Hepatitis C virus genotyping assays are usually based on the sequence analysis of an amplified segment of the genome, which is commonly the 5' untranslated region. Currently, 5'UTR based assays are reasonably accurate with more than 95% concordance with genotypes that have been identified by nucleotide sequencing of the NS5B region or other coding regions of the HCV genome . In our study, LiPA and NS5B sequencing showed 100% agreement at the type level. These results are similar to those reported by another study using the 5'UTR with a type agreement of 99.5% . Another study with 357 samples from French blood donors has demonstrated 100% agreement between 5'UTR and NS5B sequence analysis regarding type classification . However, depending on the geographical region, genotype identification based on 5'UTR may be unreliable because some genotype 6 variants that have been found in Southeast Asia have identical 5'UTR sequences to those of genotypes 1a or 1b . In Brazil, the most prevalent HCV genotypes are 1, 2 and 3, whereas genotypes 4 and 5 are rarely identified and genotype 6 has not been previously described .
When we compared the subtyping results of LiPA and partial NS5B sequencing, we found 47.9% (82/171) of misclassification among samples, including non-subtypable discrepant and inconclusive results. LiPA could not discriminate 39.6% (40/101) of the genotype 1 samples at the subtype level. LiPA misclassified three samples of genotype 1b as genotype 1a and four samples of genotype 1a as genotype 1b. LiPA did not accurately discriminate between genotypes 1a and 1b, which may be attributed to the lack of discriminating power of LiPA probes 5 and 6. The difference between these probes covers a single nucleotide change in the HCV genome, which is an A to G transition at position -99 within the 5'UTR. There is evidence that this transition represents a sequence polymorphism that cannot be used to differentiate between subtypes 1a and 1b . Accordingly, we found that LiPA yielded inconclusive results for 11 genotype 1 samples, which were classified as genotype 1a/1b.
Concerning genotype 2, 26 samples were investigated in our study. The small sample size is due to the low prevalence of this genotype in Southern Brazil (less than 4%). In our study, only five of the 26 genotype 2 samples were simultaneously subtyped as 2b by LiPA and NS5B sequence analysis. The 21 remaining genotype 2 samples were not accurately subtyped by LiPA. After performing the sequence analysis of the NS5B region, 11 of these 21 samples were classified as genotype 2b, nine samples as genotype 2 c and one sample as genotype 2a. This poor resolution was expected because LiPA targets 5'UTR, which may result in incorrect identification of subtypes 2a and 2 c due to the lack of nucleotide polymorphisms . Similarly, a previous study has reported that 61% of the subtypes of genotype 2 were misclassified by 5'UTR-based genotyping in France . Another study has compared genotyping using real-time PCR targeting the NS5B region and LiPA v.1.0 and has shown that LiPA did not subtype 68 of the 295 samples (23%) .
The low reported prevalence of mixed HCV genotype infection suggests that this type of infection is a rare event and is influenced by the population that is studied and the genotyping method that is employed . A study with 600 patients has shown that 2.2% of patients displayed evidence of mixed infections. These sera have been screened using sequencing and serological assays in parallel . Accordingly, only two laboratories have correctly identified one of the three samples containing mixed HCV genotypes in a national evaluation study that was conducted in France . Cases of mixed HCV genotypes have not been confirmed in our study. Although all of the samples were characterized as infected based on a single genotype, direct sequencing is not the gold standard method to detect mixed infection. The best approach to reliably detect mixed-genotype infections is PCR amplification followed by cloning of the PCR products or the use of next-generation sequencing methods. Unfortunately, this approach is unfeasible in a routine clinical setting . The results that are reported herein are similar to those found in another study that has been conducted in Brazil using sequencing to genotype 1,688 samples from chronic HCV patients, who did not show evidence of mixed infection .
In our study, the 5'UTR-based genotyping method (LiPA) showed higher efficiency than the NS5B region to generate PCR amplification products. Primers targeting the 5'UTR, which is a highly conserved region, produce fewer false negative results in PCR reactions for HCV detection. The major factor to be considered when evaluating the generation of false negative PCR results is the existence of different strains or subtypes of HCV . Because the polymorphic positions are not homogenously distributed within the NS5B region, difficulty to generate amplicons may be due to inadequate primer design, choice of highly polymorphic annealing sequences or low viral load .
Currently, there is a new available commercial version of LiPA (v.2), which contains probes targeting both the 5'UTR and the core regions of the HCV genome, which improves the accuracy of subtyping, particularly regarding the discrimination between subtypes 1a and 1b. By the time this study was conducted, this version was not available for use in Brazil.
In the future, a protective vaccine and the availability of HCV-specific antiviral drugs that are based on proteases or polymerase inhibitors will probably require genotyping at the subtype level. This genotyping approach will improve our understanding of the genetic diversity of HCV. Genotyping methods that are based on the 5'UTR are important tools for routine clinical purposes because they are sufficiently precise at the genotype level. Our results show that HCV genotyping using partial NS5B sequence analysis is an efficient method that allows accurate discrimination of subtypes and might be an effective tool to study the molecular epidemiology of HCV.