We used a novel HRM diversity assay to compare HIV diversity in adults with different stages of HIV disease. Adults with acute HIV infection had uniformly low HRM scores (low levels of HIV diversity). However, the median HRM scores for those individuals were significantly higher than those obtained for plasmid controls for all regions except GAG2. This indicates that the HRM diversity assay can detect a low level of HIV diversity very early in HIV infection. In adults with acute HIV infection, the highest median HRM score was obtained for the ENV2 region, which contains IDR cluster I of gp41. HRM scores for all regions except for ENV2 were significantly higher in seropositive adults with recent HIV infection than in adults with acute HIV infection.
In all six regions analyzed, we found significantly higher levels of HIV diversity in adults who were infected for at least two years (non-recent group) than in adults near the time of HIV seroconversion (recent group). An important finding of this study was that viruses from individuals with non-recent infection often exhibited low diversity in at least one of the regions analyzed; in each region, there was some overlap in the HRM scores from adults with recent and non-recent infection. A similar finding was reported in a study that used a heteroduplex mobility assay to compare HIV diversity in the V3-V5 region of HIV env
in adults with likely recent vs. likely non-recent infection (classified using a detuned enzyme immunoassay strategy) 
. These findings and ours indicate that diversity-based measures (obtained using the HRM diversity assay or another method for viral diversity analysis) are not likely to be useful for HIV incidence analysis if they rely on analysis of a single genomic region. Our study extends the previous report 
by comparing diversity in multiple genomic regions and by using sample sets from individuals with known recent and known non-recent infection. This expanded analysis revealed that individuals with non-recent HIV infection rarely had low diversity in all regions that we examined. Furthermore, multivariate logistic regression showed that higher HRM scores in three regions (GAG2, ENV1, and ENV3) were independently associated with non-recent HIV infection. Independent diversification in different regions of the HIV genome is likely to reflect several factors, including: (1) different selective forces act on different HIV gene products, inducing diversification of different genomic regions, and (2) the very high frequency of genetic recombination of HIV lowers genetic linkage of different subgenomic regions. Our findings suggest that HIV diversity may be a useful biomarker for HIV incidence determination, provided that multiple HIV genomic regions are analyzed. The HRM diversity assay is simpler and less expensive than many other laboratory approaches used to measure HIV diversity, and it is particularly well-suited to multi-region analysis. The HRM diversity assay can be used to measure diversity in any RNA or DNA sample, including HIV RNA and proviral DNA. For HIV incidence applications, it makes most sense to analyze the actively replicating pool (e.g., plasma HIV RNA) rather than proviral DNA, which is likely to include archived sequences from viruses that were circulating earlier in infection.
Advanced HIV disease is associated with misclassification of individuals with non-recent infection as recently infected using serologic incidence assays 
. Our results indicate that this is not likely to confound the use of the HRM diversity assay for HIV incidence testing. In this study, almost half (48%) of the samples in the non-recent group were from individuals with CD4 cell counts <50 cells/mm3
. In the non-recent group, HRM scores were not significantly different among adults with CD4 cell counts above vs. below 50 cells/mm3
, and all of the unusually high HRM scores in this group (outlier and extreme values) were from adults with CD4 cell counts <50 cells/mm3
. These data show that advanced HIV disease is not associated with misclassification using the HRM diversity assay, and suggest that the HRM diversity assay may be useful for identifying samples from adults with advanced HIV disease who were misclassified as recently infected using serologic incidence assays.
Viral suppression is also associated with misclassification of individuals with non-recent infection as recently infected using serologic incidence assays 
. In this study, samples from adults on ART had detectable HIV RNA (to permit amplification of HIV RNA for analysis); it is not known whether those individuals were non-adherent to their treatment regimens or were failing ART. Because the HRM diversity assay uses different primer pairs for amplification of various regions of the HIV genome, and because the primers are designed to bind to relatively conserved sequences, sequence differences in HIV samples are not likely to impair binding of all of the relevant primer pairs. Therefore, amplification failure for all primer pairs is likely to indicate low viral load. We do not feel that it is necessary to screen samples for viral load prior to testing with the HRM diversity assay. If desired, a viral load assay could be used to confirm viral suppression in samples with multi-region amplification failure. We recognize that recently-infected individuals who have very low viral loads could be misclassified as non-recent if virologic suppression is used as a biomarker for non-recent infection. However, because infected individuals are not likely to have natural or ART-induced viral suppression early in infection, this type of misclassification should be infrequent and should have very little impact on HIV incidence estimates.
For samples that do amplify, we recognize that there is a potential to underestimate diversity when fewer copies of HIV RNA are used in the analysis. However, in a previous study 
, we demonstrated that results from the HRM diversity assay were not significantly affected by differences in sample volume (0.1 vs. 0.5 ml), HIV viral load (range: 2,000 to 50,000 copies/ml), or the number of HIV RNA copies used to prepare DNA templates for amplification (range: 100 to 5,000 copies of HIV RNA). Those results support the use of the HRM diversity assay for analysis of clinical samples with variable viral loads.
The HRM diversity assay provides data that is likely to be independent of data from serologic incidence tests. Therefore, use of the HRM diversity assay in combination with serologic testing is likely to improve the precision of multi-assay algorithms for HIV incidence, lowering misclassification rates. shows an example of an existing multi-assay algorithm that combines four assays for HIV incidence determination: a BED screening assay and an avidity screening assay (using a high cut-off for recent HIV infection for both assays), CD4 cell count, and viral load 
. While CD4 cell count is a useful biomarker for reducing misclassification, inclusion of CD4 cell count data in incidence algorithms presents certain logistical challenges. First, because CD4 cell counts must be obtained in real-time (before other HIV incidence testing has been performed), CD4 cell count testing must be performed for all HIV-infected individuals evaluated, rather than the smaller subset who appear to be recently infected based on serologic testing. Second, many valuable sample sets from clinical trials and surveillance studies include only stored serum or plasma. Unless CD4 cell counts were obtained at the time of sample collection, it is not possible to assess incidence in those sample sets using an algorithm that includes CD4 cell count data.
Use of the HRM diversity assay as part of a multi-assay algorithm for HIV incidence determination.
We are exploring whether inclusion of the HRM diversity assay as part of a multi-assay algorithm will eliminate the need for CD4 cell count data, allowing all of the testing to be performed using a single plasma or serum sample. shows a possible alternative multi-assay algorithm that incorporates the HRM diversity assay. To reduce the cost and effort needed for analysis, it would be most effective to screen samples for recent infection using serologic HIV incidence assays (e.g., BED and avidity screens). The subset of samples that are classified as “recent” based on serologic testing could then be tested using the HRM diversity assay to improve the precision of HIV incidence estimates. With this analytic plan, the number of samples that would require HRM diversity analysis would be relatively small. Therefore, the cost of the HRM diversity assay would not greatly impact the overall cost of the incidence assessment. Our preliminary data suggest that HIV from recently-infected individuals usually lacks diversity across the HIV genome, while HIV from individuals with non-recent infection is genetically diverse in one or more genomic regions. Studies are underway to identify genomic regions and assay cutoffs in each region that optimally discriminate between recent and non-recent infection. Once those parameters are set, it might be possible to use a simple approach in which samples are characterized as non-recent if they have a high HRM score in at least one of several regions tested.
In summary, this study provides proof of principle that HIV diversity can be used as a biomarker to distinguish between adults with recent vs. non-recent HIV infection. Further studies are needed to evaluate the performance of multi-assay algorithms for HIV incidence determination that include the HRM diversity assay.