C, respectively, classified genetically in the absence of serological cross-neutralization data (2–4). RVs are responsible for the majority of upper respiratory tract infections (common colds) and also can cause more severe illnesses of the lower respiratory tract, such as pneumonia and asthma exacerbations. The species affects the virulence, as RV-A and RV-C cause more severe illnesses in infants and are more likely to cause exacerbations of childhood asthma (5, 6).

RV is a positive-sense single-stranded RNA virus with a genome of 7.1 to 7.2 kb, consisting of a single gene that codes for a long polyprotein of about 2,100 amino acids. The translated polyprotein is cleaved by viral proteases 2A and 3C to yield 11 mature proteins. The viral capsid is composed of 60 copies each of the VP4, VP2, VP3, and VP1 proteins. There are two untranslated regions (UTRs), i.e., the 5′ UTR, which is typically 610 to 630 bases long and precedes the open reading frame, and the 3′ UTR, which consists of 40 to 45 bases upstream of the poly(A) tract. The 5′ and 3′ UTRs contain a number of structural and sequence elements necessary for viral genome translation and replication. Sequence analyses of RVs reveal high genetic diversity, especially in the capsid-coding region, and evidence for recombination events mapped mainly in the 5′ UTR and the protease 2A gene (7, 8).

Molecular methods for viral genome detection and sequencing directly in clinical samples are essential for RV diagnosis. The typical workflow for PCR-based RV diagnosis includes RNA extraction from clinical samples (nasal lavage fluid or sputum specimens) followed by reverse transcription (RT) and either single (9, 10) or multiplex (11–13) PCRs and partial sequencing to determine virus type. RV-specific diagnostic RT-PCR assays can target the 5′ UTR, structural genes (e.g., VP4/VP2 [1A/1B] and VP1 [1D]), or both. Primers that anneal to highly conserved motifs in the 5′ UTR provide the most sensitive assays for detection of both prototypic strains (14–16) and novel RV variants (9, 10, 17, 18). The majority of the published primers (and probes) target these sequence motifs, with minor differences in length, location, or assay type (e.g., conventional versus quantitative or one-step versus nested PCR).

One disadvantage associated with the use of RV 5′-UTR primers is nonspecific amplification of human genomic DNA (chromosome 6) or RNA (large regulator noncoding RNA B2) sequences, yielding a nonspecific 424-bp product that is similar in size to the virus-specific amplicon (390 bp) (19–21). The nonspecific results are typically found in a small subset of nasal lavage fluid samples but are more common when clinical samples contain high concentrations of human RNA or contaminating genomic DNA (e.g., cases of intense cellular inflammation in airways or nasal brushing samples). In addition, the majority of 5′-UTR sequences of species C (RV-Ca clade) are genetically similar to those of some RV-A types due to putative historical interspecies recombination (7, 8). Structural genes (e.g., VP4/VP2 or VP1) are less suitable for universal diagnostic primers due to their greater sequence variability, but phylogenetic analysis of the capsid-coding regions clearly segregates RV-A, RV-B, and RV-C species and types, and these characteristics make them preferred for unequivocal RV species and type assignment (22–25).

The major goal of this study was to design a RV diagnostic protocol that is sensitive, specific, able to assign viral species and type, efficient, and cost-effective. To accomplish this goal, we modified the RV 5′-UTR diagnostic assay to improve specificity and sensitivity and coupled it with a high-throughput sensitive assay targeting the VP4/VP2 region for unequivocal confirmation of species and type assignments for RV-C isolates. In addition, we updated and revised the reference sequence database to enable more accurate typing of clinical isolates of RV-A, RV-B, and RV-C species.

MATERIALS AND METHODS

Virus strains and clinical samples.

RV-A16 and RV-B14 are laboratory strains that were grown in HeLa cell suspension and purified as described previously (26, 27). RV-C15 is a clinical isolate that was produced by transfection of HeLa or WisL cells with viral RNA transcribed in vitro, as described previously (28). Clinical samples for validation of the typing assay were obtained from children 4 to 12 years of age who were participating in a study of RV epidemiology (29). Samples were collected weekly for a total of 5 consecutive weeks during peak RV seasons (spring and fall) in 2007 to 2009. Additional nasal samples were also obtained from children in a clinical study of the role of respiratory viruses in sinusitis. Both studies were approved by the University of Wisconsin Human Subjects Committee, and informed consent was obtained from study participants.

RNA extraction and reverse transcription.

Total RNA was extracted from 350 μl of nasal secretions or laboratory virus preparations using TRIzol LS reagent (Life Technologies), as described previously (9). RNA (20 μl) was reverse transcribed using a high-capacity cDNA reverse transcription kit with RNase inhibitor (Life Technologies). Detailed protocols are provided in the supplemental material.

PCR and sequencing.

The two-step PCR protocol and diagnostic primers for amplification of the 5′ UTR of RV were described previously (6, 9). Modified primers targeting the 5′ UTR and VP4/VP2 are listed in Table 1. The first PCR amplification was performed in a 26-μl volume using 2.5 μl of cDNA template and Platinum PCR Supermix High Fidelity (Life Technologies); 5 μl of the first PCR product was then reamplified in a second PCR, in a 52-μl reaction volume, with either the same or seminested primers (as indicated), using GoTaq Green master mix (Promega). Reactions were run in a programmable DNA Engine thermocycler (Bio-Rad). The PCR products were visualized by electrophoresis on a 1.5% agarose gel, treated with exonuclease I and shrimp alkaline phosphatase (ExoSAP-IT mixture; Affymetrix) to remove residual primers and deoxynucleoside triphosphate (dNTPs), and sequenced using reverse primer P3, 5′UTR-rev, 5′UTR-revseq, 5′UTR-f, or VP2-C252-5r. Sequencing was carried out using the BigDye Terminator v. 2.1 cycle sequencing kit (Life Technologies), at the DNA Sequencing Facility of the University of Wisconsin-Madison Biotechnology Center (Madison, WI). If the quality of sequences obtained by direct sequencing of PCR products was not sufficient for type assignment (e.g., due to the presence of multiple RV types in one sample), then the amplicons were cloned into the pGEM-T Easy vector (Promega). The nucleotide sequences of 6 to 10 positive clones for each isolate were determined using plasmid-specific T7 (forward) and M13 (reverse) primers (Promega). RV isolates with putative recombinant sequences were retested using type-specific primers (Table 1) designed to specifically amplify only the recombinant sequences and not the parental sequences in dual-RV clinical samples, using the two-step PCR procedure and reaction conditions described above. Clinical samples that were positive with the 5′-UTR primers and contained RV-Ca were further tested with the 5′UTR-f and VP2-C252-5r primers, amplifying VP4 and partial VP2, in a two-step PCR assay followed by sequencing.

TABLE 1.

Diagnostic PCR primers used in the study

Primer namea	Sequence (5′ to 3′)b	Degeneracy level	Assay
5′UTRn-A1	GCACTTCTGTTTCCCCGGY	2	Seminested PCR
5′UTRn-A2	GCACTTCTGTTACCCCGGT	1	Seminested PCR
5′UTRn-A3	GCACTTCTGTTTCCCCCG	1	Seminested PCR
5′UTRn-B1	ACTTCTGTTTCCCCGGAGC	1	Seminested PCR
5′UTRn-B2	ACTTCTTGTTCCCCGGAGC	1	Seminested PCR
5′UTRn-Cc	TTCTGTTTCCCCGGRYGTG	4	Seminested PCR
5′UTR-rev	ACGGACACCCAAAGTAGT	1	Seminested PCR
5′UTR-revseq	TCAGGGGCCGGAGGA	1	Direct sequencing
A53-f	CTAATCGTTATCCGCAAGGTGC	1	Recombination test
B6-f	ACCGTTATCCGCCAACCAA	1	Recombination test
A53-r	CTCGTTACGACCAAGAATATGCT	1	Recombination test
B6-R1164-r	TGCTCATTACGACCTTAATATCAC	1	Recombination test
A82-f	GATCGTTATCCGCAAGATGC	1	Recombination test
B27-f	AACCGTTATCCGCCAATCAA	1	Recombination test
A82-r	TACCCACTGGATTGTGAGCAAT	1	Recombination test
B27-r	ACAACATTGGATTATGTTGCAAG	1	Recombination test
A21-f	GATCGTTATCCGCAAAGTGC	1	Recombination test
A21-r	ATTGCTCATTACGACTAGCAACAT	1	Recombination test
B6-R1912-r	AGAATTACTCATTACGACCTTAATGTC	1	Recombination test
5′UTR-f	ACTACTTTGGRTGTCCGTGT	2	Two-step PCR
VP2-C252-5r	AGTGATTTGYTTIAGCCTATC	2	Two-step PCR

^af, forward primer; r, reverse primer.

^bDegenerate nucleotides are designated with the standard ambiguity code.

Sequence analysis.

The nucleotide sequences (.ab1 format) were assembled and aligned using SeqMan software in DNAStar v. 9.0 (DNAStar). The quality of direct sequences generated with the 5′UTR-revseq and P3 primers was assessed by Phred base-specific quality score analysis (30, 31) integrated into CodonCode Aligner software (CodonCode Corp.). The phylogenetic analysis was performed with the neighbor-joining method of MEGA 5.1 software (32), with the pairwise-deletion option for gaps and missing data. The evolutionary distances were computed using the p-distance method and are presented as the number of base differences per site. The confidence probability that the interior branch length was >0 was estimated using a bootstrap test of 500 replicates. The phylogenetic trees were drawn to scale, with branch lengths in the same units as those of the evolutionary distances used to infer the phylogenetic tree.

Nucleotide sequence accession numbers.

All updated sequences were deposited in GenBank under the accession numbers KF695389 to KF695402.

RESULTS

Expanded 5′-UTR sequence database and refined diagnostic primers.

To design improved primers for pan-RV diagnosis, we analyzed a comprehensive 5′-UTR sequence data set that included a total of 156 reference strains of RV-A, RV-B, and RV-C (see Fig. S1 in the supplemental material) and multiple clinical isolates. This analysis confirmed a high degree of sequence conservation of the previously identified short motifs in the 5′ UTR (P1, P2, and P3) used for primer design (Fig. 1A). RNA secondary structure prediction of the RV 5′ UTR mapped the P1 and P3 primers to the bases of two stem-loop domains (domains 2 and 5) of the RV internal ribosomal entry site (IRES), while sequence conservation in P2 between domain 4 and domain 5 was not associated with a specific structural motif (Fig. 1B).

FIG 1.

Locations of primers used for detection and typing of RV. (A) Schematic representation of the previously published (9) and modified diagnostic primers in the virus genome. (B) Secondary structure of the 5′ UTR and cis-acting replication element (cre) of RV-C15 (GenBank accession number GU219984) predicted using the Mfold web server (38) with mapped primer annealing sites. Major stem-loop domains are numbered 1 to 6 below the structure, and the open reading frame (ORF) start codon (position 608) is indicated. Published primers are shaded, and modified primer annealing sites are underlined, boxed (position 529), or shown in bold (VP2-C252-5r). (+), forward primer; (−), reverse primer.

We next compared nucleotide sequence alignments of RV strains to human chromosome 6/RNA B2 sequences from the 424-bp PCR product (Fig. 2A), which revealed up to 100% identity between the 3′ ends of P1 and P3 primers (Fig. 2B), consistent with the observed nonspecific amplification of human sequences in the RV diagnostic tests. The sequence analysis also revealed short (3- to 7-base) motifs that were conserved among RV species but were different from the human sequences. In an attempt to eliminate nonspecific amplification and to design primers specific for RV-A (and RV-Ca), RV-B, and RV-Cc, rather than degenerate primers, these motifs were incorporated into the forward diagnostic primer set 5-UTRn (Fig. 2B).

FIG 2.

Nonspecific amplification product (A) and consensus sequences of RV-A, RV-B, and RV-C reference strains and clinical isolates in the primer annealing sites in 5′ UTR (B) and VP2 (C). Alignment positions correspond to the RV-C15 sequence (GenBank accession number GU219984). RV-C 5′-UTR sequences are divided into Ca and Cc subgroups. Human RNAB2/Chr6 sequences (GenBank accession numbers GQ497714 and NT007592) matching RV sequences in the 5′ UTR are shaded. Degenerate nucleotide positions are designated by the standard ambiguity code and shown in bold. Reverse primers are shown as reverse complementary sequences to match the alignments. Annealing sites of the original primers targeting the 5′ UTR (B) and novel primers targeting the 5′ UTR and VP2 (C) are boxed.

Since RV differentiation at the species level is sufficient for some large-scale epidemiological studies, we first validated the new primers by testing a set of previously typed RV-positive clinical samples in three separate “species-specific” seminested reactions (primers A1 plus A2 plus A3, primers B1 plus B2, and primer Cc), using the P1-P3 products for reamplification. The primers revealed overall comparable or even greater sensitivity, in comparison with P1 and P3, but cross-reacted with samples of all three RV species (see Fig. S2 in the supplemental material). ExoSAP-IT treatment of the first PCR products to prevent carryover of excess P1 and P3 universal primers did not completely eliminate the cross-reactivity of the modified primers (see Fig. S3 in the supplemental material); for this reason, we used the primers as a single forward primer set (n = 6; total degeneracy level of 10) in the second PCR, rather than separately, in further experiments.

Evaluation of the new 5′-UTR diagnostic primers.

In order to evaluate the RV specificity and analytical sensitivity of the modified diagnostic assay, we first selected a panel of RV-positive nasal lavage fluid samples containing a variety of known RV-A, RV-B, and RV-C types (n = 7), as well as a false-positive sample that had previously generated the nonspecific 424-bp amplicon. We compared the performance of two enzyme-buffer combinations (Platinum PCR Supermix [Life Technologies] and GoTaq Green master mix [Promega]). The latter was selected for better performance and also had the advantage of generation of an adenine overhang suitable for direct TA cloning of amplicons, at lower cost. The combination of the GoTaq mixture and the set of modified 5′UTRn-A, -B, and -C and 5′UTR-rev primers in the second PCR produced strong and distinct bands but did not prevent nonspecific amplification of human sequences (see Fig. S4A and B in the supplemental material).

We also compared the sensitivity of the modified protocol using serial 10-fold dilutions of purified RV-A16, RV-B14, and RV-C15 strains with known titers. The sensitivity of the first PCR was 10² to 10³ PFU (see Fig. S4C in the supplemental material), and the second PCR increased the sensitivity to 1 to 10 PFU (see Fig. S4D in the supplemental material). The two-step PCR protocol utilizing the original P1 and P3 primers in the first PCR and the redesigned primer set in the second seminested PCR was selected for use in further experiments.

Next, we attempted to eliminate nonspecific amplification of human sequences by using the modified reverse primer 5′UTR-rev, containing an additional conserved RV-specific nucleotide (thymine) at the 3′ end (Fig. 2B and Table 1), in the first and second PCRs and removing unincorporated primers from the first PCR (ExoSAP-IT; Affymetrix) before reamplification. Use of the modified primer in both PCRs, or ExoSAP-IT treatment regardless of the reverse primer used in the first PCR, eliminated nonspecific amplification (Fig. 3A). Direct testing of the modified and original reverse primers in the first and second PCRs showed similar sensitivity for RV detection and improved specificity of 5′UTR-rev (Fig. 3B), which was confirmed by testing a panel (n = 17) of clinical specimens (Fig. 3C and B). Of note, a sample (RV-A) that produced only the nonspecific product after amplification with the P3 primer (Fig. 3B) was shown to be RV positive using 5′UTR-rev, confirming improved specificity and sensitivity of the modified assay for testing samples with high human RNA concentrations and very low copy numbers of viral RNA. Reverse primer 5′ UTR-rev was selected for use in both the first and second PCRs in the modified assay.

FIG 3.

Elimination of nonspecific amplification of human sequences in the RV 5′-UTR assay. (A) Effect of exonuclease I and shrimp alkaline phosphatase (ExoSAP-IT) treatment of the first PCR products on nonspecific amplification of human sequences, using the original (P3) or modified (5′UTR-rev) primers in the first and/or second PCR. (B) Effect of the reverse primer in the first PCR on detection of RV-A, RV-B, and RV-C isolates and nonspecific amplification of human sequences (−). (C and D) Comparison of the original and modified reverse primers in the first PCR in testing of a set of clinical samples. A total of 17 clinical samples with high human RNA concentrations were amplified with P3 (C) generating nonspecific product or 5′UTR-rev (D) primers in the first PCR. Products of the second PCR using the modified primers are shown in panels B to D. L, 1-kb Plus DNA ladder (Life Technologies).

In order to minimize the number of samples that require cloning of the PCR products due to insufficient quality of the sequences obtained with the direct sequencing method, we designed a new reverse primer, 5′UTR-revseq (Fig. 1B), internal to the amplicon. We compared its efficacy in parallel direct sequencing reactions with the original sequencing primer P3 (6, 9), using a set of PCR products from RV-A, RV-B, and RV-C clinical isolates. The 5′UTR-revseq primer generated slightly shorter (∼230-bp versus 300-bp) but more reliable sequences than did P3 (see Fig. S5 in the supplemental material), with higher sequence quality and lower error probability being confirmed by Phred quality score analysis (see Table S1 in the supplemental material).

Phylogenetic analysis and type assignments based on 5′-UTR sequences.

Our RV molecular typing assay is based on phylogenetic analysis of an ∼270-bp-long fragment internal to the P1-P2 PCR amplicon (Fig. 1A) (9). The original sequence data set that included all of the prototype sequences of RV-A and RV-B and 52 novel W types (6, 9) was augmented with the newly identified RV-A101 to RV-A106, RV-B101 to RV-B104, and available RV-C types (n = 47) to determine phylogeny and to reanalyze intertype divergence. RV 5′-UTR reference sequences clustered into 3 major clades, which included RV-B, RV-Cc, and combined RV-A and RV-Ca types (see Fig. S1 in the supplemental material). Their segregation was supported by high bootstrap values (≥90% of 500 replicates). RV-Ca sequences segregated into at least 4 major subclusters relatively distant from most RV-A branches but, based on the 5′-UTR amplicon sequence alone, they could not be clearly separated from a number of RV-A types (e.g., A12, A45, A51, A65, A71, A78, A101, A102, A103, and A106).

Comparison of the frequency distributions of the pairwise nucleotide distances between the 5′-UTR sequences of all RV-A (n = 83) and RV-B (n = 29) types and available RV-C (n = 47) types revealed bimodal patterns for species A and B (reflecting distances within and between major branches). There was overall greater divergence between RV-C types (from 4 to 42%), in comparison with RV-A or RV-B, with a wider multipeak distribution corresponding to distances between types within and between distant Ca and Cc clusters (see Fig. S6A in the supplemental material). Our analysis showed intertype divergence values of <7% for a number of A (n = 25), B (n = 17), and C (n = 10) type pairs, including some newly assigned types (see Table S2 in the supplemental material). Some of these type pairs were proposed to be reclassified as single types (4), due to serological cross-reactivity and/or high levels of sequence identity in the coding regions (e.g., RV-A8/A95, RV-A29/A44, and RV-A54/A98).

Development of a RV-C-specific typing assay targeting VP4/VP2.

The 5′-UTR typing assay has excellent sensitivity but, as shown in the previous section, cannot reliably assign species of RV-Ca types. In order to develop a sensitive and specific assay for species and type assignments for RV-C clinical isolates, we evaluated published PCR primers targeting RV structural genes (8, 22, 25) and also performed multiple sequence alignments of coding sequences of available prototypic strains of RV-A, RV-B, and RV-C, to design new primers. In preliminary tests with RV-positive clinical samples, the sensitivity of the published degenerate primers was suboptimal (data not shown). Therefore, we designed and evaluated several new reverse primers with minimal degeneracy (n = 2) targeting the 5′ end of VP2 coupled with the 5′-UTR-specific forward primer (Fig. 2C). The VP2-C252-5r primer, which anneals to the terminal loop of the predicted cis-acting replication element (cre) in the RV-C15 genome (Fig. 1B), demonstrated high analytical sensitivity (10 PFU after reamplification in the second PCR), generating a 330-bp product ready for direct sequencing (Fig. 4A). Strong virus-specific bands were found only for samples containing RV-C (Fig. 4B); this was confirmed with a larger panel of previously typed clinical samples (Fig. 4C). In contrast to 5′-UTR analysis, the nucleotide distance distributions between partial (252- to 258-bp) VP4/VP2 sequences of A, B, and C types revealed single peaks (see Fig. S6B in the supplemental material) with fewer violations of the proposed thresholds of 10% sequence divergence for type assignments (see Table S2 in the supplemental material).

FIG 4.

Sensitivity and specificity of the VP4/VP2 PCR assay for detection and molecular typing of RV-C. (A) Amplicons obtained after the first (left) or second (right) PCR with serial 10-fold dilutions (10⁵ PFU to 10⁻¹ PFU equivalents) of purified RV-C15. (B) Test of primer specificity using clinical samples containing RV-A, RV-B, or RV-C isolates. (C) PCR amplification of a larger set of RV-C and novel RV-A isolates from clinical samples confirms VP4/VP2 assay sensitivity and specificity. *, potentially novel RV-A types (W22, W48, and W49 isolates) that do not match any of the reference types. L, 1-kb Plus DNA ladder (Life Technologies). Products of the second PCR are shown in panels B and C.

Typing of clinical isolates.

Using sequence information from the 5′-UTR fragment, the pairwise nucleotide distances between previously assigned RV W types (Wisconsin clinical isolates) ranged from 6 to 42% (6, 9). Four pairs (W2/W24, W4/W9, W18/W40, and W3/W21) had minimal distances between pairs less than the previously proposed threshold of 7%; however, their assignments as separate types were confirmed by VP4/VP2 sequence analysis. Phylogenetic analysis of W types and published reference types based on partial 5′-UTR (see Fig. S7 in the supplemental material) and VP4/VP2 (see Fig. S8 in the supplemental material) sequences showed different tree topology, as expected, with distinct grouping of RV-C sequences into one cluster only in the VP4/VP2 tree. The 5′-UTR-based assignments in most cases matched VP4/VP2 assignments performed by finding the closest sequence among the published reference types, with only a few exceptions (Fig. 5).

FIG 5.

Neighbor-joining phylogenetic trees based on partial 5′UTR and VP4/VP2 nucleotide sequences of selected RV-C reference types and newly identified or putative recombinant Wisconsin (W) types, constructed using MEGA 5.1 software. All major nodes are labeled with bootstrap values (% of 500 replicates). Branch lengths are proportional to nucleotide similarity (p-distance). RV-A16 and RV-B14 were included to represent RV-A and RV-B clusters, respectively, and poliovirus (PV-3l) was included as an outgroup. Accession numbers for the 52 W types have been published (6, 9); GenBank accession numbers for the newly identified W types (shown in bold) are KF695389 to KF695402. W type numbers are followed by corresponding RV-C type designations (in parentheses) based on VP4/VP2 partial sequences (http://www.picornastudygroup.com/types/enterovirus/hrv-c.htm) except for the four putative recombinant W types (underlined), which are followed by the closest reference 5′-UTR sequence type in the VP4/VP2 tree and by the closest reference VP4/VP2 sequence type in the 5′-UTR tree, to show incongruent clustering.

A total of four types, i.e., W21 (C33 in 5′ UTR/Cpat22 [provisionally assigned type] in VP4/VP2), W32 (C51/C37), W47 (C8/Cpat28), and W50 (C20/C34), reveal incongruent clustering in these two regions, suggesting intertypic recombination. Although the published 5′-UTR reference sequence set for RV-C is not yet complete (e.g., C34, C37, Cpat22, and Cpat28 sequences are lacking in GenBank), the putative recombinant origin of these viruses is evident from analyses of clinical isolates potentially representing the parental 5′-UTR sequences (e.g., W47-R954, W51, and W59). Human RV-C8 and Cpat28 were recently proposed to be reclassified as a single C8 type (4); however, a full-length Cpat28 sequence analysis is necessary for final assignment.

We expanded our sequence database with seven RV types (W53 to W59) (Fig. 5) that had not been detected in Wisconsin previously. 5′-UTR sequences of three clinical isolates (W55, W56, and W58) clustered with reference types (e.g., C1, C10, C32, and C39) that had <7% p-distances in the 5′ UTR but had distinct VP4/VP2 sequences (see Tables S2 and S3 in the supplemental material). The other four types (W53, W54, W57, and W59) matched novel RV-A (A106) or RV-C (C17, C33, and Cpat28) 5′-UTR and/or VP4/VP2 published reference sequences. All of the updates to the current sequence database are summarized in Table S4 in the supplemental material.

Artificial recombination during PCR amplification.

Sequence analysis of the cloned 5′-UTR sequences of recent clinical samples containing more than one RV type found several putative recombinant isolates (n = 8) with potential crossover sites in the central part of the PCR amplicon (Fig. 6A to C). All of these sequences (1 or 2 of 6 clones that were sequenced for each sample) were found along with the parental sequences of certain RV-A, RV-B, or RV-C types. To test for artificial recombination during PCR amplification, we selected three samples and designed three sets of primers (Table 1) specifically matching each single RV type and tested them in three possible combinations (e.g., A82-f plus A82-r, B27-f plus B27-r, and A82-f plus B27-r for isolate R1742). PCR products were produced only with the homologous primer pairs, indicating the absence of recombinant sequences in these samples before amplification (data not shown). In addition, we suspect that the previously assigned W14 type that lacks the matching VP4/VP2 sequence is also an artificial recombinant (Fig. 6D).

FIG 6.

Nucleotide sequence alignments of three recent RV isolates (A to C) and the W14 type (D) with putative artificial recombinant 5′-UTR sequences and parental reference strains. Potential recombination sites are boxed and shaded.

DISCUSSION

PCR assays targeting highly conserved sequence motifs in the 5′ UTR (9, 10, 17) feature the highest sensitivity for RV genome detection, whereas those amplifying structural gene sequences (8, 22, 25) provide more accurate species and type assignments but typically utilize highly degenerate primers that detect RV only in about 46 to 80% of RV-positive samples (7, 8, 33–36). In this study, we combined elements of both assays to develop a practical diagnostic and typing protocol that has excellent specificity and sensitivity, reduced sample-processing time, and accurate RV species and type assignments. We revised and updated our reference sequence database, redesigned 5′-UTR primers, and developed new primers for amplification of the VP4/VP2 fragment, to enable more accurate typing of clinical isolates of RV-C species. Through parallel analysis of partial 5′-UTR and VP4/VP2 sequences, we identified seven RV types (W53 to W59) that had not been detected in Wisconsin previously, and we found four putative natural recombinants (W21, W32, W47, and W50) among the previously assigned types.

Since sequence analysis of the updated and more comprehensive 5′-UTR data set confirmed high sequence conservation of the previously identified short motifs, we used the same locations for primer modifications. We tried to develop species-specific primers (Fig. 2B) that could quickly classify isolates as RV-A, RV-B, or RV-C; however, this was not possible due to the cross-reactivity of the primers.

It was interesting and surprising to find that both the forward (P1) and reverse (P3) primers from the original typing assay almost completely matched the human long noncoding RNA B2 sequence and generated an amplicon that was close in size to the RV-specific product (Fig. 2A). Until now, there has been no information on how 5′-UTR primer design affects rates for nonspecific amplification products. We experimentally confirmed that the addition of one virus-specific base (thymine) to the 3′ end of the reverse primer helped to eliminate nonspecific amplification of the 426-bp product when used in both PCRs and combined with the new 5′UTRn set in the seminested PCR. The modified primer is only slightly different (lacking 1 to 5 bases at the 5′ end) from the published reverse primers (10, 25), but effects on nonspecific amplification in those studies were not provided. Although some low-yield nonspecific products (human sequences) are still generated in about 30% of tested samples (Fig. 3D), they can be distinguished by size and intensity from the stronger virus-specific amplicons, without sequence confirmation.

We further improved direct sequence quality by using an internal primer, 5′UTR-revseq, that allowed us to type >90% of single-type RV samples without cloning. The remaining 10% of the samples and those containing multiple types (≤6% of RV-positive samples) must be sequenced, using a simple TA cloning procedure with commercially available reagents. Of note, we have found that artificial recombination can occur during PCR amplification of heterologous viral RNA templates in dual-type samples; therefore, careful verification is needed for evolutionary analysis of RV for recombination events when both parental and recombinant sequences are detected in the same specimen.

It has been shown that the combination of molecular diagnostic assays targeting both conserved 5′-UTR (in different formats) and capsid-coding regions improves typing results and also minimizes the risk of missing novel RV variants (25, 33–35, 37). We supplemented our modified 5′-UTR-based assay with the new primers amplifying short fragments of the VP4/VP2 coding sequence for unequivocal species and type assignments for RV-C isolates and identification of putative recombinant sequences.

The 5′-UTR sequences are not sufficient for new RV type assignments for clinical isolates that appear to be novel. Analysis of intertype divergence among a total of 156 published 5′-UTR reference sequences (∼300 bp) of RV-A, RV-B, and RV-C types showed that low values for p-distances between distinct types make it difficult to set up clear divergence thresholds. In addition, phylogenetic analysis of the 5′ UTR does not segregate the majority of RV-C types (RV-Ca clade) from RV-A types, due to historical interspecies recombination (7, 8). We agree with the previous proposals (3, 4) that sequencing of the complete VP1 (minimally) and other capsid-coding regions or full genome sequencing is absolutely necessary for genetic assignments for new RV types when partial 5′-UTR and/or VP4/VP2 analysis indicates low sequence identity with all known reference types.

On the other hand, the modified 5′-UTR assay has the greatest sensitivity. With the availability of a comprehensive set of partial 5′-UTR reference sequences (currently including 156 reference types and multiple isolates), we routinely use the modified RT-PCR assay targeting the 5′ UTR, followed by direct sequencing, as the first step in RV type identification. Overall, the 5′-UTR typing assay does a good job of differentiating RV-A and RV-Ca types. The smallest nucleotide p-distance (13%) found between the closest RV-A and RV-Ca partial 5′-UTR sequences (RV-A78 and RV-C22) is substantially greater than the proposed type assignment threshold of 7%; therefore, RV-A and the majority of RV-Ca types can be differentiated and accurately assigned through 5′-UTR analysis alone. However, there are more limited amounts of sequence data for some RV-Ca types, compared to RV-A types, and those types cannot be assigned with certainty using 5′-UTR information alone. Under such circumstances, VP2/VP4 sequence analysis is helpful.

Furthermore, if the results of sequence comparisons are equivocal (e.g., putative recombinant), then the sample is retested with the VP4/VP2 assay. Most (if not all) of the isolates with interspecies recombination described to date possess the RV-A 5′-UTR sequence and the RV-C capsid-coding sequence. For this reason, our VP4/VP2 assay should be efficient for analysis of putative recombinant RVs in most cases. Finally, if sequences from more than one RV or a recombinant sequence are detected in a sample, then cloning can be attempted for definitive results. This approach has allowed us to type clinical isolates with >97% matching results from the two regions, and this success rate is likely to improve as additional RV-Ca sequences are added to the reference database.

Our protocol (Fig. 7) has been used to type RVs from >3,800 clinical samples (see Fig. S9 in the supplemental material) that tested positive in multiplex RT-PCR assays (13). Of note, our current results in combination with previously published data (6, 9) demonstrated that, since 1999 in one geographical area (Madison, WI), we detected a total of 52 of 63 types of the recently discovered RV-C species (including provisionally assigned types) identified worldwide.

FIG 7.

Workflow for the modified RV molecular typing assay. Novel RV isolates required VP1 or full-genome sequencing for definitive type assignment.

In summary, we report a modified RV molecular typing assay targeting the 5′-UTR and VP4/VP2 regions, with improved specificity and sensitivity. This RT-PCR assay, followed by partial sequencing and phylogenetic analysis, demonstrated efficient detection and reliable species and type assignments directly in clinical samples using a comprehensive set of reference sequences for both genomic regions. Our fast accurate method has proved useful for typing RVs in large epidemiological studies involving hundreds of samples.