Resistance-associated substitutions (RASs) in hepatitis C virus (HCV) appear upon failure of treatment with direct-acting antivirals (DAAs). However, their origin has not been clarified in detail. Among 11 HCV genotype 1b patients who experienced virologic failure with asunaprevir (ASV)/daclatasvir (DCV), 10 had major NS5A L31M/V-Y93H variants after treatment. L31M/V-Y93H variants were detected as a minor clone before therapy in 6 patients and were the most closely related to the post-treatment variants by phylogenetic tree analysis in 4 patients. Next, to consider the involvement of a trace amount of pre-existing variants below the detection limit, we analysed human hepatocyte chimeric mice infected with DAA-naïve patient serum. L31V-Y93H variants emerged after treatment with ledipasvir (LDV)/GS-558093 (nucleotide NS5B inhibitor) and decreased under the detection limit, but these variants were dissimilar to the L31V-Y93H variants reappearing after ASV/DCV re-treatment. Finally, to develop an infection derived from a single HCV clone, we intrahepatically injected full-genome HCV RNA (engineered based on the wild-type genotype 1b sequence) into chimeric mice. A new Y93H mutation actually occurred in this model after LDV monotherapy failure. In conclusion, post-treatment RASs appear by 2 mechanisms: the selection of pre-existing substitutions among quasispecies and the generation of novel mutations during therapy.