All patients gave written informed consent to their study treatment and to having their data analysed. One-hundred and thirty patients with viral load data available at set point, and before and after treatment interruption lasting at least 7 days, were included in the study. Mean pretreatment set-point viral loads were calculated, if more than one value was available within 6 months before initiation of antiretroviral treatment. If patients underwent more than one STI, only the first

interruption was considered in the analysis. Demographic data for the patient population are summarized in Table 1. The KIR genotype was assessed using sequence-specific primer (SSP) polymerase chain reaction (PCR) [17]. Alleles of KIR3DL1 Smad inhibitor differing in cell surface expression Panobinostat mouse were discriminated by intermediate resolution allele-specific PCR into those carrying alleles with high (*h; 3DL1*001, *002, *003, *006, *008, *009, *015 and *020), low (*l; 3DL1*005 and *007), or no surface expression (3DL1*004) using a combination of PCR SSP protocols previously

described [18-20]. Patients were grouped into those expressing two high expression alleles (*h/*y) and those expressing at least one low-expressing allele (*l/*x) [21]. The KIR3DL1*004 allele, which is not expressed on the cell surface, was analysed separately. KIR3DL1 ligands were typed using a real-time PCR method that much discriminates between the two types of HLA-Bw4, Bw4-80Thr and Bw4-80Ile [22]. The HLA-C35 single nucleotide polymorphism (SNP) (rs9264942) was typed using a pre-designed custom assay using TaqMan chemistry (Applied Biosystems, Foster City, CA). The SNP in HCP5 (rs2395029) was typed by direct sequencing (forward primer 5′-3′ ACGATTCTCCTCACACTTACA; backward primer 5′-3′ TCTCTCCCAAAACCACACTC). Viral load data were compared using nonparametric tests (the Mann–Whitney U-test and the Kruskal–Wallis test). Values are reported as medians and interquartile ranges (IQRs). Correlations

between variables were assessed by calculating Spearman’s rho. Generalized linear models were used to test the impact of the polymorphisms on the control of viral replication in multivariate fashion. All P-values reported are two-sided. To account for multiple testing, we considered associations of P ≤ 0.01 as significant. The distribution of pretreatment set-point viral loads, as well as viral loads before and after STI, is shown in Figure 1. The median pretreatment viral load was 4.73 log HIV-1 RNA copies/ml (interquartile range 4.14–5.74 copies/ml). Immediately before treatment interruption, viral load was below the detection level in the majority (71%) of patients. Viral load increased after STI to a median of 3.06 log copies/ml (IQR 1.46–4.61 log copies/ml; Fig. 1a). The median interval between treatment interruption and viral load assessment off ART was 21 days (IQR 18–43 days).