Full-spectrum HIV drug resistance mutation detection by high-resolution complete pol gene sequencing

Abstract

Background: Drug resistance mutation testing is a key element for HIV clinical management, informing effective treatment regimens. However, resistance screening in current clinical practice is limited in reporting linked cross-class resistance mutations and minority variants, both of which may increase the risk of virological failure.

Methods: To address these limitations, we obtained 358 full-length pol gene sequences from 52 specimens of 20 HIV infected individuals by combining microdroplet amplification, unique molecular identifier (UMI) labeling, and long-read high-throughput sequencing.

Results: We conducted a rigorous assessment of the accuracy of our pipeline for precision drug resistance mutation detection, verifying that a sequencing depth of 35 high-throughput reads achieved complete, error-free pol gene sequencing. We detected 26 distinct drug resistance mutations to Protease Inhibitors (PIs), Nucleoside Reverse Transcriptase Inhibitors (NRTIs), Non-Nucleoside Reverse Transcriptase Inhibitors (NNRTIs), and Integrase Strand Transfer Inhibitors (INSTIs). We detected linked cross-class drug resistance mutations (PI+NRTI, PI+NNRTI, and NRTI+NNRTI) that confer cross-resistance to multiple drugs in different classes. Fourteen different types of minority mutations were also detected with frequencies ranging from 3.2% to 19%, and the presence of these mutations was verified by Sanger reference sequencing. We detected a putative transmitted drug resistance mutation (TDRM) in one individual that persisted for over seven months from the first sample collected at the acute stage of infection prior to seroconversion.

Conclusions: Our comprehensive drug resistance mutation profiling can advance clinical practice by reporting mutation linkage and minority variants to better guide antiretroviral therapy options.

Keywords: Cross-resistance; Drug resistance mutations; HIV; Minority variants.

Fig 1.. Maximum likelihood tree.

Maximum likelihood tree of all 358 full-length pol gene (consensus) sequences from the CDC seroconversion cohort (denoted by SC) and LAC-USC Rand Schrader Clinic cohort with the HXB2-pol sequence. Each study participant’s cluster was highlighted by a unique colored box.

Fig 2.. Sequencing depth for error-free sequencing and UMI validation.

A. Full-length HIV pol gene sequencing precision was tested by comparing the consensus sequence of a given number of SC24–1’s raw reads with the corresponding Sanger reference sequence. A bootstrap test was conducted by randomly sampling a given number of length-filtered raw reads 100 times. The proportion of consensus sequences that were identical to the reference sequence was plotted with its 95% confidence interval, as the number of raw reads increased. All 100 bootstrap runs produced the correct consensus sequence without any base errors when the number of raw reads was 35 or greater. B. Sequence composition of SC24–1’s 11 UMIs, GAGTAAAATC (30.5), ATCGTAATAT (18.7), AGTTATTCTG (11.0), GTAAGCACTA (9.3), AACTCTCTGT (9.0), CGGGGCACAA (8.0), AATACCAAGT (6.7), TTCCCACCAA (4.7), ACTGGAAATA (4.3), CTGATACCAG (4.2), AAGCACAATC (1), plotted by WebLogo3. Here the number in the parenthesis denotes the relative frequency of raw reads with each UMI. The size of each letter is proportional to the frequency of each nucleotide. C. Sanger chromatogram of SC24–1’s UMI region, plotted by UGENE. Each of nucleotide bases, A, T, G, and C is represented by green, red, black and blue, respectively.

Fig 3.. Cross-class linked drug resistance mutations.

A. Amino acid sequence alignment of the 20 pol gene sequences from NK9147 in the LAC-USC Rand Schrader clinic cohort, plotted by AliView. The top sequence showed linkage among the cross-class drug resistance mutations, V32M (PI), M41L (NRTI), and T215E (NRTI). The remaining 19 sequences showed linkage between M41L (NRTI) and T215E (NRTI) mutations. B. Amino acid sequence alignment of the 15 pol gene sequences obtained from SC5–7 in the CDC seroconversion cohort. The top sequence showed the linkage between I54S (PI) and V106I (NNRTI) mutations. C. Amino acid sequence alignment of the nine pol gene sequences from SC19–10. The top five sequences had linkage between V108I (NNRTI), V179E (NNRTI), T215L (NRTI), and H221Y (NNRTI) mutations. The next two sequences had linkage between V108I (NNRTI), V179E (NNRTI), and T215L (NRTI) mutations. The last two sequences had linkage between V179E (NNRTI) and T215L (NRTI) mutations. D. Amino acid sequence alignment of the 11 pol gene sequences from SC24–1. The top sequence showed linkage between D67d (NRTI), S68d (NRTI), T69d (NRTI), K70d (NRTI), and K103N (NNRTI) mutations.

Fig 4.. Validation of minority drug resistance mutations.

A. Sanger sequencing chromatogram data for specimen NK9147. The V32M region showed double peaks of the wild-type nucleotides (GTA) and mutant-type nucleotides (ATG). This mutation was detected in one out of 20 pol gene sequences (5% frequency). The proportion of raw reads for this mutation was around 9.7%. B. Chromatogram data for specimen EC8287. The V108I mutation, resulting from a G to A base substitution, was detected in 4 out of 21 pol gene sequences (19% frequency). The proportion of raw reads that contributed to these 4 sequences was 10%. A minor peak in the Sanger chromatogram confirmed the presence of this mutation. C. The presence of the A128T drug resistance mutation resulting from a G-to-A base substitution was confirmed by chromatogram data for specimen SC23–7. The frequency of this mutation was 5% and the proportion of raw reads for this mutation was 7.5%. D. Sanger chromatogram peaks and pol gene sequence alignment for specimen SC24–1. A total of 11 unique pol gene sequences were obtained from this specimen with 10 wild-type sequences (AGACAGCACTAAATGGA) and one mutant-type sequence with a 14-nucelotide deletion (AGAGAAAATTAGTAGAT). The frequency of this 14-nuleotide deletion was 9% (one out of 11 pol gene sequences) and the proportion of reads that contributed to this consensus sequence was around 28%. The deletion signature was detected in the chromatogram peaks.

Fig 5.. Pretreatment drug resistance mutations of the CDC seroconversion cohort.

A. The frequency of pretreatment drug resistance mutations in study participant SC5 across three time points. The number of pol gene sequences collected from each time point was presented by horizontal grey bars (3, 2, and 15 sequences). The V106I mutation persisted in 100% of the pol gene sequences between December 2008 and June 2009, while the frequency of the I54S mutation increased from 0% to 6.7% in June 2009. B. The frequency of the E157Q mutation in study participant SC8 across four time points. The frequency the E157Q mutation increased from 0% to 7.1% and 11.1% in April 2009 and May 2009, respectively. The frequency of E157Q mutation subsequently decreased to 0% in July 2009. C. Drug resistance mutation frequency of study participant SC19 across four time points. The V179E and T215L mutations were detected in 100% of the pol gene sequences collected from October 2009 to July 29th, 2010, indicating their persistent presence during that time period. The frequency of V108I and H221Y mutations increased from 0% to 100% between initial time point and March 2010, but decreased to 77.8% and 55.6% in July 29^th, 2010, respectively. D. The K103N mutation was observed in all pol gene sequences from SC24 over a period of seven months since the first time point sample. The frequencies of the D67d, S68d, T69d, and K70d mutations decreased from 9.1% to 0% for all time points after the initial time point.