Coevolution of HIV-1 and broadly neutralizing antibodies

Abstract

Purpose of review

Exploring the molecular details of the coevolution of HIV-1 Envelope with broadly neutralizing antibodies (bNAbs) in infected individuals over time provides insights for vaccine design. Since mid-2017, the number of individuals described in such publications has nearly tripled. New publications have extended such studies to new epitopes on Env and provided more detail on previously known sites.

Recent findings

Studies of two donors – one of them an infant, the other with three lineages targeting the same site – has deepened our understanding of V3-glycan-directed lineages. A V2-apex-directed lineage showed remarkable similarity to a lineage from a previously described donor, revealing general principles for this class of bNAbs. Understanding development of CD4 binding site antibodies has been enriched by the study of a VRC01-class lineage. Finally, the membrane-proximal external region is a new addition to the set of epitopes studied in this manner, with early development events explored in a study of three lineages from a single donor.

Summary

These studies provide templates for immunogen design to elicit bNAbs against a widened set of epitopes, generating new directions in the quest for an HIV vaccine.

INTRODUCTION

The last decade was marked by the groundbreaking discovery of HIV-1 broadly neutralizing antibodies (bNAbs) with exceptional potency and breadth, neutralizing most circulating HIV-1 strains [1]. Their effectiveness at preventing infection and reducing viral load in passive transfer animal models renewed hope that an HIV-1 vaccine might be achievable [2]. However, unusual features exhibited by many bNAbs including long heavy chain complementarity determining region 3s (CDRH3s), high levels of somatic hypermutation (SHM), insertion–deletion events (indels), and auto- or poly-reactivity, suggested long and complex maturation pathways not easily reproducible by vaccination. These observations, along with the failure of traditional vaccine strategies, led the field to consider vaccine regimens based on detailed studies of bNAb ontogeny during infection [3]. This approach relies on the identification and characterization of antibody precursors (unmutated common ancestors or UCAs) and key intermediates, as well as the relevant viral Envelope (Env) variants eliciting and driving the maturation of bNAb lineages. Enabled by technologies include rapid mAb isolation from single B-cells, next-generation sequencing (NGS) of the B-cell repertoire and viral Env populations, and functional and structural analyses of antibody/Env interactions, such studies reveal a roadmap for vaccine design.

Given the limited number of donors for whom samples exist allowing full Env/bNAb coevolution studies – the small fraction of individuals that develop bNAbs, the small number of longitudinal cohorts of untreated individuals, and the lack of new cohorts with the advent of early antiretroviral treatment – many studies lack one or more of these elements. Some use NGS samples from chronic-infection timepoints, or cloned antibodies only, to reconstruct precursors known as germline revertants [4-10]. Such constructs generally use mature CDRH3s and are therefore less relevant than true UCAs, yet some have provided important insights later validated by coevolution studies. This review focuses on studies of donors with longitudinal samples from the time of infection, in which both Envs and bNAb lineages were evaluated at a richly detailed molecular level.

INSIGHTS INTO SPECIFIC EPITOPES

The Env/bNAb coevolution studies reported to date and reviewed here describe the development of bNAb lineages targeting four epitope regions [11-14] (Fig. 1 and Table 1): the CD4-binding site (CD4bs) (CH103 and CH235 from donor CH505; PCIN63 from donor PC63), the V2-apex (CAP256-VRC26 from donor CAP256; PCT64 from donor PC64), the V3-glycan high-mannose patch (PCDN from donor PC76; DH270 from donor CH848; BF520.1 from infant donor BF520; PCIN39 from donor PC39), and the membrane-proximal external region (MPER) (VRC42, VRC43, and VRC46 from donor RV217-40512).

FIGURE 1.

Sites of vulnerability on the HIV-1 Env trimer. Model shows Env trimer as determined by cryo-electron microscopy and X-ray crystallography. Epitopes of known bNAbs are shown with their footprints color-coded. Viral membrane is shown in light gray. bNAbs from lineages described in the text are color-coded by epitope. N-terminal and C-terminal portions of MPER are distinguished by color. The CAP256-VRC26.09 bNAb (cyan) is shown for scale. bNAbs, broadly neutralizing antibodies; MPER, membrane-proximal external region. Source: Image adapted from Refs. [11-14].

Table 1.

Characteristics of broadly neutralizing antibody lineages that have been the subject of Env/antibody coevolution studies

Epitope	Characteristics of bNAbs	Donor	Sampling	Ab lineage	Breadth median IC50	VH-gene SHM (nt)	VL-gene	CDR3 length	indels	Poly- auto- reactivity	Lineage emergence	Breadth in lineage	Serum breadth	Viral subtype	Viral sequencing	UCA binds/ neutralizes T/F	Escape	References
CD4bs	Multiple classes:-VH-generestricted: conserved angle of approach * VRC01-class: VH1-2, 5aa CRDL3, flexible/short CDRL1 loop *VH1-46 restricted - CDRH3 binder: Various angles of approach, gene usage and structural characteristics Highly somatically mutated (mean 28.4%] Need to accomodate surrounding glycans (N276 in loopD, N462 in V5-loop]	CH505	4–160 wpi	CH103	55% 4.54 μg/ml	VH4-59 14%	VL3-1	H: 15 aa L: 10 aa	Yes Del CDRL1	++	14 wpi	144 wpi	84 wpi	C	Yes	Neutralization	LoopD V5 loop CD4bs loop	[24-25]
		CH505	4–323 wpi	CH235	90% 0.6 μg/ml	VH1-46 28%	VK3-15	H: 15 aa L: 8 aa	No	++	14 wpi	152 wpi	84 wpi	C	Yes	Weak binding	LoopD V5 loop CD4bs loop	[23,24]
		PC063	7–322 wpi	PCIN63	80% 0.24 μg/ml	VH1-2 15%	VK1-5	H: 15 aa L: 5 aa	No	Variable	168 wpi	289 wpi	266 wpi	C	Yes	No binding	LoopD V5 loop CD4bs loop	[26■■]
V2-glycan	Long (25–39aa) anionic CDHR3 Majority VH3 + DH3 genes Limited angles of approach Glycan binding (N160, N156) V2 C-strand K/R residues (166, 169, 171]	CAP256	24–206 wpi	VRC26-CAP256	63% 0.003 μg/ml	VH3-30 12%	VL1-51	H: 36 aa L: 13 aa	No	-	38 wpi	119 wpi	84 wpi	C	Yes	Neutralization	166 and 169	[11,28,30]
		PC064	14–200 wpi	PCT64	37% 0.42 μg/ml	VH3-15 13%	VK3-20	H: 25aa L: 8 aa	No	+/−	43 wpi	80 wpi	80 wpi	A	Yes	Binding to 293S-expressed Env	166 and 169	[34■■,35■■]
V3-glycan	Diverse gene families Frequent indels Diverse angles of approach N332-glycan + GDIR dependant V3-glycans promiscuity	PC076	23–192 wpi	PCDN	47% 0.53 μg/ml	VH4-34 16%	VK3-20	H: 22 aa L: 8aa	No	+/−	72 wpi	120 wpi	120 wpi	C	Yes	No binding	N328, N332	[41]
		PC039	15–265 wpi	PCN39	45% 0.03 μg/ml	VH4-34 16%	VK3-20	H: 22 aa L: 10aa	Yes Ins CDRH1	++	37 wpi	72 wpi	84 wpi	C	Yes	No binding	2 pathways–GDIR, 301, 332–V3 deletion	Landais, Poignard, Murrell, Wilson et al., unpublished
		BF520	14–26 wpi	BF520.1	58% 1.95 μg/ml	VH1-2 7%	VK3-15	H: 20 aa L: 7aa	No		26 wpi	62 wpi	62 wpi	A	Yes	No autologous Env binding but binds BG505-SOSIP		[39,43■■]
		CH848	11–246 wpi	DH270	55% 0.08 μg/ml	VH1-2 13%	VL2-23	H: 20aa L: 9 aa	No	+/−	186 wpi	205 wpi	n.a.	C	Yes	Weak binding to V3-peptide mimic	V1-loop GDIR N332 > N334	[42■■]
MPER	Auto/poly-reactivity Narrow angle of approach Bind viral membrane	RV217-40512	4–163 wpi	RV217-VRC42	96% 4.7 μg/ml	VH1-69 11%	VK3-20	H: 15 aa L: 9 aa	No	+	22 wpi		69 wpi	AE	Yes	Binding to scaffolded MPER peptide on 60-mer		[44■■]
		RV217-40512	4–163 wpi	RV217-VRC43	65% 1.7 μg/ml	VH4-4 11%	VL7-23	H: 19 aa L: 9 aa	No	+/−	22 wpi		69 wpi	AE	Yes			[44■■]
		RV217-40512	4–163 wpi	RV217-VRC46	30% 11 μg/ml	VH1-69 9%	VK1-39	H: 14 aa L: 9 aa	No	++	22 wpi		69 wpi	AE	Yes			[44■■]

CD4bS

BNAbs to the CD4bs typically develop after more than 5 years of infection and are typically highly somatically hypermutated – often 30% or more [15] – suggesting a long and complex maturation process. Two types of CD4bs bNAbs have been identified: CD4bs bNAbs classically using CDRH3 for epitope recognition; and CD4bs bNAbs with heavy chain variable gene (VH)-restricted recognition, including the VRC01 multidonor class, mostly relying on V-gene encoded structural features mimicking CD4 [15,16]. Immunogens eOD-GT8 and 426c, designed to bind VRC01-class germline revertants [5,6] have already succeeded in eliciting VRC01-like antibody responses in transgenic mouse models [17-21]. However, defining boosting immunogens capable of driving lineage maturation toward neutralization breadth remains a challenge. A major roadblock is the acquisition of mutations in the light-chain complementarity determining region 1 (CDRL1) loop to avoid structural clashes with Env V5-loops and Loop D glycans [22].

The detailed development of CD4bs bNAbs was first described in a series of articles demonstrating how two CD4bs bNAb lineages, CH103 (CDRH3 binder) and CH235 (VH1-46 gene restriction) cooperated in donor CH505 by restricting the possibilities of escape through distinct binding modes [23-25]. Each lineage induced an intense epitope diversification supporting the maturation of the other lineage. These groundbreaking studies were the first to trace Env/bNAb coevolution, and several features observed in the most recent studies were first noted for CH505.

The genetic and structural similarities shared by VRC01-class bNAbs, along with their breadth and potency, make them a particularly attractive target [15]. The first coevolution study of a VRC01-class bNAb, PCIN63 [26^■■], showed that unlike most VRC01-class bNAbs, the PCIN63 antibodies achieved equivalent neutralization with only 13% SHM, mostly focused on residues previously shown to be minimally required for VRC01 neutralization. Although the heavy chain UCA sequence was determined with near certainty, several unmutated light chain sequences could correspond to a putative UCA; only two of the reconstructed UCA candidates bound with low affinity to eOD-GT8. One Env variant lacking both the N276-glycan and V5-glycans isolated ~6 months before the first detection of the lineage appeared to be the eliciting variant. However, in contrast to most VRC01-class bNAbs, PCIN63 antibodies appear to be partially dependent on the presence of the N276-glycan. These data suggest a shorter pathway more compatible with vaccination.

V2-apex

The V2-apex bNAbs are characterized by long CDRH3s (24–39 amino acids) which are typically very rare in the human repertoire. Nevertheless, such bNAbs are detected in ~10% of broad neutralizers [27] and include some of the most potent bNAbs isolated to date [28,29]. These bNAbs target V2 loop-strand C peptide positively charged residues and typically require binding to the N160 and/or N156-glycans as well as accommodation to V1-loop glycans. Located at the trimer apex, their epitopes often span more than one protomer such that these antibodies prefer (PG9, VRC38, and CH01) or require (PGT145 and CAP256) intact trimer for binding.

The initial study of V2-apex bNAb development described one of the most potently neutralizing bNAb lineages, CAP256-VRC26 [11,28,30]. The CAP256 donor was infected and superinfected by clade C viruses but the Env variants implicated in the elicitation and development of breadth derived entirely from the superinfecting virus [30,31^■]. Viral diversity – overall and at the epitope – coincided with diversification in the antibody lineage and the appearance of breadth. Early escape mutations at key epitope residues 166 and 169 were tolerated by the broader lineage members, suggesting a mechanism for acquisition of breadth [30,31^■]. Additional studies suggested a role for sialic-acid binding [32] and Env escape fitness cost [33] in supporting CAP256-VRC26 lineage maturation, although the association with breadth was unclear.

The study of PCT64 bNAb lineage development [34^■■,35^■■] confirmed the critical role of localized diversity at key V2 strand C residues in driving bNAb maturation and highlighted the role of glycans. The Env evolution pattern at key V2 C-strand 166/169 residues supporting PCT64 bNAbs maturation was remarkably similar to the one described in CAP256, suggesting a roadmap for immunogen design. Evaluation of PCT64-UCA binding to early autologous Env trimers suggested that Env glycoform heterogeneity played a critical role in the lineage initiation. With its 25-aa CDRH3 and lower SHM, PCT64 may also represent a more achievable target for vaccine design.

V3-glycan high-mannose patch

The high-mannose patch at the base of the V3-loop, also called the V3-glycan supersite, is the most commonly targeted epitope in broadly neutralizing sera [27]. This maybe associated with reduced structural constraints for this epitope, as bNAbs targeting this region show multiple angles of approach, gene usage, and subtle differences in epitope and glycan requirements. These properties suggest that V3-glycan bNAbs might represent a more achievable target. Many [36] but not all [37-39] of these bNAbs have indels in CDR loops which actively engage the glycan shield to access the V3-loop peptide containing the co-receptor binding site.

The development of N332-glycan dependent bNAbs in donors PC076 (PCDN lineage) and PC039 (PCIN39 lineage) was characterized by parallel maturation to breadth in multiple branches, supporting the hypothesis that several structural solutions to neutralization breadth exist even within a single lineage. While PCIN39 bNAbs all independently acquired insertions in the CDRL1 [40], the PCDN lineage achieved breadth without indels in any branches [41]. This not only confirms that indels may not be strictly required but also suggests that indels may not be as rare as anticipated and could readily be selected by the appropriate immunogens. Interestingly, both PCDN76 and PCIN39 used the VH4-34 gene, previously associated with autoimmune disorders but both lineages matured away from autoreactive features.

Three lineages that all targeted the high-mannose patch were isolated from donor CH848 [42^■■]. The combined selective pressure of the non-bNAb DH475 and DH272 antibody lineages on the autologous Env led to a significant shortening of the V1-loop, which supported the DH270 bNAb lineage maturation. Interestingly, DH270 early intermediates acquired an improbable mutation in the CDRH2 introducing a coldspot for SHM and thereby fixing a mutation ultimately critical for development of heterologous breadth.

All three studies isolated bNAbs with relatively low levels of SHM (13–16%), driven by variation of V3-loop (mostly anchored at the V3 base) and in the V1-loop (glycans). Most recently, the maturation of BF520.1, an infant bNAb [39,43^■■] was found to require as little as 2% SHM without any indels to achieve neutralization breadth within 6 months of infection.

Membrane-proximal external region

Similar to the CD4bs, the MPER is a major target of antibody responses in infected individuals although but broad MPER responses are rare. bnAbs to the C-terminus of MPER can be extremely broad [1] and possess lipid-binding properties allowing access to their epitope adjacent to the viral membrane, but also associated with auto and polyreactivity, raising questions about their compatibility with vaccination.

In the first coevolution study for MPER bNAbs, three MPER-targeting bNAb lineages were identified from a single superinfected individual (RV217-40512) in the prospective RV217 cohort [44^■■]. Antibodies RV217-VRC42.01, VRC43.01, and VRC46.01 all had modest levels of somatic hypermutation (~10%) and normal antibody-loop lengths but used distinct recognition modes. The broadest lineage, VRC42, was similar to the known bNAb 4E10 in gene usage and mode of binding to the MPER peptide. A multimeric immunogen containing the founder virus MPER sequence could activate B cells bearing the VRC42-UCA. Furthermore, a VRC42 intermediate with only 3% SHM showed 50% breadth, demonstrating that a few changes can have an outsized impact. Although autoreactivity was detected for all three lineages, VRC42 and VRC43 were less autoreactive than 4E10 and were not associated with autoimmune disorder in this individual, raising cautious optimism for vaccination.

PRINCIPLES IN COMMON FROM DIFFERENT STUDIES

While each lineage has unique features, several common principles have emerged from studies of lineages targeting the same site and even across epitopes.

The naive B cell engages the early virus

Although this may seem an obvious conclusion, several studies have suggested an alternative origin for HIV antibodies via activation of memory B cells directed against gut microflora that cross-reacted with Env [45,46]. Indeed, for several lineages, UCA autologous binding and neutralization were not observed (PCT64, PCDN, DH270, and PCIN63), although this may be a limitation of the assayed binding partners. In contrast, some studies make the case for a more classical origin of these antibodies via engagement of naive B cells by early viruses. This has been observed for CH103, CAP256-VRC26, and VRC42 [11,25,44^■■]. The eliciting Env is not necessarily the transmitted-founder virus, but may be a variant arising weeks or months after infection [26^■■,30,41].

Furthermore, the availability of longitudinal samples does not completely alleviate the uncertainty of UCA determination as reported for the DH270, VRC42, and PCIN63 lineages [26^■■,42^■■,44^■■]. These studies also highlighted how single nucleotide variations can have substantial functional impact, warranting caution for using functional evidence to determine which candidate UCA represents the true precursor.

Relatively few mutations are needed to gain breadth

The high levels of SHM detected in the first extremely broad bNAbs [47,48] raised concerns that only highly somatically mutated antibodies could exhibit breadth. However, subsequent discoveries have shown that antibodies with 7–15% mutation [35^■■,36] can be broad and potent. In addition, some early intermediates from the BF520.1 and VRC42 lineages [43^■■,44^■■] are surprisingly broad despite having just 2–3% SHM. Virtually all of these studies have shown that shorter pathways compatible with vaccination are achievable. In fact, not only are not all mutations necessary for breadth [26^■■,49] but the level of SHM does not translate to breadth: mature antibodies with varying degrees of breadth were interspersed along the phylogenetic trees of CAP25-VRC26, PCDN, PCT76, and PC64 [26^■■,28,34^■■,41]. This observation inspired the concept of ‘off-track’ antibodies [30] that are selected during infection for binding to contemporaneous viral variants but not for heterologous neuralization breadth. Thus, coevolution studies not only provide insights on which immunogens to choose, but also which to avoid, for eliciting bNAbs.

Improbable mutations are selected

Somatic hypermutations are called ‘improbable’ [50^■] when, prior to selection, they: occur at unfavorable sequences (coldspots) for activation-induced cytosine deaminase (AID), the enzyme responsible for SHM; require multiple base substitutions for a specific amino-acid change; or require multiple mutations simultaneously to be structurally viable (e.g. a CDHR3 disulfide bond in CAP256-VRC26 [11]). In some lineages, amino-acid changes that are critical for the development of breadth are encoded by improbable mutations (DH270, CH235, and BF520.1) [42^■■,50^■]. The low frequency of such mutations may contribute to the rarity of bNAb lineage development in natural infection. As for rare precursors, the significance of this low frequency for vaccination is uncertain as relevant immunogens will be designed to increase the odds of positive selection.

Antibody lineages mature in the context of virus diversification by overcoming early viral escape

Several longitudinal studies noted that breadth is concomitant with viral diversification [25,30,34^■■,42^■■] suggesting the ability to tolerate amino-acid variation at critical epitope sites as a mechanism for imparting breadth [24,30,34^■■] in successive waves of Env escape and antibody adaptation. In several cases (CH505-CH235, CH848, and PCDN), the epitope diversity was driven by escape from different cooperating autologous neutralizing antibody responses [23-25,41,42^■■]. A previous observation that development of bNAbs is also associated with overall higher anti-Env IgG responses suggests a more global mechanism in which a diverse autologous antibody response may limit the Env overall escape landscape [25,51]. Immune pressure from maternal antibodies may similarly favor the rapid development of bNAbs in infants [39]. Superinfection, which may be associated with increased neutralization breadth [52,53], may also contribute to epitope diversity via recombination between strains. Of note, this may only aid the development of breadth when the targeted epitope is present in both the primary and superinfecting viruses allowing selection of cross-reactive B cells [31^■].

CONCLUSION

Although features of the first bNAbs – high levels of SHM, indels, long CDR loops, and autoimmunity – were initially thought to be unattainable by vaccination, these apparent roadblocks are beginning to fall with the studies reviewed herein and the discovery of bNAbs with more normal features. Comparisons of the developmental pathways in multiple individuals with shared or different specificities are beginning to reveal similarities in bNAb maturation and in the nature of the Env triggering the broad lineages, indicating potential paths for immunogen design. Given the level of variability observed for some epitopes, it still remains highly important to study the development of bNAbs targeting various epitopes as additional lineages may refine immunogen strategies. Importantly, no such studies have yet addressed the gp120–gp41 interface, fusion peptide, and silent face regions. Finally, analysis of the developmental pathway of non-bNAbs would shed light on the molecular events limiting the acquisition of neutralization breadth. Immunization studies in animals have begun evaluating immunogens based on sequences and structures identified in coevolution studies. Overall, these studies have not only been valuable to the HIV vaccine effort but provided unprecedented details of immune mechanisms that could be applied to other pathogens.

KEY POINTS.

• The coevolution of HIV-1 virus and broadly neutralizing antibodies has been studied in nine HIV-infected individuals.

• While each lineage has unique features, several common principles have emerged from studies of lineages targeting the same site and even across epitopes.

• These studies reveal pathways that are less complex than originally thought, providing optimism for eliciting them via vaccination.