Metagenomic analysis of viral diversity in respiratory samples from patients with respiratory tract infections in Kuwait.

A metagenomic approach based on target independent next-generation sequencing has become a known method for the detection of both known and novel viruses in clinical samples. This study aimed to use the metagenomic sequencing approach to characterize the viral diversity in respiratory samples from patients with respiratory tract infections. We have investigated 86 respiratory samples received from various hospitals in Kuwait between 2015 and 2016 for the diagnosis of respiratory tract infections. A metagenomic approach using the next-generation sequencer to characterize viruses was used. According to the metagenomic analysis, an average of 145, 019 reads were identified, and 2% of these reads were of viral origin. Also, metagenomic analysis of the viral sequences revealed many known respiratory viruses, which were detected in 30.2% of the clinical samples. Also, sequences of non-respiratory viruses were detected in 14% of the clinical samples, while sequences of non-human viruses were detected in 55.8% of the clinical samples. The average genome coverage of the viruses was 12% with the highest genome coverage of 99.2% for respiratory syncytial virus, and the lowest was 1% for torque teno midi virus 2. Our results showed 47.7% agreement between multiplex Real-Time PCR and metagenomics sequencing in the detection of respiratory viruses in the clinical samples. Though there are some difficulties in using this method to clinical samples such as specimen quality, these observations are indicative of the promising utility of the metagenomic sequencing approach for the identification of respiratory viruses in patients with respiratory tract infections.

INTRODUCTION

Historically, the word metagenome was first used in 1998 to describe the collection of microbial genome found in a soil sample, including microorganisms that could not be cultured by conventional methods.1 Early metagenomics studies on environmental samples yielded the identification of metabolic characters, the classification of organisms and the discovery of antibiotics and enzymes.2, 3 Now, metagenomic studies include a broad range of research including marine ecological research, plant and agricultural, human genetics and diagnostics of human diseases. The first application of metagenomic in virus discovery was in the analysis of viruses in soil samples from two marine sites in San Diego.4 Later, the approach was used to analyze viruses in different fields.5 The traditional methods of virus detection involved filtration, tissue culture, electron microscopy, and serology.6 Among these, the standard gold method of virus detection was cell culture. However, many viruses cannot be easily cultivated, and two milestone innovations solved this problem; polymerase chain reaction (PCR) and DNA sequencing (Sanger method).6 By using these methods, several important emerging viruses, such as hendra virus,7 nipah virus,8 menangle virus,9, 10 melaka virus,11 and reston ebola virus,12 were discovered. Despite the sensitivity of PCR, it can only detect one virus at a time. Multiplex PCR, on the other hand, is used to identify multiple targets for the identification of more than one virus in a single test. However, it is often difficult to standardize the assay by using various primers. Hence, novel approaches that overcome the difficulties of viral detection with the conventional molecular methods are needed to discover novel human viruses.13 Not surprisingly, virologists were the first to explore the use of sequence‐independent, a metagenomic approach using Next Generation Sequencer (NGS) to detect human‐associated viruses. The genome of DNA and RNA viruses can be detected directly after extraction from samples through metagenomic sequencing.14

Worldwide, respiratory tract infection (RTI) is an important cause of hospitalization among young children and elderly, with significantly high mortality and morbidity.15 RTI is a group of diseases of both upper or lower respiratory tract. Upper respiratory tract infections (URTIs) include laryngitis, common cold, rhinitis, pharyngitis/tonsillitis, otitis media and rhinosinusitis/sinusitis. On the other hand, lower respiratory tract infections (LRTIs) include bronchitis, bronchiolitis, tracheitis and pneumonia. These respiratory tract infections increase the extent of the problem in patients with chronic comorbidities and asthma,16 chronic obstructive pulmonary disease (COPD),17 very young, elderly,17 and immunocompromized patients. Respiratory viruses account for over 65% of all respiratory infections and 90% of URTIs.18, 19 Bacteria, however, only represent 10% of all URTIs.19 Despite the improvement in the molecular techniques for viral discovery, the viral aetiology of RTI is still unknown in a high number of cases either because the tests are ineffective or the virus is unrelated to any of the currently known respiratory viruses. Current diagnostic methods to detect respiratory viruses are mainly based on sequence‐dependent molecular amplification techniques such as PCR, which identify a panel of known viruses, and therefore, new respiratory viruses are missed. To overcome the pitfalls associated with PCR approaches based on NGS techniques such as viral metagenomic will become a logical step as routine viral diagnostics on clinical samples. Broaden not only the detection range of viruses but also provide an additional characterization of the detected viruses such as genotypes and subtypes of viruses. However, the efficiency and viability of using such techniques in diagnostic setting require further study.

Since there are no studies on the use of metagenomic approach for the identification of pathogens responsible for different diseases, the novelty of this study was to develop a metagenomic approach using NGS for the detection of known and unknown viruses associated with respiratory tract infections among patients in Kuwait.

METHODOLOGY

Study population and sample collection

A total of 86 patients (56 were females and 30 males) with signs and symptoms of URTIs and LRTIs between 2015 and 2016 were enrolled in this study. The patients were admitted to different hospitals in Kuwait including Al‐Sabah Hospital, Al‐Farwaniya Hospital, Al‐Adan Hospital, Mubarak Al‐Kabeer Hospital, Infectious Diseases Hospital, and Al‐Amiri Hospital. The respiratory diseases of the patients were pharyngitis/tonsillitis, rhinitis, bronchitis, bronchiolitis, and pneumonia. The patients ranged in age from one day to 89 years, with a median age of two years. Summary of clinical samples is shown in Table 1. The respiratory samples included nasopharyngeal aspirates/wash, nasopharyngeal swab, bronchoalveolar lavage, tracheal aspirates, sputum, throat swabs, and nasal swabs. They were collected in the hospital and processed at the Virology Unit, Faculty of Medicine, Kuwait University for the presence of viral nucleic acids using multiplex Real‐Time PCR and metagenomic analysis based on next‐generation sequencing.

Table 1

Summary of metagenomic sequencing and multiplex real‐time PCR data

Sample no.	Age	Gender	Total read a	% of genome coverage b	Metagenomic results c	Multiplex real‐time PCR results	CT d
1	4 yr	M	94 071	‐	‐	Human bocavirus	12.28
2	6 mo	M	94 183	‐	‐	Rhinovirus	28.10
3	5 yr	M	135 071	‐	‐	Adenovirus	31.77
4	1 mo	M	181 304	‐	‐	Respiratory syncytial virus	19.40
5	4 mo	M	219 673	1.85	Influenza A virus (H9N2)	Rhinovirus	27.81
6	3 yr	M	119 744	‐	‐	Respiratory syncytial virus	18.65
7	6 mo	F	210 747	6.69	Torque teno virus 19	Enterovirus	24.33
8	1 yr	M	428 426	5.4	Respiratory syncytial virus	Respiratory syncytial virus	25.87
9	7 yr	M	184 918	2.69	Rotavirus F chicken	Adenovirus	18.22
10	1 yr	M	154 571	23.24	Human bocavirus	Adenovirus + human bocavirus	AdV: 14.5056 HBoV: 13.0494
11	4 yr	M	205 218	4.16	Respiratory syncytial virus	Respiratory syncytial virus	22.58
12	11 yr	M	209 072	‐	‐	Enterovirus	34.02
13	23 yr	M	126 583	‐	‐	Human metapneumo virus	32.63
14	1 mo	M	76 761	‐	‐	Influenza A+ human bocavirus	Flu A:25.0305 HBoV: 15.1864
15	4 yr	F	269 314	3.56	Respiratory syncytial virus	Respiratory syncytial virus	26.43
16	68 yr	M	192 990	‐	‐	Influenza B	16.51
17	2 yr	F	202 955	‐	‐	Respiratory syncytial virus	17.78
18	1 yr	M	136 219	1.47	Respiratory syncytial virus	Respiratory syncytial virus + Adenovirus	AdV: 17.3483 RSV: 15.2738
19	2 yr	M	259,599	‐	‐	Rhinovirus	11.09
20	58 yr	M	127 399	2.84	Rotavirus A	Influenza A virus	15.31
				1.92	Influenza A virus		‐
21	1 yr	M	439 927	‐	‐	Parainflunza 1+ adenovirus	PIV1: 18.6867 AdV: 25.1402
22	3 yr	M	166 168	3.44	Respiratory syncytial virus	Respiratory syncytial virus	25.43 ‐
				1.86	Rotavirus strain E
23	5 yr	F	158 839	‐	‐	Influenza A virus	33.27
24	57 yr	M	168 581	2.43	Adult diarrhoea rotavirus strainJ19	Respiratory syncytial virus	5.98
				1.4	Influenza A virus (H2N2)
25	13 yr	F	40 774	‐	‐	Respiratory syncytial virus	29.78
26	3 yr	M	39 964	‐	‐	Human bocavirus	35.27
27	13 yr	M	164 746	‐	‐	Influenza A virus	17.32
28	32 yr	F	112 605	1.01	Torque teno midi virus 2	Influenza A virus	14.55
29	1 mo	M	85 792	‐	‐	Respiratory syncytial virus + parainfluenza 3 virus	RSV: 32.9743 PIV3: 30.0936
30	2 yr	M	164 227	‐	‐	Respiratory syncytial virus	19.8748
31	4 mo	M	40 948	‐	‐	Rhinovirus	11.5602
32	4 yr	F	4 074	‐	‐	Influenza A virus	14.1571
33	1 mo	F	159 476	‐	‐	Respiratory syncytial virus	15.0062
34	3 yr	F	40 894	‐	‐	Respiratory syncytial virus	12.5713
35	1 mo	F	175 935	3.01	Rotavirus B	Respiratory syncytial virus	12.6546
				1.95	Influenza C virus
36	4 mo	M	40 795	‐	‐	Respiratory syncytial virus	12.3435
37	1 mo	M	119 447	44.43	Human rhinovirus 14	Rhinovirus + human metapneumo virus	HRV: 11.5732 HMPV: 14.87
38	67 yr	M	40 993	1.76	Human respiratory syncytial virus	Influenza B + human respiratory syncytial virus	Flu B: 12.5212 RSV: 12.1386
39	59 yr	M	209 709	‐	‐	Influenza A virus	15.6758
40	89 yr	M	117 703	‐	‐	Influenza A virus	15.3729
41	25 yr	F	107 994	2.06	Human respiratory syncytial virus	Respiratory syncytial virus	12.5934
42	2 yr	M	257 959	99.23	Respiratory syncytial virus	Respiratory syncytial virus	10.36
43	70 yr	F	100 377	‐	‐	Adenovirus	12.37
44	1 mo	M	173 720	2.03	Influenza C virus	Rhinovirus	11.96
45	4 mo	M	33 958	‐	‐	Respiratory syncytial virus	21.91
46	66 yr	F	22 444	3.01	Human rotavirus B strain	Influenza A virus	13.48
47	1 mo	M	301 529	‐	‐	Respiratory syncytial virus	19.0166
48	2 mo	M	187 769	85.88	Bovine corona virus	Coronavirus OC 43	16.1313
				72.52	Human metapneumo virus
				23.73	Coronavirus HKU14
49	2 yr	F	24 825	‐	‐	Human metapneumo virus	24.3051
50	47 yr	M	24 726	‐	‐	Human metapneumo virus	11.8083
51	1 yr	M	168 915	‐	‐	Rhinovirus	33.9199
52	6 yr	F	18 995	‐	‐	Resipratory syncytial virus	12.574
53	28 yr	F	100 404	‐	‐	Human metapneumo virus	12.4938
54	16 yr	M	121 120	2.57	Human rhinovirus C	Rhinovirus	11.6254
55	68 yr	F	116 653	4.05	Respiratory syncytial virus	Influenza A virus	18.5788
56	2 mo	M	221 941	‐	‐	Rhinovirus	13.0581
57	1 yr	M	74 807	1.97	Human respiratory syncytial virus	Adenovirus	12.466
58	1 mo	M	167 451	3.33	Respiratory syncytial virus	Respiratory syncytial virus	12.525
59	11 mo	M	88 049	‐	‐	Human metapneumo virus	12.1959
60	63 yr	F	60 882	‐	‐	Rhinovirus	21.5916
61	27 yr	F	141 599	5.2	Rotavirus C	Rhinovirus	28.7566
62	9 yr	M	177 528	‐	‐	Rhinovirus	11.6255
63	10 yr	F	164 350	77.21	Respiratory syncytial virus	Respiratory syncytial virus	17.1614
64	1 mo	F	159 543	2.47	Human rhinovirus C	Rhinovirus	22.5983
65	2 mo	F	306 199	‐	‐	‐	‐
66	4 yr	F	156 745	1.37	Human respiratory syncytial virus	‐	‐
67	14 dy	F	5 806	‐	‐		‐
68	2 mo	M	302 314	4	Rotavirus strain E	‐	‐
69	4 mo	M	327 618	‐	‐	‐	‐
70	15 dy	F	137 021	2.63	Respiratory syncytial virus	‐	‐
71	5 mo	M	258 050	2.13	Influenza A virus (H1N1)	‐	‐
72	73 yr	M	323 340	2.82	Rotavirus D chicken	‐	‐
				1.85	Influenza A virus (H5N1)		‐
73	1 yr	M	32 382	‐	‐	‐	‐
74	1 yr	F	18 984	‐	‐	‐	‐
75	2 yr	F	142 437	‐	‐	‐	‐
76	2 mo	F	39 441	‐	‐	‐	‐
77	1 mo	M	232 798	2.94	Rotavirus G chicken	RSV	13.45
				2.35	Hantaan virus
78	2 mo	M	187 421	1.99	Rotavirus strain E	‐	‐
79	9 yr	M	43 808	‐	‐	‐	‐
80	49 yr	M	35 596	‐	‐	‐	‐
81	12 yr	M	267 407	‐	‐	‐	‐
82	4 yr	M	114 665	2.33	Rotavirus strain E	‐	‐
83	3 dy	F	176 085	‐	‐	‐	‐
84	7 mo	M	180 162	2.92	Rotavirus strain E	‐	‐
85	7 mo	F	3	‐	‐	‐	‐
86	65 yr	M	45 363	‐	‐	‐	‐

Total number of MiSeq reads after removal ow quality, short or adapter containing reads.

All read mapped to deno novo assembled genome using Burrows‐Wheel Aligner and the value reports the percentage of times each position sequenced.

As per Burrows‐Wheel Aligner results.

CT Real‐Time PCR cycle threshold value.

Nucleic acids extraction and multiplex real‐time PCR assay

Total nucleic acids were isolated from clinical samples using Roche ®MagNA Pure LC system (Roche Diagnostics, Indianapolis, IN), according to the manufacturer's instructions and stored at −80°C for further processing. Multiplex Real‐Time PCR assay using Fast Track Kit (Fast −Track Diagnostic, Luxembourg, Germany) was used to detect common respiratory viruses from the extracted RNA according to manufacturers' instructions. This assay is Multiplex Real‐Time PCR for detection of 21 respiratory viruses by TaqMan® technology. It required five tube multiplex for detection of influenza A, influenza A (H1N1) swl, influenza B, human rhinovirus (HRV), human coronavirus NL63 (HCoV‐NL63), HCoV‐229E, HCoV‐OC43, HCoV‐HKU1, parainfluenza (PIV) 1, 2, 3, 4, human metapneumovirus (HMPV) A/B, respiratory syncytial virus (RSV) A/B, bocavirus, adenovirus (AdV), parechovirus, enterovirus, mycoplasma pneumonia, and internal control. This PCR assay is a routine diagnostic test for the detection of respiratory viruses which is performed at the Virology Unit, Faculty of Medicine, Kuwait University.

Next generation sequencing and metagenomic analysis

Fresh nucleic acids (RNA and DNA) were extracted from each respiratory sample processed for metagenomic analysis using the Illumina MiSeq (San Diego, CA) platform for NGS according to standard procedures.20 Briefly, after nucleic acid extraction step, genomic and host DNA was removed from each sample using Ambion DNA‐free (Life Technologies) according to manufacturer's instructions to obtain metagenome RNA. Then, 10 ng of DNA‐ free RNA was integrated into first‐strand cDNA synthesis primed by random hexamers and then amplified using whole transcriptome amplification kit (Qiagen, Valencia, CA). An exact concentration of 1 μg of DNA was prepared after DNA quantification step using Qubit® 2.0 Fluorometer and Qubit™ dsDNA BR Assay kits (Invitrogen, CA). Illumina TruSeq DNA library preparation kit V2 (Illumina San Diego, CA) was used to prepare DNA libraries, followed by sequencing of 150‐bp paired‐end reads on an Illumina MiSeq instrument at the OMICS Research Unit, Health Science Centre, Kuwait University.

Bioinformatics

The quality of the sequence data from a total of 86 Miseq runs was checked using FastQC (version 0.10.1; http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). The Sequence fastq files were trimmed where the average quality score was <30 using a fast toolkit (http://hannonlab.cshl.edu/fastx_toolkit/index.html), and from bases 18‐145 were kept for all files. High‐quality paired‐end sequences were retained for downstream analysis. Host and bacterial sequences were removed from these sequences by mapping them to a database containing all human and bacterial DNA/RNA sequences. A database of references containing about 5800 strains available on National Center for Biotechnology Information (NCBI) was created by merging all sequence into one FASTA file. The host removed reads were aligned against a viral reference using Burrows‐Wheeler Aligner (BWA) and sam files were created. Option samtools flagstat provided the percentage of host and viral reads in sam files. A statistic summary providing the information regarding some reads corresponding to the individual viral strain and the percentage of viral genome covered for all viruses was extracted from sam files using pileup option in bb map program (http://jgi.doe.gov/data-and-tools/bbtools/bb-tools-user-guide/bbmap-guide/).

Statistical analysis

The data were analyzed using a computer software “Statistical Package for Social Sciences,” SPSS version 24.0 (IBM Corp, Armonk, NY). The descriptive statistics are presented as frequencies and percentages. The Cohen's Kappa (k) was applied to find the measure of agreement on detecting respiratory viruses by both multiplex Real‐Time PCR and metagenomics sequencing techniques. The two‐tailed probability value “p” <0.05 was considered statistically significant.

RESULTS

Virus detection by multiplex real‐time PCR

Eighty‐six respiratory samples from patients with URTIs and LRTIs who visited different hospitals in Kuwait between 2015 and 2016 were analyzed for respiratory viruses by multiplex Real‐Time PCR assay. The results showed that 65 out of 86 samples (75.6%) were positive for respiratory viruses. Among the positive samples, 23 (35.4%) were positive for RSV, 12 (18.5%) samples were positive for HRV, eight (12.3%) samples were positive for influenza A, four (6.2%) samples were positive for AdV, four (6.2%) samples were positive for HMPV, two (3%) samples were positive for bocavirus, two (3%) samples were positive for enterovirus, one (1.5%) sample was positive for influenza B, one (1.5%) sample was positive for PIV, one (1.5%) sample was positive for HCoV‐OC43, and seven (10.7%) samples were positive for dual viruses (Figure 1). The combinations of dual virus samples were as follow: adenovirus and human bocavirus; human bocavirus and influenza A virus; RSV and AdV; AdV and PIV‐A virus; RSV and PIV‐3; HRV and HMPV; RSV and influenza B virus. The remaining 21 (24.4%) samples were negative for any respiratory viruses (Figure 1). The threshold cycle value (CT) for each detected virus is shown in Table 1.

Figure 1

Clustered Bar chart of the percentage of patients positive for respiratory viruses detected by multiplex RT‐PCR assay (n = 86)

Virus detection by metagenomic sequencing

From the 86 samples, an adequate amount of RNA was obtained to perform whole‐transcriptome sequencing using the Illumina Miseq sequencer for the detection of RNA viruses. Sequencing of cDNA libraries from all samples generated an average of 145, 019 reads (range; 3‐439 927), after quality trimming and filtering. On an average, 64% (range, 5.72‐98.16%) of the reads were derived from host (human) genome, and 2% (range, 0.04‐7.87%) of the reads were derived from viruses (Figure 2). The rest of the reads (34%) were derived from bacteria and other viruses that did not have any significant match in the known databases. The identified viral sequences belonged to the following respiratory viruses; respiratory syncytial virus (15% of patient), influenza A virus (4.7% of patient), rhinovirus (3.5% of patient), influenza C virus (2.3% of patient), torque tenoviruses(2.3% of patient), human bocavirus (1.2% fo patient), mixed viral infection consisted of bovine coronavirus, human metapneumovirus, and coronavirus HKU14 (1.2% of patient) (Figure 3). Also, non‐respiratory viruses were detected in 14% of the patients. These viruses were rotaviruses A, B, C, D, E, F, G, J19, and hantavirus. The rest of the identified viral sequences belonged to plenty of non‐human viruses, such as plant associated viruses, were found in 55.8% of the patients (Figure 3). The average genome coverage of the identified viruses after genome sequence assembly was 12%. The highest genome coverage was for RSV (99.23%), and the lowest for torque teno midi virus 2 (1.01%) (Table 1). Table 1 demonstrates a summary of the metagenomic sequencing and multiplex Real‐Time PCR results for each clinical sample. It is worth mentioning that sequences from human endogenous retrovirus, which is a non‐pathogenic virus, was common shared viral sequences found in 44 sample (51.2%) with an average genome coverage of 5.2% (range, 1.2‐11.6%).

Figure 2

Pie chart of the taxonomic distribution of metagenomic sequencing reads from clinical samples. Data are average values of reads

Figure 3

Clustered Bar chart of the percentage of patients positive for viruses detected by metagenomic sequencing assay (n = 86)

To evaluate the feasibility of the metagenomic approach for the detection of respiratory viruses in clinical samples, a comparison was made between multiplex Real‐Time PCR and metagenomic sequencing results. Of the total 86 clinical specimens, multiplex Real‐Time PCR detected 65 (75.6%) clinical samples as positive for respiratory viruses and 21 (24.4%) as negative, while metagenomic sequencing technique detected 28 (32.6%) clinical samples positive and 58 (67.4%) negative. Overall, 24 (27.9%) clinical samples were detected as positive, and 17 (19.8%%) as negative, by both the techniques, giving an absolute agreement on 41 (47.7%) clinical samples. However, applying the Cohen's Kappa statistics for a measure of agreement (k = 0.112, P = 0.129), only a slight agreement was found between the two techniques.

DISCUSSION

The respiratory tract is a target of many human viruses, especially RNA viruses. The current molecular techniques for the detection of respiratory viruses are largely targeted dependent tests which detect a limited number of viruses. On the other hand, NGS‐based metagenome approaches are target independent which can detect common, unexpected pathogens, and novel viruses in a given sample.21

To explore the utility of the metagenomics approach in the identification of respiratory viruses in clinical samples, we first performed multiplex Real‐Time PCR testing for the initial identification of respiratory viruses in respiratory samples from patients with URTIs and LRTIs. RSV was the most predominant respiratory virus detected in the patient's samples (35.4% of the patients), while HRV was the second prevalent virus (18.5% of the patients). Our observation is in agreement with others who showed that RSV is the most prevalent virus in patients with respiratory tract infections.22, 23, 24 In addition to single viral infections, our result demonstrated that mixed viral infections are a frequent phenomenon; it was observed in seven samples indicating that more than one virus can trigger respiratory tract infections and that may influence the severity and outcomes of the respiratory disease. One patient with mixed viral infection presented with severe bronchopneumonia while the rest of the patients had mild respiratory symptoms. These results suggested the absence of an association between multiple viral infections and the severity of the respiratory disease. In contrast to our results, studies showed that patients with respiratory tract infection due to dual viral infections were hospitalized significantly more often than those with respiratory disease due to a single virus.25, 26 In agreement with our results, others did not demonstrate any correlation between the presence of multiple viruses and the severity of the respiratory diseases.27, 28, 29, 30

Respiratory samples from the 86 patients were analyzed further for the presence of viruses by metagenomic sequencing using NGS. It should be noted that this study is the first of its kind to perform metagenomic analysis of the virome of respiratory tract specimens. Sequencing of cDNA libraries from the clinical samples revealed an average of 145 019 of 150‐bp paired‐end reads which considered relatively lower than expected. We anticipated that this problem is due to the low concentration of the cDNA obtained by the whole transcriptome amplification step. Analysis of the sequencing reads revealed that over half of the reads (64%) were host‐derived reads, however, only 2% of the reads derived from viruses. The residual of the reads (34%) included bacterial reads and other reads that did not have any significant match in the database. Other researchers have also reported high percentage (90%) of human‐derived sequence in nasal specimens analyzed by metagenomics sequencing.31, 32, 33 Therefore, it appears that the presence of human‐derived reads is an intrinsic problem when nucleic acids are extracted directly from respiratory samples. In comparison to our results, Nakamura and coworkers identified lower average (0.76%) of virus‐associated sequences, and Yang and colleagues identified only 0.05% of virus‐associated sequences.32, 34

Corresponding with multiplex Real‐Time PCR assay, RSV was the most predominant virus detected in the respiratory samples by the NGS‐based metagenomic approach, while influenza A virus was the second most prevalent virus. Interestingly, two torque teno viruses were detected by the metagenomic approach in two respiratory samples; torque teno virus 19 and torque teno midi virus 2. Torque teno viruses (TTV) are new, emerging infectious agents that are exceptionally high prevalence, and relatively uniform distribution worldwide.35, 36 It was suggested that TTV is related to numerous diseases, such as hepatitis, respiratory diseases, cancer, haematological and autoimmune disorders. However, their direct involvement is under arguments.37, 38, 39 Metagenomic sequencing in this study revealed mixed viral infection (bovine coronavirus, HMPV, and HCoV‐HKU14) in only one clinical sample.

Many non‐respiratory viruses were detected in 14% of the clinical samples. Different strains of rotaviruses were among these viruses. Indeed, it is not unusual to detect rotaviruses in respiratory specimens. It was shown in the earlier studies the detection of rotavirus in nasopharyngeal secretions of children with acute gastritis. These studies speculated the association of diarrhoea due to rotavirus with respiratory signs and symptoms.40, 41 Another prominent observation in our study was the detection of a vast number of non‐human viruses such plant viruses in all clinical samples. We speculate that the presence of plant viruses in the respiratory samples is a natural phenomenon. They come from food, water, or dust and eventually become be part of normal flora of the upper respiratory tract which may/or may not be linked to any disease. Another prominent observation was the detection of sequences from a human endogenous retrovirus (HERV) approximately half of the patients. This observation reveals the high prevalence of human endogenous retrovirus in the population of Kuwait. According to many investigations, it was found that human endogenous viruses are old viruses that started to integrate into the human genome around 30‐40 million years ago and now about 8% of the human genome is derived from sequences with similarity to retroviruses.42

One of the major questions is how to intercept metagenomic analysis results by NGS in term what is clinically relevant for the patient. The comparison of the metagenomic sequencing data to RT‐Real time PCR data revealed absolute agreement of 47.7%. Although metagenomic approach failed to detect all the multiplex PCR positive samples, it did offer several advantages over the multiplex Real‐Time PCR method. For examples, the metagenomic sequencing identified several viruses causing respiratory infections that are usually not detected using routine diagnostic assays (eg, torqueteno viruses and coronavirus HKU14). Also, it delivered more detailed typing information for the detected viruses including subtype for some viruses. Furthermore, it is possible that multiplex RT‐PCR amplifies viruses present in minimal concentrations that have no significance in the disease process. It should be noted that two nucleic acid extracts for each clinical sample were prepared. The first extract was used to perform RT‐PCR and the second extract was used to perform metagenomics sequencing using NGS. This, in turn, will generate differences in nucleic acid concentration and quality which may further affect the results.

The analytical sensitivity of metagenomics sequencing is highly influenced by the quality of the samples. To improve the sensitivity of this approach is by increasing the depth of sequencing, which in turn can be overcome by improving the sample preparation.43 One of the solutions is to perform additional steps of ample‐filtering before sequencing to eliminate host cells and enhance the microbial genome to ensure more efficient microbial pathogen detection. Some studies employed a pre‐treatment step before nucleic acid extraction which required a combination of filtration and density‐dependent centrifugation to enhance the majority of viruses.4, 44, 45 Others used an enzyme cocktail containing DNasI, RNaseI, and benzonase to digest naked host nucleic acid.46 One of the limitations of pre‐treatment step to remove host DNA contents from clinical samples is the possibility to introduce a bias and decrease the sensitivity for DNA viruses such as adenovirus which considered as vital virus causing respiratory diseases. Also, filtering steps may reduce significantly the quality of DNA/RNA produced, which may further effect metagenomics sequencing results. Therefore, it is recommended to keep pre‐treatment steps to a minimum to avoid any particular bias. We have used a virus identification algorithm, BWA and BLAST from NCBI for the identification, clustering and read removal. One noticeable step for future improvement lies in the treatment of the BLAST undefined reads. BLAST search did not classify a large portion of the sequence data generated here. Future efforts will include an emphasis on alternative virus identification to characterize these novel sequences.

Massive sequencing projects and databases based on NGS technologies are not safe from contamination issues.47, 48 It is problematic when contaminant sequence reads being lost in the numerous of reads from the target sample, and consequently difficult to detect and clean out. In NGS library assembly protocols, it is essential to have one or multiple PCR amplification steps that could elevate the concentrations of DNA, thereby increasing the risk of contamination. Also, cross‐contamination of starting material is a devastating issue. There is a chance that samples may contaminate each other when they are prepared at the same time, or multiple libraries are produced in parallel. To avoid this type of contamination is to perform all sample and library preparations individually. If it is not feasible to do all sample and library preps separately, then strict measures are needed to be employed to keep samples pure. Certainly, it is essential to identify at which steps of the experimental protocol the contaminations are most often happening, and to deliver guidelines on how to avoid it.

CONCLUSION

In conclusion, we conducted a metagenomic sequencing analysis of the viruses in patients with RTI. The results presented in this study demonstrated that metagenomic approach is a promising diagnostic tool in clinical virology in which RT‐Real‐Time PCR couldn't detect many respiratory viruses that were detected by metagenomic approaches. Such sequence independent detection method will increase the chance to detect the causative agent of viral RTI and to gain information of the viruses that cannot be obtained by the current diagnostic tool. Although many obstacles to the routine use of metagenomic approaches do exist, which include cost, labor intensity and turnover time, it was proved to be highly sensitive in many studies, and it theoretically provides more information regarding virus species/type for virus diagnosis with the ability of the detection of unknown viruses. The question is probably when metagenomics sequencing will become a routine test for detecting infectious viruses? This approach requires improvement in sample preparation, validated pipelines for read sorting and taxonomic assignation and soon will substitute the current diagnostic tools. Indeed, this novel study provided outstanding information on the association of different viruses with RTIs in Kuwait. Moreover, metagenomics queening approach will be used in the near future to study the viral diversity in other syndromes such as pyrexia of unknown origin and gastroenteritis.