Simultaneous Genotyping of All Hemagglutinin and Neuraminidase Subtypes of Avian Influenza Viruses by Use of Padlock Probes

Abstract

A subtyping assay for both the hemagglutinin (HA) and neuraminidase (NA) surface antigens of the avian influenza virus (AIV) has been developed. The method uses padlock probe chemistry combined with a microarray output for detection. The outstanding feature of this assay is its capability to designate both the HA and the NA of an AIV sample from a single reaction mixture. A panel of 77 influenza virus strains was tested representing the entire assortment of the two antigens. One hundred percent (77/77) of the samples tested were identified as AIV, and 97% (75/77) were subtyped correctly in accordance with previous examinations performed by classical diagnostic methods. Testing of heterologous pathogens verified the specificity of the assay. This assay is a convenient and practical tool for the study of AIVs, providing important HA and NA data more rapidly than conventional methods.

Respiratory viruses such as influenza viruses consistently present serious problems for human and animal health worldwide. Both human and animal disease outbreaks have significant economic costs from medical treatment, loss of working days, loss of animal stocks, preventative or stamping-out measures, and the potential to cause zoonotic infections, among other things. In the United States, the annual direct medical costs related to influenza epidemics are estimated to be 10.4 billion U.S. dollars. The total economic burden, including other factors such as lost earnings, amounts to 87.1 billion U.S. dollars (13). The economic losses from animal outbreaks are sizeable as well. The poultry industry has been hit the hardest. Between 1998 and 2005, more than 200 million birds have been involved in avian influenza outbreaks. In the European Union between December 1999 and April 2003, more than 50 million birds died or were culled because of highly pathogenic avian influenza disease events (6). Although vaccines are used for both human and animal disease control programs, this is far from an adequate solution. The high variability of the viral genome means that long-term immunization is not possible (21).

Influenza viruses belong to the Orthomyxoviridae family and are divided into types A, B, and C according to the structural (matrix [M] and nucleoprotein [NP]) genes. Type B and C influenza viruses are restricted to a few hosts, including humans, and generally cause mild symptoms. Type A viruses are more virulent and have a broad host range. The viral genome is segmented; types A and B contain eight segments, and type C contains seven segments, encoding 11 different proteins. Influenza A viruses are classified into subtypes according to two surface antigens, hemagglutinin (HA) and neuraminidase (NA). Currently, there are 16 known HA and 9 known NA subtypes. Although they can combine by reassortment, not all possible variants have been observed (21, 23).

Subtyping is essential for the identification of circulating strains, investigation of viral reassortment, and epidemiological characterization of recently emerging viruses. The “gold standard” for the identification of subtypes is serological testing (http://www.oie.int/eng/normes/mmanual/A_00037.htm), but it is time-consuming and needs a high viral titer. There are also several assays using nucleic acid detection techniques for this purpose (7, 8, 16, 22, 24), but none of them use the approach described herein. In this study, a method using padlock probes (PLPs) for the general and simultaneous detection of each of the 16 HA and 9 NA subtypes of avian influenza virus (AIV) was developed.

A PLP is a linear oligonucleotide that contains half of a unique target recognition sequence at its 5′ end and the other half at its 3′ end, referred to as the two probe arms (12, 14). In the presence of an appropriate DNA or RNA target sequence, the probe becomes circularized by ligation while unreacted probes remain linear. The sequence in the middle of the probe is the amplification region, which is designed for double amplification (rolling-circle amplification [RCA] and PCR) with a general primer pair. It also contains a customized tag sequence, which is unique for each particular probe. This design allows PLPs to be multiplexed by designing probes with different arms to target different regions of DNA or RNA. Since all of the PLPs have the same primer recognition sites, only one primer pair is required in the assay. To identify whether targets are present requires the simple correlation of the target recognition sites (arms) with the unique tag sequence that is detected on a solid-surface microarray, which contains sequences complementary to each unique tag located in the amplification region of the probes (Fig. 1). Amplification of the circularized probe takes place in two steps, the first of which is RCA. This step produces a large linear concatemer composed of approximately 1,000 repeats of each of the circularized probe molecules (2), which then acts as a template for the second stage, a traditional PCR with a labeled fluorescent primer. Remaining linear, unreacted probes fail to be amplified in either of the two amplification steps. By first amplifying reacted probes by RCA, the risk of amplification artifacts is reduced, increasing the sensitivity of the assay.

FIG. 1.

Schematic representation of the PLP assay. Panel a shows the ligation step, in which the DNA ligase seals the adjacent 5′ and 3′ ends of the PLP when it is base paired to the appropriate target, thereby creating a circular molecule (red indicates the target-complementary regions of the probe, light blue shows the forward primer site, dark blue shows the reverse primer site, and green symbolizes the microarray tag). In panel b, the circularized probes are amplified by an RCA reaction, resulting a long stretch of concatemer products. The RCA products are further amplified by a PCR in which one of the primers is fluorescently labeled (c), and then the labeled PCR products are loaded onto a solid-surface microarray (d) in which the fluorescence of each spot can be measured.

The design, which allows all circularized probes to be amplified by the same primer pair and then sorted by unique microarray tags, makes the PLP system more suitable for use in highly multiplexed assays compared to the traditional multiplex PCR, which is limited by the number of primer pairs that can be combined in individual reaction mixtures before cross-reactivity becomes a problem (9). Another advantage of the PLP design in virology relates to the variability of (viral) genomes, especially in the case of RNA viruses. PCR-based approaches require two conserved regions (one for each primer). Contrarily, PLPs need only one recognition site on the target sequence, where the two probe arms hybridize. The PLP design can therefore provide the benefit of speed and sensitivity derived from using a nucleic acid-based method while the amount of information is greatly increased by the high level of multiplexing.

Our results show that the method presented here is capable of differentiating among all of the known HA and NA subtypes and identifying AIV with high accuracy and repeatability. The analysis is performed in a single reaction tube with just a few sequential additions of reagents before the readout is performed on an oligonucleotide array, which helps to decrease the risk of contamination and allows detection in a short time.

MATERIALS AND METHODS

PLPs and oligonucleotides.

Sequences of the HA, NA, and M genes were downloaded from the National Center for Biotechnology Information Influenza Virus Resource (http://www.ncbi.nlm.nih.gov/genomes/FLU/). The sequences were aligned by using ClustalW (19), and the most conserved parts of the M, HA, and NA sequences were selected by using the BioEdit Sequence Alignment Editor software (v7.0.5.2) (9, 20). PLPs (see Table S1 in the supplemental material) were designed with the ProbeMaker software (18), and oligonucleotides were purchased from Biomers (Ulm, Germany). All probes were chemically 5′ phosphorylated. All other oligonucleotides (see Table S1 in the supplemental material) and primers were also ordered from Biomers (Ulm, Germany). Oligonucleotides for microarray probe spot assays, complementary to the tag sequences of PLPs, were attached to the array by a 5′-NH₂-(TC)₇T-3′ residue at their 5′ ends. Tag sequences were selected from the GeneFlex Tag Array collection (Affymetrix, Santa Clara, CA).

Virus samples and cDNA synthesis.

Allantois liquids containing virus isolates were obtained from The Swedish National Veterinary Institute (Uppsala, Sweden). Viral RNA was isolated in Trizol by following the manufacturer's instructions for RNA extraction (Invitrogen). Prior to reverse transcription, 5 μl of RNA solution was denatured in the presence of 0.02 U of random hexamers (Amersham Biosciences) for 10 min at 65°C and then chilled on ice. The reverse transcription mixture comprised 1× First Strand buffer (Invitrogen), 200 nM deoxynucleoside triphosphate (dNTP) mix (Amersham Biosciences), 100 nM dithiothreitol (Amersham Biosciences), 32 U of RNAguard (Amersham Biosciences), and 200 U of Moloney murine leukemia virus reverse transcriptase (Invitrogen) in a total volume of 20 μl. The reaction mixtures were incubated at 42°C for 90 min, followed by inactivation of the enzyme at 95°C for 5 min.

PLP ligation and amplification.

Two microliters of cDNA was combined with 5 μl of a ligation mixture containing 1× Ampligase buffer, 0.5 U of Ampligase (Epicentre), and 0.15 nM PLP mix. Reaction mixtures were placed in a thermal cycler at 95°C for 1 min and 58°C 60 min and then left at 8°C. Eight microliters of an RCA mixture containing 2× Phi29 buffer (Fermentas), 375 μM dNTP, 0.75 μg bovine serum albumin, 12.5 nM Fwd primer, and 1 U of Phi29 DNA polymerase (Fermentas) was added to the ligation reaction mixtures, resulting in a 15-μl reaction mixture. The tubes were incubated at 37°C for 30 min, followed by 85°C for 5 min, and stored at 8°C. Thirty-five microliters of a PCR mixture containing 1× Platinum Taq buffer (Invitrogen), 180 μM dNTP, 0.7 mM MgCl₂, 290 nM each Fwd primer and Rew-Cy5 primer, and 2 U of Titanium Taq DNA polymerase (ClonTech) was added for a final volume of 50 μl. Reaction mixtures were placed in a thermal cycler at 94°C for 2 min; cycled 20 times through 95°C for 15 s, 58°C for 20 s, and 72°C for 10 s; and finally kept at 8°C until later use.

Oligonucleotide arrays and hybridization.

Microarray slides and reusable silica masks were used to create hybridization cassettes (as described in reference 4) with 16 wells per glass slide (CodeLink; Amersham Biosciences). Spot oligonucleotides were dissolved to a 50 μM concentration in 100 mM NaPO₄ and printed on the slides with a Piezorray instrument (Perkin-Elmer, Shelton, CT). All wells contained the entire set of probes in triplicate. For sample hybridization, 45 μl of PCR product was transferred to individual wells and the hybridization cassette was placed at 52°C for 1 h. The silica mask was removed in a 0.075xTNT wash solution (11.25 mM NaCl, 0.75 mM Tris-HCl [pH 8], 0.004% Tween 20), and the slide was transferred to a 0.075xTNT wash for 10 min at room temperature, followed by a quick rinse in distilled H₂O, and dried by centrifugation. Microarrays were scanned in a Genepix 4000B scanner (Axon Instruments), and images were analyzed with GenePix Pro 6.0 software (Molecular Devices Corp.). The local mean background of each spot was subtracted from the recorded mean signals, and an average was calculated from the triplicates. Spots with resulting negative or zero values were disregarded. Nontarget control signals were subtracted from the corresponding sample signals, and the results were set to zero if negative. A cutoff threshold of 10% of the maximal signal was set to disqualify lower signals.

Sensitivity of the assay.

Assay sensitivity was calculated in two ways. The first was with synthetic oligonucleotide target sequences in 10-fold dilutions (10⁻¹ to 10⁻¹²) that matched the target sequence for the recognition sites of the H13_1, H13_2, and H14_1 probes. The second sensitivity test consisted of 10-fold dilutions (10⁻¹ to 10⁻⁷) of the H1N1 and H5N1 isolates which had titers calculated by a hemagglutination inhibition test.

RESULTS

A subtyping assay was developed with the goal of detecting the presence of AIV and determining the HA and NA subtypes. Seven thousand five hundred sixty sequences were used for designing the recognition sites of PLPs targeting conserved regions of the 16 HA and 9 NA genes. The probes for the general detection of influenza virus were designed to target the most conservative part of the influenza virus genome, the M gene. Originally, probes for the NP gene were used for general detection as well but later their use was discontinued because of the better efficacy of the M probes. The primer sites (identical in all probes) and the tag regions (unique for each probe) of the PLPs were assembled with the ProbeMaker software, which generated oligonucleotide probes with similar melting temperatures and checked for lack of cross-reactivity and dimer formation. A BLAST (1) search was performed for the designed probes, but no significant similarity to irrelevant target sequences was found.

Each well on the microarray contained all of the spots in triplicate. During the data evaluation procedure, the fluorescence values were qualified in two ways. They were first compared to the nontarget control sample and then compared to any other reaction spots in the same subarray. Calculations were made on an Excel spreadsheet. In most cases, the correct signals were significantly higher, allowing direct evaluation after scanning, while further data processing allowed quantitative diagnosis.

Seventy-seven samples were used to test assay performance. All were detected as AIV positive according to the M-based probes; 75 were subtyped correctly (Table 1). Two samples did not give interpretable subtyping signals (A/Mallard/Sweden/S900229/03/H10N8 failed for both HA and NA; for A/Mallard/Sweden/S900820/03/H12N5, NA was not detectable), probably because of the diversity of the HA and/or NA in the target region of these samples, and their sequences have not matched any assay probes. None of the samples were falsely subtyped. Ten different avian and human heterologous pathogens were tested by investigating cross-reactions, and all proved to be negative (see Table S2 in the supplemental material).

TABLE 1.

Summary of the results of this study^a

Sample subtype	No. of samples	Positivity ratio (%)b/avg ratio of positive probes for each subtype (%)c
H1N1	3	100/50
H2N3	5	100/86
H3N2	1	100/62
H3N8	10	100/50
H4N3	3	100/86
H4N6	11	100/71
H5N1	10	100/58
H5N2	3	100/57
H5N3	2	100/55
H6N2	1	100/73
H6N8	5	100/63
H7N1	1	100/67
H7N7	5	100/63
H8N4	2	100/75
H9N2	1	100/75
H10N4	2	100/71
H10N7	1	100/50
H10N8	1	0/50
H11N9	4	100/86
H12N5	2	50/90
H13N6	1	100/75
H14N5	1	100/100
H15N9	1	100/83
H16N3	1	100/71

^aFor extended results, see Table S2 as supplemental material.

^bIndicates how many of the samples were subtyped correctly.

^cShows how many probes were positive compared to the total number of probes (sum of HA, NA, and M) from each subtype.

The smallest detectable amount of artificial target molecules was 600 copies, and the lowest detectable dilution of the H1N1 and H5N1 strains (both with a 1:32 hemagglutination inhibition titer) was 10⁻⁴ (data not shown). For comparison, the H5N1 sample was also run on a diagnostic PCR (17), where the lowest detectable dilution was 10⁻⁶. The repeatability of the assay was examined by triplicate trials with an average coefficient of variation below 15% in the Swedish National Veterinary Institute laboratory. Thirty randomly chosen cDNAs were repeated in a blind study with identical results at the Rudbeck Laboratory (Uppsala, Sweden).

DISCUSSION

The method developed in this study combines the multiplexing capability of PLP chemistry with microarray detection to identify samples positive for AIV with HA and NA subtyping simultaneously in approximately 4 h.

The assay presented contained the following reaction steps. During the ligation, the probes which were base paired to their appropriate target became circularized by DNA ligase and only these were amplified in the following RCA reaction. Since RCA is a linearly processing amplification, the probes were further amplified and fluorescently labeled in the following PCR. The PCR products were loaded onto a microarray to which oligonucleotides complementary to the tag sequences of the probes had been attached previously. Finally, the microarray slide was scanned to measure and quantify the intensity of the fluorescence of each spot.

This method is able to detect isolates positive for influenza virus and simultaneously subtype the HA and NA surface proteins. The PLP methodology was chosen because of its high multiplexing capacity, which constrains more conventional PCR methods (10, 15). PLPs also reduce the number of conserved regions needed on target nucleic acids because both probe arms bind to a single region (2, 3). Amplification is then based on the circularized PLP and is independent of the viral sequences. Furthermore, the DNA ligase only functions for precise matches that help avoid a false diagnosis (11, 13). However, the high variability of the influenza virus genome can produce false-negative results as a consequence of even a single nucleotide change at the site of the ligation. It may be difficult to find probes for all variations of the HA and NA genes, but the system incorporates redundant probes, targeting each gene in more than one position—this reduces the risk that mutations will result in assay failure during subtyping. The aim during the design of the probes was to cover all variants with the sequence information available, and this resulted in various numbers of probes for various subtypes. The main reason for that was the different genetic diversity of each subtype, which is strongly related to the number of accessible sequence data.

The use of orthogonal tag sequences amplified by common primers renders the assay flexible, as new probes can be added if new AIV variants emerge that are not covered by the current set of probes. This assay format also provides the option of broadening the assay's detection capability to include PLPs for differential diagnosis involving pathogens that produce similar clinical symptoms.

The sensitivity of the assay was evaluated with a dilution series of artificially synthesized target molecules. The detection limit was found to be 600 copies. To further investigate its sensitivity, the assay's performance was compared to that of a diagnostic PCR (17) and classical virus isolation. The PLP assay gave reliable signals down to a 10⁻⁴ dilution, whereas the diagnostic PCR had a detection limit of 10⁻⁶. The virus isolation gave barely readable signals after just one dilution step. This work, as another study (25) has already suggested, showed that although the PLP assay cannot match the sensitivity of a real-time PCR, it still has reasonably high sensitivity. Furthermore, it offers advantages in multiplexing and target region selection that conventional PCR methods cannot, providing a powerful alternative for the determination of AIV subtypes.

For sample testing, mostly Scandinavian AIV strains were available in this study, but in the system design a worldwide collection of sequences was used, ensuring that the assay can be used on samples from any location. Allantois fluids containing cultivated viruses were used in this study, but five H5N1 clinical samples (V428 to V599) were also tried out with appropriate results. This study was focused on avian viruses, but during the design stage it was unavoidable to involve strains from other species as well. Preliminary test results suggest that the method, after further extension and evaluation, can be applicable for the detection of human and swine influenza viruses.

Although PLPs are long oligonucleotides (>100 bp), which cost more than regular primers, 1 nmol of probes allows almost a million reactions, which makes the individual assay cost very low. The other reagents involved in the reaction are standard molecular assay chemicals. Although the equipment used is currently uncommon in many diagnostic laboratories and the lack of automation of the assay is a certain limitation in this area, similar approaches are becoming widely accepted (5, 7, 8, 16, 22, 24).

The subtyping assay presented here takes 4 h to perform (excluding cDNA preparation), which is comparable to diagnostic PCRs. In summary, subtyping of AIV by PLPs can be accomplished in a comparably short and straightforward process that has the potential to be applied for the clinical diagnostic differentiation of AIV strains.