**2. Results**

#### *2.1. Characteristics of the Study Population*

A total of 174 patients were included in the survey in 2016 (Table 1). Only 60 (34.5%) patients were followed up a median 10 months later, with 49 still in ART interruption and 11 re-initiating ART. Among the 174 patients, 88.5% (154/174) were aged 30 and above; 66.1% (115/174) were male; 65.5% (114/174) were illiterate or had a primary school education; 67.2% (117/174) were married or living with a partner; 61.5% (107/174) and 4.0% (7/174) of the patients were infected through heterosexual and homosexual contacting, respectively and 27.6% (48/174) were infected through injection drug using. The numbers of patients with CRF01\_AE, CRF07\_BC and CRF08\_BC HIV-1 strains were 35, 61 and 58 (20.1%, 35.1% and 33.3%), respectively. Seventy (40.2%) patients had a CD4+ T cell count of <200 cells/mm<sup>3</sup> at investigation. A total of 150 (86.2%) patients had received nonnucleotide reverse transcriptase inhibitor (NNRTI)-based first-line antiretroviral regimens (stavudine (d4T)/azidothymidine (AZT)/tenofovir (TDF) + lamivudine (3TC) + efavirenz (EFV)/nevirapine (NVP)) before ART interruption. The median duration of treatment before ART interruption was 16 (interquartile range (IQR): 7–26) months, while the median duration of ART interruption was 12 (IQR: 6–24) months at survey.

#### *2.2. Detected HIV Drug Resistance Mutations*

Drug resistance mutations (DRMs) were detected at 12 positions in the partial *pol* region by Sanger sequencing (Figure 1), including one in the protease (PR) region, an accessory protease inhibitor (PI)-related resistance mutation (Q58E), and eleven in the reverse transcriptase (RT) region, including one nucleotide reverse transcriptase inhibitor (NRTI)-related (M184V), and ten others were NNRTI-related. The most common drug resistance mutation was K103N (13.2%), followed by V179D (6.3%) and E138AGK (3.5%). At the 20% detection threshold, NGS detected all the mutations identified by SS except

a V106M mutation in one patient, as there were mixtures in the first (G to R) and third nucleotides (A to R) in the codon at Sanger sequencing. In addition, NGS detected three more mutations in three patients: Y188C, E138A and K103N, at frequencies of 22.96%, 27.38% and 43.68%, respectively. There were no additional DRMs detected at the frequencies between 15% and 20%. Low-frequency DRMs (<15% frequency) were only detected by NGS, but not by SS. K65R was the most common low-frequency DRM with frequencies between 1% and 9%, concentrated at frequencies from 2% to 5%. It is interesting that this low-frequency K65R mutation is significantly unevenly distributed among subtypes; 5.7% (2/35) in CRF01\_AE, when compared with 95.1% (58/61) in CRF07\_BC and 93.1% (54/58) in CRF08\_BC (*p* < 0.001). Other low-frequency NRTI-related mutations were K70QE, F77L, T215AI and K219QE. M46LI was the common low-frequency PI-related mutation, with frequencies ranging from 1% to 13%. Other PI-related mutations such as L10F, I47V, I50V, F53L, I54VT and N83D have the mutation frequency of about 2%.



a includes CRF 55\_01B (2, 1.1%), CRF 62\_BC (1, 0.6%), Unknown (1, 0.6%). HIV, Human immunodeficiency virus; ART, antiretroviral therapy; IQR, interquartile range; d4T, Stavudine; 3TC, Lamivudine; AZT, Azidothymidine; TDF, Tenofovir; EFV, Efavirenz; NVP, Nevirapine; LPV/r, Lopinavir/r.

**Figure 1.** Frequency and pattern of HIV drug resistance (HIVDR) mutations detected by Sanger sequencing (SS) and next-generation sequencing (NGS) at different detection thresholds. Note: HIVDR, HIV drug resistance; SS, Sanger sequencing; NGS, Next-generation sequencing; PI, Protease inhibitor; NRTI, Nucleoside reverse-transcriptase inhibitor; NNRTI, non-nucleoside reverse-transcriptase inhibitor.

#### *2.3. HIV Drug Resistance Interpretations*

Based on Sanger sequencing, 19.5% (34/174) of the patients were shown to have drug resistance variants. With NGS, the rate of resistance was the same; 20.7% (36/174) at the detection thresholds 20% and 15%. It climbed up to 21.3% (37/174), 24.1% (42/174), 45.4% (79/174) and 79.9% (139/174) at thresholds 10%, 5%, 2% and 1%, respectively. Compared with SS, NGS go<sup>t</sup> significantly higher rates of drug resistance at the 1%, 2% and 5% thresholds (*p* < 0.05). For PI and NRTI, the prevalence of HIVDR was the same at 0.6%, identified by SS and NGS at the 15% detection threshold. However, NGS at the 1% and 2% thresholds identified more NRTI-related drug resistance variants (69.0% and 24.7%, respectively) than SS. Compared with NGS, a slightly lower percentage of HIVDR was found by SS for NNRTI-related drug resistance (19.0%, Table 2). For the efavirenz (EFV) or nevirapine (NVP) in first-line NNRTI, the difference between NGS and SS in the identification of drug resistance levels was relatively small. EFV- or NVP-related resistance rates were identified in 15.5% (27/174) and 16.1% (28/174) of patients by SS and NGS (>15% frequencies), respectively.


#### *2.4. Relationship between CD4+ T Cell Count, Viral Load and HIVDR Mutation Frequency*

The patients were divided into three groups across the mutation frequencies detected by NGS: 35 patients without HIVDR variants, 103 patients with mutation frequencies lower than 15% and 36 patients with mutation frequencies higher than 15%. Their median CD4+ T cell counts were 140 cell/mm<sup>3</sup> (18–289), 265 cell/mm<sup>3</sup> (IQR: 172–378) and 222 cell/mm<sup>3</sup> (IQR: 82–302), respectively (*p* < 0.05). In addition, their median viral loads were 4.6 log10 copies/mL (4.0–4.9), 4.3 log10 copies/mL (IQR: 3.6–4.7) and 4.0 log10 copies/mL (IQR: 3.6–4.4), respectively (*p* < 0.05). In addition, there was a statistically significant difference in the viral loads between patients with high-frequency variants and patients without variants (*p* = 0.0198, Figure 2).

**Figure 2.** The relationship between the CD4+ T cell count, viral load and HIV drug resistance mutation frequency. Note: The patients were divided into three groups according to the mutation frequency detected by next-generation sequencing. High-frequency variants mean that their mutation frequencies were higher than 15% and low-frequency variants mean that their mutation frequencies were less than 15%. HIVDR, HIV drug resistance; ns, no significance.

#### *2.5. Changes of HIVDR Mutations at Follow-Up*

Forty-nine patients were still with ART interruption and were available for follow-up after the median of ten months (IQR: 8–11). At baseline, mutations with a frequency of 20% and above were NRTI-related, such as M184VI (2.0%, 1/49), and NNRTI-related like K103N (14.3%, 7/49), E138AG (4.1%, 2/49), V179D (2.0%, 1/49) and P225H (2.0%, 1/49). Although these variants still existed at follow-up, the frequencies of the mutations M184VI, K103N and P225H decreased over time, and most of them remained at frequencies of more than 20%. However, the frequency of K103N in one patient (GX064) had dropped from 43.7% to 15.3%, and the mutation K103N in another patient (CQ046) with a frequency of 36.9% disappeared (Supplementary Materials). Within a year, some minority DRMs at frequencies 1%–10% remained unchanged, including: PI-related D30N, M46LI, I54VT and N88D, NRTI-related K65R and NNRTI-related Y188CHL. Moreover, K65R was still the most common low-frequency mutation at the follow-up. However, some minority DRMs at frequencies of 1%–5% disappeared, including N83D with PI-related, K70E, T215A, and K219E with NRTI-related and K101E, Y181C, H221Y and K238T with NNRTI-related, while others emerged, such as NNRTI-related V106A in patient GX088 at a frequency of 8.7%, and L23I, I47V and I84V with PI-related, D67N and F77L with NRTI-related and L100V and F227L with NNRTI-related, which appeared at frequencies of 1%–5% (Figure 3).

**Figure 3.** Changes of HIV drug resistance mutations during baseline and follow-up. Note: 49 of the 174 patients were followed up nearly a year later. Mutations and mutation frequency were detected by next-generation sequencing. PI, Protease inhibitor; NRTI, Nucleoside reverse-transcriptase inhibitor; NNRTI, non-nucleoside reverse-transcriptase inhibitor.
