Background
HIV-1 is highly genetically variable and on this basis, falls into four distinct groups: M, O, N and P [
1,
2]. Group M viruses account for the majority of global HIV-1 infections and displays a tremendous amount of genetic variability as well, with nine divergent (pure) subtypes (A, B, C, D, F, G, H, J, and K), over 55 circulating recombinant forms (CRFs) and several unique recombinant forms (URFs) [
3,
4]. These viruses also differ phenotypically in their co-receptor use preferences and entry to cells.
HIV-1 entry into host cells requires cooperate engagements of both the viral envelope and host cell surface (CD4) receptor, in a process requiring the engagement of one or more of a group of seven-transmembrane chemokine receptors (co-receptors), the CCR5 and CXCR4 being the most common [
5,
6]. Viral tropism (co-receptor usage) has important consequences and relationships with infection and disease outcomes. Thus, viruses have been characterized phenotypically on the basis of tropism into syncytium inducing (SI) and non syncytium inducing (NSI) [
7]. CXCR4 (X4) tropic viruses are largely of SI phenotype. These T-Cell tropic viruses arise during late stages of disease progression or infection [
8]. CCR5 (R5) tropic viruses are of NSI phenotype of macrophage lineage and dominate the early stages of infection. HIV has demonstrated ability to switch tropism during the course of disease, a process that the virus uses adaptively to propagate in the presence of antiviral immune or therapeutic pressure [
8]. Viral tropism can be determined either by using the more rigorous but expensive cell based phenotypic test or assigned on the basis of the relatively inexpensive genotyping sequence analysis that however, suffer from reduced sensitivity [
9]. To improve robustness of genotypic tropism assignments, a number of tools have been developed including T-Cup 2.0, WebPSSM
x4r5, WebPSSM
sinsi and Geno2Pheno, as well as rule-based methods like Esbjörnsson rule and Raymond’s rule with the objective of expanding the scope of sensitivity to include more non-B subtypes [
9].
HIV is further adapted to its environment on the basis of envelope diversity, which exists in the form of variable glycosylation patterns and densities, as well as genetic variability affecting both immune and therapeutic functions [
10]. The envelope consists of alternating constant regions (C1 - C5) and variable regions (V1 – V5) that provide critical structural and functional integrity for the virus, including co-receptor binding [
11]. The overall amino acid charge of the HIV-1 V3 influences viral phenotype selection: a higher positive charge favors the SI phenotype and CXCR4 (X4) utilization while the loss of an N-linked glycosylation event in the V3 region together with a higher positive charge is associated with the virus switching from R5 to the X4 phenotype [
6]. Overall, HIV-1 gp120 is heavily glycosylated by the infected host, with the glycans accounting for up to 50% of its total mass [
12]. The position and number of N-Linked oligosaccharides attached to a protein have a profound effect on viral structure, protein expression and function [
13], and specifically, tropism attributable to variations in the V3 loop [
14]. These patterns also influence receptor binding and the phenotypic properties of the virus [
6]. A change of sequons in the HIV envelope protein gp41, for example, can induce a conformational change in the associated gp120 that will dramatically diminish the binding of gp120-specific antibodies [
15], and potentiating T cell immune escape [
16,
17].
Studies have shown that genetic differences at HIV subtype level reflect in co-receptor usage and in the N-linked glycosylation pattern. Subtype C preferentially uses CCR5 and rarely induces syncytia [
18]. Moreover, there is an apparent selection for subtype A and C variants that are less glycosylated and with shorter V1-V2 loop sequences [
19], potentially enhancing the transmission of these strains. The faster rate of disease progression often accompanying HIV-1 subtype D can be linked to a higher frequency of syncytium formation and X4 usage [
20].
Kenya’s HIV landscape features multiple genetic variants with at least 9 subtypes and recombinants [
21‐
26], but only a few of local studies also looked at co-receptor and glycosylation characteristics across subtypes [
27,
28]. In this paper, we sought to assess both phenotypic and genotypic differences in the context of expanded HAART so as to rapidly and readily generate comparative association matrix of potential N-linked glycosylation (PNG) pattern, co-receptor usage and viral subtype diversity in relation to clusters of virologic treatment outcome. Partial
C2V3 envelope sequences generated previously from our published work was used for these analyses [
29].
Discussion
The HIV virus establishes lifelong infection despite the use of HAART, and in many instances, ‘escape’ therapeutic intervention because of continuous viral evolution and formation of quasi species with distinct genetic and phenotypic variability. In these patients on extensive HAART in a context of highly dynamic and heterogeneous virus populations, different viral subtypes and recombinants exhibit diversity in coreceptor tropism as well as glycosylation patterns. We discuss these results that have important implication to HIV pathogenesis and therapeutic targeting [
38‐
41].
Viral tropism and co-receptor usage
HIV-1 tropism is critical to host-cell interactions and is implicated widely in disease or infection processes. Our data revealed a significant association between viral subtype and coreceptor tropism using G2P platform (all FPR algorithms), Raymond’s rule and WEBPSMM. Between 68% and 81% of all the isolates were R5-tropic using G2P’s three FPR algorithms. None of the isolates were determined to be dual tropic across all criteria used. FPR10 is routinely applied as the standard algorithm for HIV-1 tropism determination by G2P, but is biased in favor of subtype B and likely to over-represent R5-tropism in settings where HIV is highly genetically variable like Kenya. Therefore, we applied different phenotyping approaches to assess tropism of the different isolates. There was still however, substantial discordance among the alternative phenotyping algorithms in assigning tropisms to the different subtypes. Specifically, output from Raymond’s and Esbjörnsson’s platforms tended to have higher proportions of X4 variants, while output from WebPSSM had lower proportions of X4 variants relative to other algorithms used. Actual phenotyping tests were not done to validate the accuracy of these phenotyping platforms, thus limiting the breadth of interpreting the efficiencies of the different platforms for correct phenotype assignment.
Evidence suggests that different HIV subtypes may have specific preferences for coreceptor usage [
42]. Subtype C for example has been shown to preferentially use CCR5 and to rarely induce syncytia [
43,
44]. A few studies involving Kenyan subjects have shown majority of circulating HIV to be R5 tropic, and mostly due to the predominance of R5-tropic HIV subtype A [
27,
28,
45]. Significant variations became apparent when tropism is disaggregated by viral subtype, with majority of subtype A1, C and A1 recombinants being R5 tropic. The converse was true for subtype D viruses, nearly all of which were X4-tropic. CXCR4 usage is preferred by subtype D viruses and is generally associated with a more rapid decrease of CD4 counts and rapid disease progression [
20,
27]. Subtype D, C and recombinant HIV strains have shown specific dynamism in Kenya over the past decade, partly due to population and demographic cross-border patterns, with the ultimate effect on a shrinking predominance of subtype A viruses [
21,
29,
46,
47]. In this setting of highly variable viral genetic strains and where adherence and virologic treatment failure is prevalent [
48], the outcomes of long-term treatment is likely to be additionally influenced by the changing phenotypic property that promotes rapid disease processes. In particular, a rise in X4-tropic subtype D would imply increased risk of rapid disease and potential dampening of antiretroviral treatment (ART) response. Thus, treatment plans will need to be structured to account for both viral genetic diversity and the associated phenotypic characteristics.
Apart from subtype C, there has not been much success in coming up with highly specific and sensitive phenotyping or genotyping tools for the other non-B subtypes. For subtype A for example, one study showed that for the multiple tools used, specificity was high but sensitivity was very low [
9]. The low sensitivity was likely attributable to other regions outside the V3 (i.e. V1, V2 and V4) which also impact tropism [
9,
49,
50]. Thus accurate assignment of tropism would require an expanded analysis of the envelope, beyond the V3 as considered in this paper. Such accuracy becomes significant in the design and management of therapy, given certain CCR5 blockers have failed in patients co-infected with X4-tropic viruses [
51].
Potential N-Linked Glycosylation Sites (PNGs)
The HIV-1 is heavily coated with glycans that make the virus ‘invisible’ to the host’s immune system [
52]. An increase in the length and number of PNGs in the V1V2 region for example has been shown to play a role in HIV-1 resistance to neutralizing antibodies [
53]. On average, we found that, the NXT PNG pattern was significantly compartmentalized in blood, with heavy NXT glycosylation resident in the cellular compartment of patients that responded effectively to HAART. This preferential glycosylation for the NXT relative to NXS pattern has been documented by others as well [
54]. Contrary to our observations, independent data from the analysis of C2-V5 gp120 envelope has shown more glycosylation of RNA isolates from plasma than of proviral DNA from cellular compartment in Australian patients on HAART [
55]. Our data was derived from a much shorter V3 region and in addition, varies from the Australian finding in that all our isolates from cellular blood were obtained from patients who had suppressed virus to undetectable levels hence experiencing long-term virologic treatment success (VS), and all the cell-free isolates were from patients with virologic treatment failure (VF). We hypothesized that long-term exposure to HAART that results in VS would dim the need for robust humoral response against the virus that is already under intense antiretroviral pressure. The result is a reduced selection pressure for antibody escape, and the propagation of viruses that are more likely to lose their PNGs than those from VF patients where robust antibody response is needed to raise alternative antiviral defense. This line of argument is supported by earlier observations that HIV-specific antibody responses are depressed in both chronically infected and in HAART responsive patients [
56]. Similarly but unlike the previous study, our population was infected with multiple virus subtypes with the main strain being HIV subtype A1, which is unlike HIV infection environment in most western countries where subtype B predominates with little to no genetic heterogeneity of virus populations [
57].
N-linked glycosylation constitute both structural and functional adaption of the virus that affect a range of host infection outcomes including transmissibility and immune evasion [
58,
59]. Another likely but more cautious argument is that accumulation of HIV isolates with fewer PNGs in cell free plasma could suggests a ‘molecular learning’ and protective process by which HIV-1 sheds off specific glycosites under therapeutic pressure. Ultimately, such an association of reduced PNG sites in plasma virus isolates retain credibility when considered in the context of accompanying antibody neutralization [
59]. When we further disaggregated isolates by viral subtype and by tropism, HIV-1D and X4-tropic isolates tended to have higher average PNGs per isolate than other subtypes or R5-tropic strains. Since both subtype D and X4 tropism are associated with adverse disease outcomes, increased glycosylation of these strains may be associated with specific protective amino acid sites that reinforce the mechanisms for HIV pathogenesis. In deed our data supports this argument, as there was absolute (100%) glycosylation of X4 tropic viruses at N262 relative to other strains. This position was recently reported to provide steric hindrance in antibody neutralization experiments that protected the virus from antibody targeting [
60]. Subtype D also had the highest number of PNG sites glycosylating 100% of the isolates compared to other subtypes. Some evidence from mainly subtype B and C viruses have suggested that R5-tropic strains tend to have more PNGs than their X4 counterparts [
61,
62]. There is certainly a need to interpret these data within the various specific contexts of HIV infection. In populations with highly heterogeneous genetic diversity of HIV, infections with multiple strains and variable treatment approaches and outcomes, it is not unlikely for the virus to acquire different adaptation strategies.
Glycosylation at position 301 (302) at the stem of the V3 loop is thought to play a role in stabilizing the V3 loop and in modulating co-receptor binding. Mutation at this point results in a drastic decrease in viral infectivity probably due to reduced co-receptor binding [
12]. In addition to positions N296 and N366, our analyses revealed significant differences by subtype at glycosylation position N302. At position 296, only 20% of the subtype C isolates were glycosylated compared to 64.7% for subtype A1, 80% for subtype D and 100% for subtype A2 or of recombinants of A1. Conversely, 100% of subtype C sequences were glycosylated at N366 compared to only 20% of subtype D, 57% of A1 recombinants, 75% of subtype A2 and 82% of subtype A1. At position N302, all the subtypes had 100% glycosylation except for subtype C that had 80% glycosylation. Hence, envelope molecular patterns that affect HIV glycosylation varies across viral genotypes, and may have association with disease or treatment outcome. Overall, our data should be interpreted cautiously within the context limitations of the study; the sample size was not large enough to warrant population-level inferences. Similarly due to limitation in available resources, we did not sequence DNA of patients experiencing virologic treatment failure (VL > 1000) nor generate RNA derived sequences of those with virologic treatment success (VL < =1000). Including these sequences in future comparisons may add important dimensions as regards viral diversity between proviral reservoirs and circulating clones. This study however, provides useful basis for deeper investigations of larger representative samples from long-term HAART patients.
Conclusions
Kenya and most countries where HIV treatment was for a long time limited by the availability of resources to fund basic antiretroviral medication, has in the last 10-12 years scaled up treatment without equivalent programs to monitor important outcomes and effects of the same. We and a few others have recently dedicated research to evaluate these effects, in respect of adherence, treatment outcomes and genetic diversity of the HIV in the presence of HAART [
29,
48,
63]. The present set of data demonstrates that in these patients who have received suppressive HAART over an extended period, viral tropism and glycosylation follows a trajectory of viral genetic diversity, compartmentalization and virologic response. Specifically, genetic variability at subtype level was significantly associated with virus preference for co-receptor usage and with envelope glycosylation pattern and density- subtype D being preferentially X4-tropic, extracellularly compartmented in blood and more heavily glycosylated than other subtypes. Furthermore, envelope glycosylation is significantly associated with virus tropism, compartmentalization in blood and virologic treatment response. As discussed elsewhere in this paper, these relationships are not just consequential to treatment options, but also to therapeutic and preventive vaccine designs.
Acknowledgements
We thank Grace A. Ochieng, Kevin Onyango, Paul Onyango, Nancy Lagat, JoycelineKinyua, and all field associates and peer councilors who facilitated patient recruitment and sample collection. Dr. Florence Oloo coordinated study activities. The data presented in this publ;ication builds on the work of the authors as published previously by Kitawi et al [
29].