The Panther Fusion System with Open Access Functionality for Laboratory-Developed Tests for Influenza A Virus Subtyping

Nucleic acid amplification tests, such as PCR, are the method of choice for respiratory virus testing, due to their superior diagnostic accuracy and fast turnaround time. The Panther Fusion (Fusion; Hologic) system has an array of highly sensitive in vitro diagnostic (IVD) real-time PCR assays for respiratory viruses, including an assay for influenza A (FluA) virus, influenza B (FluB) virus, and respiratory syncytial virus (RSV) (FFABR assay). The Fusion system has Open Access functionality to perform laboratory-developed tests (LDTs) alongside IVD assays.

I nfluenza viruses are classified into three distinct types, A, B, and C, with subdivisions in type A based on antigenic characterization of the surface glycoproteins. Currently circulating strains of influenza A (FluA) virus include H3N2 and H1N1pdm09. Since the 2009 influenza pandemic, H1N1pdm09 quickly became the dominant H1 strain, displacing the former seasonal H1N1 (fsH1) strain; however, rare occurrences of fsH1 have been observed. This is fortunate, because the fsH1 strain had become universally resistant to oseltamivir (Tamiflu). The clinical significance of the H3N2 strain is its association with higher attack rates, hospitalization rates (1), and mortality (2,3), as well as the suboptimal vaccine efficacy against this strain (4). Because coinfection with multiple strains can occur, influenza strain typing is important for proper assignment of patients to cohorts. Furthermore, FluA virus subtyping not only has therapeutic and clinical benefits; it is an important epidemiological tool with public health benefits.
Panther Fusion (referred to below as Fusion) is a fully automated high-throughput system with on-demand testing capabilities. This system integrates nucleic acid extraction with either transcription-mediated amplification (TMA) or real-time PCR. The in vitro diagnostic (IVD) respiratory virus assays on the Fusion instrument, which include the FluA-FluB-respiratory syncytial virus (FluA/B/RSV [(FFABR]), Paraflu, and adenovirus-human metapneumovirus-rhinovirus (AdV/hMPV/RV) assays, have been shown to have exquisite sensitivity (5). Furthermore, the Fusion system has Open Access functionality, which enables the performance of laboratory-developed tests (LDTs) with full automation. Not only can LDTs be run alongside IVD tests; they can be run from the same nucleic acid extract. Our laboratory developed and compared two LDTs on the Fusion instrument for the subtyping of FluA virus, to complement the respiratory virus assays. The results were compared to those of the FFABR assay for sensitivity in viral detection and to those of the Prodesse ProFASTϩ (PFAST) assay for subtyping accuracy. Furthermore, the clinical specimens used were analyzed previously by the following methods, in addition to the FFABR and PFAST assays: the cobas Influenza A/B (cIAB) test, FilmArray Respiratory Panel 1.7 (RP), and the Xpert Flu (Xpt) assay (5,6). The PFAST and RP assays provided subtype analysis for H3, 09H1, and fsH1 strains, while the Xpt assay provided a 09H1 call-out result. This sample set provided insight into the sensitivities and accuracies of the LDTs relative to those of commercial assays.
(This study was presented in part at both the 2019 ASM Clinical Virology Symposium and ASM Microbe 2019.)

Specimens.
Clinical specimens comprised 110 nasopharyngeal swab (NP) specimens, in 3 ml of viral transport medium (VTM), previously determined to be positive for FluA virus by the FFABR assay (Hologic, San Diego, CA). FluA virus subtypes had been determined by the PFAST assay (Hologic) for 104 specimens. These samples were used in previous studies comparing influenza virus PCR assays (5, 6) and were also analyzed by the following methods: the cIAB test (Roche Diagnostics, Indianapolis, IN), RP (BioFire, Salt Lake City, UT), and the Xpt assay (Cepheid, Carlsbad, CA). Of the six nontyped FluA virus specimens, two had been FluA virus positive by both the cIAB and FFABR tests and four by the FFABR assay only. Specimens were stored at -80°C but had undergone multiple freeze-thaw cycles.
Viruses in VTM and the dilutions tested six to eight times for limit-of-detection (LOD) and precision analyses. The 1:10 dilutions for the fsH1, 09H1, and Swiss strains were also diluted 1:4 to meet assay validation requirements in New York State. These dilutions were tested simultaneously by the LDT-FAST, exWHO-FAST, and FFABR assays, using the same nucleic acid extract. Determinations of viral nucleic acid concentrations were based on quantified control viral RNA (Hologic, Inc.) using either the Prodesse ProFluϩ (Hologic) or the PFAST assay, depending on strain/clade-based amplification efficiency (6). The Swiss, 09H1, and fsH1 strains were combined at concentrations of 3.77, 4.83, and 3.81 log 10 copies/ml, respectively, for use as a positive control and for additional precision and reagent stability studies.
Test methods. The test methods investigated in this study are listed in Table 1. Nucleic acid extraction and amplification. The LDTs were performed on the fully automated Panther Fusion system with Open Access functionality. This instrument utilizes universal nucleic acid extraction reagents and adds internal control (IC) RNA at the initiation of extraction. Following the standard protocol for respiratory virus detection, 500 l of specimen was added to a Panther Fusion specimen lysis tube containing 750 l buffer and 360 l of the mixture used for extraction. The nucleic acid was subsequently eluted into 50 l, and 5 l was amplified for an effective sample volume of 14.4 l.
High-performance liquid chromatography (HPLC)-purified primers and probes for the LDT-FAST test (Biosearch Technologies, Petaluma, CA) were combined with 8 mM Tris (pH 8.0; Hologic), 4 mM MgCl 2 (Hologic), 50 mM KCl (Hologic), and RNA-IC primers and probes (Hologic), and the mixture was overlaid with Panther Fusion oil reagent (Hologic). exWHO-FAST (Integrated DNA Technologies, Inc., Coralville, IA) ( Table 2) primers and probes were combined with the same concentrations of Tris and KCl and with 4.5 mM MgCl 2 , and the mixture was overlaid with oil. The Fusion system rehydrates the enzyme and nucleotide pellet with the primer, probe, and buffer mixture and adds it to the nucleic acid extract in 25-l reaction mixtures. Real-time reverse transcriptase PCR (RT-PCR) was performed using the following amplification profile: 1 cycle of 8 min at 46°C, 1 cycle of 2 min at 95°C, and 45 cycles of 95°C for 5 s, with 60°C for 22 s.
The primers and probes from both the WHO's one-step duplex real-time RT-PCR (Annex 2B, Protocol 2) and its one-step real-time RT-PCR for the H1 gene of influenza A(H1N1)pdm09 virus (Annex 2B, Protocol 1) (7) were modified to have melting temperatures (T m ) more compatible with the Fusion system's optimal amplification profile. Based on assay optimization studies, it was determined that the the modified H3 primers and probe from WHO Protocol 2 and the modified H1N1pdm09 primers and probe from WHO Protocol 1 demonstrated superior performance in combination (data not shown). This multiplex was expanded into a triplex to include a target for the former seasonal H1 (fsH1) strain for a complete FluA virus strain-typing assay.
HA sequence analysis. For in silico analysis of primer and probe homology, hemagglutinin (HA) gene sequences were obtained from the Global Initiative on Sharing Avian Influenza Data (GISAID) EpiFlu database. This database comprises influenza sequences for the semiannual vaccine strain selection uniquely submitted from contributors such as the World Organisation for Animal Health (formerly the Office International des Epizooties), national reference laboratories, and all WHO Collaborating Centers for Surveillance, Epidemiology, and Control of Influenza (13). Included were all unique human isolate sequence data from the United States available as of 28 June 2019 but collected during the following periods: H3 and 09H1 isolates from 1 September through 30 April 2019, sH1 isolates from 1 September Statistical analysis. Sensitivities and confidence intervals (CI) were determined using Microsoft Excel 2016 (Microsoft, Redmond, WA) (9). Probit analyses for the LOD with a 95% probability of detection were performed using SPSS, version 13.0 (IBM, Armonk, NY).

RESULTS
Of the 104 previously subtyped FluA virus-positive clinical samples, 54 were H3, 46 were 09H1, and 4 were fsH1. All samples were appropriately subtyped by both the LDT-FAST and the exWHO-FAST assay (Table 3). Of the six previously untyped samples, three were subtyped as H3 by both assays. Another two were subtyped by the LDT-FAST test only as H3, and one sample was not subtyped by any method. The original threshold cycle (C T ) values with the FFABR assay were 37.1 and 39.7 for the two samples subtyped by the LDT-FAST test only. The original FFABR assay C T value for the sample not subtyped by either LDT was 37.9. This sample was retested with the FFABR assay and was found negative upon repeat testing. Although these samples were not previously subtyped, they were all collected during a 2-week period in 2015 when the incidence of FluA virus infection was 25% and the subtype was exclusively H3 (5, 6). Based on this information, the sensitivities were 100% (95% CI, 100%) and 98.2% (95% CI, 95.6% to 100.7%) for the LDT-FAST and exWHO-FAST assays, respectively. Furthermore, these assays demonstrated higher sensitivity than the cIAB, PFAST, RP, and Xpt assays based on results from previous studies with the same samples (6). The LODs for the LDT-FAST assay were comparable to those for the FFABR assay with all FluA virus isolates tested, including the currently circulating 09H1 isolate and the H3 clade 3C.3a isolates. The LODs for the exWHO-FAST assay were comparable to those for the FFABR and LDT-FAST assays with the 09H1, sH3, and H3 clade 3C.1 isolates but were higher with the clade 3C.3a and H3N2v isolates ( Table 4). The LOD studies were also designed to demonstrate the intra-and interassay reproducibilities; three different concentrations of the samples were tested multiple times in one run of testing and were then retested on different days. The coefficients of variation (CV) for the intra-assay and interassay C T values, for individual targets, across three concentrations of virus, ranged from 0.8% to 2.7% with the LDT-FAST assay and from 0.7% to 2.7% with   the exWHO-FAST assay (Table 5). Similarly, for the positive control in which targets are combined, the CV ranged from 2.7 to 2.9% for the LDT-FAST assay and from 0.8 to 1.0% for the exWHO-FAST assay over 20 days of testing. Since the same master mix preparations were on board the Panther Fusion instrument and were used for the duration of the 20-day reproducibility study, this analysis also demonstrated that the master mixes were stable onboard for at least 20 days. Because LOD performance appeared to be strain dependent, primer/probe homology was analyzed against circulating strains. Unique isolates from the United States uploaded to GISAID were aligned and the number of mismatches counted for each primer and probe. More than 97% of circulating H3 isolates demonstrated exact homology with the LDT-FAST primers and probes, and there was not a significant difference upon analysis by clade (Table 6). However, the exWHO-FAST assay demon-  strated a number of mismatches that were clade associated. H3 primer/probe sets for both assays demonstrated mismatches with the H3N2v strains, as expected, since these targets were not designed to detect these isolates.

DISCUSSION
The Panther Fusion respiratory virus IVD assays (the FFABR, Paraflu, and AdV/ hMPV/RV assays) are a modular approach to syndromic testing on a fully automated platform. This system and associated IVD assays have numerous benefits, such as exquisite sensitivity and onboard reagent stability. However, this system does not include a FluA virus subtyping assay, which is often part of a respiratory virus panel.
With the Open Access system, LDTs can be used alongside IVD assays, with the same full automation. Our laboratory developed and compared two LDTs on the Fusion instrument for the subtyping of FluA virus: one assay using Hologic's proprietary primers and probes from their Prodesse ProFASTϩ assay and one using modified versions of primers and probes recommended by the WHO (7).
Both subtyping LDTs demonstrated clinical sensitivities comparable to that of the FFABR assay and were superior to many other commercial assays for FluA virus detection. The sensitivities of the cIAB, PFAST, RP, and Xpt assays in this sample set were 91%, 94%, 90%, and 85%, respectively, also by comparison to the FFABR assay. The sensitivities of 100% and 98.2% for the LDT-FAST and exWHO-FAST assays, respectively, were an impressive finding, because these samples had been subjected to repetitive freeze-thaw cycles for previous studies with the other assays. Another important point is that 97% of the H3 samples were taken from a previous prospective study involving consecutive specimens received in the lab during a 2-week period in the winter of 2015 (6). Analyses with such study populations predict assay sensitivities better than analyses of selected samples. During this 2-week period, the incidence of FluA virus was 25% and the subtype was exclusively H3. Indeed, the six previously untyped samples were collected during this time, and five of the six were typed as H3 by at least one of the typing LDTs. Initially, it appeared that there was one falsenegative sample by both typing assays. However, repeat FFABR testing on the aliquot used for the subtyping assays was negative, indicating that either the virus titer was near the limit of detection for all three assays or the viral RNA had degraded.
We developed two assays due to an initial lack of certainty that our lab would continue to have access to the proprietary sequences from Hologic. Although both subtyping assays demonstrated sensitivities superior to those of many other commercial assays for FluA virus detection, the LDT-FAST assay was a little more sensitive than the exWHO-FAST assay. Interestingly, the FluA virus samples missed by the exWHO-FAST assay were from the 2-week period in 2015 in which the FluA virus population consisted primarily of clade 3C.2a viruses (10,11). This clade had at least one mismatch with the exWHO-FAST forward primer and probe. Indeed, the WHO subsequently published an update to its recommended primers and probes for influenza virus detection, with an alternate target region for H3 viruses, for enhanced coverage (12). Despite the mismatches seen in the H3 primer and probe regions, the exWHO-FAST assay was still more sensitive for the H3 target than many other commercial assays. However, genetic drift does present a problem for any FluA virus typing system, because the HA gene is under the greatest environmental (immune) pressure to drift. It is necessary to watch diligently for frequent untypeable influenza virus-positive specimens, particularly if the C T values with the diagnostic FluA test are low, since assay modification may be needed. Fortunately, the risk associated with a false-negative typing result (FluA virus positive, subtype unknown) is minimal and is not an uncommon result with other IVD assays.
The value of including primers and probes for subtype fsH1, at an added cost of approximately $1, is certainly questionable. Our lab chose to include them because we did have two cases in 2018. Similarly, one may assert that viral clade determinations are needed, particularly since vaccine efficacy can differ with circulating viral clades, as seen with influenza virus A/H3 and FluB clades. However, the value in a clinical lab is currently minimal. Also, with regard to value, the benefit of running a subtyping assay concurrently versus sequentially with a FluA diagnostic assay would have to be addressed at each institution. Many of the experiments in this study were run concurrently, providing proof of concept that an LDT can be performed side by side with an IVD assay on the same sample and extract. It should also be noted that the current instrument software allows for only three assays from the same extract; however, Hologic is currently in the process of updating the system to allow for five assays per extract.
Besides the exquisite sensitivity seen with these FAST assays, the automation associated with the Fusion system has numerous other benefits. First, this is a fully automated process from the placement of the sample on the instrument to the reporting of results, which is certainly an improvement and reduces labor needs significantly. The instrument is designed for high-throughput testing with on-demand testing capabilities, and it complements Hologic's modular approach to syndromic respiratory disease testing. Last, the system has reagent and consumable tracking with advance warning when more is needed.
In conclusion, both FluA virus typing assays were successfully adapted to run on the Panther Fusion instrument, with sensitivities comparable to that of the FFABR assay, providing a valuable complement to the Panther Fusion respiratory menu. This study also demonstrated proof of concept that LDTs and IVD assays can be processed side by side with full automation from sample to answer, either from the same sample or from the same eluate.