Characterization of mutation spectra with ultra-deep
pyrosequencing: Application to HIV-1 drug resistance
Chunlin Wang,1,3Yumi Mitsuya,1,3Baback Gharizadeh,2Mostafa Ronaghi,2and
Robert W. Shafer1,4
1Division of Infectious Diseases, Department of Medicine, Stanford University, Stanford, California 94305, USA;2Stanford
Genome Technology Center, Stanford University, Stanford, California 94305, USA
The detection of mutant spectra within a population of microorganisms is critical for the management of
drug-resistant infections. We performed ultra-deep pyrosequencing to detect minor sequence variants in HIV-1
protease and reverse transcriptase (RT) genes from clinical plasma samples. We estimated empirical error rates from
four HIV-1 plasmid clones and used them to develop a statistical approach to distinguish authentic minor variants
from sequencing errors in eight clinical samples. Ultra-deep pyrosequencing detected an average of 58 variants per
sample compared withan averageof eightvariants
dideoxynucleotide sequencing. In the clinical sample with the largest number of minor sequence variants, all 60
variants present in ?3% of genomes and 20 of 35 variants present in <3% of genomes were confirmed by limiting
dilution sequencing. With appropriate analysis, ultra-deep pyrosequencing is a promising method for characterizing
genetic diversity and detecting minor yet clinically relevant variants in biological samples with complex genetic
per sample detectedby conventionaldirect-PCR
[Supplemental material is available online at www.genome.org. The raw data from this study are available online at
Dideoxynucleotide (Sanger) sequencing of non-clonal PCR prod-
ucts (direct PCR sequencing) of plasma viral cDNA is widely used
to detect more than 50 drug-resistance mutations in the molecu-
lar targets of HIV-1 therapy—reverse transcriptase (RT) and pro-
tease—in clinical settings (US Department of Health and Human
Services Panel on Clinical Practices for Treatment of HIV Infec-
tion 2006). A major limitation of direct PCR sequencing, how-
ever, is its inability to detect low proportions of drug-resistant
variants in the heterogeneous virus population existing in a pa-
tient’s plasma sample (Palmer et al. 2005). Several studies have
shown that minor drug-resistant variants that are not detected by
population-based sequencing are clinically relevant in that they
are often responsible for the virological failure of a new antiret-
roviral treatment regimen (Jourdain et al. 2004; Kapoor et al.
2004; Lecossier et al. 2005; Palmer et al. 2006b). Multiple ap-
proaches have been developed to detect minor HIV-1 variants in
research settings; however, no single approach has proved useful
for clinical settings.
The 454 Life Sciences GS20 sequencing platform allows mas-
sively parallel picoliter-scale amplification and pyrosequencing
of individual DNA molecules (Margulies et al. 2005). Simons and
colleagues described two cases in which ultra-deep pyrosequenc-
ing detected minority variant drug-resistance mutations in a pre-
viously treated patient in whom mutations were no longer de-
tectable by standard direct PCR sequencing (Simons et al. 2005).
Tsibris and colleagues demonstrated that ultra-deep pyrose-
quencing could accurately quantify a mixture of three HIV-1 en-
velope variants pooled in defined proportions of 89%, 10%,
and 1% (Tsibris et al. 2006). Here, we systematically investigate
the potential of ultra-deep pyrosequencing to detect minor vari-
ants in the HIV-1 RT and protease genes from clinical plasma
samples. We characterize the types of sequence errors associated
with such ultra-deep pyrosequencing of HIV-1 virus populations,
develop statistical methods for handling these errors, and vali-
date our approach using molecular and limiting dilution clonal
Plasmid DNA clones and clinical RT-PCR products
We performed ultra-deep pyrosequencing on four plasmid DNA
clones of cultured HIV-1 isolates and eight RT-PCR products from
RNA extracted from cryopreserved plasma samples. Each se-
quenced sample encompassed all 99 HIV-1 protease codons and
the first 241 reverse transcriptase codons. Sample preparation
and ultra-deep pyrosequencing are explained in detail in the
Methods section and illustrated in Supplemental Figure 1. The
plasmid clones included the laboratory strain NL43 (M19921;
which was sequenced twice) and two recombinant viruses in
which RT codons 24–312 were replaced with cloned cDNA from
two multidrug-resistant clinical virus isolates (AY35744,
AY351750). The RT-PCR products were obtained from an un-
treated HIV-1-infected patient in 1992 and from seven antiretro-
viral-treated patients from 2000 to 2005. The plasma HIV-1 RNA
levels in the eight plasma samples were each >100,000 copies/
mL. The median number of cDNA copies prior to sequencing was
100 with an inter-quartile range of 75–180.
Sequence data and alignment
The GS20 sequencing platform generated an average of 6827
reads per sample (mean length of 105 nucleotides [nt]) on four
3These authors contributed equally to this work.
E-mail email@example.com; fax (650) 725-2088.
Article published online before print. Article and publication date are online at
http://www.genome.org/cgi/doi/10.1101/gr.6468307. Freely available on-
line through the Genome Research Open Access option.
17:1195–1201 ©2007 by Cold Spring Harbor Laboratory Press; ISSN 1088-9051/07; www.genome.org
sequencing technologies: Methods and goals. Nat. Rev. Genet.
Simons, J.F., Egholm, M., Lanza, J.R., Turenchalk, G., Desany, B., Ronan,
M.T., Knight, J.R., Du, L., Leamon, J.H., Rothberg, J.M., et al. 2005.
Ultra-deep sequencing of HIV from drug-resistant patients [abstract
142]. In 14th International HIV Drug Resistance Workshop. Québec
Thomas, R.K., Nickerson, E., Simons, J.F., Janne, P.A., Tengs, T., Yuza,
Y., Garraway, L.A., LaFramboise, T., Lee, J.C., Shah, K., et al. 2006.
Sensitive mutation detection in heterogeneous cancer specimens by
massively parallel picoliter reactor sequencing. Nat. Med.
Tsibris, A., Russ, C., Lee, W., Paredes, R., Arnaout, R., Honan, T., Cahill,
P., Nusbaum, C., and Kuritzkes, D.R. 2006. Detection and
quantification of minority HIV-1 env V3 loop sequences by
ultra-deep sequencing: Preliminary results. In 15th International HIV
Drug Resistance Workshop. International Medical Press, Sitges, Spain.
Tsongalis, G.J., Gleeson, T., Rodina, M., Anamani, D., Ross, J., Joanisse,
I., Tanimoto, L., and Ziermann, R. 2005. Comparative performance
evaluation of the HIV-1 LiPA protease and reverse transcriptase
resistance assay on clinical isolates. J. Clin. Virol. 34: 268–271.
US Department of Health and Human Services Panel on Clinical
Practices for Treatment of HIV Infection. 2006. Guidelines for the
use of antiretroviral agents in HIV-1-infected adults and adolescents
(The living document, October, 2006). http://aidsinfo.nih.gov/.
Vahey, M., Nau, M.E., Barrick, S., Cooley, J.D., Sawyer, R., Sleeker, A.A.,
Vickerman, P., Bloor, S., Larder, B., Michael, N.L., et al. 1999.
Performance of the Affymetrix GeneChip HIV PRT 440 platform for
antiretroviral drug resistance genotyping of human
immunodeficiency virus type 1 clades and viral isolates with length
polymorphisms. J. Clin. Microbiol. 37: 2533–2537.
Received March 6, 2007; accepted in revised form April 27, 2007.
Ultra-deep pyrosequencing for HIV drug resistance