Background:  ART has reduced the incidence and severity of HIV neurocognitive impairment but its prevalence remains high. HIV drug resistance is a major determinant of treatment failure, and several studies of subjects undergoing lumbar puncture for clinical indications have demonstrated frequent discordance of drug resistance patterns between cerebrospinal fluid (CSF) and plasma virus. Such studies can be confounded by selection bias and by blood-CSF barrier injury due to central nervous system (CNS) co-morbidities. The objective of this study was to determine the prevalence of discordant resistance in subjects undergoing lumbar punctures in the absence of CNS co-morbidities.

Methods:  Subjects were 137 volunteers prospectively enrolled in CHARTER, a North American, observational study of the effects of HIV and ART on the nervous system. All subjects who consented to lumbar puncture and had viral loads in CSF and plasma >200 copies/mL were eligible for this cross-sectional analysis. CSF and plasma virus were genotyped with the Viroseq Genotyping system version 1.5. Pair-wise genetic distances were estimated and mutational patterns compared.

Results:  Protease and reverse transcriptase sequences were successfully obtained from 135 of 137 paired CSF and plasma specimens; 65% of subjects demonstrated drug resistance in 1 or both fluids. Of samples with resistance, patterns differed between CSF and plasma samples in 32%. The frequency of discordance did not differ between ART classes, including protease inhibitors (PI). Of samples without resistance, genetic distances between CSF and plasma pol sequences were often relatively high, supporting partitioning of HIV populations in the absence of overt resistance.

Conclusions:  Discordant resistance between CSF and plasma virus were frequently observed in this North American cohort in the absence of active CNS co-morbidities. These findings indicate that compartmentalization of HIV replication is not wholly dependent on variable ART penetration because classes of poorly penetrating ART (i.e., PI) did not differ from others, and because genetic distances between CSF and plasma virus sequences were at times large even in the absence of drug resistance mutations. The clinical implications of these findings remain to be determined.