Molecular Epidemiology of HIV-1 Subtype B Infection across Florida Reveals Few Large Superclusters with Metropolitan Origin

Incidence of HIV has remained relatively stable in the United States in recent years (1). Nevertheless, new diagnoses are not homogeneously distributed across the United States and some regions are disproportionately affected more than others (2). In 2017, population rates of new HIV diagnoses were highest in the South, where the state of Florida had the highest number of new diagnoses (1). In 2019, the U.S. Department of Health and Human Services (DHHS) released the federal plan for Ending the HIV Epidemic (EHE) within 10 years, identifying 48 counties with high incidence of HIV diagnoses, including seven urban Florida counties (Broward, Duval, Hillsborough, Miami-Dade, Orange, Palm Beach, and Pinellas), for initial funding (3). The EHE plan comprises multiple strategic approaches to reduce new HIV infections by 90% in the next 10 years, including building the capacity to detect and respond to ongoing and emerging clusters of HIV infection (3). Molecular epidemiology techniques (e.g., phylogenetics and phylodynamics) applied to viral genomic data can be used to identify genetic transmission clusters to prioritize for intervention (4, 5). Previous phylodynamic studies have identified external lineage introductions that may respond to drug regimens differently (6), revealed hidden transmission chains (7), and detected rapidly growing clusters of public health concern (8, 9).

The application of molecular methods to characterize the origin, spread, and infection dynamics of HIV in Florida remains to be explored. Per the enhanced HIV/AIDS Reporting System (eHARS), as of the end of 2020, approximately 117,000 people with HIV (PWH) are living in Florida (10), a highly diverse state with frequent tourism and domestic and foreign relocation. The Florida Department of Health (FDOH) has been collecting partial HIV-1 polymerase (pol) sequences from surveillance laboratories to monitor antiretroviral resistance since 2007, with the aim of reaching greater than 60% of persons with diagnosed HIV per year having an analyzable HIV nucleotide sequence within 12 months of diagnosis. The objective of this study was to apply molecular epidemiology techniques to identify HIV-1 clusters with high infection rates, evaluate evidence of imported lineages from outside geographic regions, and explore the phylogeographic spread of the largest clusters. These results were used to consider how the EHE plan could be improved in Florida by the FDOH.

We report, for the first time, an in-depth molecular epidemiology and spatiotemporal analysis of the HIV epidemic in Florida. We identified factors associated with infection cluster status and size, assessed cluster demographic features, and inferred the origin and putative geographic spread of the largest clusters across the state. Considering that Florida is a popular state for tourism and domestic and foreign relocation, we investigated whether the largest clusters were connected to recent introductions from other U.S. states or countries, as has previously occurred in other U.S. regions (6, 14). The lack of recent links between Floridian sequences and those from other U.S. states and international cases suggests an epidemic independently evolving from external influences. Yet, the uncertainty surrounding the time of cluster origin could indicate that epidemiological links among the sampled individuals are missing and that the large clusters may only be revealing a portion of even larger networks. Nevertheless, the detected clusters included exclusively Floridian strains suggesting that for the past several years, the Florida epidemic has been mainly driven by within state transmission rather than frequent outside introductions.

Overall, only 18% of HIV-1 subtype B sequences in Florida were linked in our study, which is comparable with the 22.1% clustering observed in New York City with similar sequence completeness (9). A study in Washington reported a similar clustering rate of 18% among prevalent infections with 49% sequence completeness (15). Yet, our linkage rate is much lower than studies conducted in other states, including North Carolina (50%) using a phylogeny-based approach, and Washington (46%) and Michigan (54%) using identical genetic distance-based methods but with high sequence completeness (16, 17). Although the proportion clustered in our study increased to 32% after removing individuals diagnosed before 2010, these findings were still lower than expected for the large number of sequences analyzed. Compared with simulations by Dasgupta et al., the low level of clustering observed in our study implies that only about 15% of PWH diagnosed between 2012 and 2017 in Florida have received a genotype (17). Yet, our data show that 44% received a genotype during this period (Table 1) and 41.3% received a genotype within 12 months of diagnosis (Table S1). Therefore, the large number of unlinked sequences is likely indicative of issues related to data completeness, rather than slowed transmission (18). While sequence completeness in Florida has improved, it is still below the CDC recommended rate of ≥60%. Despite the fact that molecular epidemiologic inferences are sensitive to data completeness and cannot account for undiagnosed infections (17), the results still provide actionable public health information for health officials (19).

The populations with the lowest odds of clustering in Florida were those with older diagnoses, living in a rural county, and female and Black PWH. These differences may be indicative of disparities in genotype coverage in these vulnerable groups. Cluster size was inversely associated with the age of cluster members—with a greater prevalence of younger PWH detected in the largest clusters. A similar trend was observed in North Carolina (16) and may be due to younger people having more recent diagnoses which increases the likelihood of capturing linkages. Our results are consistent with the epidemiological characteristics of the most at-risk groups for HIV infection in Florida (10). The lack of clustering among women with HIV-1 warrants further research, however, as Florida has the second highest number of women diagnosed with HIV in the nation as of 2017 (1, 2). Persons with mother-to-child (MTC) transmission had lower odds of clustering, which may indicate low rates of genotyping among pregnant women living with HIV, despite engagement in the health care system. Hence, genotyping among viremic pregnant women should be recommended. The reduced odds of clustering among Black PWH who accounted for the largest proportion (42%) of new HIV diagnoses in Florida in 2017 (10) is concerning, and likely a result of receiving suboptimal care. Lower odds of clustering among Black PWH has been observed in previous transmission cluster studies in the United States and may be linked to older or delayed diagnoses, or less genotypic drug resistance testing in this population (5, 9). Our assortative analysis is consistent with prior literature (20). Black PWH make up one of the largest percentages of undiagnosed PWH in the country and are more likely to have lower viral suppression (20, 21). In Florida, Black PWH are least likely to initiate care and have higher odds of drug resistance compared with White and Hispanic/Latino PWH (22). Persons living in rural counties also had lower odds of clustering. Clusters were highly assortative by geography, implying that the missing genetic links are living in the same geographic regions. Southern U.S. states have the highest rates of new HIV infections in nonmetropolitan areas as of 2018 (23). Almost half of PWH in priority clusters in 19 states, including 10 in the south, were not in EHE counties in a 2021 study (24). Recent outbreaks in rural areas driven by the opioid crisis highlight the increased risk for HIV transmission in rural America (25). Several barriers exist in rural communities for HIV prevention and care, including prolonged poverty, stigma, and lack of transportation, which may have contributed to the low clustering we observed in these populations (26). It is important to enhance outreach and public health efforts to help lessen the burden of infection among these groups.

This study revealed significant undersampling in key, possibly vulnerable, populations leading to more than expected unclustered sequences. Undiagnosed infections, lack of health care coverage, distrust in health care systems, HIV criminalization laws, and provider refusal may be among the reasons for decreased genotype testing in these populations. Restrictions on data sharing between states prevented the ability to investigate the degree to which interstate transmission is occurring. However, the CDC notifies states if there are rapidly growing clusters that have members from other states observed, because they have the deidentified data for all jurisdictions. States have their own reporting and data sharing laws, and not all states have implemented molecular HIV surveillance activities. In 2018, the CDC released the notice of funding opportunity, “PS-18-1802 - Integrated Prevention and Surveillance for Health Departments,” which paved the way to improve and increase molecular HIV surveillance activities across funded jurisdictions (27). Departments of Health across the country and the CDC could consider implementing strategies to increase genotyping from providers, while also working to address barriers to testing, and having conversations with the community to address privacy and ethical concerns.

Our phylogeographic analyses show that the Florida epidemic has been largely driven by within-state transmission and that most of the detected clusters have been well established in Florida for a relatively long time, suggesting that missing sequences are likely from Floridian PWH who are undiagnosed, out of care, or whose providers did not order a genotype test. Given the high rates of tourism across the state, it is possible that links to external introductions might missing due to the unavailability of sequences for the vacationers, or to the high proportion of unclustered individuals. The EHE plan prioritizes seven urban Florida counties for heightened HIV prevention services (3). These counties represented significant transmission hot spots in our study, and therefore, our findings support this approach. However, our study also highlights how phylogenetic analysis can provide information on health disparities that needs to be addressed. Our findings revealed low clustering frequency in vulnerable populations which may hinder the success of EHE and further widen disparities in access to HIV care and preventive services. The demographic diversity of PWH in the United States and the disproportionate epidemic among Black PWH necessitates approaches that are both equitable and tailored to key populations (28). Further, HIV transmission is not limited to high incidence areas but can result from influx and efflux of infections to and from these locations limiting success of geographically focused interventions (29). Thus, directing resources to rural Florida counties, in addition to women and Black PWH, will be important to achieve the EHE goals.

When performing our analyses, we considered the ethical discussions recently raised by Tordoff et al. (30), including the inference of transmission directionality among individuals and vulnerable populations, and assortativity of transmission categories. To this effect, our analyses were careful not to infer any individual- or demographic (age, gender, and race/ethnicity) group-level transmission directionality and we exclusively reported virus flow across large geographic regions (i.e., counties) rather than individual groups. Cluster analysis included both geographic and demographic strata, and we focused on differences among clustered and unclustered sets. The findings confirm structural disparity, but also pose new research questions, such as the lack of linkage among women. In the assortativity analysis, we elected to report only spatiotemporal and nonspecific cross-demographic ranges. We acknowledge the lack of theory on how phylogenetic-derived indices are influenced by structural causes of HIV disparity, and that the understanding of such causal pathways at both individual- and community-level is critical to design better interventions. Nonetheless, one of the EHE operational pillars is geographic prioritization, and our objective was to confirm if the current set of Florida counties should be reconsidered. Our findings are of great public health utility as they provide the evidence needed to reconsider additional counties in future iterations of the EHE, with beneficence to the population, to ultimately help achieve health equity and reach vulnerable populations more effectively. In the context of HIV stigma and criminalization, we recognize that there is a need to conduct in parallel ethical discussions on the usage of molecular surveillance data to reduce any potential direct harm to individuals or reiteration of systemic discrimination, and to learn more about the concerns of the community.

This project was supported through contracts and grants from the Florida Department of Health (CODNY-P-01), the National Institute of Allergy and Infectious Diseases (R21-AI138815-01 and R01AI145552-01A1), the Stephany W. Holloway University Chair in AIDS Research, and the University of Florida’s “Creating the Healthiest Generation” Moonshot initiative, which is supported by the University of Florida Office of the Provost, University of Florida Office of Research, University of Florida Health, University of Florida College of Medicine, and University of Florida Clinical and Translational Science Institute. The funders had no role in the writing of the manuscript or the decision to submit it for publication. The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the FDOH or the other funders. We have not been paid to write this article by any agency. We also acknowledge the extremely valuable contributions made by William Switzer, Ellsworth Campbell, and Sergey Knyazev.

S.N.R., M.C.F.P., C.M., and M.S. had full access to the data while all authors had access to all data outputs and contributed to data interpretation; C.M., M.C.F.P., and M.S. conceived and designed the study; S.N.R. and C.M. developed the methodology, with input from S.D. and B.V.; M.C.F.P. and M.S. were responsible for funding acquisition; E.S. was responsible for data curation and acquisition; S.R. and C.M. were responsible for writing-original draft; C.M., S.N.R., M.C.F.P., M.S., S.D., B.V., R.L.C., and E.C.S. were responsible for writing-review & editing. The corresponding authors (C.M. and M.S.) had final responsibility for the decision to submit for publication.

We declare they have no competing interests.