Next-generation sequencing analysis of new genotypes appearing during antiviral treatment of chronic hepatitis C reveals that these are selected from pre-existing minor strains

Coinfection with more than one hepatitis C virus (HCV) genotype is common, but its dynamics, particularly during antiviral treatment, remain largely unknown. We employed next-generation sequencing (NGS) to analyse sequential serum and peripheral blood mononuclear cell (PBMC) samples in seven patients with transient presence or permanent genotype change during antiviral treatment with interferon and ribavirin. Specimens were collected right before the therapy initiation and at 2, 4, 6, 8, 12, 20, 24, 36, 44 and 48weeks during treatment and 6months after treatment ceased. A mixture of two different genotypes was detected in the pretreatment samples from five patients and the minor genotype constituted 0.02 to 38%. A transient or permanent change of the predominant genotype was observed in six patients. In three cases genotype 3 was replaced as the predominant genotype by genotype 4, in two cases genotype 3 was replaced by genotype 1, and in one subject genotype 1 was replaced by genotype 4. The PBMCand serum-derived sequences were frequently discordant with respect to genotype and/or genotype proportions. In conclusion, pre-existing minor HCV genotypes can be selected rapidly during antiviral treatment and become transiently or permanently predominant. In coinfections involving genotype 3, genotype 3 was eliminated first from both the serum and PBMC compartments. The PBMCand serum-derived HCV sequences were frequently discordant with respect to genotype and/or genotype proportions, suggesting that they constitute separate compartments with their own dynamics. INTRODUCTION Hepatitis C virus (HCV) is characterized by extreme genetic diversity resulting from a lack of proofreading activity of viral RNA-dependent RNA polymerase and an extremely high replication rate, with over 10 new viral particles being released per day [1, 2]. Consequently, the HCV populations in infected hosts exhibit a high heterogeneity of circulating molecular variants termed quasispecies, or genotypes and subtypes when they differ by more than 30% [3, 4]. A number of studies have revealed that the high molecular variability of HCV may affect such important mechanisms of viral–host interaction as escape from specific immune response, determination of cell tropism and replication capacity [5–7]. HCV variability also correlates with spontaneous elimination of the virus [8, 9] and antiviral treatment outcome [10, 11]. The dynamics of infection with more than two distinct HCV strains were investigated in clinical settings such as blood recipients exposed to virus-positive blood from two different donors [12] and HCV-positive transplant recipients superinfected by infected grafts [13]. Studies on the occurrence and dynamics of coinfection or superinfection with different HCV strains revealed the transient coexistence of ‘new’ and ‘old’ variants followed by rapid relative increase of the new or pre-existing strain and suppression of the other strain [12, 14, 15]. This competition among HCV strains probably reflects their different replication fitness in a particular host and/or different sensitivity to immune response [16, 17]. Although the HCV genotypes are more variable than the viral strains, their dynamics in the setting of coinfection or superinfection are likely to be affected by similar driving forces. However, little is known about the dynamics of HCV Received 22 January 2018; Accepted 21 September 2018 Author affiliations: Department of Immunopathology of Infectious and Parasitic Diseases, Medical University of Warsaw, 3C Pawinskiego Street, 02106 Warsaw, Poland; Institute of Physiology and Pathology of Hearing, 17 Mokra Street, Kajetany 05-830 Nadarzyn, Poland; Hospital for Infectious Diseases, 37 Wolska Street, 01-201 Warsaw, Poland; Department of Infectious Diseases, Warsaw Medical University, Warsaw, Poland; Department of Medical Genetics, Medical University of Warsaw, 3C Pawinskiego Street, 02-106 Warsaw, Poland. *Correspondence: Bukowska-Ośko Iwona, ibukowska@wum.edu.pl


INTRODUCTION
Hepatitis C virus (HCV) is characterized by extreme genetic diversity resulting from a lack of proofreading activity of viral RNA-dependent RNA polymerase and an extremely high replication rate, with over 10 12 new viral particles being released per day [1,2].Consequently, the HCV populations in infected hosts exhibit a high heterogeneity of circulating molecular variants termed quasispecies, or genotypes and subtypes when they differ by more than 30 % [3,4].
A number of studies have revealed that the high molecular variability of HCV may affect such important mechanisms of viral-host interaction as escape from specific immune response, determination of cell tropism and replication capacity [5][6][7].HCV variability also correlates with spontaneous elimination of the virus [8,9] and antiviral treatment outcome [10,11].
The dynamics of infection with more than two distinct HCV strains were investigated in clinical settings such as blood recipients exposed to virus-positive blood from two different donors [12] and HCV-positive transplant recipients superinfected by infected grafts [13].Studies on the occurrence and dynamics of coinfection or superinfection with different HCV strains revealed the transient coexistence of 'new' and 'old' variants followed by rapid relative increase of the new or pre-existing strain and suppression of the other strain [12,14,15].This competition among HCV strains probably reflects their different replication fitness in a particular host and/or different sensitivity to immune response [16,17].
Although the HCV genotypes are more variable than the viral strains, their dynamics in the setting of coinfection or superinfection are likely to be affected by similar driving forces.However, little is known about the dynamics of HCV Downloaded from www.microbiologyresearch.orgby IP: 54.70.40.11On: Wed, 24 Jul 2019 11:20 :17 coinfection with more than two genotypes among patients undergoing antiviral treatment.This is largely due to the limitations of routine methods, which are not capable of detecting minor genotype populations [18].Moreover, genotype determination is usually not repeated during therapy or follow-up.The analysis of HCV coinfections is also hindered by the rapid decrease of viral replication after the initiation of the therapy [19].
In the present study we employed next-generation sequencing (NGS) to perform in-depth longitudinal analysis of HCV genotypes in a group of patients who revealed the presence of a newly predominant genotype during antiviral treatment.In particular, we attempted to determine whether the observed transient presence or permanent genotype switch is an effect of superinfection or results from the selection of a minor pre-existing strain.In addition, we analysed peripheral blood mononuclear cells (PBMCs) to elucidate their potential role as a reservoir of different genotypes.

Detection of genotypes in patients' samples
Only four patients had a sufficiently high viral load during therapy to permit genotype analysis by the commercial assay.The genotype was the same as baseline in three subjects (patients #14, #16 and #75), while in patient #9 there was a switch from genotype 3a to genotype 4. Additional genotypes were not identified by the Versant HCV Genotype 2.0 assay in any of the serum samples.
The NGS performed on all HCV RNA-positive serum and PBMC samples provided a mean of 82 241 reads per sample (range from 665 to 181 341).After filtering, the mean number of reads per sample was 40 980 (range from 212 to 95.806).The results of NGS analysis of the 5¢UTR in all patients at baseline, during antiviral treatment and at 6 month follow-up are shown in Table 1.
Two different HCV genotypes were found in every one of the seven subjects in at least one sample.A mixture of genotypes was detected in 9 out of 18 (50.0%) HCV RNApositive serum samples and in 8 out of 19 (42.1 %) HCV RNA-positive PBMC samples.The proportions of the minor HCV genotype ranged from 0.01 to 37.89 % of the total.In 11 of the 17 (64.7 %) samples containing more than 1 genotype, the contribution of the minor strain was below 1 %.More than two genotypes were not detected in any of the samples.
In the pretreatment specimens mixed infection was identified in five subjects; in three cases it was in both the serum and the PBMCs and in two cases it was only in the serum.During treatment two different genotypes were detected in sera from four subjects: #9, #14, #17 and #75 (in a single sample in each) and in PBMCs from three (in two samples in patient #2 and in one sample each in patients #14 and #16).
Six months after the completion of therapy viral sequences were detected in three patients: in one case it was in both the serum and the PBMCs (patient #9) and in two cases (patients #17 and #75) it was only in the PBMCs.Patient #17 was infected with a mixture of genotypes 3 and 1, and the now predominant genotype 1 initially constituted a minor (0.12 %) genotype.All three subjects were diagnosed as SVR-positive based on the results of commercial assay.
A change of the predominant genotype in at least one sample, when compared with the pretreatment sample, was observed in six patients (#2, #9, #16, #17, #29 and #75).In three cases (patients #2, #9 and #16) the predominant genotype changed from genotype 3 to genotype 4, in two cases (patients #17 and #29) it changed from genotype 3 to genotype 1 and in one subject (patient #75) genotype 1 was replaced by genotype 4 as the predominant genotype.In two patients (#17 and pt.#75) the newly emerging predominant genotype persisted for 6 months after therapy, becoming the major viral strain.In both of these cases the presence of viral RNA was restricted to PBMCs.
The presence of HCV RNA and the proportions of different genotypes in the serum and PBMC samples are presented in Table 1 and Fig. 1, while the actual 5¢UTR sequences are shown in Fig. 2.

Verification of PCR and NGS for the detection of genotype mixtures
To determine whether the detected proportions of different genotypes are significantly influenced by the number of initial viral template copies (viral load) a serum sample containing a mixture of genotypes 1 and 4 was serially diluted in HCV-negative serum to contain 10 4 , 10 3 , 10 2 and 10 viral copies per amplification (Table 2).While the last concentration turned out to be insufficient, analysis of three different viral loads gave similar proportions of genotypes (genotype 4 ranged from 41 to 47 % of the total, while genotype 1 ranged from 53 to 62 %).Thus, the assay seems to be semiquantitative over a range of viral loads.
In the second set of experiments we tested the efficacy of our approach to determine different proportions of genotypes.For this purpose clones containing sequences of genotypes 1, 3 and 4 were mixed in a wide range of proportions.As shown in Table 3, both clones were always detected when present and the proportions of genotypes were accurate enough to consider the assay as being at least semiquantitative.
Since the genotyping was conducted on conserved 5¢UTR, which shows relatively few differences between various genotypes, additional experiments and calculations were performed to determine whether the genotypes could have been wrongly assigned.MMLV-based reverse transcriptases have an error rate in the range of 3.7Â10 À5 -6.7Â10 À5 per base [20], Taq polymerase has an error rate of 2Â10 À5 -1Â10 À5 [21] and Illumina sequencing produces by far the highest number of errors: 0.1-1Â10 À2 per base sequenced Neg.
Neg.  On: Wed, 24 Jul 2019 11:20:17 [22].Thus, in the worst case scenario the expected error rate for the whole analysis would be about 1Â10 À2 per site.
When analysing the 5¢UTR genotype differences we determined that the minimal number of nucleotide substitutions necessary to change the genotype identity was three (genotype 4 would be read as genotype 5), while for all other genotype pairs the required number of changes was four and more (not shown).Thus, while the probability of nucleotide change at any particular site was as high as 1 : 100, the probability of changes at three specific sites was only 1 : 10 6 , and the lowest concentration of a minor genotype detected in our study was 0.01 %.The latter is at least two logs away from the theoretical error threshold for genotype misidentification.
This low probability of genotyping error was confirmed empirically by NGS analysis of cloned HCV DNA representing three different genotypes (1, 3 and 4).As seen in Table 4, sequences differing by at least three nucleotides from the cloned sequence were extremely rare.In addition, PCR products from all of the cloned sequences representing different genotypes as well as their mixes (Table 3) were analysed for the presence of discordant genotypes and none were found.

DISCUSSION
Concurrent infection with two or more HCV genotypes (dual or mixed infections) was previously studied in chronically infected patients [23] as well as in transmission settings [24].However, because there is often an extremely low proportion of minor strains, detailed analysis of mixed infections is difficult.Commercial assays and such wellestablished methods as restriction fragment length polymorphism (RFLP) analysis, line probe reverse hybridization or cloning followed by sequencing (usually limited to 10-20 clones) provide only limited sensitivity [18,25,26].NGS offers a new approach to the analysis of mixed infections, as its sensitivity largely surpasses all the above methods [27].
In the current study we employed NGS to analyse sequential serum and PBMC samples in seven patients with transient presence or permanent genotype change during antiviral treatment.A mixture of two different genotypes was detected in the pretreatment samples from five patients and the minor genotype was determined to constitute 0.02 to 38 %.During treatment, a transient or permanent change of the predominant genotype was observed in six patients.In one patient (patient #75) the minor genotype only constituted 0.01 % of the total virus during treatment, but was absolutely predominant (100 %) after treatment.Six months after the cessation of therapy three patients were found to be HCV RNA-positive; while the genotype that had been dominant at pretreatment was still predominant in onepatient, in the remaining two patients the minor genotype was now predominant.The PBMC-and serum-derived sequences were frequently discordant with respect to genotype and/or genotype proportions.
The appearance of a new HCV genotype during treatment could result from the selection of a pre-existing strain or, alternatively, it could be the result of superinfection.NGS analysis of pretreatment serum and PBMC samples identified the presence of a minor genotype in five out of seven patients, demonstrating that the former scenario was very likely.In the remaining two patients, who presented with an additional genotype during treatment, the superinfection could not be totally excluded.However, it seems unlikely as the patients were not injecting drugs at that time and had no other identifiable risk factors.Thus it is possible that the minor strain was present even below the detection limit of the already sensitive NGS assay.
Our findings suggest that HCV mixed-genotype infections could be much more common than is indicated by routine detection.Depending on the study group and the techniques employed, the frequency of such infections was reported to range from less than 1 % [28] up to more than 50 %, and was highest among HIV-coinfected patients, patients with bleeding disorders and intravenous drug users (IDUs), all of whom have a high risk of repeated HCV exposure [29,30].Coinfection or superinfection with different HCV genotypes is possible as the immune response after primary infection is not protective against subsequent exposure to different viral strains or or even the same viral strain, and previous reports have demonstrated that IDUs can acquire and clear multiple consecutive HCV infections [15,31].
Our results show that different HCV genotypes are often present in very low proportions and thus cannot be detected by routine methods.In four of our patients the minor genotype population represented less than one per cent (0.01 to 0.17 %).Interestingly, in all four patients the minor variant overtook the major strain to become the dominant genotype in serum and/or PBMCs.Without knowledge of the presence of preexisting infection this could be misdiagnosed as a superinfection event.The changes in the predominant genotypes in our patients are likely to have been the result of the direct effect of the antiviral treatment, as in all but one of the cases they took place during the therapy.The rapid effect of antiviral treatment on HCV population is evident from the almost immediate precipitous drop of viral load observed in treated patients [19].
The observed genotype switch itself could be due to an interferon-boosted immune response against the more treatment-sensitive, and consequently more rapidly eliminated, genotype 3. The different sensitivity of HCV genotypes to IFN and ribavirin treatment is well known and it was recommended that genotypes 1 and 4 be treated for twice as long as genotype 3 [32].In four patients from our study (#2, #16, #17 and #29) who initially showed dual infection with genotype 3 and either genotype 1 or 4, the former was the first to be suppressed or eliminated.However, in patient #9 the surpassing of genotype 3 by genotype 4 was only temporary, as the former returned and became the only genotype to be detected.Our study is the first to document the likely higher sensitivity of genotype 3 to treatment compared to genotypes 1 and 4 in the setting of coinfection, where competition between genotypes is direct.
In patient #2 the initial proportion of genotypes 3 and 4 in the serum was 37.9 to 62.1 %, respectively, while in the PBMCs collected at the same time it was 98 to 2 %, respectively.Interestingly, during therapy only the PBMC samples remained HCV RNA-positive in this patient and the initially minor genotype 4 took over, representing nearly 100 per cent of the total viral population.Similarly, inverted proportions of genotypes 3 and 4 were found in the serum and PBMCs from patient #16 at the eighth week of treatment.These observations are compatible with the existence of viral compartmentalization in cells of the immune system and it was previously found that PBMC-and serum-derived HCV populations often differ from each other, particularly in the setting of HIV coinfection [6,33].
The replication of HCV in cells other than hepatocytes is a generally accepted fact [34] and this phenomenon seems to be facilitated by HIV coinfection [35].As our patients were all HIV-positive, the replication and subsequent compartmentalization of HCV in PBMCs was not unexpected.
It was previously reported that HCV is able to persist longer in the cells of the immune system than in serum [36].While in the current study we have not observed a substantially longer HCV persistence in PBMCs compared to serum during therapy, 6 months after the end of treatment two patients had HCV RNA in their PBMCs but not in their serum.It is noteworthy that in the latter two patients the persisting genotype was different than the one that had initially been dominant.
NGS offers unparalleled sensitivity for the analysis of mixed genotype infections.In patient #75 we were able to detect a minor genotype that constituted as little as 0.01 % of the total viral population.This result is similar to the results reported by Buckton et al. [37], who identified minor HCV variants representing 1 : 10 000 of the total population using real-time pyrosequencing of the digested 5¢UTR amplicon.
Thus the sensitivity of NGS with regard to the detection of mixture of HCV variants is almost as high as that described for strain-specific NS5B region amplification, which ranges from 0.01 to 0.001 % [12].However, NGS has some important limitations, such as common environmental contamination and index mispriming, which may result in the assignment of sequences to the wrong samples.Furthermore, it may produce a bias of genotype ratios due to the stochastic phenomena inherent in any low-copy-number analysis [38].
In conclusion, using highly sensitive NGS, we found that pre-existing minor HCV genotypes can be selected rapidly during antiviral treatment to become transiently or permanently predominant.Whenever present, genotype 3 was the first to be eliminated from both serum and PBMC compartments.PBMC-and serum-derived HCV sequences were frequently discordant with respect to genotype and/or genotype proportions, suggesting that they constitute separate compartments with their own dynamics.

METHODS Subjects
The study group encompassed seven HCV/HIV-coinfected subjects selected from a cohort of 37 patients described in detail elsewhere [39], who were treated for chronic hepatitis C between June 2005 and July 2007.We studied patients in whom single-strand conformational polymorphism (SSCP) analysis revealed changes in the 5¢UTR during treatment or follow-up, and subsequent cloning and sequencing showed the presence of a new genotype.Serum and PBMC samples were collected right before the commencement of the therapy (time '0'), and at 2, 4, 6, 8, 12, 20, 24, 36, 44 and 48 weeks, as well as 6 months after treatment.The characteristics of all seven patients are presented in Table 5.In six patients HCV infection was probably contracted by injecting illicit drugs, while one patient was most likely infected by her sexual partner.Patients received what was then routine 24-week (when infected by genotype 3) or 48-week (when infected by genotype 1) therapy with peginterferon alpha-2b (Peg-Intron; Schering-Plough Corp, Nenilworth, NJ, USA), 1.5 µg/kg/ weekly combined with ribavirin (Rebetol; Schering-Plough) at a dose of 1000 mg day À1 for patients weighing <75 kg or 1200 mg day À1 for patients weighing !75 kg.The infecting genotype at baseline was determined by the Versant HCV Genotype 2.0 assay (LiPA, Siemens Healthcare, Germany).The same assay was used for the analysis of all subsequent serum samples.In addition, all HCV RNA-positive serum and PBMC samples were analysed for HCV genotype by NGS as later described.The HCV viral load was determined using the Cobas Amplicor HCV monitor test v2.0;(Roche Diagnostics, Pleasanton, CA, USA).

5¢-UTR amplification and NGSg
The serum and PBMC samples were subjected to total RNA extraction with Trizol LS Reagent (Life Technologies; Carlsbad, CA, USA) and the HCV 5¢-UTR was amplified as described previously [40].Total RNA was extracted from 250 µl of serum or approximately 10 6 cells and suspended in 20 µl of water, 5 µl of which was then incubated for 30 min at 36 C in a 15 µl reaction containing 50 pmol of the antisense primer (5¢ TGA/GTGCACGGTCTACGAGACCTC 3¢; nt 342-320), 1Â RT buffer (Thermo Fisher Scientific), 5 mmol l À1 dithiothreitol, 5 mmol l À1 MgCl 2 , 1 mmol l À1 dNTP and 20 units of Moloney murine leukaemia virus RT (Thermo Fisher Scientific).After being heated to 99 C for 5 min, 50 pmol of the sense primer (5¢ A/GAC/ TCACTCCCCTGTGAGGAAC; nt 35-55), 3.5 µL 10Â polymerase chain reaction (PCR) buffer II (Thermo Fisher Scientific) and 5 units of Taq DNA polymerase (Thermo Fisher Scientific) were added, and the volume was adjusted to 50 µL.Amplification was run as follows: initial denaturing at 94 C for 4 min, followed by 50 cycles of 94 C for 1 min, 58 C for 1 min and a final extension at 72 C for 7 min.The PCR products were indexed for NGS analysis performed on MiSeq Sequencer (Illumina, San Diego, CA, USA).The sequences of the specific primers have been published previously [41].PCR indexing was performed in 25 cycles, each comprising three steps: 94 C for 1 min, 58 C for 1 min and 72 C for 1 min.The amplified products, which were 400 bp in length, were extracted from the agarose gel (QIAquick Gel Extraction kit; Qiagen, Hilden, Germany) and quantified on a Qubit 2.0 Fluorometer (Life Technologies, Carsbad, CA, USA).The quality and the average length of the sequence library of the indexed samples were assessed using a bioanalyser (Agilent Technologies, Santa Clara, CA, USA).Each sample was sequenced twice.
To detect contamination, at least one HCV-negative sample was included in each reaction series.Additionally, to prevent index mispairing during demultiplexing, the forward and reverse index primers were only used once in each run.HCV sequences were not detected in any of these negative control samples.
Control cloned templates 5¢UTR amplicons of genotype 1, 3 and 4 were cloned into a plasmid vector (pCR2.1;TOPO, Invitrogen, Waltham, USA) and verified by bidirectional Sanger sequencing.These plasmids were mixed in different proportions, amplified and subjected to NGS sequencing as described earlier.

Fig. 1 .
Fig. 1.Distribution of HCV genotypes as determined by NGS analysis of 5¢UTR amplicons in sequential serum and PBMC samples in seven patients receiving antiviral treatment with peg-interferon and ribavirin.0 w, right before start of therapy; w, weeks of treatment; 6 m, 6 months after completion of therapy.

Fig. 2 .
Fig. 2. NGS nucleotide sequence alignment of the 5¢UTR of HCV amplified from sequential serum and PBMC samples from seven patients who revealed a new genotype during antiviral treatment.0 w, right before commencement of therapy; w, weeks of treatment; 6 m, 6 months after completion of therapy; s, serum; p, PBMCs; G, genotype.The number of reads representing a particular sequence is shown in parentheses.

Table 1 .
Detection by NGS of HCV genotypes in 5¢UTR amplicons from 7 patients who revealed a newly predominant genotype in serum and/or in PBMCs during or after antiviral treatment.

Table 2 .
Results of NGS of a serially diluted serum sample containing a mixture of HCV genotype 4 and genotype 3

Table 4 .
Errors within the 5¢UTR sequence of HCV genotypes (G) 1, 3 and 4 generated during PCR reaction and NGS

Table 5 .
Baseline characteristics of seven HCV/HIV-coinfected patients undergoing therapy with interferon and ribavirin in whom a new genotype was detected [42]rior to therapy.c,sustainedvirological response (SVR) was defined as the absence of detectable HCV RNA in the serum 6 months after the end of therapy using a high-sensitivity commercial assay[42].IDU, intravenous drug user.NA, not applicable.Downloaded from www.microbiologyresearch.orgby IP: 54.70.40.11On: Wed, 24 Jul 2019 11:20:17