The European Prevalence of Resistance Associated Substitutions among Direct Acting Antiviral Failures

Background: Approximately 71 million people are still in need of direct-acting antiviral agents (DAAs). To achieve the World Health Organization Hepatitis C elimination goals, insight into the prevalence and influence of resistance associated substitutions (RAS) is of importance. Collaboration is key since DAA failure is rare and real-life data are scattered. We have established a European collaboration, HepCare, to perform in-depth analysis regarding RAS prevalence, patterns, and multiclass occurrence. Methods: Data were extracted from the HepCare cohort of patients who previously failed DAA therapy. Geno—and subtypes were provided by submitters and mostly based on in-house assays. They were reassessed using the Comet HCV subtyping tool. We considered RAS to be relevant if they were associated with DAA failure in vivo previously reported in literature. Results: We analyzed 938 patients who failed DAA therapy from ten different European countries. There were 239 genotypes (GT) 1a, 380 GT1b, 19 GT2c, 205 GT3a, 14 GT4a, and 68 GT4d infections. Several unusual subtypes (n = 15) (GT1b/g/l, GT3b, GT4k/n/r/t) were present. RAS appeared in over 80% of failures and over a quarter had three or more RAS. Multiclass RAS varied over target region and genotype between 0–48%. RAS patterns such as the Q30R + L31M and Q30R + Y93H in GT1a, the L31V + Y93H and L31V + Y93H for GT1b, and A30K + L31M and A30K/V + Y93H for GT3a all occurred with a prevalence below 5%. Conclusion: RAS occur frequently after DAA failures and follow a specific genotype and drug related pattern. Interpretation of the influence of RAS on retreatment is challenging due to various patterns, patients’ characteristics, and previous treatment history. Moving towards HCV elimination, an ongoing resistance surveillance is essential to track the presence of RAS, RAS patterns and gather data for a re-treatment algorithm.


Introduction
Approximately 71 million people worldwide are infected with the hepatitis C virus (HCV), from which only a small proportion received curative direct-acting antiviral agents (DAAs) [1]. In addition, the HCV epidemic is still ongoing in several population groups, such as men-who-have-sex-with-men and people-who-inject-drugs [2][3][4].
Moving forwards to achieving the World Health Organization (WHO) HCV elimination goals, the highly effective DAAs play a major part in obtaining elimination. Although DAAs are well tolerated and have an outstanding efficacy, in rare cases patients fail to obtain sustained virologic response (SVR). When failure occurs, this is often in the presence of resistance associated substitutions (RAS) [5][6][7][8][9].
As many patients are still in need of DAA therapy, a substantially large number of patients can fail first-generation DAA treatment [10,11]. Although newer, more robust, and pan genotypic regimens are available, failure remains present. Especially in resource limited settings where regimens are often older (more prone to resistance) and limited by the number of DAA compounds, and geno-/subtypes are more challenging [12]. RAS may jeopardize the elimination goals resulting in continuous transmission and disease progression [13].
RAS exist in two different forms: as polymorphism, or as mutations that emerge under the pharmacological pressure of the DAAs. In the former case, RAS are characteristics of that specific HCV virus, and will persist indefinitely. In the latter, reversion to wild-type could occur in a variable percentage of patients, within months in the NS3 region and years in the NS5A [14,15].
Several studies show that the existence of certain RAS can impact SVR rates and complicate retreatment options, especially among treatment experienced patients [16][17][18][19][20]. Although newer and more potent DAA regimens are available as first or second-line therapy, these are not always accessible or used [8,21]. This requires more knowledge regarding RAS to tailor first-line therapy and guide second-line therapy. Currently, available epidemiological data are scattered among different centers and therefore often generated from single center studies. Moreover, sample sizes are small. In addition, to our knowledge only a single study assessed the magnitude of RAS in a real-life cohort [22].
Without aggregated data, the identification of clinically relevant RAS circulating in the real-life population the first years after the wide availability of the DAAs, and their detailed analysis in relation to treatment data, is extremely difficult. To examine the prevalence of known and putative RAS after DAA failure, with a particular focus on RAS patterns and type of multiclass RAS, we used data from the European HepCare database from 2015 until 2019 [13]. HepCare involves multiple centers from 16 countries in Europe and the Mediterranean region.

Study Population
HepCare is an ongoing observational multicentre study including several clinical sites within Europe and the Mediterranean. HepCare covers 16 different countries including Belgium, Cyprus, Denmark, France, Germany, Italy, Israel, Luxembourg, the Netherlands, Poland, Portugal, Romania, Russia, Slovenia, Spain, Turkey.
In HepCare, baseline (prior to receiving DAA treatment) or failure (after failing DAA therapy) HCV sequences were included from adults (≥18 years of age). Sequence data were combined with available clinical, virological, and demographical data and stored in a secure database. In addition, classification of genotype and subtype was performed at the clinical study site and mostly through in-house assays. As a quality control check all genoand subtypes were reassessed using the COMET HCV subtyping tool [23].

Ethics Statement
All sequences were derived at the clinical study sites based on local protocols. Ethical approval for the study protocol was reviewed by the Medical Ethics Committee Erasmus MC (MEC-2018-1271).

Inclusion and Exclusion Criteria
For this analysis we subtracted data from HepCare of individuals, who failed DAA therapy and who had sufficient DAA treatment data available (name of treatment) from 2015 until the beginning of 2019. In addition, a viral sequence of NS3/NS5A and/or NS5B at failure should have been available covering the protein region of the failed DAA compound. We excluded failure sequences from individuals with insufficient or unknown treatment data, who discontinued treatment for several reasons (e.g., cytotoxicity, lost to follow-up, diseased) and who had a reinfection and therefore no virological failure related to resistance.

Analysis of RAS
All sequences were aligned with their reference strain before analysis [24]. RAS at failure were defined as clinically relevant amino-acid changes, based on previously reported literature, at the analyzed positions compared to the reference strain [25][26][27]. In addition, to known RAS, we analyzed all amino-acid changes at clinically relevant positions across different geno-and subtypes; 36,41,43,54,55,80,122,155,156,168 Table S1) [25,28]. The cut-off value for variant calling was set at 15-20% for population sequencing and 5% for next generation sequencing. The overall RAS prevalence was calculated and listed with a 95% confidence interval (CI). For every gene segment (NS3, NS5a, and NS5B) our analysis was limited to patients who received an inhibitor target to that gene. In addition, we made a distinction between patients who failed an NS5B nucleoside analogue or an NS5B non-nucleoside analogue. Categorical data were statistically compared using the chi-square test or fisher's exact, if appropriate, using R.

European Prevalence of RAS after DAA Failure
Among the most common genotypes (n = 925) 82% of patients failed on DAA therapy with at least one RAS. Consequently, only 18% of patients failed with no RAS. Over one third of the patients (36%) had one RAS, 22% had two, and almost a quarter (24%) of the failures had three or more RAS in any of the target genes.
No individuals with GT2c, GT3a, GT4a/d patients failed on NS5B non-nucleoside analogues with an available NS5B sequence.

RAS in Unusual Subtypes Defined as GT1 Non-A/B, GT3 Non-A, and GT4 Non-A/D
Although rare, several unusual subtypes were identified in our cohort. Unusual subtypes are defined as GT1 non-a/b, GT3 non-a/d, and GT4 non-a/d. In two cases the geno-subtype could not be determined (Table 2). In the first case, the submitter provided GT1a while COMET mentioned 1b and 1l. In the second case, the submitter provided GT1d while COMET identified a GT1b.
Several RAS were detected after failure (Table 2). In addition, in unusual subtypes RAS in the NS5B region are uncommon. Sample HC_06 only had an NS3 sequence available but failed a SOF + LDV regimen. Therefore it is unclear if any RAS emerged after failure. Although no longer used in clinical practice in Europe, in total 16 (1.7%) individuals failed (GT1a n = 3, GT1b n = 12, GT3a n = 1) an ASV + DAC regimen of which 10 relapses, two non-responders, and three breakthroughs. All GT1a patients had RAS in both NS3 + NS5A (n = 3). In one patient a V36M + Q80K + D168E and Q30E + L31C pattern emerged. In GT1b half of patients had an NS3 RAS (n = 12) and in all the patients with available NS5A sequence (n = 2) RAS were present. In NS3 the pattern of Y56F/H + D168E/V often occurred. In NS5A we identified the combination of L28G + L31M and L31M + Y93H. In the GT3 patient only the NS5A was sequenced with the A30K + P58A + A62S pattern after failure.
There were three patients with a GT2c who failed a (OMB + PTV/r + DAS) regimen, from which two were non-responders. In the first sample the L36V + V158M + D168V NS3 RAS pattern was found combined with the F28C + P58S in NS5A. In the second sample the D168V RAS occurred in the NS3 region combined with the F28C + C92S in NS5A. Additionally, the third failure also had a F28C mutation in NS5A.
Only one GT4a patient failed a (OMB + PTV/r + DAS) regimen in our cohort. No NS3 and NS5A RAS were found after failure and no NS5B sequence was available. There were six GT4d failures of which in one the Y56H + D168A and L28ALSV + M31V + T58P pattern occurred. In 83% (43.6-97.0) of failures the T58P was present. RAS at position 93 occurred with a prevalence of 33% (9.7-70) to C and CS.

Simeprevir + Daclatasvir (SIM + DAC)
Although no longer used in clinical practice, this combination was used at the beginning of the NS5A-era and in our collection four patients with a GT1a infection failed on SIM + DAC. After failure, the multiclass pattern R155K (NS3) + Q30E (NS5A) pattern occurred with a prevalence of 50% (15.0-85.0) and the R155K + Q30K in 25% (4.6-69.9). In addition, in GT1b patients RAS occurred both in NS3 and NS5A after failure: Y56F + D168V + L31M + Y93H with a prevalence of 40% (11.8-76.9). Two patients with a GT4d infection failed on a SIM + DAC regimen. The first patient had a breakthrough with RAS both present in NS3 (A156G + D168E) and NS5A (L28V + R30S + T58P) after failure. The second patient relapsed with no RAS in NS3 and the M31V present in the NS5A region at failure.
Solely one GT2c patient failed a SOF-DAC regimen with a L31M mutation in the NS5B region and no other RAS were detected. One GT4a patient was treated with SOF + DAC, no NS5B mutations were detected and unfortunately no NS5A sequence was available.
One GT4a patient failed a SOF + DAC treatment. RAS were only present in the NS5A region (L30H) and none in the NS5B region. In the GT1l sample of the patient who failed SOF + LDV only the Q30R mutation was present after failure ( Table 2).
Among GT3a failures no A30K + Y93H occurs in the NS5A region. The A30V + Y93H was rare with a prevalence of 3.6% (0.6-17.7). Other RAS were mainly present at position 62 (A/P/T/V) all with a prevalence < 10%. GT3A patients who failed an LDV-containing regimen had a significantly lower RAS prevalence than those who failed daclatasvir (DAC) (p-value < 0.001). In the NS5B region no RAS were detected. In the GT3b sample one RAS was detected in the NS5A region (V31M).
Ten GT4a patients were treated with SOF/LDV. In one out of the two patients with an available NS5B sequence the S282T RAS was found. In 14% the NS5A RAS pattern L30R + L31M and L30H + Y93H occurred.
One patient had a breakthrough on SOF + SIM with a GT4a infection. The patient had multiclass RAS after failure as the Q80R was found in NS3 and the S282T was present in NS5B. From the GT4d patients who failed a SOF + SIM regimen (n = 14) RAS only occurred in the NS3 region and mostly at position 168 (A/E/V).

Sofosbuvir + Velpatasvir (SOF + VEL)
There were five patients with a GT1a infection that failed SOF + VEL. From those one had a relapse. In 60% of failures the M28T mutation was present of which in 20% combined with the Q30L + Y93H and in 20% with the L31M. From only two patients the NS5B sequence was available from which in one case the C316Y RAS was present.
There were 4 GT1b with a relapse. Unfortunately, no NS5A sequences were available. In the NS5B region 50% of samples had the L159F and in one case this was combined with the C316N.

Discussion
Although DAAs have a high efficacy resulting in a low number of virological failures, the collaboration of several clinical and virological centres within the HepCare consortium allowed the analysis of a high number of real-life DAA failures [13]. Our results showed that RAS are highly frequent after failure and mostly occur in NS3 and NS5A regions. In various cases, complex RAS patterns were present in multiple target genes. Our results identified a great variety of RAS patterns after failure and more specific patterns over geno-/subtype. NS5A failures tend to have the highest number of RAS after failure; NS3 RAS were common, and NS5B RAS uncommon. As NS5A RAS have a very long persistence they are of the hardest potential challenge for retreatment [14,15]. However, despite frequent NS5A-RAS detection many individuals still have retreatment options. For example, individuals with the major Y93H RAS (73% GT1b and 57% GT3A), can still reach a 95% SVR rate with the SOF + VEL + VOX combination [32].
Most of our samples included the most common genotypes and a limited number (~1%) of unusual subtypes, defined as GT1 non a/b, GT2, GT3 non a, and GT4 non a/d. These unusual subtypes are more common in low-middle income countries, however, do appear occasionally in European clinical practices. Within our cohort there are six countries which report unusual subtypes among their patient population. Unusual subtypes are more challenging to treat as SVR rates tend to be lower potentially due to the natural variation at baseline of these subtypes [12,33,34]. Countries where these unusual subtypes are highly prevalent among the countries with the highest HCV prevalence in the world [11]. Moreover, they are low-middle income countries with limited to no resources for HCV sequencing at baseline or after failure. Therefore, aggregating available data are needed to obtain additional knowledge regarding these subtypes.
In many cases RAS were identified which do not hamper the efficacy of second-line regimens. Moreover, significant relevant NS5A RAS patterns to these regimens (SOF + VEL + VOX or G/P) such as the Q30R + L31M and Q30R + Y93H in GT1a, the L31V + Y93H and L31V + Y93H for GT1b, and A30K + L31M and A30K/V + Y93H for GT3a all occurred with a prevalence below 5%. Nevertheless, for patients with RAS who fail a second-line regimen there are still treatment options. Recently, Dietz et al. showed that even after failure of SOF + VEL + VOX, an 81% SVR rate can be obtained using rescue regimens [35]. These results are highly promising and assuring for clinics where RAS testing after failure is not embedded in the standard of care. The newer regimens should be made available in these countries as a rescue regimen. As they are only needed for a minority of the population this should be financially feasible.
When sequencing is available and performed, interpretation of RAS after failure can be quite complicated as HCV has many geno-/subtypes and RAS patterns. Especially, as based on current knowledge and data, no algorithm is constructed to guide decisionbased re-treatment. Therefore, retreatment often requires input of an experienced team. Moving forwards to elimination, this could jeopardize a simplification and decentralization of the HCV care cascade in certain regions, as the population of failures (~2-5% of 71 million who require HCV treatment) still require RAS testing and a team experienced with retreatment [11]. These regions will be the areas where the robust second-generation DAA regimens are unavailable and SOF + LDV or SOF + DAC are pan-genotypically used, or the unusual subtypes are highly epidemic. These are often also the regions where simplification and decentralization of the care cascade are highly needed to treat all the numbers of patients still infected with HCV.
Our study has several limitations. First, due to the retrospective nature of this cohort some data is lacking, e.g., patients could have been treated elsewhere without the knowledge of the current treating physician. Moreover, individuals who failed with lower fibrosis scores were less common in our cohort (13% with F0 and F1) related to the previously installed treatment restrictions (>F2/F3 only). Secondly, most reported data are from the more difficult to treat patients, since these patients are in care of treatment centres with available HCV assays, instead of small hospitals in which DAA treatment and HCV assays were unavailable. Thirdly, failures from countries which do not have the organization or financial resources to perform sequencing are missing in our cohort. Additionally, the study cohort is based on convenience sampling and therefore the study only includes data that was shared. Consequently, not all European countries are included. Nonetheless, the genotype distribution of our cohort is comparable with the European genotype distribution and similar DAA regimens were used in Europe. Fourthly, variation of the number of RAS to other studies and in our data can be due to a different sampling time. In our study samples were drawn with a median time of 15 weeks after end of treatment. Since NS3 RAS tend to disappear mostly within a year after the end of therapy, the RAS prevalence will likely be lower when a second sample would be drawn [14]. Our study found a variable NS3 RAS prevalence among different genotypes, the lowest in GT4. Other studies varied between 53-78% [36,37]. Lastly, our study included limited data on failures of second-generation DAA therapy. However, in many countries GLE + PIB or SOF + VEL + VOX are not present or reimbursed so first-generation regimens are continuing to be used. Furthermore, our cohort reflects the available regimens in several countries. For instance, OBV/PTV + rDAS was the single available regimen in Romania for many years. Our results should, therefore, be interpreted with caution.
Although resistance is rare, RAS prevalence after failure ranges from 0-100% depending on geno-/subtype and target region. Interpretation of the influence of RAS on retreatment is quite challenging due to the various patterns, patients' characteristics, and previous treatment history. Moving forwards to HCV elimination, ongoing HCV resistance surveillance is essential to track the presence of RAS, RAS patterns, and gather data for a re-treatment algorithm. Additionally, more data and knowledge are required on the unusual subtypes so that the best HCV care can be provided, also in low-middle income countries.

Supplementary Materials:
The following are available online at https://www.mdpi.com/article/ 10.3390/v14010016/s1. Table S1. Analyzed amino-acid positions for the different hepatitis C geno-/subtypes. Figure S1. distinct RAS pattern after direct-acting antiviral failure per region per Hepatitis C genotype. Table S2. European RAS prevalence after failure specified over different geno-/subtypes. Table S3. European prevalence of NS5A RAS patterns.  Institutional Review Board Statement: The study was conducted according to the guidelines of the Declaration of Helsinki. All sequences were derived at the clinical study sites based on local protocols. Ethical approval for the study protocol was reviewed by the Medical Ethics Committee Erasmus MC (MEC-2018-1271).

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.