Natural polymorphisms in HIV-1 CRF01_AE strain and profile of acquired drug resistance mutations in a long-term combination treatment cohort in northeastern China

Two thousand and thirty-four HIV-1 CRF01_AE-infected patients were selected from a long-term ART cohort (follow-up every 3 to 6 months) at the First Affiliated Hospital, China Medical University in Shenyang between January 2002 and December 2017. Partial HIV-1 pol sequences (HXB2: 2253–3269) obtained by Sanger sequencing based on HIV drug resistance genotyping assays [24] for each participant at baseline were used to analyze the natural polymorphisms of CRF01_AE. One thousand three hundred and thirty patients received first-line ART (two nucleoside reverse transcriptase inhibitors [NRTIs] + one NNRTI), of which 105 patients experienced TF, defined by a persistently detectable viral load exceeding 1000 copies/ml after 6 months of ART according to the Consolidated Guidelines on the Use of Antiretroviral Drugs for Treating and Preventing HIV Infection of WHO in 2016 [25]. Forty-two TF patients receiving tenofovir/lamivudine/efavirenz (TDF/3TC/EFV) treatment, the first-line ART regimen in China, were further selected to analyze the acquired DRM profile of CRF01_AE, based on the detection of at least one major DRM (Stanford HIVdb algorithm v8.8) in Sanger sequencing involving HIV drug resistance genotyping assays. The study was approved by the Ethics Committee of the First Affiliated Hospital of China Medical University and all patients signed informed consent forms. The flow chart of participant selection and analysis is shown in Additional file 1: Figure S1. Data on the demographic and clinical characteristics of all participants were collected from clinical records and are shown in Additional file 2

For phylogenetic analysis, the pol sequences of 2034 CRF01_AE-infected patients at baseline were aligned with reference sequences downloaded from the Los Alamos HIV database (https://​www.​hiv.​lanl.​gov/​) using the ClustalW tool in Mega v7.0 software, and then were manually edited. The models package in Mega v7.0 was used to determine the best nucleotide substitution model for this dataset. The reference sequences included twelve CRF01_AE strains from Africa and Thailand sampled between 1990 to 2001 and the representative sequences from seven major CRF01_AE lineages in China previously reported [17]. FastTree v2.1.9 was used to estimate an approximately maximum-likelihood phylogenetic tree based on the GTR + G + I nucleotide substitution model. The reliability of the phylogenetic tree was determined with local support values based on the Shimodaira–Hasegawa (SH) test with 1000 replicates. The phylogenetic tree was displayed using FigTree v1.4.3. Node SH-like support value ≥0.9 indicated a lineage [26].

A maximum-likelihood tree was reconstructed with the pol sequences of 42 TF patients at both baseline and TF using Mega v7.0. Bootstrap resampling (1000 datasets) of multiple alignments was performed to test the statistical robustness of the trees with the GTR + G + I nucleotide substitution model. A bootstrap value > 70 was identified as a cluster [27].

DRMs were identified using the Stanford University HIV Drug Resistance Database (https://​hivdb.​stanford.​edu/​) and interpreted using the Stanford HIVdb algorithm (HIVdb v8.8, Sierra v2.3.0; https://​hivdb.​stanford.​edu/​hivdb/​by-mutations/​).

The mutation rates of amino acids at sites 1 to 240 of reverse transcriptase (RT) region of the pol gene were compared between the 2034 treatment-naïve CRF01_AE sequences and subtype B sequences from treatment-naïve patients in the Stanford University HIV Drug Resistance Database, with an average of 46,118 isolates (one isolate per person) analyzed at each site (https://​hivdb.​stanford.​edu/​cgi-bin/​RTMutSummary.​cgi; accessed on 04/08/2019). The mutation rates were also compared between 1148 TS patients and 105 TF patients. The HIV-1 strain HXB2 was used as the reference standard. The sites with a different amino acid (compared to the corresponding site in HXB2), and with a prevalence > 1%, were defined as natural polymorphism sites.

The CorMut package [28] v1.25.0 based on the R Project for Statistical Analysis (R v3.5.2) was used to analyze co-variation. The HIV-1 strain HXB2 was used as the reference sequence of location. Positively selected mutations (PSMs) were determined using selection pressure (Ka/Ks ratio), with Ka/Ks > 1 and log odds ratio (LOD) > 2 [29]. Conditional selection pressure (conditional Ka/Ks, cKa/Ks) was used to measure the correlation between PSMs, with cKa/Ks > 1 and LOD > 2 indicating the presence of directional co-variation.

Longitudinal plasma samples between baseline and TF from four cases with Y181C and L228R mutations were selected. Viral RNA was extracted from the plasma samples using a QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer’s protocol and reverse transcribed using a Transcriptor First Strand cDNA Synthesis Kit (Roche Diagnostics, Indianapolis, IN, USA) with the specific primer Rev2–1 (5′-TCCTGCCATRGRAGATGCCTAA-3′). A 453-bp fragment (HXB2: 2868–3320) in the RT region of the pol gene was then amplified by two rounds of nested polymerase chain reaction (PCR) using a KOD-Plus-Neo kit (TOYOBO, Osaka, Japan) with the following outer and inner primers, respectively: MAW26/RT-21n (5′-GTATTTCTGCATTAAGTCTTTTGATGG-3′), 3-3F (5′-ACAGTACTAGATGTGGGAGATGC-3′)/3-3R (5′-TATATCATTGACAGTCCAGCT GTC-3′). The reaction conditions are shown in Additional file 3.

The PCR products were purified with Agencourt AMPure XP beads (Beckman Coulter, Brea, CA, USA) and then quantified using a Qubit 3.0 Fluorometer (Life Technologies, Carlsbad, CA, USA). The fragment length was accurately evaluated using an Agilent 2100 Bioanalyzer (Agilent Technologies, Waldbronn, Germany). Subsequently, the purified PCR products were adjusted to 2.5 ng/μl and indexed with an adaptor using a TruSeq Nano DNA LT library preparation kit (Illumina, San Diego, CA, USA) according to the manufacturer’s protocol. The indexed DNA libraries were analyzed using the Agilent 2100 Bioanalyzer and accurately quantified using a Roche LightCycler® 480 (LC480) Real-Time PCR system (Roche, Risch, Switzerland) and normalized to 10 nM, then pooled, denatured, and diluted to 15 pM, and finally mixed with 50% PHIX Control Libraries (Illumina, San Diego, CA, USA) to create a final volume of 600 μl.

Deep sequencing was performed using an Illumina MiSeq System (Illumina, San Diego, CA, USA). Oracle VM Virtual Box-5.2.22 software was used to build a virtual environment for running QIIME 2 Core-2018.4 (http://​qiime.​org/​) in Windows operating system. According to the data quality assessment using FASTQC v0.11.7 software, paired-ended sequences were trimmed by 10–15 bp and truncated to 280–285 bp, and the other parameters were set to the default values. The data were denoised and dereplicated using dada2 plugin v2018.4.0 [30]. The sequences and numbers of HIV-1 quasispecies in each sample were reported with feature-table plugin v2018.4.0, and were then aligned using the ClustalW tool in Mega v7.0.

The mutation rate of each amino acid site in RT was compared between treatment-naïve CRF01_AE and subtype B, between TS and TF CRF01_AE-infected patients, and between baseline and TF time point in 42 TF patients using the binomial distribution. The mutation rates and the number of DRMs of the 42 CRF01_AE TF patients between baseline and TF were also compared using the McNemar test and the Wilcoxon test, respectively. The statistical calculations were performed using SPSS software v20.0. P < 0.05 was used as the cutoff for significance.

In this study, 40 out of 2034 (1.97%) treatment-naïve CRF01_AE-infected patients had transmitted DRMs, with the common DRMs comprising K103 N, G190S, K101E, T215S, K65R, and K219Q. In addition to above DRMs, natural polymorphisms of amino acids with a prevalence > 1% were detected at 53 (53/240, 22.1%) sites in RT, of which nine sites (40, 68, 69, 98, 103, 118, 179, 210, and 238) were known drug resistance-associated sites. Moreover, 31 sites (4, 5, 6, 8, 11, 28, 32, 35, 36, 39, 40, 43, 88, 103, 104, 105, 111, 118, 123, 135, 172, 173, 174, 177, 179, 200, 203, 207, 211, 214, and 238) in CRF01_AE had higher mutation rates than subtype B HIV-1 strains in the Stanford HIV Drug Resistance Database (|Z value| ≥ 3) (Fig. 1). These 31 sites were defined as CRF01_AE-specific polymorphism sites, which included five known drug resistance-associated sites, site 238 (73.8%), site 118 (26.1%), site 179 (21.2%), site 103 (8.1%) and site 40 (3.1%), as well as 26 other sites that were not known to be associated with drug resistance (Fig. 1).

According to the phylogenetic analysis, the 2034 sequences mainly belonged to two CRF01_AE lineages, including 416 (20.5%) sequences of lineage 4 and 1522 (74.8%) sequences of lineage 5 (Additional file 4: Figure S2). Fifty-one and forty-four natural polymorphism sites in lineages 4 and 5 were detected, respectively, with differences in 35 sites between the two lineages (|Z value| ≥ 3). Both lineages had 26 polymorphism sites with higher mutation rates than in subtype B HIV-1globally (|Z value| ≥ 3), including two known drug resistance-associated sites (sites 179 and 238) (Fig. 1).

A total of 1330 out of 2034 CRF01_AE-infected patients received first-line ART, among which 105 (7.9%) patients experienced TF. We found 13 sites with differences between TF and TS patients (1148, 86.3%), comprising the polymorphisms at sites 75 and 189, which were only found in TF patients, and the polymorphisms at sites 4, 5, 8, 21, 32, 49, 105, 165, 169, 171, and 204, which were only found in the TS patients. The mutation rate of site 75 in TF patients was significantly higher than in TS patients (|Z value| ≥ 3) (Fig. 2).

Forty-two CRF01_AE-infected patients with TDF/3TC/EFV TF were selected according to the flow chart presented in Additional file 1: Figure S1 to determine the acquired DRM profile of CRF01_AE. The time between baseline and TF sampling time point among the 42 TF patients was 184 days (interquartile range: 177.0–236.5). The number of DRMs at TF time point were significantly increased compared to baseline (Z = -5.604, p < 0.001). The sequences of the baseline and TF time point from each patient of the 42 TF patients clustered with bootstrap value higher than 85 in the phylogenetic tree (Additional file 5: Figure S3). The mutation rates of 14 sites increased significantly at TF time point, with increase ranging from 9.5 to 66.7% (Table 1). Of these 14 sites, 13 were known drug resistance-associated sites, including seven NRTI-associated sites and six NNRTI-associated sites. The NRTI-associated DRMs detected at TF time point in descending order included K65R (57.1%), M184 V/I (47.6%), S68G (26.2%), A62V (14.3%), K70E/R (9.5%), and Y115F (9.5%). The NNRTI-associated DRMs detected at TF time point included G190S/C (66.7%), K101E/N/Q (52.4%), V179D/I/A/T/E (45.2%), Y181C (42.9%), K103R/N/S (42.9%), and V106 M (23.8%) (Table 1). It was noted that an unknown mutation (V75 L) was detected at site 75, a drug resistance-associated site, which increased from 4.8% at baseline to 16.7% at TF time point (Z value = 2.494, p < 0.05; p _{McNemar test} = 0.008). Moreover, a new mutation (L228R) was detected at site 228, a non-DRM site in the Stanford HIVdb algorithm, which increased from 0% at baseline to 11.9% at TF time point (Z value = 2.306, p < 0.05; p _{McNemar test} = 0.063). We speculated that both V75 L and L228R might be potential new DRMs in CRF01_AE.

To explore the role of potential new DRMs, the mutations at 14 sites with significantly increased mutation rates at TF were used for co-variation analyses. Nine known DRMs (K65R, V106 M, Y115F, V179 T/E/D, Y181C, M184 V, and G190S) and two potential new DRMs (V75 L and L228R) were demonstrated to be under positive selection pressure (Ka/Ks > 1, LOD > 2). Twenty-eight links were detected among these mutations (cKa/Ks > 1, LOD > 2) (Table 2). Among them, the known DRMs Y181C and G190S showed the strongest correlation (cKa/Ks_Y181C-G190S = 22.86, LOD = infinity). V75 L was correlated with known DRMs G190S (cKa/Ks_V75L-G190S = 3.24, LOD = infinity), K65R (cKa/Ks_K65R-V75L = 2.00, LOD = 5.04), and M184 V (cKa/Ks_V75L-M184V = 1.25, LOD = 4.03). L228R was correlated with known DRMs G190S (cKa/Ks_L228R-G190S = 2.25, LOD = infinity) and K65R (cKa/Ks_K65R-L228R = 2.00, LOD = 3.46), and strongly correlated with Y181C (cKa/Ks_Y181C-L228R = 6.00, LOD = 4.09) (Table 2).

To further explore the temporal association and the evolutionary dynamics between Y181C and L228R, longitudinal plasma samples of four CRF01_AE-infected patients with Y181C and L228R mutations were studied using deep sequencing. The first case demonstrated a time lag between the Y181C and L228R mutations; Y181C occurred in 53.4% of the sequences at 1-month post treatment, which increased to 100% at 3 months post treatment, and L228R did not appear until 6 months post treatment, when 87.1% of sequences carried both Y181C and L228R mutations. The second and third cases had Y181C and L228R only at TF. For the second case, 100% of sequences carried both Y181C and L228R simultaneously while, for the third case, 80% of sequences carried both Y181C and L228R simultaneously, and the remaining 20% carried only Y181C (Fig. 3). The fourth case could not be analyzed due to sequencing failure.