Rapid Method for Detection of Influenza A and B Virus Antigens by Use of a Two-Photon Excitation Assay Technique and Dry-Chemistry Reagents

Abstract

New separation-free assay methods for the rapid detection of influenza A and B virus antigens are presented. The methods employ dry-chemistry reagents and the recently developed two-photon excitation (TPX) fluorescence detection technology. According to the assay scheme, virus antigens are sandwiched by capture antibody onto polymer microspheres and fluorescently labeled antibody conjugate. Consequently, fluorescent immunocomplexes are formed on the surface of microspheres in proportion to the concentration of the analyte in the sample. The fluorescence signal from individual microspheres is measured, separation free, by means of two-photon excited fluorescence detection. In order to demonstrate the applicability of the new assay technique for virus antigen detection, methods for influenza A and B viruses were constructed. The assay method for influenza A virus applied a molecular fluorescent label, whereas the method for influenza B virus required a nanoparticle fluorescent reporter to reach sufficient clinical sensitivity. The new methods utilize a dry-chemistry approach, where all assay-specific reagents are dispensed into assay wells already in the manufacturing process of the test kits. The performance of the assay methods was tested with nasopharyngeal specimens using a time-resolved fluoroimmunoassay as a reference method. The results suggest that the new technique enables the rapid detection of influenza virus antigens with sensitivity and specificity comparable to that of the reference method. The dose-response curves showed linear responses with slopes equal to unity and dynamic assay ranges of 3 orders of magnitude. Applicability of the novel TPX technique for rapid multianalyte testing of respiratory infections is discussed.

Over the last decades, influenza virus-specific antiviral drugs such as neuraminidase inhibitors (9) have been developed to alleviate symptoms and to shorten the symptomatic period of influenza virus infections. Currently, however, antiviral drugs are expensive, and their use without reliable diagnosis is not recommended. Therefore, prescription of influenza virus-specific drugs should be limited only to patients diagnosed as being positive within a few days from the onset of symptoms. Preferably, results from a pathogen identification test should be available during the same patient visit. This makes conventional laboratory methods impractical, as their turnaround time and cost structure do not support point-of-care testing. Rapid assay methods (such as lateral flow) are commercially available (4, 19, 23, 43), but their performance is compromised compared to standard clinical in vitro techniques.

Rapid virus diagnoses help to manage virus outbreaks and to reduce the empirical use of antibiotics, as positive virus test result can often be used to exclude bacterial etiology. Many of the patients with pharyngitis, bronchitis, or common flu are still prescribed antibiotics, just in case, although it is now well known that most cases are caused by viral pathogens. The superfluous use of antibiotics has lead to increasing antibiotic resistance of bacteria, which is commonly recognized as a major threat to public health (24, 42). The unnecessary use of antibiotics could often be avoided if a reliable positive virus test result or negative bacterial test result was available at the point of care. It has been shown that rapid virus diagnosis is advantageous for both society and the patient, as it shortens hospitalization times, decreases unnecessary prescription of antibiotics, and speeds patient recovery (3, 28, 44). To cope with future challenges, new and improved analytical methods for point-of-care testing of infectious diseases are needed.

An important property of viral antigen detection methods is good specificity (19). A false-positive virus test result may lead to inappropriate medication and severe infection complications by other, undetected, disease-causing microorganisms (19). Precision requirements of virus assays are not strict since the tests are qualitative in nature; the patient is either infected or not infected. Moreover, methods for virus detection from respiratory specimens are prone to result variation due to inconsistent specimen composition and collection procedures (13). An ideal assay method for pathogen detection from respiratory specimens would be rapid, simple, and cost-effective. The method should also allow automated and quantitative result readout. So far, these objectives have not been reached by a single assay technique.

In this paper, we present a new immunoassay technique for rapid antigen detection of influenza A and B viruses. The new technique is based on a separation-free bioaffinity assay technique, ArcDia TPX, and the use of dry-chemistry reagents. The assay technique employs microspheres as a solid-phase reaction carrier, fluorescent antibody conjugates, and the detection of two-photon excited fluorescence from individual microspheres (10, 30, 41). The technique allows quantitative separation-free bioaffinity assays from a volume of a few microliters in subpicomolar sensitivity (15, 29). The applicability of the technique for the detection of serum antigens (10, 15) and antibodies (17), for the detection of antigens bound on the cell surface (32), for competitive binding assays (33), and for recognition of nucleic acid sequences (22, 35) has been demonstrated. The aim of the present study was to develop methods with short turnaround times for rapid point-of-care testing for influenza A and B viruses.

MATERIALS AND METHODS

Reagents.

ArcDia assay buffer consisted of 50 mM Tris-HCl, 150 mM NaCl, 0.01% Tween 20, 0.5% bovine serum albumin (fraction V), and 10 mM NaN₃ (pH 8). Dry-chemistry buffer consisted of 10 mM Tris-HCl, 50 mM NaCl, 10 mM NaN₃, 0.01% Tween 20, 0.5% bovine serum albumin, and 5% sorbitol (pH 8.0). Arctic Diagnostics (Turku, Finland) provided the fluorescent labeling reagents ArcDia BF 530 succinimidyl ester (20, 21) and the corresponding methyl ester (20, 21). Monodisperse, carboxyl-modified microspheres (diameter, 3.22 ± 0.08 μm; 11.3% [wt/vol]; 1.2 carboxyl acids/nm²) made of cross-linked polystyrene were purchased from Bangs Laboratories (Fishers, IN). Nanoparticles (copolymer of styrene and acrylic acid; diameter, 55 nm; solid content, 5%) were a generous gift from Harri Härmä (University of Turku, Turku, Finland) and were prepared as described previously by Huhtinen et al. (14). Anti-influenza virus monoclonal antibodies A3, A1, B2, and B4 (38, 39), as well as 5H and 6H (25), were obtained from the Department of Virology, University of Turku, Turku, Finland. Anti-influenza virus monoclonal antibodies of clone B2 were also obtained from Medix Biochemica Oy (Kauniainen, Finland), and antibodies 2/3 (immunization strain B/Beijing/184/93), IB633, and 1/22 were obtained from HyTest Ltd. (Turku, Finland). Purified influenza B virus preparation (strain B/Qingdao/102/91) was obtained from HyTest Ltd. Microtitration plates (384-well plate with black walls and a clear bottom) were obtained from Greiner Bio-One (Frickenhausen, Germany) (catalog no. 788096), and plate sealing film (adhesive PCR film) was obtained from Abgene House (Surrey, United Kingdom).

Clinical specimens.

Pretreated nasopharyngeal specimens (collected by aspiration or by swabs) (n = 65) were obtained from the specimen library of the Diagnostic Unit, Department of Virology, University of Turku, Turku, Finland. Influenza A virus-positive samples represented influenza virus subtypes of H1N1 and H3N2. The sample pretreatment technique was reported previously in the literature (39). In brief, the samples were diluted (1:5) in phosphate-buffered saline containing 2% Tween 20 and 20% newborn calf serum and sonicated to disrupt the mucus. During the study, each patient sample was measured at least four times with new two-photon excitation (TPX) methods.

Preparation of immunoassay reagents.

Microspheres were coated with monoclonal anti-influenza virus antibodies (clone A3 for influenza A virus and clone B2, B4, 5H, 6H, 2/3, IB633, or 1/22 for influenza B virus) by using passive coating and EDAC [1-ethyl-3-(3-dimethylaminopropyl)carbodiimide] fixation as described previously (41). Monoclonal anti-influenza virus antibodies (clone A1 for influenza A virus and clone B2, B4, 5H, 6H, 2/3, IB633, or 1/22 for influenza B virus) were labeled with succinimidyl ester of the fluorescent labeling reagent by using a method described previously (41). The average number of label molecules per antibody after conjugation (labeling degree) was determined by photometry similarly to a procedure described previously by Waris et al. (41). Nanoparticle tracer (NPT) for influenza B virus was prepared from nanoparticles, ArcDia BF 530 dye, and antibodies (clones B2, 2/3, IB633, 1/22, and B4) as described previously in the literature (31).

Assay principle of the TPX assay technique.

The new assay methods on the TPX detection platform followed the immunometric assay principle. The immunometric binding reaction takes place on the surface of monodisperse polystyrene microspheres, which have been coated with monoclonal antibodies specific for influenza A or B virus antigens. These microspheres work as a solid-phase reaction carrier for the immunocomplex formation. Fluorescence from the surface of the individual microspheres is measured separation free, directly from the reaction mixture, using the ArcDia TPX detection technique (29). The optical configuration of the fluorometer (15) and the physical phenomena related to the measurement technique were described previously (30).

Immunoassay procedure.

An assay reagent cocktail was prepared by mixing equal volumes of microparticle suspension (2 × 10⁷ pieces/ml in dry-chemistry buffer) and fluorescent tracer in dry-chemistry buffer (molecular tracer for influenza A virus, 8 nM; NPT for influenza B virus, 2.4 × 10¹³ nanoparticles/liter). The cocktail was dispensed (5 μl) in the assay wells of a 384-well format microtiter plate with Tecan dispensor automate MiniPrep 60 (Tecan Systems Inc., CA). Pretreated clinical nasopharyngeal specimens were thawed and bath sonicated for 20 s. The samples were diluted with assay buffer by factors of 3 and 15, if not otherwise stated. For wet-chemistry assays, the addition of the reagent into the assay wells was immediately followed by the manual addition of prediluted samples (15 μl), and the wells were sealed with a plate-sealing film. For dry-chemistry assays, the wells were evaporated to dryness in a desiccator over silica gel (22°C overnight). The prediluted samples were dispensed manually (20 μl) in the assay wells containing the reagents in a dry state, and the wells were sealed with the film. The wells were incubated at room temperature under continuous stirring (Eppendorf ThermoShaker at 1,400 rpm). The reaction mixtures in wells of 384-well plates were measured with an ArcDia TPX plate reader (PR6-001; ArcDia Ltd., Turku, Finland) using a measurement time of 25 s per well. During this measurement time, typically 20 to 40 individual microspheres were measured. Microsphere-specific data obtained from single assay wells were subjected to a data reduction algorithm similar to that described recently by Glotsos et al. (8). In order to compensate for the effect of fluorescent sample matrix components, the microsphere-specific signal was normalized to the solution fluorescence count. Analytical sensitivity of the method was defined as the signal level exceeding the negative control by three times the intra-assay standard deviation (3SD), whereas the limit for a positive test result (cutoff) was calculated from the interassay signal variation of negative samples.

Reference methods.

Time-resolved fluoroimmunoassay (TR-FIA) (also known as dissociation-enhanced lanthanide fluoroimmunoassay) methods for influenza A and B viruses were used as reference methods (25, 40). The TR-FIA technique has been extensively described and is known to provide subpicomolar sensitivity, which exceeds that of conventional colorimetric enzyme-linked immunosorbent assay (ELISA). The TR-FIA methods utilize clone A3 as the capture antibody and clone A1 as the tracer antibody for influenza A virus (38, 39) and clone B2 as both the capture antibody and the tracer antibody for influenza B virus (38, 39). Performance of the methods in comparison to ELISA and culture has been demonstrated in the literature (25, 40). Borderline samples were confirmed with PCR methods (similarly to methods described previously in references 6 and 18) for influenza A or B virus. The TR-FIA and PCR methods are in daily clinical diagnostic use in the Diagnostic Unit, Department of Virology, University of Turku, Turku, Finland.

Statistical analyses.

The microsphere-specific data were analyzed using an R package (version 2.2.1; R Foundation for Statistical Computing) based on a method described by Glotsos et al. (8). Pearson correlations were calculated using SAS Enterprise Guide (version 3; SAS Institute Inc.). Coefficients of variation (CVs) (percent) were calculated from sets of at least 18 repeated measurements (four to six different analyte concentrations).

RESULTS

Method development. (i) Method for influenza A virus.

The method for influenza A virus was optimized with respect to the labeling degree of the tracer and tracer concentration. The best performance (signal-to-background ratio, reaction rate, and lowest limit of detection) was obtained with a labeling degree of four fluorophores per antibody and with a final tracer reaction concentration of 1 nM (data not shown). The average signal imprecision (CV) of the optimized method was 8%. The method was not optimized with respect to antibody clones since the clones of the reference method (A3 as a capture antibody and A1 as a tracer antibody) also gave excellent method performance with the new technique.

(ii) Method for influenza B virus.

Assay methods for influenza B virus were constructed using antibody clones of B2, 2/3, IB633, and 1/22, all recognizing the nucleoprotein antigen, and 5H, 6H, and B4, all recognizing the hemagglutinin antigen. These clones were tested in all reasonable combinations as capture and as tracer antibodies using the molecular label approach. The reaction components were further optimized with respect to labeling degree and tracer concentration. The study revealed, however, that the performances of the new influenza B virus methods were markedly compromised compared to the reference method (data not shown).

In order to improve the performance of the influenza B virus method, a nanoparticle fluorescent tracer approach was used. NPTs were prepared by impregnating polymer nanospheres with fluorescent dye to obtain a label reagent with fluorescence intensity 100 times higher than that of the corresponding molecular label. The nanoparticles were then coated with antibodies and analyzed with photon correlation spectrometry. The average sizes (CVs) of the B2, 2/3, IB633, 1/22, and B4 NPTs were 210 nm (10%), 200 nm (8%), 180 nm (8%), 250 nm (5%), and 150 nm (7%), respectively.

The assay methods employing NPTs were optimized with respect to antibody clones and tracer concentrations (data not shown). The results showed that the highest signal-to-background ratios and reaction rates were obtained using clone B2 as the capture antibody and clone 2/3 as the tracer antibody in a final reaction concentration of 0.3 × 10¹³ pieces/liter (5 pM). The improvement in the signal-to-background ratios provided by the NPT reagent, compared to the corresponding optimized method using the molecular tracer, was in the order of 3- to 10-fold, depending on the analyte concentration and incubation time (see results in Fig. 1). The corresponding improvement in sensitivity was slightly compromised due to elevated signal imprecision (15% CV, on average) compared to that of the molecular tracer method (8% CV, on average) (see results in Fig. 1).

FIG. 1.

Dose-response curves for influenza B virus methods using molecular tracer (dashed line, 3SD of signals of negative control reactions) and NPT (dotted line, 3SD of signals of negative control reactions) at the 120-min time point.

Performance of the new method for the detection of influenza B virus was compared to the reference method using purified influenza B virus preparation as an analyte. Both methods showed a detection limit a dilution of approximately 1:20,000, corresponding to a concentration of 50 ng/ml.

(iii) Dry-chemistry reagent approach.

In order to allow simple assay protocols and to prolong the shelf life of the assay reagents, dry-chemistry reagents are widely used in commercial test kits. Recently, we introduced such dry-chemistry methodology to separation-free immunoassays on the TPX platform (16). In the present study, for the first time, we applied the same methodology to a method that employs NPT reagent. This was expected to be very challenging, as nanoparticle reagents are characterized by a tendency for aggregation. Any significant change in conditions of the NPT colloid can trigger the aggregation process. The performance of the dry-chemistry method was studied in comparison to that of a corresponding wet-chemistry method. Surprisingly, no difference in signal level, imprecision, or sensitivity between wet- and dry-chemistry methods was observed, as shown in Fig. 2. In order to evaluate the stability of the dry-chemistry reagents (for influenza A and B viruses), the dry-chemistry assay wells were stored refrigerated (+6°C) for variable times (0 to 12 weeks), after which they were used for the assays with clinical samples. The results of this study showed that at each time point and for each patient sample, the test indicated the same clinical result (negative or positive) as the original wet-chemistry assay (Table 1). This was the case, although the samples were freeze-thawed between the measurements at different time points, which tends to lower signal levels in the immunoassay. Thus, the results suggest that the storage time did not affect the performance of the dry-chemistry reagents.

FIG. 2.

Dose-response curves for influenza B virus NPT method using wet-chemistry (sample volume, 15 μl) and dry-chemistry (sample volume, 20 μl) reagents. Dashed and dotted lines are 3SD (of negative control reactions) levels of the wet-chemistry and the dry-chemistry methods, respectively.

TABLE 1.

Assay results for clinical samples using dry-chemistry assay methods for influenza A virus and influenza B virus at different time points of reagent storage

Sampleb	TPX method test resulta
	0 wk	6 wk	12 wk
Influenza A virus
1	++	++	++
2	+++	++	++
3	+++	+++	+++
4	++	++	++
5	+++	+++	++
6	+	+	+
7	+	+	+
8	+++	+++	+++
9	+++	+++	+++
10	+++	+++	+++
11	+++	+++	+++
12	+++	+++	+++
13	+++	+++	+++
14	+++	+++	+++
15	++	++	++
16	++	++	++
17	++	++	++
18	+++	+++	+++
19	+++	+++	+++
20	++	++	++
21	+++	+++	+++
22	+++	+++	+++
23	++	+++	++
24	++	+++	+++
25	+++	+++	+++
26	+++	+++	+++
27-59	−	−	−
Influenza B virus
1	−	−	−
2	+/−	+/−	+/−
3	−	−	−
4	++	+	++
5	++	++	++
6	+++	+++	+++
7	+++	+++	+++
8	+++	+++	+++
9	+++	+++	+++
10	+++	+++	+++
11	+++	+++	+++
12	+++	++	++
13	+++	+++	+++
14	+++	+++	+++
15	+++	+++	++
16	+/−	+	+/−
17	+++	+++	+++
18	+++	+++	+++
19	+++	+++	+++
20	+++	+++	++
21	+++	+++	+++
22	+++	+++	++
23	+++	++	+++
24	++	+++	++
25	+++	+++	++
26	++	++	++
27	+++	+++	+++
28	+++	+++	+++
29	++	++	++
30-65	−	−	−

^a+, ++, and +++ correspond to low-positive, positive, and high-positive samples defined as signal in the ranges of interassay 3SD to 300, 300 to 1,000, and over 1,000 TPX units, respectively. +/− corresponds to samples which exceeded the detection limit but fell below the method-specific cutoff.

^bSamples detected as being borderline by TR-FIA are marked in boldface type.

Correlation and method performance studies.

The results obtained from clinical samples with the new TPX methods were compared to those of the TR-FIA reference methods. For influenza A virus, 24 positive, 2 borderline, and 33 negative samples were analyzed, and for influenza B virus, 25 positive, 4 borderline, and 36 negative samples were analyzed with both techniques (TPX and TR-FIA). A borderline sample is defined as a sample which gives in the TR-FIA reference method a signal within the range of 0.5 to 2 times the cutoff level set for positive detection. In this study, all borderline samples were subjected to PCR confirmation. The positive samples were cross-analyzed with the TPX methods to reveal possible false-positive results caused by cross-reactions between influenza A and B viruses.

The linearity of the assay methods for influenza A and B viruses was studied using a dilution series from highly positive clinical samples (three for both influenza A and influenza B viruses) or purified influenza B virus. Both methods showed excellent linearity (R² ≥ 0.99) and a dynamic assay range of 3 orders of magnitude (Fig. 1). During the study, 15 hemolytic and 5 turbid samples were analyzed, and the new methods provided same clinical result for these samples as the reference TR-FIA methods. In addition, these samples located close to the mean line in TPX/TR-FIA correlation plots (Fig. 3 and see Fig. 5). Signal levels of the negative samples were measured at different incubation time points, while, typical for the TPX assay technique, neither signal levels nor imprecision was affected by the incubation time (data not shown).

FIG. 3.

Scatter plot presenting the correlation between the TPX method (x axis, arbitrary signal units) and the TR-FIA method (y axis, TR-FIA units) for influenza A virus. The figure shows signals of influenza A virus-positive and -negative samples and signals given by influenza B virus-positive samples (for the cross-reactivity study, all signals were below the cutoff level). The total number of samples is 59. Dashed lines present the cutoff values set for a positive test result. The sample on the bottom right () was negative by TR-FIA but positive by TPX and PCR. The sample at the middle left (×) was positive by TR-FIA but negative by TPX and PCR.

FIG. 5.

Scatter plot presenting the correlation between the TPX method (x axis, arbitrary signal units) and the TR-FIA method (y axis, TR-FIA units) for influenza B virus detection. The figure shows signals of influenza B virus-positive and -negative samples and signals given by influenza A virus-positive samples (for the cross-reactivity study, all signals were below the cutoff level). The total number of samples is 65. The dashed lines present the cutoff values set for a positive test result.

(i) Method for influenza A virus.

The correlation between the new assay method (TPX, at an incubation time of 2 h) and the reference method (TR-FIA) for influenza A virus is shown in Fig. 3. The correlation between the methods was good, with a Pearson correlation coefficient of 0.92 (P < 0.0001). All samples that tested positive or borderline with the reference method deviated from negative samples with the new method (interassay 3SD rule cutoff). One of the TR-FIA-negative samples gave repeatedly positive results by TPX (sample marked with an asterisk in Fig. 3). PCR testing of this sample also gave a positive result, and thus, the repeatedly obtained negative TR-FIA result was considered to be a false-negative result. One of the samples that was borderline with TR-FIA was PCR and TPX negative, and thus, the repeatedly obtained positive TR-FIA result was considered to be a false-positive result (sample marked with × in Fig. 3).

The kinetic measurement option of the TPX technique (separation-free detection) provides an easy means for determinations of performance as a function of time. Here, we determined the clinical sensitivity of the TPX method relative to that of the reference method (constant assay time of 2 h) at different time points of incubation. At 20 min, 23 out of 24 positive samples (3SD criteria) were determined to be positive by the new method, while two borderline samples remained negative at this time point. At 60 min, all of the positive and borderline samples were deemed positive (Fig. 4). The clinical sensitivity (including borderline samples) of the new method at 20 and 60 min in comparison to the TR-FIA method (PCR confirmed) were 88 and 100%, respectively. In addition, one extra positive sample was found by the TPX method. The correlation between assays carried out with wet and dry chemistries was excellent (Pearson correlation coefficient of 0.97; P < 0.0001), and both approaches indicated the same clinical result for each patient sample. None of the influenza B virus-positive (cross-reactivity study) or -negative samples gave positive results by the influenza A virus method (for data, see Fig. 3), indicating specificity close to 100% (n = 33). Intra-assay imprecision was studied by repeated measurement (n = 20) of the clinical samples. Samples in the low end of the response curve (n = 4) showed CVs between 6 and 13%, samples in the mid range (n = 4) showed CVs between 7 and 11%, and samples in the high end (n = 4) showed CVs between 7 and 16%.

FIG. 4.

Performance of the TPX method for influenza A virus detection at the 60-min time point. +, negative samples (n = 33); , borderline samples by TR-FIA (n = 2); ×, positive samples (n = 24). The dashed line presents the cutoff value set for a positive test result.

(ii) Method for influenza B virus.

The correlation between the new assay method (TPX) and the reference method (TR-FIA) for influenza B virus is shown in Fig. 5. The correlation between the methods was good, with a Pearson correlation coefficient of 0.82 (P < 0.0001). Positive and negative samples deviate from each other by the TPX method. At 25 min of incubation, 25 out of 25 positive samples were determined to be positive, and four borderline samples (called samples a, b, c, and d hereafter) were determined to be negative. At the time point of 120 min, two of the four borderline samples (samples a and b) barely cut the analytical detection threshold (intra-assay 3SD) set for positive detection in TPX, while these samples did not reach the method-specific cutoff (interassay 3SD) (results shown in Fig. 6). One (sample b) of the two samples was found to be positive by PCR. The sample (sample c) that was negative by TPX was found to be positive by PCR. The other two borderline samples (samples a and d) gave an uninterpretable result by PCR, as the PCR products showed different retention by gel electrophoresis compared to the band of the positive control. Sample a was deemed positive based on TR-FIA and TPX (false-negative result by PCR), whereas sample d was deemed negative based on TPX and PCR (false-positive result by TR-FIA). Hence, the clinical sensitivity of the new method at 25 min of incubation was 89%. In addition, two samples, samples a and b, cut the threshold set for the analytical detection limit (intra-assay 3SD) at the 120-min time point. The correlation between assays carried out with dry and wet chemistries was excellent (Pearson correlation of 0.96; P < 0.0001), with both approaches indicating the same clinical result for each patient sample. None of the influenza A virus-positive (cross-reactivity study) or -negative samples gave a positive result by the influenza B virus method (for data, see Fig. 5), indicating specificity close to 100% (n = 36). Intra-assay imprecision was studied by repeated measurement (n = 20) of the clinical samples. Samples in the low end of the response curve (n = 4) showed CVs of 22 to 26%, samples in the mid range (n = 4) showed CVs of 8 to 22%, and samples in the high end (n = 4) showed CVs of 6 to 11%.

FIG. 6.

Performance of the TPX method for influenza B virus detection at the 120-min time point. +, negative samples (n = 36); , borderline samples by TR-FIA (n = 4); ×, positive samples (n = 25). The dashed line presents the cutoff value set for a positive test result.

DISCUSSION

In the laboratory diagnosis of influenza virus infections, antigen detection and PCR are preferred over serological methods and virus isolation (1, 6). Despite improvements in PCR-based techniques and their automation during the last decade, molecular biology methods have not made a clear breakthrough in the way they were expected to. The inherent limitations of PCR-based techniques were recently discussed (37, 45). Although PCR has the potential for higher sensitivity than immunoassay techniques, including TPX, point-of-care applications of PCR-based methods do not currently look feasible due to their technical complexity and high assay cost.

In this paper, we have described new rapid immunoassay methods for influenza A and B virus antigen detection that are based on the ArcDia TPX technique and the use of dry-chemistry reagents. The new technique is characterized by an extremely simple (one-step) and separation-free assay procedure, and it was shown to provide sensitivity and specificity comparable to those of TR-FIA reference methods and excellent method-to-method correlation. When considering the nature of collection of the respiratory specimen, and the precision requirements for virus antigen detection in general, the precision obtained with the new methods can be considered to be excellent. The results of this study show that the new assay technique fulfills the requirements for an efficient rapid assay method to detect virus antigens in respiratory specimens. According to the result of the present paper and those reported in our previous study (16), the new technique seems to be adaptable to point-of-care settings, still providing performance comparable to that of centralized laboratory methods.

The assay method for influenza A virus was constructed on the TPX platform by using the same nucleoprotein antibody clones as those used in the TR-FIA reference method. This antibody pair provided good assay performance on both method platforms. Development of the assay method for influenza B virus, however, was less straightforward. In order to improve the performance of the assay method for influenza B virus, several antibody clones were tested using the molecular label approach (clones B2, 2/3, IB633, 1/22, 5H, 6H, and B4) (data not shown). In our study, the best results were obtained by using B2 as the capture antibody and B2 or 2/3 as the tracer antibody. Overall, nucleoprotein antibodies showed better assay performance than hemagglutinin antibodies. Despite the extensive optimization work, the molecular label approach failed to reach sufficient clinical sensitivity required for influenza B virus testing.

It has been reported repeatedly in the literature that methods for influenza B virus antigen detection are characterized with lower sensitivity than corresponding methods for influenza A virus detection (19, 26). This was also the case in our study. This suggests that antigen and/or virus expression or secretion levels associated with influenza B virus infections are lower than those of influenza A virus infections. Prevalence of human anti-influenza antibodies in nasal secretions (5, 7) could also explain the low signal levels for influenza B virus due to competition with reagents. In the case of influenza A virus, however, nasopharyngeal specimens gave normal immunoassay signal levels. This finding refutes the notion of the prevalence of competing antibodies, as nasal secretions in influenza A and B virus infections probably contain comparable amounts of anti-influenza virus immunoglobulins.

To overcome the sensitivity limitations of the molecular label, fluorescent nanoparticle labels have been developed for several applications (12, 27). The use of NPT in relation to the separation-free TPX assay technique has also been reported (15). The use of NPTs can enhance the fluorescent signal yield per three-component bioaffinity complex to several orders of magnitude. Furthermore, each fluorescent nanoparticle is coated with tens to thousands of antibodies, leading to a reagent of multivalent binding properties. This multivalency can result in an increased binding degree, as demonstrated in the literature (15, 31, 36).

In pursuit of improved sensitivity, we applied the nanoparticle approach to influenza B virus antigen detection using the TPX platform. Clones B2, 2/3, IB633, 1/22, and B4 were tested as NPTs in combination with variable antibody-coated surfaces. The pair B2 (capture antibody) and 2/3 (tracer antibody) provided the best assay results. On average, the nanoparticle approach resulted in several-times-higher signal levels and signal-to-background ratios (up to 10 times) than the molecular label approach. The multivalent binding properties (avidity) of NPT seem to enhance the tracer binding degree since the optimal NPT concentration is remarkably lower (5 pM) than the optimal molecular tracer concentration (1 nM). In conclusion, the results show that the NPT can also successfully be used with the dry-chemistry approach to enhance the sensitivity of antigen detection. The use of NPT was not considered to be necessary for the influenza A virus method, since the dynamic range of the method with the molecular tracer matched well with the concentration range of the influenza A virus antigens in clinical samples.

We studied the specificity of the methods by cross-analyzing influenza A and B virus-positive samples. As shown by the data in Fig. 3 and 5, no cross-reaction was found. This result is in line with results described previously by Walls et al. (38), who studied the cross-reactivity of antibody clones A3, A1, and B2 by testing for cross-reactivity with influenza A and B viruses, parainfluenza virus types 1, 2, and 3, adenovirus, and respiratory syncytial virus. Walls et al. did not observe cross-reactivities. On the other hand, we have found that this particular antibody pair for influenza A virus also detected inactivated H5N1 virus nucleoprotein antigen (our unpublished data).

Most laboratory methods with an automated result readout (e.g., TR-FIA or ELISA) must apply separation steps (washes), which inevitably lead to multistep assay protocols and fixed incubation times. The performance of such methods is a compromise between the sensitivity and the turnaround time. In contrast to conventional laboratory methods, the TPX assay technique enables on-line monitoring of reaction kinetics (11, 15, 17). This makes the determination of a positive detection rate as a function of incubation time straightforward. With the new technique, highly positive samples can be detected reliably after a few minutes of incubation, and the positive test results are thus rapidly available for clinical diagnosis by the physician. In case the sample shows a negative result at this time point, the incubation is continued, and the reaction is measured again at a later time point. The kinetic measurement option thus allows the rapid (in 10 to 20 min) identification of moderately positive and highly positive samples, still not compromising the overall sensitivity of detection. Moreover, the new technique is characterized by an automated result readout, which allows connections to laboratory databases and networks (19, 43). In case virus subtype information is required, the test can be modified by implementing subtype-specific antibodies, or alternatively, aliquots of positive samples can be referred to a specialized laboratory for subtyping. Follow-up of negative samples by nucleic acid amplification methods can be applied if considered necessary.

In primary health care, the identification (or exclusion) of the pathogen causing respiratory symptoms (diagnosis) should preferably be done during the same patient visit. In order to make the laboratory diagnosis cost-efficient, an integrated multianalyte testing panel is needed. Such a testing panel would comprise reagents for six to nine different bacterial and viral pathogens, which would be sufficient to cover a significant proportion of respiratory infection cases. The use of such a diagnostic product in primary health care would enable the differentiation between bacterial and viral infections, enable correct treatment, and decrease the unnecessary use of antibiotics.

Neither conventional laboratory methods nor modern high-throughput random-access analyzers allow rapid and cost-effective multianalyte point-of-care testing (2, 16, 34). The detection platform presented here, in contrast, seems to fulfill the given requirements for point-of-care testing. The challenge associated with the development of the respiratory disease panel relates to specimen collection and pretreatment methods. The same specimen type may not be optimal for detecting viral and bacterial antigens. In continuation of this work, we aim to study the applicability of the TPX assay technique to the testing of respiratory infections in the above-described panel format.