Background: HAART can control viral replication in HIV-1 infected individuals, but can’t eliminate the virus due to viral persistence in latent reservoirs. Eradication strategies have focused on reactivation of latent HIV-1 in the reservoir without global T cell activation. Our recent study suggested reactivation of latent HIV-1 alone might not be enough, and stimulation of HIV-1-specific CTL response facilitated clearance of the latent reservoir. However, plasma HIV-1 develops CTL escape mutations to evade immune response. CTL escape of HIV-1 has been studied extensively through the analysis of plasma virus, but little is known regarding the potential presence of CTL escape variants in the latent reservoir. This raises a critical issue for CTL-based eradication strategies: if CTL escape variants are archived in the latent reservoir, the host CTL response may fail to recognize and clear infected cells after latency is reversed. Methodology: Proviral DNA from HAART-treated patients was extracted, Gag was amplified by PCR and deep-sequenced in search for CTL escape mutations. Latent HIV-1 was recovered from patients’ resting CD4+ T cells using a viral outgrowth assay. Gag from the outgrown viruses was sequenced and compared with the proviruses. Outgrown viruses were used to infect CD4+ T cells from the same patients. Autologous pre-stimulated or nonstimulated CD8+ T cells were cocultured with infected CD4+ T cells to access killing. Results: CTL escape variants dominate in the latent reservoirs of chronic phrase-treated patients, but not in acute phrase-treated patients. After reactivation, outgrown viruses still carried the same CTL escape variations with provirus in the reservoir. CD4+ T cells infected by these escape variants were insensitive to the relevant epitope-specific CTL clones, but can be killed by broad spectrum CTL response. More importantly, we were able to identify and stimulate subdominant CTL clones in chronically infected patients which can recognize and kill autologous target cells infected with these escape variants. Conclusions: Our study demonstrates that the dominant presence of CTL-resistant viruses in the latent reservoir poses great hurdles to the viral eradication. Boosting broad spectrum or more specifically, subdominant CTL responses is required for the elimination of viral latent reservoir.