Study of hepatitis B virus quasispecies by ultra-deep pyrosequencing: Resolving the nitty-gritty

Abstract

Rodriguez et al.1 described the dynamics of hepatitis B virus (HBV) resistance to adefovir by ultra-deep pyrosequencing (UDPS), a validated technique to study HBV quasispecies.2 Their article reports relevant results, which, however, require further examination. HBV polymerase variants resistant to nucleos(t)ide (NUC) treatment (NUCsRV) were detected in 83% of NUC-naïve cases, in keeping with previous reports, with prevalence restricted by sensitivity: 83%-100% at 0.03% cutoff, 12%-20% at 1% cutoff (Table 1). The authors describe an interesting computational method assuming exponential outgrowth of selected variants. Nonetheless, the relatively low coverage of each sample, at best 4,853 sequences, is questionable with respect to the lowest cutoff reported (0.03%) because variants represented by a single sequence (e.g., patient AH, with viral load 2.1 log IU/mL and 0.34% of rtT184S/A/I/L) would be considered correct, a singular notion that is not backed by previous experience (Table 1). Thus, UDPS standardization (e.g., minimal viral load and sequence coverage for standardized sensitivity, computational algorithms) must still be resolved. The authors state that some previous UDPS studies assessing preexisting NUCsRV, including ours,3 have methodological flaws: lack of sensitivity, no error rate consideration, and/or no linkage studies. This is surprising, because we reported the lowest cutoff to date (0.03%), established by clonal sequence processing and Poisson filtering.3 Rodriguez et al. report different cutoffs for each variant, 0.03% being the lowest, whereas in our study 0.03% was used for all positions (0.007% average clonal error). So sensitivity was not lacking. Furthermore, we did perform a linkage study. The analysis demonstrated HBV quasispecies dynamics in a sequentially NUC-treated patient, with emergence of complex combined variants, particularly after entecavir. Lastly, although it is a secondary issue, questions also arise about patient classification in their study. The 9.57% of rtM204V/I at baseline in “naïve” patient C suggests prior lamivudine treatment (as the authors indicate). Similarly, the 11% rtA181T/V in “naïve” patient AA, 100 times greater than previously UDPS-detected at baseline (Table 1), also suggests prior NUC therapy. As the study investigates pretreatment HBV variants, inclusion of these cases, detectable by routine methods, raises the concern that some variant populations may have been overestimated. In conclusion, UDPS is consolidated for HBV quasispecies research, as supported by Rodriguez et al. However, certain factors, such as those mentioned here, should be kept in mind to avoid misinterpretation of the state-of-the-art in UDPS studies. Francisco-Rodriguez Frias, M.D.1,2 David Tabernero, M.S.1,2 Rafael Esteban, M.D.1,3 Maria Buti, M.D.1,3 1Centro de Investigación Biomédica en Red de Enfermedades Hepáticas y Digestivas (CIBERehd) Instituto de Salud Carlos IIIMadrid, Spain 2Biochemistry Department Hospital Universitari Vall d'Hebron (HUVH)Vall d'Hebron Institut de Recerca (VHIR)Universitat Autònoma de Barcelona (UAB)Barcelona, Spain 3Liver Unit Hospital Universitari Vall d'Hebron (HUVH)Vall d'Hebron Institut de Recerca (VHIR)Universitat Autònoma de Barcelona (UAB)Barcelona, Spain