629    

Background:  Most HIV-1 resistance-associated reverse transcriptase mutations are found within the N-terminal 240 residues of the RT enzyme. To search for additional mutational signatures outside the conventional range of many commercial genotyping assays, we applied a recently published statistical approach to a large clinical database of sequences spanning RT residues 1-400.

Methods:  HIV-1 RT mutations spanning positions 1-400 from 41,122 de-identified clinical samples sequenced at Quest Diagnostics between 1/2002 and 6/2003 were examined. Viruses with at least one resistance mutation were compared with those with no associated resistance using Chi-square statistics with the Benjamini-Hochberg correction for multiple comparisons. The binomial correlation coefficient Phi was calculated for all pairwise combinations of significant mutations. Clusters were identified with the Neighbor-Joining and the Fitch-Margoliash methods.

Results:  The dataset contained 26,655 (64.8%) viral sequences with at least one predicted ARV resistance (PR or RT) and 14,467 (35.2%) sequences with no predicted resistances.  Mutations at 53 RT positions (present in >1% of the ARV-resistant group) were found >2 times more often than in the latter group ( p <0.01 corrected for 400 tests).  Of 53 positions, 37 were known primary or accessory resistance mutations, including 8/9 recently described NRTI-associated mutations (39,43,203,208,218,221,223,228); 16 novel mutations (28,31,40,68,90,172,224,227,237,238,284,348,350,359,370,371) were also more frequent in the ARV-resistant group.  Binomial correlation coefficients for 1378 pairs identified 3 principal NRTI-related clusters and 1 NNRTI-related cluster.  Residues 68,203,208,218,223, and 228 were assigned to the NRTI clusters and residue 221 grouped with the NNRTI cluster.  Other significant correlations included residues 359+371 (f = 0.372), 181+221 (f = 0.323), and 106+227 (f = 0.268).

Conclusions:  We have identified several novel RT mutations and associated them with known clusters of resistance-related mutations in a large clinical dataset.  Several mutations did not group with known clusters, including five mutations in the RT connector domain.  Statistical analysis of extended-length clinical RT sequences is a useful technique to uncover genotypic correlates of resistance.  Phenotypic and clinical studies are needed to further elucidate the impact of these mutations on clinical response.

Keywords: RT; mutations; resistance