Introduction
Human rhinoviruses (HRV) are small, positive-sense RNA viruses within the family Picornaviridae. Over 150 genotypes of this important human pathogen have been recognized within species HRVA, HRVB, and HRVC of the genus Enterovirus (http//:www.picornaviridae.com). HRV infections occur throughout the year and throughout the world. HRV are responsible for a high incidence and wide range of respiratory infections in all populations, including one-half to two-thirds of all common colds and many cases of otitis media and sinusitis in the upper respiratory tract. Lower tract infections include bronchiolitis, pneumonia and exacerbation of disease in children with asthma and cystic fibrosis, and in adults with chronic obstructive pulmonary disease. Cases of serious illness and even death due to HRV have been reported, especially in immunocompromised individuals, the elderly and infants [1, 2].

Laboratory detection of HRV is important for diagnosis and surveillance, especially in high risk populations. HRV are frequently detected as co-infections with other respiratory viruses and in individuals with long-term and asymptomatic shedding [3]. In addition to qualitative detection, accurate quantification of HRV RNA in clinical samples is needed for studies on the association of HRV viral load with viral transmission and with patient symptoms and outcomes. Viral-load studies of other respiratory viruses have shown that a correlation exists between quantity of virus and disease severity. HRV viral-load determinations may also be important for patient management, especially in asymptomatic patients who test positive for HRV at low levels. More importantly, accurate HRV viral-load assessments will be necessary for evaluating the performance of potential HRV antiviral drugs [4].

Detection
HRV were initially detected by growth in cell culture. Approximately 100 serotypes of HRV grown in cell culture were antigenically characterized by their reactions with various antisera. The serotypes were subsequently classified into two groups, A and B, according to their sensitivity towards antivirus agents [5] and are now included in HRV species A (80 genotypes) and B (32 genotypes) based on genetic sequencing. Cell culture is sensitive for detection of many, but not all HRV genotypes; 55 HRV that do not grow in the cell culture lines normally used in the clinical laboratory and have been detected only by molecular methods are classified in HRV species C (http//:www.picornaviridae.com).

The use of molecular methods for the detection of HRV in clinical specimens has provided more accurate information about the disease burden and epidemiology of these ubiquitous viruses. The molecular method most often used to detect HRV is real-time reverse-transcription (RT)-PCR [3]. RT-PCR assays, when accompanied by amplification of serially diluted standards of known RNA copy numbers (RT-qPCR), can be used to quantify the number of viral copies in a sample. By comparing the PCR Ct value (the PCR cycle at which fluorescence reaches a certain threshold) of a clinical specimen to the standard curve, the relative quantity of the analyte can be calculated [6].
Within the HRV genome, the region most frequently targeted for RT-PCR by clinical assays is the 5’ non-coding region (NCR), which exhibits the most sequence homology among the HRV genotypes. However, even in this region, there is a lot of sequence diversity, which makes it challenging to design a single, consensus PCR primer and probe set to amplify all HRV genotypes with equal efficiency. In order to amplify HRV genotypes with diverse sequences in the prime/probe binding regions, consensus PCR primer and probe sets have been designed with degenerate and modified bases or multiple oligonucleotides [7–10]. However, consensus RT-qPCR assays may not give accurate quantitative results for all HRV genotypes due to amplification inefficiency caused by base mismatches between the consensus primers and probe and the viral sequences [11].

Quantitation by RT-qPCR
To determine if a consensus RT-qPCR assay [7] could be used to accurately quantify all genotypes of HRV, including those with sequence differences in the primer and probe binding regions, we compared the efficiency and sensitivity of a consensus RT-qPCR assay to that of genotype-specific RT-qPCR assays [4]. In Figure 1(a), the results of RT-qPCR assays using type-specific primers and probes, which exactly match the target sequences, show standard curves indicating accurate and sensitive quantification of RNA transcripts from six specific HRV genotypes. However, RT-qPCR using a consensus HRV primer and probe set did not give accurate or sensitive quantification for some HRV genotypes, especially types A33 and A88 (Fig. 1b). RNA from HRV genotypes with base mismatches between the consensus primer and probe sequences and the specific viral sequences was inaccurately quantified using the consensus assay, most likely due to poor amplification efficiency.

Quantitation by RT-dPCR
Digital RT-PCR (RT-dPCR), which provides absolute nucleic acid quantification without the need for PCR Ct values and standard curves and is less affected by poor amplification efficiency, may perform better than RT-qPCR for quantification of HRV RNA. In dPCR, an amplification reaction, which contains fluorescent dye to measure amplification, is divided into 12?000 to 200?000 independent partitions, each ideally containing no more than one target molecule. The reaction is amplified to end point and the number of fluorescent (positive) and non-fluorescent (negative) partitions is counted. In specimens with more targets than partitions, Poisson statistics are used to calculate the average number of targets per positive partition and thus, the number of targets in the original sample [12, 13]. Compared to qPCR, dPCR is less susceptible to amplification inefficiency caused by primer/probe sequence mismatches because quantification derives from a PCR reaction that cycles to endpoint rather than from an amplification curve as in qPCR. Accurate quantification by dPCR is also not dependent on a well-calibrated standard [14]. These characteristics make dPCR especially useful for quantifying viral targets with many subtypes and high sequence diversity that leads to mismatches between targets and PCR primer and probe sequences, such as HRV.
To determine if consensus RT-dPCR would perform better than consensus RT-qPCR for quantification of HRV genotypes, we similarly tested RNA transcripts of HRV genotypes, including some with sequence variation in the consensus primer and probe binding region, by RT-dPCR using both type-specific and consensus primers and probes. In Figure 2(a), the results of RT-dPCR assays using type-specific primers and probes show good correlations between the expected number of RNA copies/reaction and the observed number. When amplified by RT-dPCR using the consensus assay (Fig. 1b), in contrast to RT-qPCR, the observed number of RNA copies/reaction was also closely correlated with the expected number for most of the HRV genotypes tested.

In a previous study [4], data from 16 HRV genotypes that represented the consensus primer and probe binding sequences of 128 genotypes indicated that, when using consensus primers and probe, RT-dPCR quantification of HRV RNA was more accurate than that of RT-qPCR for some genotypes. We found that although the consensus RT-qPCR did accurately quantify many HRV genotypes, it did not accurately quantify all genotypes of HRV due to sub-optimal amplification of genotypes with sequences that do not exactly match those of the primers and probe. Consensus RT-dPCR, however, did not overcome all sequence mismatch-induced amplification inefficiency, as evidenced by genotype A88 (Fig. 2b), which has a single mismatch near the middle of the forward primer.

Although RT-dPCR has been shown to be more accurate than RT-qPCR for quantification of HRV and may be applicable to other viruses with high sequence diversity, like HIV and HBV, it has some disadvantages for routine use in a clinical laboratory. RT-dPCR has a more limited dynamic range compared to RT-qPCR (104 for RT-qPCR compared to 108 for RT-qPCR), which would require dilution and retesting of samples with high viral loads. Running an RT-dPCR assay requires more hands-on technician time and has a lower throughput than current RT-qPCR assays. Digital PCR instruments and reagents are also currently more expensive than most qPCR systems.

Conclusion
In conclusion, dPCR was a better alternative to qPCR on RNA templates known to have significant sequence diversity that cannot be avoided during primer and probe design and should be considered the better molecular method for quantification of HRV in respiratory specimens.

References
1. Brownlee JW, Turner RB. New developments in the epidemiology and clinical spectrum of rhinovirus infections. Curr Opin Pediatr 2008: 20: 67–71.
2. Gern JE. The ABCs of rhinoviruses, wheezing, and asthma. J Virol 2010: 84(15): 7418–7426.
3. Mackay IM. Human rhinoviruses: The cold wars resume. J Clin Virol 2008: 42: 297–320.
4. Sedlak RH, Nguyen T, Palileo I, Jerome KR, Kuypers J. Superiority of digital RT-PCR over real-time RT-PCR for quantitation of highly divergent human rhinoviruses. J Clin Microbiol 2017; 55(2): 442–449.
5. Andries K, Dewindt B, Snoeks J, Wouters L, Moereels H, Lewi PJ, Janssen PA. Two groups of rhinoviruses revealed by a panel of antiviral compounds present sequence divergence and differential pathogenicity. J Virol 1990: 64: 1117–1123.
6. Mackay IM, Arden KE, Nitsche A. Real-time PCR in virology. Nucleic Acids Res 2002: 30: 1292–1305.
7. Lu X, Holloway B, Dare RK, Kuypers J, Yagi S, Williams JV, Hall CB, Erdman DD. Real-time reverse transcription-PCR assay for comprehensive detection of human rhinoviruses. J Clin Microbiol 2008: 46(2): 533–539.
8. Granados A, Luinstra K, Chong S, Goodall E, Bahn L, Mubareka S, Smieja M, Mahony J. Use of an improved quantitative polymerase chain reaction assay to determine differences in human rhinovirus viral loads in different populations. Diagn Microbiol Infect Dis 2012: 74: 384–387.
9. Tapparel C, Cordey S, Van Belle S, Turin L, Wai-Ming L, Regamey N, Meylan P, Mühlemann K, Gobbini F, Kaiser L. New molecular detection tools adapted to emerging rhinoviruses and enterviruses. J Clin Microbiol 2009: 47(6): 1742–1749.
10. Bochkov YA, Grindle K, Vang F, Evans MD, Gern JE. Improved molecular typing for rhinovirus species A, B, and C. J Clin Microbiol 2014: 52(7): 2461–2471.
11. Hoffman NG, Cook L, Atienza EE, Limaye AP, Jerome KR. Marked variability of BK virus load measurement using quantitative real-time PCR among commonly used assays. J Clin Microbiol 2008: 46(8): 2671–80.
12. Vynck M, Trypsteen W, Thas O, Vandekerckhove L, De Spiegelaere W. The future of the polymerase chain reaction in virology. Mol Diagn Ther 2016: 20: 437–447.
13. Huggett JF, Cowen S, Foy CA. Considerations for digital PCR as an accurate molecular diagnostic tool. Clin Chem 2015: 61: 79–88.
14. Sedlak RH, Jerome KR. Viral diagnostics in the era of digital polymerase chain reaction. Diagn Microbiol Infect Dis 2013: 75(1): 1–4.

The author
Jane Kuypers PhD
Department of Laboratory Medicine, University of Washington, Seattle, WA, USA
Vaccine and Infectious Disease Division, Fred Hutchinson Cancer Research Center, Seattle, WA, USA


E-mail: kuypers@uw.edu

Background
The term ‘acute respiratory tract infection’ (ARTI) encompasses a spectrum of conditions ranging from the common cold to pneumonia. Multiple organisms may infect the human respiratory tract, such as bacteria and fungi, but the majority of episodes are thought to be viral in origin [1]. The diagnosis of ARTI is made clinically but the lack of pathognomonic features means etiology cannot be determined clinically. Identifying causative organisms is challenging owing to the number of possibilities but is important for many reasons. In severe cases identification of an organism will guide therapeutics, although specific treatments for viral ARTI are generally limited to influenza. At the other end of the clinical spectrum identifying a viral cause in mild cases allows the physician to defer treatment with antibiotics and reassure the patient. This is an essential component of antimicrobial stewardship as high rates of antibiotic use are associated with circulating antimicrobial resistance [2]. The majority of antibiotic prescriptions issued in the UK are for respiratory tract infections [3] yet antibiotic use puts patients at risk of adverse drug reactions and in many cases will not lessen the duration of symptoms [4].

Respiratory infection diagnostics
Until recently viral diagnostics relied on cell culture or animal/egg inoculation. These time-consuming and laborious methods provided only a retrospective diagnosis and were, therefore, of little use in the management of acute infections. Nucleic acid amplification tests (NAAT), directly detecting the RNA or DNA of pathogens, have largely superseded these.

Many methods of RNA/DNA detection are available (Table 1). The most widely used is polymerase chain reaction (PCR). This has the benefit of rapid turnaround times and high levels of sensitivity and specificity in comparison to cell culture [5]. A pair of primers is required for each target but it is possible to use multiple primer sets within a single reaction (up to four) without compromising test sensitivity over the monoplex assay. This chemistry is now available as closed systems providing rapid results as a near-patient test (GeneXpert by Cepheid, Cobas by Roche).

Other examples of molecular NAATs which are available but not commonplace would be loop-mediated isothermal amplification (LAMP) and microarrays. LAMP detects nucleic acids but does not rely on thermocycling. It does, however, require multiple primers for each target (usually six) and as a consequence the sensitivity is more likely to be affected by genome mutations than standard PCR. This also makes multiplexing multiple targets within a single reaction more complicated.

Microarrays (also known as DNA chips or biochips) use a collection of oligonucleotide probes, about 70 bases in length, immobilized on a solid surface. The probes are complementary portions of DNA or RNA designed to match conserved regions of a genome; thus, if present, the target will bind to the corresponding probe which can then be quantified. Multiple probes may be attached to a single surface, screening for a large number of pathogens in a single reaction. As probes are targeted against conserved regions of the pathogen genome they may also detect related but novel pathogens.

To be used as a comprehensive diagnostic test numerous targets must be included to cover the likely pathogens. In the case of respiratory infections commercial assays are available with around 33 targets over 8 reactions [6]. Despite this approach a viral pathogen is detected in only a minority of specimens [7] and it remains the case that a pathogen will only be detected if actively sought.

Several significant respiratory viruses have been identified in recent years; those which may have circulated for many years, such as the human metapneumovirus, or emerging pathogens, such as SARS and, more recently, MERS. Whereas these are related to other known pathogens they are genetically distinct and, therefore, would evade detection with molecular methods.

What is next-generation sequencing?
The term ‘next-generation sequencing’ (NGS) refers to the practice of sequencing millions of DNA fragments in parallel. Numerous platforms are available to carry this out and the exact chemistry varies greatly between each. In practice, either all genetic material within a sample can be sequenced – metagenomics; or, hybrid capture allows a more focused approach to an area or genome of interest, this is termed ‘target enrichment’.

Advantages of NGS
Applying metagenomic NGS to clinical samples would allow an untargeted approach to identify all the genetic material contained within. This method has demonstrated potential for use in a diagnostic setting [8, 9].
The lack of pathogen targeting means that multiple pathogens can be detected without selection (Fig. 1), including novel or emerging or divergent pathogens. In the case of many viral pathogens evolution and mutations over time can reduce detection with specific PCR reactions. Mutations affecting primer binding sites may reduce binding affinity during the reaction and, for this reason, the performance of diagnostic assays must be monitored closely and at times altered. NGS could, therefore, be used as an adjunct in the quality control of PCR assays.

It is possible to detect full genome sequences from diagnostic samples and even with partial genome sequence it is feasible to subtype viral pathogens. Real-time knowledge of the circulating viral subtype is of particular importance in the management of influenza where this informs anti-viral choice, potential resistance and vaccine efficacy. This is currently carried out using additional PCR assays and Sanger sequencing, although this is not always possible in real-time.

Laboratory workflow
Currently the identification of rare or unusual pathogens using molecular methods necessitates samples to be batched to make the process cost effective; alternatively the test is centralized to a single laboratory to which samples must be sent. Either results in an increase in turn-around time. The use of NGS without any enrichment or targeting would permit samples to be treated in the same manner irrespective of type or likely pathogen.

Challenges
A major barrier in introducing NGS to the diagnostic setting is cost. Although the cost of NGS is decreasing rapidly it remains considerably more expensive than multiplex PCR. It is, therefore, unlikely to be cost effective to use this method for pathogen detection in non-severe infections for the time being. However, any cost–benefit analysis on introducing NGS to a diagnostic setting should also consider, on the positive side of the balance sheet, the likely savings NGS would offer in reductions to epidemiological and public health testing.

Complexity and turnaround time
Current methods of library preparation are complex requiring multiple user interventions and additional equipment to that found in a diagnostic laboratory with attendant implications for the time and cost of the process. To be carried out as a routine diagnostic assay these processes would need to be simplified and, ideally, automated to reduce hands-on time and the potential for contamination and human error.
The commonly used sequencing platforms take several hours or even days to generate sequence information. It should be noted that the third generation platforms that use single-molecule real-time (SMRT) technology are rapid and, as the name suggests, can be analysed in almost real-time.

Data analysis
Data analysis and storage is a major bottleneck in the NGS process. The computational power required for analyses would be beyond the current capabilities of diagnostic services. The methods used in data analysis pose a further challenge. Currently there is no agreed method as to the best approach for data analysis; indeed this is an entire specialty in itself, bioinformatics. Development of software programmes will both make the analysis more feasible in a diagnostic service to non-bioinformaticians and will lead to standardization of data processing.

Discussion
NGS undoubtedly has potential to dramatically change the landscape of infection diagnostics. Whether it will replace current molecular methods remains to be seen. The cost and complex sample processing remains prohibitive but these novel technologies are still in an exponential phase of development. Even current methodologies are yielding promising results in this field. The lack of pathogen targeting means that there is potential for a single work flow to be applied to all specimens, no matter what the syndrome which could even be extended to non-viral pathogens, resulting in a pan-microbial diagnostic test.

The generation of virus sequence as part of a diagnostic assay has substantial management and epidemiological benefits. In terms of respiratory infections this is currently limited to resistance testing and strain analysis of influenza. However, in the management of blood-borne viruses, particularly HIV and hepatitis C virus (HCV), point mutations and minor populations may impact greatly on the management and prognosis of patients. With the introduction of novel therapies or vaccines against viral respiratory infections NGS will have an even greater clinical benefit.

Acknowledgements
I would like to thank Dr Rory Gunson and Dr Emma Thomson for reviewing the manuscript.

References
1. Clark TW, Medina MJ, Batham S, Curran MD, Parmar S, Nicholson KG. Adults hospitalised with acute respiratory illness rarely have detectable bacteria in the absence of COPD or pneumonia; viral infection predominates in a large prospective UK sample. J Infect 2014; 69(5): 507–515.
2. Linares J, Ardanuy C, Pallares R, Fenoll A. Changes in antimicrobial resistance, serotypes and genotypes in Streptococcus pneumoniae over a 30-year period. Clin Microbiol Infect 2006; 16(5): 402–410.
3. Lindbaek M. Prescribing antibiotics to patients with acute cough and otitis media. Br J Gen Pract 2010; 56(524): 164–166.
4. Butler CC, Hood K, Verheij T, Little P, Melbye H, Nuttall J, Kelly MJ, Mölstad S, Godycki-Cwirko M, Almirall J, Torres A, Gillespie D, Rautakorpi U, Coenen S, Goossens H. Variation in antibiotic prescribing and its impact on recovery in patients with acute cough in primary care: prospective study in 13 countries. BMJ 2009; 338: b2242.
5. van Elden LJ, van Kraaij MG, Nijhuis M, Hendriksen KA, Dekker AW, Rozenberg-Arska M, van Loon AM. Polymerase chain reaction is more sensitive than viral culture and antigen testing for the detection of respiratory viruses in adults with hematological cancer and pneumonia. Clin Infect Dis 2002; 34(2): 177–183.
6. FTD Respiratory Pathogens 33. Fast-track Diagnostics 2016. (http://www.fast-trackdiagnostics.com/products/ftd-respiratory-pathogens-33/)
7. Nickbakhsh S, Thorburn F, von Wissmann B, McMenamin J, Gunson RN, Murcia PR. Extensive multiplex PCR diagnostics reveal new insights into the epidemiology of viral respiratory infections. Epidemiol Infect 2016; 144(10): 2064–2076.
8. Thorburn F, Bennett S, Modha S, Murdoch D, Gunson R, Murcia PR. The use of next generation sequencing in the diagnosis and typing of respiratory infections. J Clin Virol 2015; 69: 96–100.
9. Prachayangprecha S, Schapendonk CM, Koopmans MP, Osterhaus AD, Schürch AC, Pas SD, van der Eijk AA, Poovorawan Y, Haagmans BL, Smits SL. Exploring the potential of next-generation sequencing in detection of respiratory viruses. J Clin Microbiol 2014; 52(10): 3722–3730.

The author
Fiona Thorburn PhD
NHS Greater Glasgow and Clyde, Glasgow G12 0XH, UK


E-mail: Fionathorburn@nhs.net