Multiple Infections, Recombination, and Hypermutation During a 12-Month Prospective Study of Five HIV-1 Infected Individuals

Abstract

The considerable HIV-1 genetic diversity has several implications for viral adaptive and evolutionary capabilities. Its genetic diversity is due to its high mutational rates derived from the error-prone viral reverse transcriptase activity, which generates highly heterogeneous viral populations. Moreover, genetic diversity can also increase due to intra- or intersubtype viral genomic recombination following multiple infections. This study examines HIV-1 intersubtype recombinant viruses and their increased genomic diversity over a 12-month period in five individuals from São Paulo state, Brazil. We collected peripheral blood mononuclear cells once every three months from selected participants at five distinct visits. Molecular clones of 1.15 Kbp fragments of the Pol polyprotein, spanning the protease and a portion of the reverse transcriptase (RT) genes, were generated by bulk PCR. Pol sequences were used for evolutionary analysis, including phylogenetics (using TnT), genetic diversity (using Highlighter), and hypermutation frequency (using Hypermut). Recombination detection experiments were conducted with a jumping profile-hidden Markov model (jpHMM), SimPlot++, and RDP5. We observed great genetic diversity and frequent recombination events in all patients. Furthermore, most of the patients presented hypermutations. These findings highlight the dynamic nature of HIV-1 genetic diversity, driven by frequent recombination and hypermutation, which can accelerate viral adaptation and diversification, underscoring the challenges for treatment, prevention, and disease control.

1. Introduction

HIV-1 persists and adapts through remarkable genetic diversity, which underlies its capacity for immune evasion, antiretroviral resistance, and lifelong infection [1,2,3,4]. This diversity arises from three major mechanisms: high mutation rates, recombination, and host-driven hypermutation. Reverse transcriptase (RT) operates with low fidelity, with an estimated error frequency of $3.4\times {10}^{-5}$ per nucleotide per reverse-transcribed genome, corresponding to roughly three errors for every ten replication cycles [5,6,7]. Coupled with short generation times, high viral loads, and the production of up to ${10}^{10}$ virions per day during acute infection [8,9], this leads to highly heterogeneous viral populations known as quasispecies [10,11,12].

Beyond mutation, recombination is a powerful driver of HIV-1 diversification [2,13]. Reverse transcriptase frequently switches templates 2–3 times during the replication of the viral genome [14,15], resulting in a recombination rate of $3.0\times {10}^{-4}$ per site per replication cycle [14,16]. This occurs approximately ten times more frequently than the accumulation of point mutations [17], highlighting that recombination is an intrinsic feature of HIV-1 replication [13,18,19]. By facilitating the exchange of large genomic fragments, recombination accelerates viral adaptation and enables rapid exploration of new fitness landscapes [14,20,21,22].

Recombination also shapes the global HIV-1 pandemic by giving rise to circulating recombinant forms (CRFs) and unique recombinant forms (URFs) [3,23,24,25]. According to the Los Alamos HIV-1 Sequence Database, at least 158 CRFs have been described to date [26], ranging from simple two-subtype mosaics to complex recombinants derived from multiple parental lineages [27,28]. Some CRFs represent “second-generation” recombinants containing structural segments of earlier CRFs [29]. Alongside these circulating forms, patient-specific URFs are frequently reported [30,31,32,33]. CRFs and URFs collectively account for roughly 29% of all HIV-1 strains globally, followed by subtype C (23%) and subtype A (17%) [34]. Among CRFs, CRF01_AE represents 42% of global cases, followed by CRF07_BC (28%) [34].

In Brazil, the distribution of HIV-1 subtypes and recombinants reflects this global complexity. Subtype B predominates (64.0%), followed by subtype C (13.2%), F1 (10.9%), BF recombinants (7.3%), BC recombinants (3.9%), and other rare forms (0.7%) [35]. The frequent co-circulation of multiple subtypes increases the likelihood of multiple infections, where two or more viral strains establish infection within the same individual, facilitating ongoing recombination and potentially altering disease progression [36,37,38,39,40,41]. Distinguishing single infections caused by stable recombinants from multiple infections involving actively recombining strains is therefore essential [38,42,43].

In contrast to recombination and mutation, APOBEC3-driven hypermutation generates defective proviruses lacking coding capacity [44]. Hypermutation involves excessive guanine-to-adenine substitutions introduced by the host apolipoprotein B mRNA editing enzyme catalytic polypeptide-like 3 (APOBEC3) family, particularly APOBEC3G [45,46]. While APOBEC3G activity can inactivate viral genomes, HIV-1 counters this host defense via the Vif protein, which promotes APOBEC3G degradation [47].

Here, we investigated intra-host HIV-1 diversity dynamics in five individuals from São Paulo, Brazil, over a 12-month period. We assessed the contribution of multiple infections, recombination, and hypermutation to viral genomic diversity. By analyzing molecular clones across five time points, we aimed to identify previously undetected subtypes, patient-specific recombinants, and hidden superinfections, thereby enhancing our understanding of HIV-1 evolutionary processes in a highly recombinant epidemic [48].

2. Materials and Methods

2.1. Patient Selection and Sampling

Five patients (BR1052, BR1114, BR1117, BR2028, and BR2038) were included in this project. These individuals were part of a cohort study previously conducted at the Federal University of São Paulo (UNIFESP) [49]. This research was approved by the Research Ethics Committee of the Federal University of São Paulo (accession number 1919/10). All these patients signed an Informed Consent Term and agreed to donate blood samples. All individuals were strictly treatment-naive throughout follow-up.

Blood samples obtained during the first visit (V1) were used to generate near full-length HIV-1 genome sequences [50]. The inclusion criteria for HIV-1 infected patients in this study were as follows: (1) evidence of recent infection detected by the STARHS test (serologic testing algorithm for recent HIV seroconversion); (2) identification of recombination in the HIV-1 sequence from V1 using SimPlot++ v1.3 [51]; and (3) availability of four consecutive blood samples, from visit two (V2) to visit five (V5), each spaced at three-month intervals. The total follow-up period was one year.

Sequences from the first visit are deposited in NCBI’s GenBank under the following accession numbers: PX482743 (BR1052), JN692478 (BR1114), PX482744 (BR1117), JN692437 (BR2028), and PX482745 (BR2038). All 199 clone sequences from V2 to V5 are also available in GenBank under accession numbers PX449047–PX449245.

2.2. DNA Extraction, Quantification, and Amplification

Peripheral blood mononuclear cells (PBMCs) were separated as described previously [52]. Cellular DNA from each sample was extracted using the QIAamp Viral DNA Mini Kit (Qiagen, Hilden, Germany), following the manufacturer’s instructions. The extracted DNA was quantified using a NanoDrop 2000c spectrophotometer (Thermo Scientific, Waltham, MA, USA), measuring absorbance at 560 nm.

Proviral DNA from each patient visit was amplified by nested bulk PCR, generating a 1.15 Kbp fragment spanning the protease gene and a portion of the reverse transcriptase (RT) gene. All analyses in this study were performed exclusively on this same fragment. Such region was selected due to the identification of recombination breakpoints in V1 sequences. PCR products were cloned and sequenced as described below.

For the first round of the nested PCR, reactions were set up with the following components: 25.0 mU Platinum^® Taq DNA Polymerase (Invitrogen, Carlsbad, CA, USA), 0.4 mM of dNTPs (GE Healthcare, Piscataway, NJ, USA), 2.5 mM of MgCl₂, Tris-HCl 1.0× buffer (pH 8.3), and 0.8 µM of each primer:

• POL1 (HXB2 pos. 1839–1863): 5′–GGGAGTGGGGGGACCCGGCCATAA–3′
• INBO1R (HXB2 pos. 3791–3772): 5′–CATTTGGCCTTTGCCCCTGCTTCTGTT–3′

The cycling conditions for the first round were: 35 cycles (94 °C for 45 s, 55 °C for 45 s, and 72 °C for 2.5 min), preceded by an initial incubation at 94 °C for 5 min and followed by a final extension at 72 °C for 7 min.

For the second round of the nested PCR, 5.0 µL of the first-round product was transferred to the same reaction mix as above, using the following primers:

• K1 (HXB2 pos. 2147–2166): 5′–CAGAGCCAACAGCCCCACCA–3′
• K2 (HXB2 pos. 3309–3338): 5′–TTTCCCCACTAACTTCTGTATGTCCTTGACA–3′

Cycling conditions for the second round were: 35 cycles (94 °C for 45 s, 55 °C for 45 s, and 72 °C for 1.5 min), preceded by a 5 min initial denaturation at 94 °C and followed by a final extension of 5 min at 72 °C. PCR products were confirmed by 1% agarose gel electrophoresis.

2.3. Cloning and Plasmid Preparation

Following amplification, PCR products were cloned using the Topo TA Cloning Kit (Invitrogen, Carlsbad, CA, USA) and transformed into One Shot TOP10 Chemically Competent Escherichia coli cells (Invitrogen, Carlsbad, CA, USA). At least ten clones were generated from each patient’s visit.

Plasmid DNA was extracted from positive clones using the QIAprep Miniprep Kit (Qiagen, Hilden, Germany), according to the manufacturer’s instructions.

2.4. Sanger Sequencing

Purified plasmid DNA was sequenced using the BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, Foster City, CA, USA) on an ABI 3130xL Genetic Analyzer (Applied Biosystems, Foster City, CA, USA). The primers used for sequencing were:

• RTInt (HXB2 pos. 3018–3042): 5′–CCAGCAATATTCCAAAGTAGCAGTA–3′
• DP10 (HXB2 pos. 2201–2223): 5′–CAACTCCCTCTCAGAAGCAGGAGCCG–3′
• F2 (HXB2 pos. 3301–3321): 5′–GTATGTCATTGACAGTCCAGC–3′
• K1 (HXB2 pos. 2147–2166): 5′–CAGAGCCAACAGCCCCACCA–3′
• K2 (HXB2 pos. 3309–3338): 5′–TTTCCCCACTAACTTCTGTATGTCCTTGACA–3′

Sequenced fragments were detected using a Rapid Gel XL 6% cartridge (Amersham, Little Chalfont, UK) in TBE 1.0× buffer for 30 min at 54 °C and 1300 V, with laser potency set to 25%.

2.5. Sequence Alignments

Generated sequences were collected using Sequencing Analysis Software v5.4 (Applied Biosystems, USA), and contigs were assembled with a minimum overlap of 30 bp and a minimum combination threshold of 99–100%. Consensus sequences were then determined with Sequencher Analysis Software v4.5.

All sequences had their HIV-1 identities confirmed using the BLAST tool, version 2.17.0 [53]. Raw sequences from all visits were aligned and trimmed to focus on the clone regions, selecting only the relevant fragments (Pol polyprotein) for further analysis (Supplementary Materials, alignments 1052_aligned.fasta, 1114_aligned.fasta, 1117_aligned.fasta, 2028_aligned.fasta, and 2038_aligned.fasta). To address stop codons in some polyproteins, we used a custom Python version 3.13.5 script (Supplementary Materials, program LORFS.py) to identify and extract the longest open reading frame (ORF) for each sequence. The resulting ORFs were aligned in Geneious Prime^® v2025.0.3 using translation alignment (Geneious algorithm, global alignment with free end gaps, PAM-250 cost matrix, standard genetic code, translation frame = 1, gap open penalty = 12, gap extension penalty = 3, two refinement iterations) (Supplementary Materials, alignments extended_lorfs.fasta). This alignment was used for phylogenetic inference, genetic diversity analysis, hypermutation detection, and intrapatient recombination analysis with RDP5 v5.63 [54].

In addition, a second custom Python script (Supplementary Materials, program Translate.py) was developed to translate all nucleotide sequences (Supplementary Materials, other files translated_extended_lorfs.fasta) and verify their coding capacity. Sequences identified with deleterious mutations were subsequently aligned with MAFFT v7.0 [55] to maximize overall similarity and facilitate the detection of putative recombinants (Supplementary Materials, alignment 1052_aligned.nexus, 1114_aligned.nexus, 1117_aligned.nexus, 2028_aligned.nexus, and 2038_aligned.nexus). These alignments were then analyzed with SimPlot++ v1.3 [51].

2.6. Phylogenetic Inference and Diversity Analysis

We used phylogenetic inference, along with the Highlighter online tool [56], to detect genetic diversity and multiple viral variants. To perform the phylogenetic analysis, we included clones from visits V2 to V5 of the five patients.

Four complete HIV-1 genomes were included as an outgroup: AB485641 (subtype B), AF005494 (subtype F), AY169816 (subtype O), and AF286228 (subtype C). These sequences represent major HIV-1 subtypes and were used to provide phylogenetic context for the analysis.

We employed the TNT v1.6 [57,58] without taxon limit for the phylogenetic tree search under the parsimony optimality criterion (Supplementary Materials, program Script.RUN). The parameters used for tree search were xmult= level 5 chklevel 3 replicates 1000. Each tree search experiment was replicated 10 times. The most parsimonious trees from all replicates were used to produce the final strict consensus tree (Figure 1).

The final tree was visually inspected for evidence of long-branch attraction (LBA) following the protocol described in [59]. LBA is a potential source of topological distortion under the parsimony criterion, typically recognized when unusually long branches cluster together within otherwise short-branched clades. Although most often associated with parsimony, analogous artifacts can also arise under maximum likelihood and Bayesian inference. Under parsimony, however, LBA can be directly tested by sequentially removing long branches and assessing whether this alters the inferred placement of their putative sister groups.

To evaluate how well patient identity and visit information were reflected in the phylogeny, we calculated the Consistency Index (CI) and Retention Index (RI) using TNT v1.6. CI measures the minimum number of possible steps relative to the observed steps for a character, while RI quantifies the degree to which synapomorphies are preserved despite homoplasy. Three single-character matrices were constructed: (i) patients as multistate characters, (ii) visits as multistate characters, and (iii) a compound state representing each patient–visit combination. Outgroups were coded as state 0. CI and RI were obtained using the STATS.RUN macro distributed with TNT, applied to the maximum parsimony tree inferred from the aligned HIV-1 sequences.

We also used YBYRÁ [60] to rank terminals with the potential to behave as wildcards. Wildcards are taxa with unstable phylogenetic placement that occupy multiple alternative positions across equally parsimonious trees, thereby reducing resolution and often collapsing clades into polytomies. Identifying putative wildcards highlights terminals that may represent problematic sequences, such as chimeras, contaminants, incomplete fragments, or data with excessive ambiguities. This process enables their examination under reciprocal illumination, helping to distinguish genuine lack of phylogenetic signal from localized topological instability or error, and providing a clearer view of the underlying tree structure.

For each patient, we prepared alignments that included the V1 genome as a reference along with all sequences from subsequent visits (Supplementary Materials, alignments 1052_aligned.fasta, 1114_aligned.fasta, 1117_aligned.fasta, 2028_aligned.fasta, and 2038_aligned.fasta), and analyzed them using the Highlighter online tool [56].

2.7. Recombination Detection

The programs jpHMM (jumping profile Hidden Markov Model) [61] and SimPlot++ v1.3 [51] were used to identify putative recombination events among the different HIV-1 subtypes. We selected one reference sequence for each subtype to run SimPlot++. Their GenBank accession numbers were: DQ676872 (A1), AF286238 (A2), K03455 (B), U52953 (C), K03454 (D), AF077336 (F1), AY371158 (F2), AF084936 (G), AF190127 (H), EF614151 (J), AJ249235 (K), and MN271384 (L). These references were selected from the Los Alamos HIV Sequence Database [26]. One sequence from each time point of every patient was aligned with reference sequences representing all subtypes. The length of the alignment was adjusted according to the length of the smallest sequence, as the PHI test calculation does not accept gaps at the end of sequences.

RDP5 v5.63 [54] was used to detect intrapatient recombinations with the following algorithms: RDP [62], GENECONV [63], Bootscan [64], Maxchi [65], Chimaera [66], SiScan [67], and 3seq [68]. We only accepted putative recombination events that were detected by four or more of these algorithms.

2.8. Detection of Hypermutation

Hypermutations of clone sequences from visits V2 to V5 were assessed by the Hypermut online tool [69], using V1 sequences as references. This software estimates a value for each performed test corresponding to the result of a Fisher’s exact test comparing a control sample (which must contain the lowest possible index of G-to-A substitutions) to one or more samples [69]. The patients’ sequences with a p-value lower than 5% were considered to be hypermutated.

3. Results

3.1. Demographic and Clinical Characteristics of the Study Cohort

Most of the patients were male, infected through the sexual route, and had an average age of 30.79 years at the time they were included in the cohort study. Also at that time, their average CD4 cell count was 529 cells/mL, with a viral load of 21.100 copies of RNA/mL (log = 4.32) (

Table 1. Demographic and clinical characteristics of the five patients at study entry (Visit 1). All presumed routes of infection are sexual. Patient age is given by years. The CD4+ count unit is cells/µL. Viral load is expressed by RNA copies/mL, whith scaled Log10 values inside parenthesis. Standard deviations (SD) are shown in the last line. M: Male. F: Female.

Patient ID	Sex	Age	CD4+ Count	Viral Load
BR1052	M	28	512	15,000 (4.18)
BR1114	M	32	485	28,000 (4.45)
BR1117	M	35	610	18,500 (4.27)
BR2028	F	29	498	25,000 (4.40)
BR2038	M	30	640	19,000 (4.28)
Mean ± SD		30.8 ± 2.6	529 ± 65	21,100 ± 5200 (4.32 ± 0.11)

).

3.2. Sequence Alignment, Phylogenetics, and Diversity Analysis

Most HIV-1 clones from the patient BR1052 were defective (87.5%), except for only five sequences (12.5%). The BR1114 had 25 defective clones (64.1%) and 14 translation-competent (35.9%). The BR1117 exhibited 22 defective sequences (55%) and 18 translation-competent (45%). For the BR2028, there were 28 (70%) defective clones and 12 translation-competent (30%). Finally, most sequences from the BR2038 were defective (75%), except for 10 sequences (25%). Defective proviral genomes observed here likely originate from endogenous processes such as APOBEC-mediated hypermutation, polymerase errors, and stochastic replication dynamics.

To evaluate how well patient identity and visit information were reflected in the phylogeny, we coded these labels as multi-state characters and calculated their Consistency Index (CI) and Retention Index (RI) using TNT v1.6. The patient character showed the strongest fit to the tree (RI = 0.880, CI = 0.208), indicating that viral sequences from the same patient tended to form clades, even if not perfectly. In contrast, the visit character displayed much weaker clustering (RI = 0.322, CI = 0.038), suggesting that temporal sampling within patients contributed less to the phylogenetic structure, which was expected. The compound patient–visit character yielded intermediate values (RI = 0.474, CI = 0.180), consistent with the expectation that while visits introduce additional variation, patient identity remains the primary organizing factor of viral relatedness.

Complementing this tree-based analysis, Highlighter plots (Figure 2, Figure 3, Figure 4, Figure 5 and Figure 6) revealed clusters of sequences defined by shared mutations relative to a reference sequence.

We note that, not all of the clusters observed in the Highlighter plots (Figure 2, Figure 3, Figure 4, Figure 5 and Figure 6) corresponded to sister clades in the phylogenetic tree. This difference reflects the underlying logic of the methods: while phylogenetic trees represent inferred evolutionary relationships, Highlighter compares each sequence to a master and groups them by overall similarity and mutational patterns (including transitions, transversions, synonymous/non-synonymous substitutions, or APOBEC signatures). As a result, Highlighter can highlight recombination or hypermutation events and generate similarity-based clusters that do not necessarily mirror clades in the evolutionary tree. Thus, the combination of tree-based indices (CI/RI) and similarity-based visualization provides complementary perspectives on intra-patient HIV diversity.

3.3. jpHMM Results

The jpHMM results are summarized in

Table 2. Recombination events detected by jpHMM in the five patients included in our study.

Patient	Visit	Clone	Subtype	Parent 1	Parent 2	Parent 3	Start–End
1052	1		BF1	F1	B		115-1187, 1590-2392, 8528-8607
1052	2	1	BF1	B	F1		865-1084
1052	2	2	BF1H	B	F1	H	1-42 (H), 867-1086 (F1)
1052	2	3	BF1	B	F1		866-1085
1052	2	5	BF1	B	F1		846-1086
1052	2	6	BF1	B	F1		867-1086
1052	2	7	BF1	B	F1		866-1085
1052	2	8	BF1	B	F1		846-1086
1052	2	9	BF1	B	F1		866-1085
1052	2	10	BF1H	B	F1	H	1-42 (H), 867-1086 (F1)
1052	3	1	DF1K	K	D	F1	453-857 (D), 858-1077 (F1)
1052	3	2	BF1	B	F1		859_1077
1052	3	3	BF1	B	F1		866-1085
1052	3	4	BF1	B	F1		866-1085
1052	3	5	BF1	B	F1		866-1085
1052	3	6	BF1K	B	K	F1	1-251 (K), 864-1003 (F1)
1052	3	7	BF1	B	F1		866-1085
1052	3	8	BF1	B	F1		866-1085
1052	3	9	BF1H	B	F1	H	1-42 (H), 867-1086 (F1)
1052	3	10	BF1	B	F1		866-1085
1052	4	1	BF1	B	F1		866-1085
1052	4	2	BF1	B	F1		836-1055
1052	4	3	BF1	B	F1		836-1055
1052	4	4	BF1	B	F1		866-1085
1052	4	5	BF1	B	F1		867-1086
1052	4	6	BF1	B	F1		866-1085
1052	4	7	BF1	B	F1		866-1085
1052	4	8	BF1	B	F1		836-1055
1052	4	9	BF1	B	F1		866-1085
1052	4	10	BF1	B	F1		867-1086
1052	5	1	BF1	B	F1		846-1086
1052	5	2	BF1	B	F1		867-1086
1052	5	4	BF1	B	F1		866-1085
1052	5	5	BF1	B	F1		865-1084
1052	5	6	BF1	B	F1		865-1084
1052	5	7	BF1	B	F1		865-1084
1052	5	8	BF1	B	F1		867-1086
1052	5	9	BF1	B	F1		865-1084
1052	5	10	BF1	B	F1		865-1084
1114	1		BC	C	B		1622-1082, 6104-6973
1114	2	1	BC	C	B		1-381
1114	2	2	BC	C	B		1-394
1114	2	3	BCJ	C	B	J	1-41 (J), 42-394 (B)
1114	2	4	BC	C	B		1-393
1114	2	5	BC	C	B		1-381
1114	2	8	BC	C	B		1-379
1114	3	1	BC	C	B		1-394
1114	3	2	BC	C	B		1-383
1114	3	3	BC	C	B		1-394
1114	3	4	BC	C	B		1-394
1114	3	5	BC	C	B		1-393
1114	3	6	BC	C	B		1-394
1114	3	7	BC	C	B		1-394
1114	3	8	BC	C	B		1-394
1114	3	9	BC	C	B		1-394
1114	3	10	BC	C	B		1-381
1114	4	1	BC	C	B		1-391
1114	4	2	CDJ	C	D	J	1-102 (J), 103-385 (D)
1114	4	4	BC	C	B		1-394
1114	4	5	BC	C	B		1-394
1114	4	6	BC	C	B		1-394
1114	4	7	BC	C	B		1-394
1114	4	8	BC	C	B		1-394
1114	4	9	BC	C	B		1-394
1114	4	10	BC	C	B		1-394
1114	5	1	BC	C	B		1-392
1114	5	2	BC	C	B		1-394
1114	5	4	BC	C	B		1-393
1114	5	5	BC	C	B		1-394
1114	5	6	BC	C	B		1-394
1114	5	7	BC	C	B		1-394
1114	5	9	BC	C	B		1-394
1114	5	10	BC	C	B		1-394
1117	1		BF1	B	F1		1695-1820, 3150-3796, 4189-4631, 8029-8829
1117	2	4	BF1	B	F1		794-1036
1117	3	2	BF1	F1	B		973-1086
1117	3	5	DF1	F1	D		973-1085
1117	4	1	BF1	F1	B		973-1086
1117	4	2	DF1	F1	D		973-1086
1117	4	10	BDF1	F1	B	D	1-396 (B), 970-1082 (D)
1117	5	1	DF1	F1	D		973-1086
1117	5	2	BF1	B	F1		633-1084
1117	5	3	DF1	F1	D		973-1085
1117	5	4	DF1	F1	D		973-1086
1117	5	8	BF1	B	F1		633-1084
2028	1		BF1	B	F1		952-1685
2028	5	5	BJ	B	J		1-39
2028	5	7	BJ	B	J		1-63
2038	1		BF1	F1	B		245-4381
2038	2	4	BK	B	K		1-274
2038	3	3	A2B	B	A2		1-34
2038	3	6	A2B	B	A2		1-34
2038	5	1	A2B	B	A2		1-36

(complete data in Supplementary Materials, Table S1).

In BR1052, jpHMM identified recombination in V1 and in most clones from V2–V5, except for two sequences of unidentified subtype (4.88%). The detected subtypes included 34 BF1 (82.93%), three BF1H (7.31%), one DF1K (2.44%), and one BF1K (2.44%). In BR1114, most sequences were recombinant, except six clones of pure subtype, with the following distribution: 32 BC (80%), six B (15%), one BCJ (2.5%), and one CDJ (2.5%). In BR1117, most clones were pure subtypes, with only 11 recombinants: 25 F1 (60.98%), six BF1 (14.63%), five DF1 (12.19%), four B (9.76%), and one BDF1 (2.44%). In BR2028, recombination was rare (three clones): 38 B (92.68%), two BJ (4.88%), and one BF1 (2.44%). Finally, in BR2038, most clones were pure subtypes, with only five recombinants: 36 B (87.8%), three A2B (7.32%), one BK (2.44%), and one BF1 (2.44%). Overall, jpHMM indicated that all patients had multiple infections and novel recombination events, including new subtypes, across visits. These findings corroborate the phylogenetic and genetic diversity analyses.

3.4. SimPlot++ Results

The SimPlot++ results are presented in

Table 3. Recombinations detected by SimPlot++ in the patients included in our study. PHI was calculated over 1000 permutations.

Patient	Visit	Clone	PHI	Subtype	Parent 1	Parent 2
1052	1		0.00 $\times {10}^0$	BF1	F1	B
1114	2	2	8.00 $\times {10}^{-3}$	BC	C	B
1114	2	3	1.20 $\times {10}^{-2}$	BC	C	B
1114	2	4	1.70 $\times {10}^{-2}$	BC	C	B
1114	2	5	1.20 $\times {10}^{-2}$	BC	C	B
1114	2	6	1.30 $\times {10}^{-2}$	B
1114	2	7	9.00 $\times {10}^{-3}$	B
1114	2	8	1.00 $\times {10}^{-2}$	B
1114	2	9	1.10 $\times {10}^{-2}$	B
1114	2	10	8.00 $\times {10}^{-3}$	B
1114	3	1	1.60 $\times {10}^{-2}$	BC	C	B
1114	3	2	1.60 $\times {10}^{-2}$	BC	C	B
1114	3	3	8.00 $\times {10}^{-3}$	BC	C	B
1114	3	4	1.60 $\times {10}^{-2}$	BC	C	B
1114	3	5	1.00 $\times {10}^{-2}$	BC	C	B
1114	3	6	1.40 $\times {10}^{-2}$	BC	C	B
1114	3	7	1.20 $\times {10}^{-2}$	BC	C	B
1114	3	8	9.00 $\times {10}^{-3}$	BC	C	B
1114	3	9	7.00 $\times {10}^{-3}$	BC	C	B
1114	3	10	1.60 $\times {10}^{-2}$	BC	C	B
1114	4	1	4.00 $\times {10}^{-2}$	BC	C	B
1114	4	2	4.80 $\times {10}^{-2}$	BC	C	B
1114	4	3	3.40 $\times {10}^{-2}$	B
1114	4	4	2.50 $\times {10}^{-2}$	BC	C	B
1114	4	5	3.50 $\times {10}^{-2}$	BC	C	B
1114	4	6	4.20 $\times {10}^{-2}$	BC	C	B
1114	4	7	3.30 $\times {10}^{-2}$	BC	C	B
1114	4	8	4.20 $\times {10}^{-2}$	BC	C	B
1114	4	9	3.40 $\times {10}^{-2}$	BC	C	B
1114	4	10	4.00 $\times {10}^{-2}$	BC	C	B
1114	5	1	2.00 $\times {10}^{-3}$	BC	C	B
1114	5	2	0.00 $\times {10}^0$	BC	C	B
1114	5	3	2.00 $\times {10}^{-3}$	B
1114	5	4	2.00 $\times {10}^{-3}$	BC	C	B
1114	5	5	1.00 $\times {10}^{-3}$	BC	C	B
1114	5	6	2.00 $\times {10}^{-3}$	BC	C	B
1114	5	7	0.00 $\times {10}^0$	BC	C	B
1114	5	9	3.00 $\times {10}^{-3}$	BC	C	B
1114	5	10	3.00 $\times {10}^{-3}$	BC	C	B
1117	1		0.00 $\times {10}^0$	BF1	B	F1
2028	1		0.00 $\times {10}^0$	BF1	B	F1
2028	3	1	4.10 $\times {10}^{-2}$	B
2038	1		0.00 $\times {10}^0$	BF1	F1	B

(complete data in the Supplementary Materials, Table S2).

In BR1052, the PHI test detected recombination only in V1 (BF1 subtype). Although plots suggested recombination across clones, PHI tests were not significant, likely due to short parent sequences and recombination breakpoints near the 3′ end. After adjusting the V1 sequence length to match the smallest clone and realigning to subtype references, the PHI test remained insignificant. In BR1114, BR1117, and BR2028, significant PHI values were found even for sequences classified as pure subtypes, reflecting the PHI test’s sensitivity to any recombination—including within-subtype events. In BR1114, we observed 33 BC (82.5%) and seven B (17.5%) sequences, all recombinants. BR1117 showed 30 BF1 (73.17%), one BDF1 (2.44%), five F1 (12.20%), and five B (12.20%), of which 10 sequences (seven BF1, two B, one F1) were not recombinant. In BR2028, the V1 sequence was recombinant (BF1 subtype), and all 40 clones were subtype B; of these, 12 were recombinant by PHI. In BR2038, the V1 sequence was recombinant (BF1 subtype), while all 40 clones were subtype B, none showing recombination. Thus, SimPlot++ confirmed multiple infections in all patients except BR1052.

3.5. RDP5 Results

RDP5 detected 14 intrapatient recombinants in BR1052, 10 in BR1114, 12 in BR1117, seven in BR2028, and two in BR2038 (

Table 4. Recombinations detected by RDP5 in the patients included in our study. Minor and major columns indicate the minor and major parents, respectivelly. Major: Major parents. Algorithms: R = RDP; G = GENECONV; B = Bootscan; M = Maxchi; C = Chimaera; S = SiSscan; T = 3seq. Symbols: ! = The recombinant sequence may have been misidentified (one of the identified parents might be the recombinant); ? = Only one parent and a recombinant need to be in the alignment for a recombination event to be detectable. The sequence listed as unknown was used to infer the existence of a missing parental sequence; * = The actual breakpoint position is undetermined.

						Detection Methods
Recombinants	Minor	Major	Start-End	R	G	B	M	C	S	T
!1052_V5_5, 1052_V2_3, 1052_V2_9, 1052_V3_10, 1052_V3_3-5, 1052_V3_8, 1052_V4_1, 1052_V4_4, 1052_V4_6-7, 1052_V4_9, 1052_V5_10	1052_V5_1, 1052_V2_10, 1052_V2_2, 1052_V2_5-6, 1052_V2_8, 1052_V3_2, 1052_V3_9, 1052_V4_10, 1052_V4_5, 1052_V5_2, 1052_V5_8	?(1052_V5_3), ?(1052_V2_4)	4-410	X	X	X	X	X	X	X
!1114_V4_2	1114_V3_6, 1114_V1, 1114_V2_1-2, 1114_V2_4, 1114_V2_5, 1114_V3_1, 1114_V3_3-4, 1114_V3_7-10, 1114_V4_10, 1114_V4_4-9, 1114_V5_10, 1114_V5_2, 1114_V5_5-7, 1114_V5_9	?(1114_V4_3), ?(1114_V2_10), ?(1114_V2_6-7), ?(1114_V2_9), ?(1114_V5_3)	403-1102	X	X	X	X	X	X	X
!1114_V5_3, 1114_V2_6-7, 1114_V4_3	?(1114_V4_1), ?(1114_V3_1), ?(1114_V3_3)	1114_V3_3, 1114_V4_1, 1114_V5_1	1076-80	X	X	X	X	X	X	X
!1114_V2_9	?(1114_V3_6), ?(1114_V4_1), ?(1114_V5_1), ?(1114_V5_4)	1114_V4_1, 1114_V1, 1114_V2_1-2, 1114_V2_4-5, 1114_V2_8, 1114_V3_1-10, 1114_V4_4-10, 1114_V5_10, 1114_V5_2, 1114_V5_5-7, 1114_V5_9	1061-81	X	X	X	X	X	X	X
!1114_V4_1, 1114_V5_4	1114_V2_5	?(1114_V3_6)	324-1078		X	X	X	X	X	X
1114_V5_1	1114_V2_4-5, 1114_V3_6	315-854	X	X	X	X	X	X	X
!1114_V2_5	1114_V2_8	?(1114_V2_4)	869-398	-	-	X	X	-	X	X
!1114_V2_5	?(1114_V2_10)	1114_V2_3, 1114_V2_1-2, 1114_V2_4, 1114_V2_8, 1114_V3_1-9, 1114_V4_4-10, 1114_V5_2, 1114_V5_5-7, 1114_V5_9-10	409*-450	X	-	-	X	-	X	X
!1117_V4_10	1117_V4_1, 1117_V3_4	1117_V3_3	360-1052	X	X	X	X	X	X	X
!1117_V3_4	1117_V2_8, 1117_V1, 1117_V2_1-2, 1117_V2_5-7, 1117_V2_9-10, 1117_V3_1, 1117_V3_10, 1117_V3_2, 1117_V3_5, 1117_V3_7, 1117_V4_1-9, 1117_V5_1-8	?(1117_V3_3), ?(1117_V3_6), ?(1117_V5_10)	422-1141	X	X	X	X	X	X	X
!1117_V2_4, 1117_V2_3	1117_V3_6, 1117_V3_3, 1117_V5_9-10	1117_V1, 1117_V2_10, 1117_V2_8, 1117_V3_1, 1117_V3_10, 1117_V3_7-8, 1117_V4_1, 1117_V4_4, 1117_V4_6, 1117_V5_1, 1117_V5_3-7,	42-628	X	X	X	X	X	X	X
1117_V5_8, 1117_V5_2	1117_V3_1	1117_V3_3, 1117_V3_6, 1117_V5_9-10	489-999	X	X	X	X	X	X	X
1117_V3_7	?(1117_V5_7)	1117_V3_8-10,	4*-52	X	X	X	X	X	X	X
!1117_V4_1, 1117_V3_1[P], 1117_V3_5	1117_V5_9	1117_V4_3, 1117_V4_5-9	1025-1088	-	X	X	X	-	X	X
!1117_V5_9	1117_V5_10	?(1117_V5_4)	100-628*	-	-	X	X	X	X	X
!1117_V3_3	?(1117_V3_1), ?(1117_V2_6-9), ?(1117_V3_5), ?(1117_V4_1), ?(1117_V4_3), ?(1117_V4_5-9), ?(1117_V5_7)	1117_V5_10	108-628*	-	-	-	X	X	X	X
!2028_V4_1	?(2028_V2_10)	2028_V5_8	517*-605	X	X	-	X	X	X	X
2028_V4_6	2028_V5_1, 2028_V5_8	2028_V5_8, 2028_V2_1, 2028_V2_10, 2028_V2_4-7, 2028_V3_1-3, 2028_V3_7-10, 2028_V4_2, 2028_V4_7, 2028_V4_9, 2028_V5_7, 2028_V5_9	997-359	X	X	X	X	X	X	X
!2028_V5_6	?(2028_V5_8)	2028_V2_1, 2028_V2_4-6, 2028_V3_2-3, 2028_V3_8-9, 2028_V4_4, 2028_V4_7, 2028_V5_5, 2028_V5_9	27-308	X	X	X	X	X	X	X
!2028_V5_1	2028_V2_1	?(2028_V2_10)	181-924	-	X	X	X	X	X	X
!2028_V5_8	2028_V3_3, 2028_V1, 2028_V2_1, 2028_V2_5-7, 2028_V2_9, 2028_V3_10, 2028_V3_2, 2028_V3_4-5, 2028_V3_7, 2028_V3_9, 2028_V4_10, 2028_V4_3, 2028_V4_5, 2028_V4_7, 2028_V4_9	?(2028_V4_8)	1069-359	X	X	X	X	X	X	X
!2028_V5_5	2028_V2_5	2028_V2_4, 2028_V3_10	121-871	-	X	X	X	X	X	X
!2028_V4_4	2028_V3_10, 2028_V1, 2028_V3_3-4,	?(2028_V3_3), ?(2028_V3_10)	1053-75	-	-	X	X	-	X	X
!2038_V4_1	?(2038_V2_4)	2038_V3_4	837-261	-	-	X	X	X	X	X
!2038_V2_2	2038_V5_2, 2038_V4_8	?(2038_V2_1), ?(2038_V2_3)	272-888	-	X	X	X	-	X	X

). These results suggest variable rates of intrapatient recombination among patients, reflecting differences in the evolutionary dynamics of HIV-1 within hosts.

3.6. Hypermutation Analysis

BR1052 (

Table 5. Hypermutations found in the patients included in our study. The p-values indicates the results of Fisher’s exact test. All the values are significant and bellow 5%.

Sequences	p-Value (%)	G → A	GG	GA	GC	GT
1052_V3_1	4.24	19	7	9	2	1
1114_V5_5	2.5	7	4	3	0	0
1114_V5_10	4.28	6	3	3	0	0
1114_V5_7	4.28	6	3	3	0	0
1114_V5_9	4.47	6	3	3	0	0
2028_V2_10	1.14	16	0	15	0	1
2028_V3_10	0.21	12	8	3	0	0
2028_V4_9	0.23	12	8	3	0	0
2028_V5_7	2.34	7	1	6	0	0
2038_V2_4	0.23	39	5	16	6	12

; Supplementary Materials, Table S3) and BR2038 (Table 5; Supplementary Materials, Table S4) exhibited a single hypermutated sequence in V3 and V2, respectively. These hypermutated sequences accounted for 2.5% of the total sequences obtained from these patients.

BR1114 exhibited four hypermutated clones, all detected in V5, accounting for 10.26% of the total sequences obtained from this patient (Table 5; Supplementary Materials, Table S5). Similarly, BR2028 (Table 5; Supplementary Materials, Table S6) presented four hypermutated clones (10.26%), distributed across V2 (one clone), V3 (one clone), V4 (one clone), and V5 (one clone).

Finally, BR1117 (Table 5; Supplementary Materials, Table S7) did not exhibit any hypermutated sequences.

4. Discussion

Our study followed the simultaneous processes of diversification and viral evolution in five patients over the course of one year, using real longitudinal HIV sequences obtained directly from infected individuals. Throughout this period, we repeatedly observed multiple evolutionary forces acting at the same time within the viral populations, including the accumulation of synonymous and non-synonymous mutations, recombination events, and hypermutation. This longitudinal approach clearly reveals that a single viral sampling point is insufficient to capture or explain the complexity and dynamics of the evolutionary processes taking place in vivo. To our knowledge, very few studies have examined these evolutionary forces simultaneously—synonymous and non-synonymous mutation, recombination, and hypermutation—using real patient-derived viral sequences collected over such an extended period. The dataset we generated is therefore unusually rich and provides valuable biological material for investigating HIV evolution directly within patients, while also serving as high-quality input for in silico evolutionary modeling. In the current treatment scenario, where most individuals are on antiretroviral therapy with suppressed viral loads, studies of this nature have become nearly impossible because longitudinal, naturally evolving viral populations are no longer accessible.

Accordingly with our results indicating a large portion of retrieved clones from each patient as defective, a crucial factor shaping HIV evolution is that the vast majority of proviral sequences integrated into host cells are defective, harboring large deletions, frameshifts, premature stop codons, or widespread hypermutation. As a consequence, only a small fraction of the proviral reservoir is truly replication-competent, which dramatically reduces the effective population size of HIV even when the absolute number of infected cells appears large. Because most proviruses cannot contribute to productive infection, the evolutionary trajectory of HIV is determined by the far smaller subset capable of completing the replication cycle. This reduction in effective population size intensifies the impact of stochastic forces on HIV evolution. In such constrained populations, random sampling effects and genetic drift can outweigh natural selection, allowing neutral or mildly deleterious variants to persist, rise in frequency, or even become fixed simply by chance. Recurrent population bottlenecks during replication cycles, immune-driven reductions in viral diversity further amplify these stochastic dynamics. Thus, the evolutionary landscape of HIV is shaped not only by deterministic selection but also by the randomness imposed by the limited number of competent genomes participating in replication, making the overall evolutionary trajectory more unpredictable.

Phylogenetic and genetic diversity analysis, together with recombination intersubtype results, demonstrated that all patients included in our study had multiple infections and recombinant HIV-1 sequences. Nunes et al. [48] also observed a very high frequency of intersubtype recombinant HIV-1 sequences in patients with multiple infections. Of 17 multiple-infection cases found in their study, 16 had intersubtype recombinant sequences. The results of the two tools used in this study to detect intersubtype recombinations were not identical. Across the five visits, jpHMM detected more intersubtype recombinations than SimPlot++, identifying new events in all patients, whereas SimPlot++ detected new recombinations only in BR1117. This discrepancy likely reflects the distinct algorithms employed by the two tools. SimPlot++ relies on percent identity between the query sequence and a panel of reference sequences in a sliding window [51], whereas jpHMM uses a probabilistic framework that models each subtype as a profile hidden Markov model and infers recombination breakpoints from the most probable path through the model [70,71]. Consequently, jpHMM is generally more sensitive to detecting recombination breakpoints, while SimPlot++ may overlook events that do not generate strong shifts in similarity across windows. Taken together, these results suggest that relying on a single tool may underestimate the frequency of recombination, and that combining approaches provides a more comprehensive picture of HIV-1 diversity. We also used RDP5 to detect intrapatient recombination independently of the sequence subtype. The number of intrapatient recombinants identified in each patient varied widely, from two to 14. These findings suggest that the recombination rate can vary drastically according to the host conditions. Upon considering all our recombination analyses, we argue that the HIV-1 diversity observed in our study is strongly shaped by recombination, and that its detection depends on the analytical strategy employed. While jpHMM proved more sensitive in identifying recombinant forms, SimPlot++ offered a more conservative view, reinforcing the importance of using complementary tools to capture the full extent of viral diversity.

Regarding the variation in recombination breakpoints, such variability is expected in a probabilistic framework, which naturally predicts a higher frequency of events within certain intervals without requiring each individual breakpoint to be mechanistically explained. Multiple intraviral factors can influence breakpoint positions, including secondary and tertiary RNA structures, the dynamics and abundance of the nucleocapsid protein, fluctuations in the supply and exhaustion of dNTPs within the capsid, the similarity between RNA templates during copy-choice recombination, and the relative abundance of available templates. These influences arise primarily from viral and intracellular processes. At the host-organism level, selective pressures can shape the expansion of specific recombinant lineages, but they do not directly determine the diversity of breakpoint locations themselves.

Taken together, these considerations highlight the need to clearly articulate in the manuscript how each evolutionary force impacts the viral population. Mutation introduces diversity that may be neutral, advantageous, or deleterious; deterministic natural selection constrains that diversity by favoring fitter variants; genetic drift introduces stochasticity and can drive the fixation of variants independently of their fitness; recombination reshuffles genomes, enabling rescue and escape; and hypermutation acts as a form of lethal mutagenesis. These processes are visibly reflected in our data, with synonymous and non-synonymous mutations highlighted in the alignment plots and the phylogenetic trees revealing both recombinant structure and the extent of hypermutation. Altogether, our study captures the concurrent action of multiple evolutionary forces on HIV populations in vivo over an extended period, incorporating the additional evolutionary pressure imposed by a highly defective proviral reservoir and the resulting reduction in effective population size. Such a comprehensive view of HIV evolution is increasingly rare in the literature and, under current clinical conditions, unlikely to be reproduced.

All patients harbored multiple HIV infections whether through superinfection or co-infection which significantly amplify viral diversity and accelerate evolutionary change. When distinct HIV subtypes or genetically divergent strains infect the same host, their simultaneous replication creates conditions that favor the formation of heterozygous virions containing RNA genomes from different lineages. During reverse transcription, HIV’s copy-choice mechanism allows the reverse transcriptase enzyme to switch templates between these divergent RNAs, generating recombinant genomes that integrate into host cells. This process dramatically increases the range of genetic combinations available to the virus, producing mosaic genomes that combine mutations and structural features from distinct subtypes.

The introduction of highly divergent viral lineages into the same replication environment expands the mutational space far beyond what would be achievable through mutation alone. Recombination between different subtypes can rapidly generate variants with novel phenotypic properties, including altered antigenic profiles, differences in replicative capacity, and modified susceptibility to immune responses or antiviral pressures. Because recombination can unite advantageous mutations that evolved independently in different viral backgrounds, it represents an efficient route for the virus to explore adaptive solutions that would otherwise require multiple sequential mutational steps.

One of the most consequential outcomes of such mixed infections is the acceleration of immune escape. Distinct subtypes often differ substantially in the epitopes recognized by host immune responses. Through recombination, HIV can effectively “borrow” escape mutations from one subtype and introduce them into the genomic context of another, producing variants that evade existing antibody or T-cell responses more efficiently than either parental strain. The mosaic nature of these recombinants can also make immune recognition more difficult. Thus, the presence of multiple HIV subtypes in the same host creates evolutionary opportunities that substantially increase viral diversity and generate recombinant forms with enhanced potential for immune escape. This process contributes to the emergence of circulating recombinant forms (CRFs) and unique recombinant forms (URFs), which collectively play a major role in global HIV genetic diversity. In this sense, mixed-subtype infections do not merely add to the pool of viral variants, they fundamentally reshape the evolutionary landscape of HIV by enabling rapid adaptive shifts that would be far less accessible through mutation-driven evolution alone.

Hypermutation has a profoundly deleterious impact on HIV genomes. Driven primarily by APOBEC3-mediated cytidine deamination, hypermutation introduces clusters of G-to-A substitutions that frequently generate stop codons, frameshifts, or amino acid changes incompatible with protein function. As a result, most hypermutated genomes are replication-incompetent and contribute to the large pool of defective proviruses within infected cells. Although hypermutation increases apparent sequence diversity, it effectively acts as a form of lethal mutagenesis, removing large fractions of the viral population from the replicating pool. This reduces the number of genomes available for selection and shifts the evolutionary balance toward stochastic processes, as the effective population size becomes constrained by the small subset of intact, functional genomes that remain.

The frequency of hypermutation in our study varied between 0 and 10.26%, considering all the clones of each patient separately. Kieffer et al. [72] also observed a low frequency of hypermutation in HIV-1. They found no hypermutation in the reverse transcriptase (RT) and protease sequences from plasma virus but detected hypermutations in 6.8% of RT sequences and 6.3% of protease sequences from the latent cellular reservoir. In contrast, de Lima-Stein et al. [73] observed hypermutation in 31.2% of analyzed samples. Taken together, these findings suggest that the extent of HIV-1 hypermutation varies widely depending on the genomic region analyzed, the biological source of the sequences, and the patient population studied. In our cohort, the relatively low levels of hypermutation are consistent with prior reports, reinforcing the view that while APOBEC3-mediated editing contributes to viral diversification, its overall frequency in vivo may be limited by the action of viral proteins such as Vif which directs APOBEC to the proteasome.

The current study is not free from limitations, including those imposed by the relatively small number of patients included. We expect that research involving a larger number of individuals could provide a more accurate response. Furthermore, the length of studied fragment encompassing just a portion of the pol region including partial protease e reverse transcriptase gene could be perceived as insufficient at the moment of genomic sequencing. However, sampling this limited subgenomic fragment was enough allowing for the observation of simultaneous evolutionary processes acting on HIV populations over time.

Our study provides valuable insights into the dynamic genetic diversity of HIV-1 over time, with a particular focus on intersubtype recombinant viruses. Longitudinal analyses of this kind remain scarce, yet they are critical for understanding how recombination and mutation interact within hosts to generate novel variants. By documenting these processes across multiple visits and individuals, we demonstrate the continual emergence of diverse viral populations, which has direct implications for contributing to either viral adaptation, treatment strategies, vaccine design, or the broader understanding of HIV-1 evolutionary dynamics.

Our study highlights the dynamic nature of HIV-1 genetic diversity, driven by mutation, frequent recombination, and hypermutation. Together, these processes generate a continually shifting viral landscape that complicates treatment, undermines prevention strategies, and challenges vaccine design. By capturing longitudinal evidence of these mechanisms in multiple patients, we provide direct insights into the evolutionary forces shaping HIV-1 within hosts. Future studies with larger cohorts will be essential to extend these findings, but our results already underscore the need to consider recombination and hypermutation alongside mutation as central drivers of HIV-1 evolution and persistence.