HIV-1 Diversity and Drug Resistance in Treatment-Naïve Children and Adolescents from Rio de Janeiro, Brazil

The human immunodeficiency virus type 1 (HIV-1) can be transmitted via parenteral, sexual, or vertical exposure routes. The number of HIV-1 cases detected yearly in children and adolescents in Brazil did not decrease over the last decade, representing ~5% of total cases described in the country. In recent years, the HIV-1 diversity and the prevalence of transmitted drug resistance mutations (TDRM) are moving toward a marked increase. In this study, we retrospectively evaluated the diversity of HIV-1 subtypes and the TDRM prevalence in 135 treatment-naïve HIV-1 vertically infected children and adolescents born in between 1993 and 2012. These children were assessed in either 2001–2007 or 2008–2012 when they were 0 to 17 years old. The individuals assessed in 2001–2007 (n = 38) had median CD4+ T cell counts of 1218 cells/mm3 (IQR: 738–2.084) and median HIV-1 plasma viral load of 4.18 log10 copies/mL (IQR: 3.88–4.08). The individuals (n = 97) evaluated in 2008–2012 showed median CD4+ T cell counts of 898.5 cells/mm3 (IQR: 591.3–1.821) and median HIV-1 plasma viral load of 4.69 log10 copies/mL (IQR: 4.26–5.33). A steady decrease in the median CD4 T+ cell counts was observed with age progression, as expected. The majority HIV-1 pol sequences (87%) were classified as pure HIV-1 subtypes (77% subtype B, 9% subtype F1 and 1.5% subtype C), while 13% of sequences were classified as recombinants (CRF45_cpx, n = 4; CRF28/29_BF1, n = 2; CRF02_AG, n = 1; CRF40_BF1, n = 1, CRF99_BF1, n = 1, URF_BF1, n = 8). The overall prevalence of TDRM was 14% (19/135), conferring resistance to the nucleoside reverse transcriptase inhibitors (NRTI, 13/135–9.6%), non-nucleoside reverse transcriptase inhibitors (NNRTI, 8/135–5.9%), and protease inhibitors (PI, 2/135–1.5%). The main TDRM observed for NNRTI was the K103N (n = 8), while the mutations T215I/Y/D/E (n = 7) and M184V (n = 4) were the main TDRM for NRTI. Only two TDRM were observed for PI in one individual each (M46I and V82A). Most TDRM were found in the HIV-1 subtype B (84%) sequences. This study reveals an HIV-1 epidemic with high diversity and moderate prevalence of TDRM in the pediatric population of Rio de Janeiro, indicating the existence of possible problems in the clinical management of prophylactic therapy to prevent mother-to-child transmission and future treatment options for the affected children.


Introduction
The human immunodeficiency virus type 1 (HIV-1) can be transmitted via parenteral, sexual, or vertical exposure routes [1]. The risk of HIV-1 mother-to-child transmission (MTCT) can be reduced from~40% to <2% levels using prenatal antiretroviral therapy (ART) [2][3][4] and is associated with the maternal viral load (VL) decline. The HIV-1 prevalence among pregnant women in Brazil is low (<1%) [5]; however, in the last 20 years the (2) 2008-2012, 97 HIV-1 pol sequences from children and adolescents (aged 0-17 years old) submitted to HIV genotypic drug resistance testing as part of the standard procedures of the Brazilian Network for HIV-1 Genotyping (RENAGENO). The samples assessed in 2008-2012 were further subdivided into four different groups according to the age at the sampling: <5 years (n = 63); 5-9 years (n = 22); 10-14 years (n = 10); and 15-17 years (n = 2). The differences observed in the age distribution over the two periods partially reflect changes in the Brazilian public policies for carrying out HIV-1 genotyping tests in the pediatric population.
Information on sex, age, municipality of origin, last viral load quantification, and CD4+ T lymphocyte counts were retrieved from the genotyping test form when available. There was no discerning information between intrauterine, at delivery, or breastfeeding transmission. Samples were geographically grouped into four regions based in the Brazilian Institute of Geography and Statistic (https://www.ibge.gov.br/geociencias/organizacaodo-territorio/divisao-regional accessed on 1 September 2018) procedures: (1) Metropolitan region I (includes the Rio de Janeiro capital and neighboring municipalities); (2) Metropolitan region II (comprising the municipalities from the Baixada Fluminense area, an impoverished area in the outskirts of Rio de Janeiro's capital); (3) Coastal Region; and (4) Northern region. This study was approved by the Oswaldo Cruz Institute Ethics Committee (CAAE 03925112.0.0000.5248).

Extraction, PCR Amplification and Sequencing of HIV-1 RNA
In the 2001-2007 period, the entire protease (PR) and reverse transcriptase (RT) gene sequences were obtained from plasma samples from HIV-1 positive individuals using the ViroSeq™ HIV Genotyping System (version 2.0; Applied Biosystems, Foster City, CA, USA), under conditions recommended by the manufacturers. The purified PCR products were sequenced with an ABI PRISM 3100xl Genetic Analyzer (Applied Biosystems, Foster City, CA, USA). In the 2008-2012 period, HIV-1 PR and RT nucleotide sequences were generated using the TRUGENE ® HIV-1 Genotyping Assay (Siemens HealthCare Diagnostics, Tarrytown, NY, USA), following the recommendations of the manufacturers. The ViroSeq HIV-1 kit covers the entire PR-coding region and the first 320 amino acids of RT. The TRUGENE HIV-1 sequences span the PR (amino acids 4 to 99)-and RT (amino acids 40 to 240)-coding regions. The laboratory is VQA (Virology Quality Assessment) certified for HIV genotyping using both methods, which guarantees the quality of the sequences obtained throughout the study period.

Transmitted Drug-Resistance Mutation (TDRM) Analyses
Sequences were evaluated for the presence of mutations suggestive of TDR with the Calibrated Population Resistance (CPR) Tool Version 8.0 [33]

HIV-1 Subtypes Classification and Recombination Analyses
Preliminary HIV-1 subtype classification was made with REGA subtyping tool v3.46 [35] and was intended to identify putative pure subtypes, circulating, and unique recombinant forms (CRFs and URFs, respectively). This classification was confirmed by phylogenetic analyzes after the alignment of the new 135 HIV-1 pol PR/RT sequences (covering HXB2 nucleotide coordinates 2253 to 3290) with all known HIV-1 group M subtypes and CRFs reference sequences gathered from Los Alamos HIV Sequence Database (http://www.hiv.lanl.gov, accessed on 1 March 2022), using ClustalW algorithm implemented in MEGA v7 program [36]. All reference sequences from pure subtypes and CRFs indicated by REGA and those classically described in South America were included in the phylogenetic analysis (Supplementary Table S1). Drug resistance mutation sites were retained in the final alignment.
Phylogenetic trees were built using Maximum Likelihood (ML) method implemented in the program PhyML v3.0 [37]. The nucleotide substitution model used was the GTR+I+G, selected by the jModeltest program [38]. Heuristic tree search was performed using the SPR branch-swapping algorithm and the reliability of the topology obtained was estimated with the approximate likelihood-ratio test (aLRT) based on the Shimodaira-Hasegawa-like procedure [39]. Only clusters with aLRT values above 0.8 were considered significant. Trees were evaluated in the Figtree software (http://tree.bio.ed.ac.uk/software/figtree, accessed on 1 March 2022). The sequences classified as recombinant forms were further evaluated on Simplot software v3.5.1 [40] to confirm the CRF-like and URF profiles. Bootstrap values supporting branching with HIV-1 reference sequences were determined by NJ trees constructed using the K2-parameter substitution model, based on 100 resamplings, with a 300-nucleotide sliding window moving in steps of 10 bases.

HIV-1 Reference Datasets and Phylogenetic Analyses for Subtype B, CRF02_AG and URF_BF1
HIV-1 subtype B pol PR/RT sequences from this study were aligned with subtype B reference sequences representative of the B PANDEMIC (n = 300) and the B CAR (n = 200) clades described previously [41,42]. The HIV-1 pol PR/RT sequence from this study classified as CRF02_AG was aligned with CRF02_AG reference sequences, respectively, from Africa and Brazil available on Los Alamos HIV Database. The CRF02_AG clustering pattern followed the criteria described previously [31]. The non-CRF-like recombinant sequences identified in this study were aligned with all URF BF1 Brazilian recombinants available on Los Alamos HIV Database. The clustering pattern was investigated by performing ML phylogenetic analyses as described above, and only those clusters encompassing the recombinant sequences from this study and their close relative Brazilian sequences that branched until the second ancestral node in the ML phylogenetic tree with aLRT values above 0.8 were considered significant.

Statistical Analysis
All statistical analyses were done with GraphPad Prism version 6.0 (GraphPad Software Inc., La Jolla, CA, USA) using Fisher's exact test, one-way and two-way ANOVA with Tukey's multiple comparisons tests. A p-value less than 0.05 was considered statistically significant.

Sequence Data
All the partial HIV-1 pol sequences generated in this study were evaluated by the HIV-1 sequence quality tool from Los Alamos HIV sequence database (https://www.hiv.lanl. gov/content/sequence/QC/index.html, accessed on 1 October 2021) to ensure that only sequences with high quality were used in this study and submitted to the NCBI GenBank database. The GenBank database accession numbers for the HIV-1 pol sequences described in this study are OL624883-OL625013. Ninety-seven children and adolescents were included in this study from 2008 to 2012. The main clinical and epidemiological characteristics of these individuals are shown in Table 1. Table 1. Virological, demographic, and laboratory data of HIV-infected children analyzed between 2008 and 2012.
Most subjects (65%) were younger than five-years-old when the genotyping test was conducted (median: 1 year, IQR: 0.91-2 years). The number of individuals in each age group decreased with age progression, reaching 23% in the age group 5-9 years old (median: 6 years, IQR: 5-8 years) and 12% in the age group 10-17 years old (median: 11 years, IQR: 10.75-13.25 years). In general, the pediatric population evaluated in the 2008-2012 period was comprised of 53% females, comprising most individuals sampled in the <5 and 10-17 years old age groups (52% and 83%, respectively). The frequency of male individuals was slightly higher (54%) only in the 5-9 years age group (Table 1); however, there was no statistically significant differences between the sexes (p = 0.4078).
Most recombinant strains (6%, 8/135) displayed a BF1 mosaic structure and were distributed into seven independent lineages, of which five comprised only one sequence each, while one lineage comprised two sequences and all of them were classified as URF_BF. Two BF1 sequences displayed the same mosaic structure and branched together with the CRF28/29_BF (n = 2, aLRT = 1) and was classified as CRF28/29_BF1-like. One sequence branched with CRF40_BF1 (aLRT = 1) and other with CRF99_BF1 (aLRT = 1) reference sequences, being thus classified as CRF40_BF1-like and CRF99_BF1-like recombinants. One sequence displayed a BUF1 mosaic structure and was further classified as an URF_BUF1. We also identified one AG recombinant (0.6%, 1/135) and four AUK recombinants (3%, 4/135) that branched with high support with CRF02_AG (aLRT = 1) and CRF45_cpx (aLRT = 0.97) reference sequences, respectively, and were classified as CRF02_AG-like and CRF45_cpx-like recombinants ( Figure 2).
The CRF02_AG strain identified in this study clustered together with high support (aLRT = 0.86) inside a previously described CRF02_AG Brazilian cluster (02_AGBR-II) comprising other strains from Rio de Janeiro (Supplementary Figure S2).
All non-CRF_BF1-like recombinant sequences (6%, 8/135) identified in this study were combined with all Brazilian URF_BF1 sequences available in the Los Alamos HIV Database (Supplementary Figure S3). The ML phylogenetic analyses showed that lineages BF1-I and BF1-V were not related to other Brazilian URFs. By contrast, the lineage BF1-II branched with high support (aLRT = 0.96) with three URF_BF1 strains from Rio de Janeiro and one from São Paulo, while the lineage BF1-III branched with highest support (aLRT = 1) with two BF1 sequences from Rio de Janeiro. The BF1-VI and BF1-VII lineages branched (aLRT 0.85 and 0.99, respectively) with one Brazilian BF1 sequence from Rio de Janeiro, each. The lineage BF1-IV, comprising two sequences from this study, did not branch with other Brazilian BF sequences (Supplementary Figure S3).
Despite the apparent reduction in TDRM prevalence observed between the two periods (from 24% to 10%), there was no statistical significance (p = 0.2365). The reduction in TDRM prevalence between 2001 and 2012 remained similar (from 24% to 9.5%) when only children <5 years old were considered in the 2008-2012 period; however, this difference was not statistically significant (p = 0.1278).

Discussion
This study shows a complex epidemic scenario amongst treatment-naïve HIV-1 infected children from the Rio de Janeiro Metropolitan Region, which seems to have persisted for more than a decade, between 2001 and 2012. We observed the co-circulation of pure HIV-1 subtypes and diverse and rare recombinant forms, including CRF02_AG and CRF45_cpx, coupled with a reduction in the prevalence of TDRM from 2001-2007 to the 2008-2012 period.
HIV infection in children is challenging, mainly due to the immaturity status of the immune response [43]. Disease progression in vertically infected children shows mainly two key distinct patterns: early progression, with the median age of symptom onset at 4 months of age; or late progression, with a median age of symptom onset at 6 years old [43][44][45]. A meta-analysis evaluating CD4+ T cells levels and viral load in children as predictors of disease progression [46] found that children older than two years old had an increased risk of progression to AIDS when the CD4+ T cell percentages were below 15% and when the viral load exceeded 10 5 copies per mL [46]. In our study, regardless of the period analyzed, most individuals had CD4+ T cell levels above 500 cells/mm 3 , with a gradual decrease observed in the older groups. This was expected since the number of CD4 cells in children gradually decrease with age progression until approach to adult values [47]. Until 2007, the viral load levels were slightly lower than those observed in the 2008-2012 period, with median viral loads reaching the highest levels (10 5 viral copies per mL) in children under five years. According to the PENTA guidelines [48], ART should be started urgently in all infants irrespective of clinical or immunological stage. In children aged > 12 months with no or minor symptoms (CDC clinical stage A or N or WHO stage 1 or 2), ART should be started when the CD4+ T cell count falls below 1000 cells/mm 3 for children aged 1 to 3 years or when the CD4+ T cell count falls below 500 cells/mm 3 for children aged 3 to 5 years. In children >5 years, the treatment should start when the CD4+ T cell count falls below 350 cells/mm 3 .
In 2011, the World Health Organization defined criteria for the Global Plan and eliminate pediatric HIV infections in the world by 2015 [49]. The proposed targets range from the completion of at least one prenatal consultation and an HIV diagnosis for 95% of pregnant women during prenatal care and the immediate initiation of antiretroviral therapy for 90% of HIV-infected pregnant women. The Brazilian government acts actively to reduce the HIV-1 vertical transmission by offering free prenatal and delivery coverage, as well as ensuring universal and free access to ART, HIV follow-up screenings, and breastmilk substitutes. Therefore, the follow-up of the antiretroviral naïve pediatric population tends to become increasingly rare. The presence of ARV-resistant HIV strains in untreated individuals may occur in the setting of broad ARV coverage due to factors such as poor adherence to treatment regimens, combined with the lower genetic barrier of some drugs [50]. These variants that carry antiretroviral resistance mutations may limit the future availability of therapeutic options in newly infected individuals, especially in children who acquired the infection perinatally, increasing the risk of virological failure.
In this study, we found different scenarios regarding the TDRM prevalence in the two time periods analyzed: in the 2001-2007 period, the overall TDRM prevalence was high (24%), reducing to moderate levels (10%) in the 2008-2012 period. The TDRM reported in the Brazilian pediatric population is historically low, ranging from 0% to 12.8% [13,[18][19][20]51]. One study from Manaus-Brazil, on the other hand, reported a TDRM prevalence of 16.2% in children, indicating that the TDRM prevalence scenario in Brazil may be more complex than previously thought [17]. The high TDRM prevalence found in this study in the 2001-2007 period could be partially explained by the therapy adopted in Brazil until 2007 to prevent MTCT. The preventive MTCT standard of care was initially based on the monotherapy with zidovudine, which was later modified to a dual-therapy regimen based on transcriptase inhibitors [12,52]. TDRM conferring resistance to AZT (M41L, F77L, T215I/Y/E/D) were found in 13% of the individuals assessed in 2001-2007, decreasing to 4% in the 2008-2012 sampling period. The zidovudine monotherapy for HIV-1 MTCT prevention can lead to a selection of minority resistant HIV-1 populations that can be transmitted during childbirth [53]. Furthermore, zidovudine monotherapy and a higher maternal viral load were already described as significantly associated with transmission [4]. The moderate TDRM levels observed in this pediatric population since 2008 reflect the advantages of adopting triple antiretroviral therapy, comprising two NRTIs and one NRTI as the initial regimen for children [12]. The moderate TDRM found was lower than a study conducted in Rio de Janeiro including ART-naïve patients with acute/recent HIV infection that found an overall TDRM prevalence of 16.3% [21].
A previous study associated the presence of any DRM in the mother with the presence of any DRMs in infants, supporting the recommendation that all pregnant women and HIVinfected infants undergo HIV genotypic resistance testing prior to initiation of ART [54]. In our study, regardless of the period evaluated, 19 children (14%) had some TDRM. Of these, 89% had viral strains resistant to reverse transcriptase inhibitor drugs, with 53% conferring mutations exclusively for NRTI, 21% only for NNRTI, and 16% for both. Contrasting with our results, in a study conducted on newly diagnosed untreated HIV-infected children from Rio de Janeiro sampled between 2000 and 2003, no TDRM was found. The authors suggested that as most of the children were the index case in the family, the perinatal transmission likely occurred from individuals not aware of their HIV status, and then, no prophylaxis therapy was adopted [20]. Only two individuals had TDRM for NNRTI+PI and PI. Both the M184V and the T215I mutations were the most frequent for the NRTI class. M184V is selected in therapeutic regimens that include the antiretroviral drugs 3TC/FCT and reduces susceptibility to these by up to 100-fold [55]. Furthermore, it gives a low level of resistance to ABC, while it increases susceptibility to drugs such as AZT and TDF [34]. T215I is a thymidine analog mutation (TAM) that may confer a low level of resistance to AZT [55]. For the NNRTI class, the most frequent mutation was K103N, which is known to be selected in individuals who received treatment with NVP or EFV [34]. This mutation confers high resistance to these two drugs. The prevalence of TDRM against NNRTI class, mainly K103N mutation, still poses an important threat to the ART response (virologic clearance) in the HIV-1-infected population from Rio de Janeiro in recent times [21].
The historical context is important to understanding the dynamics of the molecular diversity changing in the HIV-1 epidemics. In Brazil, several studies reported HIV-1 epidemics dominated by subtypes B, F1, C, and the recombinants among them, with a recent trend in the increase of diversity [28,29]. In this context, our study represents an important background to understanding the HIV-1 diversity of the Rio de Janeiro state in the heterosexual population, since the vertically infected children represent a proxy of the heterosexual route of transmission. Contrary to expected, there was no significant difference in HIV-1 diversity between the two periods, indicating that the HIV epidemic profile in the pediatric population of Rio de Janeiro exhibits a remarkable prevalence of non-B subtypes and recombinant forms since the beginning of the 2000s. It is important to note that the finding observed here likely reflect transmission events occurring years before sampling, since the children in our pediatric cohort were born between 1993 and 2012, and probably got infected via MTCT. In this regard, previous studies conducted with pregnant women in Rio de Janeiro found a significant increase in the HIV-1 non-B clades from 2005 to 2015 [25,26], and our results indicate that the change in the subtype B prevalence in Rio de Janeiro probably started several years before.
All subtype B sequences identified in this study clustered within the B PANDEMIC clade, confirming that this clade accounts for almost all subtype B infections in the Rio de Janeiro state [56]. The absence of subtype C in the 2001-2007 period and its low prevalence in the 2008-2012 period indicates that the sampling period is prior to the influx of this clade from the Brazilian southern region [57]. Some unique recombinant forms found in this study showed similar mosaic structures to other sequences from Brazil. Two putative new CRFs carrying the subtypes B and F1 were found, comprising isolates from Rio de Janeiro and São Paulo states. More studies are necessary to determine whether these recombinant sequences represent new CRFs or are simply URF displaying the same recombination pattern in the pol region. Besides the HIV-1 pure subtypes and CRFs classically found in the HIV-1 epidemic in Brazil [29], we also found one CRF02_AG and four CRF45_cpx sequences. The CRF02_AG strain branched within a cluster that has spread locally in Rio de Janeiro state over the last 30 years [31,57]. The four CRF45_cpx strains identified in this study were previously described as participating in an autochthonous transmission network that has spread to the Rio de Janeiro, São Paulo, and Minas Gerais states [32]. The high molecular diversity found in the HIV-1-infected pediatric population indicates that the Rio de Janeiro state has acted as a hub for the introduction and spread of new HIV-1 clades in the Brazilian HIV epidemic since the 2000s.
The main limitation of this study was the distinct age of the populations sampled in the two periods. These differences could introduce a bias since some TDRM could revert in older prenatally ART-naïve infected children [58], resulting in an underestimation of their prevalence. It is important to point out that all children included in this study were ART drug-naïve at the moment of inclusion and the majority (65%) of the children included in the 2008-2012 period were <5 years. The prevalence of TDRM was similar in the overall sample of the 2008-2012 period (10/97) compared with the subsample (6/63) of children <5 years (10% versus 9.5%, respectively). However, caution is warranted when comparing this study with others since it represents an epidemiological scenario from more than ten years ago. In that period, the universal access to ART began to be expanded in Brazil, with the offer of new pediatric therapeutic options that facilitated the initiation of global goals for the end of the AIDS epidemic by 2030 (90-90-90) established in 2014 by UNAIDS.

Conclusions
In summary, the results presented here indicate an important decrease in the TDRM prevalence in the pediatric population in Rio de Janeiro between 2001-2007 and 2008-2012 while no differences were observed in the HIV-1 molecular diversity among these two periods. The TDRM found may limit the future therapeutic options in the children infected perinatally, increasing the risk of virological failure. Our study also found high molecular diversity circulating in Rio de Janeiro since the beginning of the 2000s, with the presence of HIV-1 clades rarely detected in Brazil. More studies are necessary to evaluate the TDR prevalence in HIV-infected children in more recent periods, especially considering pretreatment genotyping assessment to determine the optimal ART regimen to be used.