Few amino acid signatures distinguish HIV-1 subtype B pandemic and non-pandemic strains

Abstract

The Human Immunodeficiency Virus Type I (HIV-1) subtype B comprises approximately 10% of all HIV infections in the world. The HIV-1 subtype B epidemic comprehends a pandemic variant (named B_PANDEMIC) disseminated worldwide and non-pandemic variants (named B_CAR) that are mostly restricted to the Caribbean. The goal of this work was the identification of amino acid signatures (AAs) characteristic to the B_CAR and B_PANDEMIC variants. To this end, we analyzed HIV-1 subtype B full-length (n = 486) and partial (n = 814) genomic sequences from the Americas classified within the B_CAR and B_PANDEMIC clades and reconstructed the sequences of their most recent common ancestors (MRCA). Analysis of contemporary HIV-1 sequences revealed 13 AAs between B_CAR and B_PANDEMIC variants (four on Gag, three on Pol, three on Rev, and one in Vif, Vpu, and Tat) of which only two (one on Gag and one on Pol) were traced to the MRCA. All AAs correspond to polymorphic sites located outside essential functional proteins domains, except the AAs in Tat. The absence of stringent AAs inherited from their ancestors between modern B_CAR and B_PANDEMIC variants support that ecological factors, rather than viral determinants, were the main driving force behind the successful spread of the B_PANDEMIC strain.

Introduction

The Human Immunodeficiency Virus Type I (HIV-1) is one of the most important human pathogens that have emerged in the 20^th century and exhibits an extraordinary degree of genetic variability, organized in groups (M, N, O, and P), subtypes (A-D, F-H, and J-L), sub-subtypes, and many recombinant forms [1]. HIV-1 subtype B comprises approximately 10% of all HIV infections in the world, being one of the most globally disseminated HIV-1 variants and the most prevalent subtype in the Americas, Europe, Oceania, as well as some Asian countries [2].

The HIV-1 subtype B shares a common ancestor with subtype D that was present in Kinshasa, capital of Democratic Republic of Congo (DRC), by the early 1940s [3]. The currently accepted hypothesis about the emergence and worldwide spread of subtype B is punctuated by two major founder events. The first event occurred when the HIV-1 subtype B ancestor strain moved from the DRC into the Caribbean around the middle 1960s, establishing the first HIV epidemic outside the African region [4–7]. The second major event took place when one subtype B strain spread from the Caribbean to the U.S. around the late 1960s and got access to high-risk transmission networks globally connected that fueled the establishment of a pandemic clade (B_PANDEMIC) responsible for most of the infections of this subtype worldwide [8, 9]. In contrast, other non-pandemic subtype B lineages (B_CAR) spread and circulates at high prevalence in several Caribbean islands and the Northern South American region, but were not successfully disseminated worldwide [4, 6, 8–11].

The introduction of the B_PANDEMIC ancestor in highly connected transmission networks in the U.S. during the very early phase of the epidemic probably accounts for the successful dissemination of this viral variant worldwide [4, 12, 13]. Differences in viral fitness, however, may also have shaped the uneven geographic distribution of distinct HIV-1 subtype B lineages. Viral transmissibility correlated with the plasma viremia during chronic infection [14], and some studies found that plasma viremia within subtype B is highly heritable, thus indicating that this trait depends strongly on the virus genotype [15–18]. Notably, a significant trend for higher viral loads among subjects infected with B_PANDEMIC relative to B_CAR strains was recently described in French Guiana [11] which may have played a role in the differential transmissibility of the two viral strains. Studies of molecular signatures in non-pandemic subtype B lineages, however, have been limited so far to the analysis of the env gene of B_CAR strains circulating in Trinidad and Tobago [19, 20].

The objective of this work is to identify amino acid signatures (AAs) that can distinguish B_CAR and B_PANDEMIC strains by the analysis of full-length (FL) and partial genome subtype B sequences representative of different Caribbean islands and American countries and by reconstructing the sequences of the most recent common ancestors (MRCA) of B_CAR and B_PANDEMIC strains. These analyses may provide crucial information to understand the potential relevance of viral genetic determinants on the epidemic spread of different subtype B variants.

Materials and methods

HIV-1 subtypes B and D datasets

HIV-1 subtype B FL genome sequences from North America (n = 330), South America (n = 151), and the Caribbean (n = 25), as well as Sub-Saharan African subtype D FL genome sequences (n = 18) that were available at Los Alamos HIV Database (http://www.hiv.lanl.gov) by November 2018, were downloaded (S1 and S2 Tables). We also downloaded HIV-1 subtype B sequences from Caribbean islands with high prevalence of B_CAR strains and that covered selected genomic regions of gag (HXB2 coordinates: 1,264 to 2,148), pol (HXB2 coordinates: 2,253 to 3,272), and env (HXB2 coordinates: 6,450 to 8,480) and HIV-1 pol sequences (HXB2 coordinates: 2,253 to 3,272) from drug-naïve individuals of Caribbean origin (S1 Table). Only one sequence for patient was selected.

Clade assignment of HIV-1 subtype B sequences

The HIV-1 subtype B sequences were aligned with the HIV Align online tool [21] and then manually curated. The presence of putative intra-subtype recombinant sequences was evaluated using the RDP4 software [22], being deemed as recombinant those sequences selected as such by three or more of the algorithms. The remaining FL and partial subtype B genome sequences were classified as either B_CAR or B_PANDEMIC based on their placement on a maximum likelihood (ML) phylogenetic tree inferred with the PhyML program [23] under the best nucleotide substitution model, selected using an online web server [24]. The heuristic tree search was performed using the SPR branch-swapping algorithm, and the reliability of the obtained topology was estimated with the approximate likelihood-ratio test [25] based on the Shimodaira–Hasegawa-like procedure. The ML phylogenetic trees were visualized using the FigTree v1.4.4 program [26].

Reconstruction of ancestral subtype B sequences

To reduce computation time while retaining most viral diversity information, we generate a “non-redundant” subtype B FL genome subset by removing very closely related B_PANDEMIC sequences. To achieve this goal, B_PANDEMIC sequences with identity above 91.5% were grouped with the CD-HIT program [27] using an online web server [28], and only one sequence per cluster (the oldest one) was selected. To map amino acid changes fixed during the evolution of subtype B, FL genome sequences of the MRCA of B_CAR and B_PANDEMIC strains were then reconstructed using a Bayesian Markov Chain Monte Carlo (MCMC) approach, as implemented in BEAST v1.8 [29, 30] with BEAGLE [31] to improve run-time. The Bayesian time-tree was reconstructed using the GTR+I+Г4 nucleotide substitution model [32], a relaxed uncorrelated lognormal molecular clock model [33], and a Bayesian Skyline coalescent tree prior [34] with non-informative default priors. MCMC chains were run for 100 × 10⁶ generations, and convergence and uncertainty of parameter estimates were assessed by calculating the Effective Sample Size (ESS) and 95% Highest Probability Density (HPD) values, respectively, after excluding the initial 10% of each run with Tracer v1.7 [35]. The convergence of parameters was considered when ESS ≥ 200. After the exclusion of sequences corresponding to the burn-in phase, the remaining ones were utilized to generate an FL consensus sequence for each MRCA using the Seaview version 4 program [36].

Amino acid signature (AAs) analyses

Nucleotide sequences were translated, and the software VESPA (Viral signature pattern analysis) [37] was used to identify positions in which the most common amino acid differs between B_CAR and B_PANDEMIC datasets as well as between subtype B and subtype D datasets. These positions were then selected, and for each a Chi-square test, as implement in R version 3.5.0 [38], was used to evaluate the statistical significance of their different amino acidic compositions. AAs were defined by positions in which both the most common amino acid was different, and the overall amino acid composition was significantly different between viral clades. For specific genomic regions corresponding to the structural gag, pol, and env genes, the number of B_CAR sequences was expanded, and the process to identify AAs between B_CAR and B_PANDEMIC datasets previously detailed was applied. The false discovery rate (FDR) method was used to correct for multiple hypothesis testing and to reduce false positives. Statistical significance was defined as FDR < 0.05.

Phenotypic prediction

We determine the frequency of genetic polymorphisms in accessory (Vif, Vpr, Nef) and regulatory (Rev) HIV-1 proteins of B_PANDEMIC and FL/expanded B_CAR datasets that were previously associated with slow HIV-1 disease progression and differential function [39–50]. The Geno2Pheno algorithm was used to predict the chemokine receptor tropism of the B_PANDEMIC and expanded B_CAR env dataset sequences based on their V3 region [51]. V3 was studied through the 11/25 rule (R or K at position 11 and/or K at position 25) [52–54], and the combination of positively charged amino acids at position 25 and an increase in total net charge [55]. The frequency of surveillance drug-resistance mutations (SDRMs) was also explored in B_CAR and B_PANDEMIC pol sequences retrieved from drug-naïve individuals of Caribbean origin using the Calibrated Population Resistance (CPR) tool (http://cpr.stanford.edu/cpr.cgi) [56]. A Chi-square test, as implement in R version 3.5.0 [38] was used to evaluate the significance of the results in both cases. Statistical significance was defined as p-values < 0.05.

Results

Classification of HIV-1 B_CAR and B_PANDEMIC FL sequences

From the 506 HIV-1 subtype B FL genome sequences of American origin initially selected, 28 sequences (6%) were identified as putative intra-subtype B recombinants and removed from the final dataset (S1 Table). The ML phylogenetic analysis of the remaining 478 subtype B FL genome American sequences revealed that most Caribbean sequences (82%) branched as basal strains and were classified as B_CAR strains, while most sequences from North (97%) and South (99%) America branched in a well-supported (SH-aLRT = 0.98) monophyletic sub-clade and were thus classified as B_PANDEMIC strains (Fig 1 and S1 Table). Despite the low number of subtype B FL genome sequences available from the Americas, the estimated relative prevalence of the B_CAR lineages in different countries was entirely consistent with previous estimates [6, 8, 9] based on much larger pol sequence datasets. Similarly, classification of additional subtype B Caribbean, covering specific regions of gag (n = 495), pol (n = 775), and env (n = 529) genes produced country ratios of B_CAR/B_PANDEMIC sequences akin to their counterparts based on the FL genome (S1 Fig and S1 Table).

Fig 1. ML phylogenetic tree of 478 HIV-1 subtype B FL genome American sequences.

Branches are colored according to their classification in pandemic (B_PANDEMIC, n = 450) and non-pandemic (B_CAR; n = 28) lineages as indicated in the legend at the bottom right. Node support (SH-aLRT) for subtype B and B_PANDEMIC monophyletic groups are indicated. The tree was rooted using HIV-1 subtype D sequences. Branch lengths are drawn to scale with the bar at the bottom indicating nucleotide substitutions per site.

AAs in B_CAR and B_PANDEMIC modern sequences

In order to identify AAs of different subtype B clades, we compared the FL genome B_PANDEMIC sequences (n = 450, sampled between 1978 and 2015) of American origin with FL genome (n = 18, sampled between 1983 and 2011), Gag (n = 28), Pol (n = 197), Env (n = 59) and Rev (n = 59, given the superposition of its CDS with Env) B_CAR sequences of Caribbean origin (Table 1). Twenty-eight sequences were originally classified as B_CAR, but 10 were removed for subsequent analyses because were sampled outside the Caribbean region. The same reasoning was used in the expanded dataset, where only B_CAR sequences of Caribbean origin were considered. The analysis of translated FL genome sequences identified nine positions that displayed compositions significantly distinct between B_CAR and B_PANDEMIC datasets, covering structural (one in P6, one in PR and one in RT), regulatory (one in Tat and three in Rev) and accessory (one in Vif and one in Vpu) proteins (Table 2). Expanded B_CAR datasets comprising partial regions of Gag, Pol, Env, and Rev encompass four out of the nine AAs previously identified. Three positions (one in PR, one in RT, and one in Rev) had their results endorsed by the additional sequences, while statistical significance in the P6 position was lost. Analyses of the expanded B_CAR datasets also identified additional AAs not detected in the FL genome dataset (Table 2), four in Gag (three in P24, positions 27, 120 and 148; and one in P7, position 12) and another in the RT (position 211).

Table 1. Sequences used for identification of AAs signatures between B_CAR and B_PANDEMIC sequences.

Country*	BPANDEMIC	BCAR
	FL	Gag		Pol		Env/ Rev		Others
		FL	Expanded	FL	Expanded	FL	Expanded	FL
DO	-	4	-	4	101	4	-	4
HT	-	5	4	5	-	5	6	5
JM	4	4	6	4	69	4	-	4
TT	-	5	-	5	-	5	35	5
BR	106	-	-	-	-	-	-	-
US	306	-	-	-	-	-	-	-
Others	34	-	-	-	9	-	-	-
Total	450	28		197		59		18

*Country codes are in accordance with ISO 3166–1. FL: Full length.

Table 2. Amino acid signatures of B_CAR and B_PANDEMIC datasets.

Location	Gene	gag				pol			vif	vpu	tat	rev
	Protein	P24	P24	P24	P7	PR	RT	RT	Vif	Vpu	Tat	Rev
	Position	27	120	148	12	41	207	211	50	24	23	57	67	102
Ancestors	BCAR	V	S	V	I	K	E	R	R	T	T	A	S	V
	BPANDEMIC	V	S	V	T	K	Q	R	R	T	T	A	S	V
Datasets	D n = 18	I89 V11	S78 N17	V100	I56 T22	K100	E88 K6	K61 R39	K89 R11	T94 S6	N94 T6	E88 A6	S67 P33	I78 V16
	BCAR n = 18	I61 V39	S50 N39	V61 T22	I39 T33	K89 R11	E50 A11	K50 R50	K56 R44	T83 S6	N44 T44	A55 E39	S83 P17	V100
	BCAR expanded	I71 V29	S50 N25	V68 T18	I54 T25	K74 R23	E48 A25	K54 R25	-	-	-	A53 E32	-	-
		n = 28				n = 197						n = 59
	BPANDEMIC n = 450	V69 I31	N55 S26	T66 V27	T40 I28	R75 K24	Q82 E9	R46 K44	R76 K24	S51 T45	T76 N22	G40 E27	P59 S39	I69 V28
p-values adj.	D/BCAR	-	-	-	-	-	-	0.7563	-	-	0.00391	0.0039	-	0.0001
	D/BPANDEMIC	0.0172	0.0461	0.0009	0.3137	0.0073	0.0002	0.7962	0.0004	0.0311	0.00229	0.0508	0.1355	-
	BCAR/BPANDEMIC	0.1084	0.3094	0.1084	0.0502	0.0200	0.0008	0.9670	0.0255	0.0015	0.0465	0.0300	0.0423	0.0300
	BCAR exp/BPANDEMIC	0.0091	0.0091	0.0376	0.0376	0.0001	0.0001	0.0001	-	-	-	0.0001	-	-

The table details positions identified as bearing amino acidic compositions significantly divergent between B_CAR and B_PANDEMIC strains. For each position are supplied (1) the residues present in the reconstructed B_CAR and B_PANDEMIC ancestors; (2) the two most common amino acid observed in that position in Subtype D, B_CAR, B_{CAR Expanded}, and B_PANDEMIC datasets accompanied by a number representing its frequency in each group; (3) the adjusted p-values. The sampling range of each sub dataset is indicated in the table. Ancestor sequences were reconstructed employing FL genomes of B_CAR (n = 18) and B_PANDEMIC (n = 69) lineages. G (Glycine), P (Proline), A (Alanine), V (Valine), L (Leucine), I (Isoleucine), M (Methionine), C (Cysteine), F (Phenylalanine), Y (Tyrosine), W (Tryptophan), H (Histidine), K (Lysine), R (Arginine), Q (Glutamine), N (Asparagine), E (Glutamic Acid), D (Aspartic Acid), S (Serine), T (Threonine). p-values < 0.05 are in bold.

By combining FL and partial genome B_CAR datasets, 13 positions were considered as signature positions differentiating B_CAR e B_PANDEMIC clades: four located on Gag (three in P24, and one in P7); three on Pol (one in the PR and two in the RT); three in Rev; while Vif, Vpu, and Tat each contributed with one position (Table 2). No signature position was identified in Vpr, Env, or Nef. None of the 13 AAs that distinguished contemporary B_CAR and B_PANDEMIC sequences were found to be invariant sites, with the exception of position Rev 102 in the B_CAR lineages. Furthermore, we also observed that for most AA positions, the most common amino acid found in a given subtype B clade corresponds to the second most frequent amino acid in the other clade (Table 2). The exceptions were position 207 in RT that displayed E₄₈/A₂₅ as the most frequent amino acids in B_CAR strains and Q₈₂/E₉ in B_PANDEMIC ones, and position 57 in Rev that displayed A₅₃/E₃₂ as the most frequent amino acids in B_CAR strains and G₄₀/E₂₇ in B_PANDEMIC ones.

AAs in B_CAR and B_PANDEMIC MRCA sequences

When the reconstructed MRCA sequences of B_CAR and B_PANDEMIC clades were compared, 21 amino acidic positions differentiated both ancestors (Table 3). Eight of them were located in Gag (one in P17, three in P24, two in P7, and two in P6); four in Pol (all of them in the RT); two in Vif, and four in Env (one in GP120, and three in GP41); while Tat, Rev, and Nef had one position each. Vpr and Vpu presented no difference in the comparisons. Of the 21 amino acidic positions that differentiated both ancestors, only four positions (three in Gag and one in Pol) displayed distinct majoritarian amino acids in the contemporary B_CAR and B_PANDEMIC datasets and only two of them (positions 12 in P7 and 207 in the RT) correspond to AAs between contemporary B_CAR and B_PANDEMIC sequences (Table 2). Thus, of the 21 amino acidic positions that differentiated the B_CAR and B_PANDEMIC ancestors only two continue to distinguish the contemporary descendant sequences. Furthermore, this analysis suggests that most (11/13) AAs that distinguished contemporary B_CAR and B_PANDEMIC sequences were probably not inherited from their ancestors. It is interesting to note, however, that the most common amino acid in 11 (including the two inherited from ancestors) out of 13 positions associated with the AAs was the same in B_CAR and subtype D sequences; while subtype D and B_PANDEMIC sequences coincide in only one position.

Table 3. Amino acid signatures of HIV-1 B_CAR and 210 B_PANDEMIC ancestors.

Location	Gene	gag								Pol				vif		tat	rev	env				nef
	Protein	P17	P24	P24	P24	P7	P7	P6	P6	RT	RT	RT	RT	Vif	Vif	Tat	Rev	GP120	GP41	GP41	GP41	Nef
	Position	84	15	91	180	12	26	30	42	35	49	122	207	47	66	74	56	363	133	182	209	10
Ancestors	BCAR	T	L	I	D	I	K	Q	R	A	R	K	E	P	V	S	G	P	N	V	L	M
	BPANDEMIC	V	I	V	E	T	R	P	K	V	K	E	Q	T	I	A	S	Q	T	I	R	V
Datasets	D	T83* V17	L50 I50	V50 I28	D61 E39	I56 T22	K72 R28	Q100	K100	T100	R72 K28	E61 K39	E89 K6	P66 H28	V89 I11	S100	G89 A11	P83 S11	N44 S44	I100	L100	I89 L11
	BCAR	T61 V39	L50 I50	I72 V22	D56 E44	I39 T33	K61 R39	P50 Q28	K72 R28	V83 T11	K89 R6	K67 P11	E50 A11	T44 P22	I67 V33	T44 A39	S78 G17	Q56 P17	T28 N22	I56 V44	H44 R39	L22 V17
	BPANDEMIC	T61 V39	I71 L28	I57 V37	E68 D32	T40 I28	R55 K45	P61 Q12	R53 K47	V79 I12	K93 R7	K62 E35	Q82 E9	T58 P15	I64 V35	T45 A41	S96 G4	Q49 H16	T45 N26	I58 V41	H45 R42	V18 L14

The table summarizes the most common amino acid for positions in which the reconstructed ancestors of Subtype B and B_PANDEMIC diverged, accompanied by a number representing its frequency in each group. G (Glycine), P (Proline), A (Alanine), V (Valine), L (Leucine), I (Isoleucine), M (Methionine), C (Cysteine), F (Phenylalanine), Y (Tyrosine), W (Tryptophan), H (Histidine), K (Lysine), R (Arginine), Q (Glutamine), N (Asparagine), E (Glutamic Acid), D (Aspartic Acid), S (Serine), T (Threonine).

Predicted phenotypic characteristics of B_CAR and B_PANDEMIC strains

Most AAs that distinguish the B_CAR and B_PANDEMIC strains were located outside domains or sites previously reported to be essential for protein function [40, 57–66] (Fig 2). The sole exception was position 23, located in Tat cysteine-rich domain (22–37), and reported as one of the three major sites of p53-derived restriction of Tat mediated by PKR phosphorylation [67]. We also evaluate the frequency of several polymorphisms in Vif, Vpr, Nef and Rev previously associated with long-term non-progressors (LTNPs) HIV-1 infected subjects and differential protein function in vitro and ex vivo [39–50] (Table 4). Analysis of amino acid composition at those positions failed to detect significant differences between B_CAR and B_PANDEMIC datasets, except for position 77 in Vpr that showed a significantly higher prevalence of the R77Q mutation in B_CAR (83%) in comparison with B_PANDEMIC (48%) sequences [45]. None of the methods here employed (Geno2Pheno algorithm, the 11/25 rule, or the combination of R at position 25 of V3 and a net charge of ≥ 5) pointed out to significant difference in the frequency of CXCR4 tropic viral variants, typically associated with a more rapid disease progression [68, 69], between the B_PANDEMIC and B_CAR env datasets (S3 Table). Finally, our analysis of HIV-1 pol sequences from drug naïve subjects from the Caribbean also failed to detect significant differences in the prevalence of SDRMs between B_CAR and B_PANDEMIC datasets (S4 Table).

Fig 2. Mapping of the identified AAs between B_CAR and B_PANDEMIC on the accessory, regulatory and structural HIV-1 proteins.

For all proteins, their functional domains are represented; (A) Vif: regions responsible for the binding to APOBEC3G (A3G), APOBEC3F (A3F), Cullin 5 (Cul5), Elongin B (EloB), and Elongin C (EloC); (B) Vpu: the transmembrane domain, the cytoplasmic domain, and the linker region between them; (C) Tat: the N‐terminal acidic domain, the cysteine-rich domain, the hydrophobic core domain, the TAR binding domain, the glutamine-rich domain, and the RGD motif; (D) Rev: the RNA binding domain (RBD), that also functions as a nuclear localization signal (NLS), the nuclear exporting signal (NES), and sequences responsible for its multimerization (Mult.); (E) P24: the N-terminal domain and the C-terminal domain, connected by the inter-domain linker, the region mainly responsible for the interaction with cyclophilin A (CypA), and the major homology region (MHR); (F) Protease (PR): its active site.

Table 4. Prevalence of polymorphisms in B_CAR and B_PANDEMIC Vif, Vpr, Nef and Rev sequences associated with slow HIV-1 disease progression and differential function.

Polymorphism	BCAR		BPANDEMIC		p-value
	# seq. Analyzed	# seq. with polymorphism (%)	# seq. analyzed	# seq. with polymorphism (%)
RevE74P	18	0	450	36 (8%)	-
RevL78I	18	0	450	43 (10%)	-
RevV104L	18	0	450	31 (7%)	-
RevS106P	18	1 (6%)	450	28 (6%)	0.908
VifI107T	18	0	450	0	-
VifR132S	18	0	450	135 (30%)	-
VprQ3R	18	3 (17%)	450	42 (9%)	0.300
VprQ44R	18	0	450	0	-
VprF72L	18	0	450	0	-
VprR77Q	18	15 (83%)	450	218 (48%)	0.003
VprI83V	18	0	450	8 (2%)	-
NefR22Q	18	3	450	49 (11%)
NefL58V	18	3	450	45 (10%)
NefK94E	18	0	450	11 (2%)	-
NefH116N	18	8 (44%)	450	123 (27%)	0.112

p-values < 0.05 are in bold.

Discussion

The current work suggests that the hypothesis that viral genetic determinants shaped the remarkable differences in the geographic dissemination pattern of the HIV-1 B_PANDEMIC and B_CAR strains is highly unlikely. Among over 3,000 positions analyzed across nine genes coded by the HIV-1 genome, we detected only 13 AAs distinguishing the B_CAR and B_PANDEMIC clades. All AAs that did differentiate the B_CAR and B_PANDEMIC clades correspond to sites with distinct degrees of polymorphism and not to invariant (or highly conserved) sites. Furthermore, for 11 out of 13 AAs positions, the most common amino acid found in a given subtype B variant corresponds to the second most frequent amino acid in the other variant. Our study also suggests that 11 out of 13 AAs that distinguished contemporary B_CAR and B_PANDEMIC sequences were probably not inherited from their ancestors and that most (19/21) amino acid differences inferred between the B_CAR and B_PANDEMIC ancestors evolved into polymorphic sites with quite comparable compositions in modern descendant sequences. Finally, nearly all AAs identified were located outside functionally relevant protein domains.

Our data supports that B_CAR and B_PANDEMIC strains are probably not distinguished by functionally relevant AAs in structural genes. Analyzes of both FL and expanded partial env sequences failed to detect AAs in this variable genomic region. Furthermore, no significant differences in the frequency of CCR5/CXCR4 tropic variants were detected between B_CAR and B_PANDEMIC sequences, supporting not great variation in the chemokine receptor usage between pandemic and non-pandemic subtype B strains. These results are fully consistent with a previous study that demonstrate that the env V3 consensus sequence of B_CAR strains from Trinidad and Tobago differs by few amino acids from the B_PANDEMIC V3 consensus and that no phenotypic features, including syncytium induction, neutralization profiles, and chemokine receptor usage, distinguish both subtype B lineages [19]. Furthermore, all AAs in Gag and Pol that distinguishing B_CAR and B_PANDEMIC sequences were located outside known conserved protein functional domains.

By contrasting, a few interesting differences between B_CAR and B_PANDEMIC strains were observed in non-structural genes. The single AAs in position 23 of Tat is located in a cysteine-rich domain and has been reported as one of the three major sites of p53-derived restriction of Tat mediated by PKR phosphorylation [67]. The presence of an N residue in that position, that is the most prevalent amino acid in subtypes A, C, D and B_CAR (44%) strains, but not in B_PANDEMIC ones (22%), have been associated with increased Tat transactivation, probably through enhanced P-TEFb binding [67, 70]. We also detect much higher frequency of the naturally occurring variation Vpr R77Q in B_CAR (83%) respect to B_PANDEMIC (48%) sequences. That mutation, that also predominates in subtypes A, C, D, G, and H, reduces apoptosis and CD4 T-cell depletion in ex vivo-infected cells and is much more prevalent in subtype B-infected LTNPs individuals (75–90%) than in subjects with progressive HIV disease (33–42%) [45]. The similar genetic composition of B_CAR and several pandemic HIV-1 clades (subtypes A, C and D) at positions Tat23 and Vpr77 argued against the hypothesis that differences at such positions resulted in a more restricted dispersion of B_CAR compared with the B_PANDEMIC strains.

Despite the very small size (n = 18) of the B_CAR FL genome dataset here used, some evidences support that the B_CAR genetic variability was not severely underestimated in this dataset and that the B_CAR consensus sequence obtained was probably not biased by the low number of FL genomic sequences available. First, the most common amino acid recovered in most positions from extended datasets was fully coincident with the one detected in the FL dataset. Expanded and FL datasets converged in 99.7% (297/298) of Gag amino acid positions, 99.1% (336/339) of Pol positions, and 97.8% (683/698) of Env positions analyzed. Second, the degree of polymorphism of the 13 AAs positions that distinguished B_CAR and B_PANDEMIC sequences was roughly comparable in FL and expanded B_CAR datasets. The paucity of FL B_CAR strains, however, might have restricted our ability to detect some additional AAs between B_CAR and B_PANDEMIC sequences. By increasing the number of B_CAR sequences we failed to recover new AAs between B_CAR and B_PANDEMIC env sequences, but we detected four additional AAs in Gag and one in Pol, increasing the overall number of AAs from three to eight in those genomic regions.

The low number of FL B_CAR sequences used might have also introduced some bias on the reconstruction of the MRCA sequences. According to our analysis, of the 13 AAs detected in modern B_CAR and B_PANDEMIC sequences, only two matched with divergent sites between the B_CAR and B_PANDEMIC ancestors. This observation support that most genetic differences between the B_CAR and B_PANDEMIC ancestors evolved toward positions with similar amino acid composition in modern subtype B sequences and that most AAs in modern B_CAR and B_PANDEMIC sequences arose during subsequent evolution and diversification of subtype B lineages. An inspection of the amino acid composition at those 13 positions in the related subtype D clade, however, supports a different scenario. We observed that B_CAR and subtype D sequences displayed the same prevalent amino acid in most (11/13) AAs positions, consistent with genetic identity inherited from the common B/D ancestor. In sharp contrast, the B_PANDEMIC and subtype D sequences displayed the same prevalent amino acid in only one AAs position. Therefore, our reconstruction of the MRCA sequences may have underestimated the number of B_CAR/B_PANDEMIC AAs inherited from the ancestors.

In summary, albeit some mutations fixed in the HIV-1 B_PANDEMIC ancestral strain could potentially have some phenotypic impact on viral transmissibility, the absence of stringent AAs distinguishing modern B_CAR and B_PANDEMIC variants and the similar amino acid composition between B_CAR and other group M subtype pandemic variants at key sites indicates that viral genetic determinants were probably not the main factor shaping the divergent pattern of geographic spread of B_CAR and B_PANDEMIC variants. The successful dissemination of B_CAR strains in some Caribbean countries that exhibit the highest HIV-1 prevalence rates outside of Africa [6, 11] also argues against the hypothesis of a reduced B_CAR viral fitness. These results support that stochastic events leading to the introduction of B_PANDEMIC ancestor into globally connected populations were the most probable driving force behind its pandemic dissemination and substantiate the crucial need for continued molecular surveillance of HIV-1 transmission on key populations worldwide.

Supporting information

S1 Fig. ML phylogenetic trees of HIV-1 subtype B American sequences on specific regions of gag, pol, and env.

Partial HIV-1 sequences covering (A) gag (1,264 to 2,148), (B) env (6,450 to 8,480), (C) pol (2,253 to 3,272), and (D) pol from drug naïve individuals (2,253 to 3,272) were classified into B_PANDEMIC (red branches) and B_CAR (blue branches) lineages according to the topology obtained in each tree. Node support (SH-aLRT) for subtype B and B_PANDEMIC monophyletic groups are indicated. The trees were rooted using HIV-1 subtype D sequences. Branch lengths are drawn to scale with the bar at the bottom indicating nucleotide substitutions per site.

(PDF)

S1 Table. Classification of the HIV-1 subtype B sequences in the B_PANDEMIC or B_CAR clades.

The table summarizes the results of the classification of the full-length (FL) and partial HIV-1 subtype B sequences in the B_PANDEMIC or B_CAR based on their placement on ML phylogenetic trees displayed in Fig 1 and S1 Fig. All sub-datasets are accompanied by their sampling range. *Country codes are in accordance with ISO 3166–1.

(PDF)

S2 Table. HIV-1 subtype D full-length genome sequences.

The table summarize the full-length (FL) HIV-1 Subtype D sequences used in our study and their sampling range. * Country codes are in accordance with ISO 3166–1.

(PDF)

S3 Table. Predicted co-receptor usage by B_CAR and B_PANDEMIC env sequences.

The table summarizes the predicted usage of chemokines receptors CCR5 and CXCR4 based on different criteria: 1) the Geno2Pheno algorithm, which classifies the sequences between R5 variants or X4 and R5X4 dual-tropic variants; 2) the 11/25 Rule, which asses the presence of arginine (R) or lysine (K) at position 11 of env V3 sequences and/or K at position 25; 3) the combination of R at position 25 of V3 and a net charge of ≥ 5.

(PDF)

S4 Table. Prevalence of transmitted drug-resistance mutations in B_CAR and B_PANDEMIC PR/RT sequences.

The table summarizes the surveillance drug-resistance mutations (SDRM) identified in PR/RT B_CAR and B_PANDEMIC sequences retrieved from drug naïve subjects. PI (protease inhibitor), NRTI (nucleoside analog reverse-transcriptase inhibitors), NNRTI (non-nucleoside analog reverse-transcriptase inhibitor). The p-values obtained in chi-squared tests are listed in the last column.

(PDF)

S1 File

(PDF)

Data Availability

The data underlying the results presented in the study are available from Los Alamos HIV Database (http://www.hiv.lanl.gov). The authors confirm they had no special access privileges to the data. A supplementary file with the accession codes of all sequences used in this study has been provided.

Funding Statement

I.A. is funded by a fellowship from Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES). G.B. is funded by fellowships from the Conselho Nacional de Desenvolvimento Científico e Tecnológico - CNPq (Grant number 302317/2017-1) and the Fundação Carlos Chagas Filho de Amparo à Pesquisa do Estado do Rio de Janeiro – FAPERJ (Grant number E-26/202.896/2018). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.