Among a representative sample of ARV-naïve subjects from the CASTLE study, HIV variants possessing TDR mutations were commonly detected by UDS. Similar rates of TDRs were identified in B and non-B subtypes. Although the study was not designed to evaluate TDRs by continent nor specific subtype, at least one sample from ARV-naïve subjects from 5 continents (Africa, Asia, Europe, North and South America) had a TDR by UDS. These data suggest that among subjects entering large international clinical trials, TDRs are common in ARV-naive persons from diverse populations and infected with B and non B subtypes. Many of these subjects harboring HIV variants with TDRs were not detected by standard genotyping. The overall TDR rate of 30.5% by UDS among ARV-naïve subjects in our study is similar to other investigations that reported rates between 28 to 33% depending on the populations studied and the sensitive genotyping methods used to identify low abundance resistant variants 
When considering TDRs both overall and by specific ARV class, TDRs were not associated with a lower virologic response for subjects initiating a boosted PI-based regimen. However, a small subset of subjects with specific N(t)RTI TDR patterns that could impact the activity of the TDF/FTC backbone used in the study did experience a higher rate of virologic failure. Interestingly, subjects with a PI TDRs identified by UDS at baseline did well on a boosted PI regimen. Possible reasons for this apparently non-intuitive finding include: 1) most PI TDRs were found in isolation (few subjects had multiple PI TDRs), 2) the PI TDRs identified had low Stanford HIVdb scores 
(e.g. 24I, 32I, 46I, and 58E and some of these have been previously described to be naturally occurring polymorphisms within the HIV-1 protease gene 
, and 3) the majority of PI TDRs were at very low levels of the viral population (<20% of the population). Thus, PI TDRs identified by UDS may have little impact on virologic response when the TDR is found in isolation and at low levels, especially in a person initiating a boosted PI regimen with a high genetic barrier to resistance. The PI TDR patterns were interpreted for all subjects with the Stanford HIVdb algorithm 
. Only one subject had high-level resistance to the PI used (ATV). This subject experienced VF and had multiple PI TDRs (46I, 58E, 84V, 85V, 90M) in high abundance (>20% levels). This subject demonstrates the previously known information that multiple PI TDRs present at high levels can impact virologic responses to boosted PI based regimens 
There was no difference in the rate of NNRTI TDRs in subjects with VS or VF with approximately 11% having a TDR NNRTI mutation. This data suggests that if a person harbors NNRTI resistant variants, a boosted PI regimen may be successful at suppressing the resistant variants. It has been previously shown that many of the low abundance resistant variants identified by UDS may not affect virologic outcomes if the resistant variants identified are only resistant to ARVs that are not part of the antiretroviral regimen being employed 
Although there was no difference in the rate of N(t)RTI TDRs in those subjects with VS or VF, there were subjects with extensive N(t)RTI TDR mutations which are known to impact responses to TDF/FTC. There was a high rate of VF in the small number of subjects with a M184V, even when the TDR was present at <20% levels of the quasispecies and missed by standard genotyping methods. Subjects with multiple TAMs also had disproportionate rates of VF. These results suggest that M184V and specific patterns of N(t)RTI mutation patterns may contribute to predicting response to a boosted PI regimen. However, the number of subjects in this study is small, and more studies should be done to investigate the impact of specific N(t)RTI TDR patterns present in low abundance on virologic responses in subjects initiating boosted PI regimens.
A strength of this study was that the patient population was recently sampled (enrolled between 2005–2006), well characterized, and represented multiple diverse ARV-naïve groups and HIV subtypes. A limitation of this study was that it was a retrospective case control study with the inherent limitations of such a design. The study design allowed us to explore the association of TDRs with clinical responses with a boosted PI regimen and the rate of TDRs in B and non B subtypes. However, investigations into the prevalence of TDRs across continents or patterns of specific TDRs by HIV subtype could not be definitely defined given that the numbers in each category were too few. Future studies should be performed to determine the prevalence of TDRs by UDS in larger retrospective and prospective studies.
Also, recent reports have suggested that the absolute number of viral variants with a TDR mutation, or “mutational load,” may impact the time to virologic failure in a person on therapy 
. Although our study was not designed to specifically determine mutational load association with virologic response, subjects with high mutational loads were generally more likely to experience VF (data not shown). Thus, several factors should be considered when evaluating minor variants effect on treatment response, such as, the specific TDR resistance conferring effect, the proportion of variants possessing the mutation (mutational load), the total number of TDRs present that affect the regimen and whether these TDRs are linked on same viral genome, and the genetic barrier of the specific ARV regimen being used. Virologic outcome based on baseline minority variants is likely multifactorial, and is not simply based on presence or absence of a mutation or a simple mutation percentage, as some mutations and mutational loads likely can be suppressed by specific regimens, in this case a boosted PI regimen 
Further studies should be performed to determine how the viral factors of mutational load, mutation linkage (mutations within the same viral genome), and specific mutational patterns interact and impact treatment responses to the many different antiretroviral regimens now used in the clinic. These important virologic parameters will need to be better defined, and the resistance interpretation algorithms accordingly adapted before sensitive genotyping technologies can be incorporated into routine HIV clinical care.
In summary, among a representative sample of ARV-naïve subjects in the CASTLE study, transmitted drug resistance mutations identified by ultra-deep sequencing were common, and B and non-B HIV-1 subtypes had a similar rate of TDRs. TDRs identified by UDS did not affect virologic response for subjects on a boosted PI. However, a small subset of subjects with extensive N(t)RTI backbone TDR patterns were likely to experience virologic failure. Further investigations should be performed to determine the prevalence of TDRs in diverse populations of HIV infected persons across the different regions of the world. Studies should also be done to determine which ARV regimens are best at suppressing HIV variants possessing TDRs at different variant levels and mutation patterns.