The trimeric envelope (Env) glycoprotein spike of human immunodeficiency virus type 1 (HIV-1) plays a crucial role in mediating host cell infection. Its exposed position on the virus surface also makes it the target for potent, broadly neutralizing antibodies (bNAbs) that are produced in a subset of infected individuals and that now influence the design of Env immunogens intended to induce similar antibody specificities. Env is extensively glycosylated, with glycans contributing to a significant proportion of the glycoprotein's mass (1). The heavily glycosylated surface of Env has been referred to as the “silent face,” with the glycans acting to shield the underlying viral protein surface from immune surveillance (2,–4). The glycan shield constantly evolves to protect the virus from newly produced neutralizing antibodies. However, compared to the highly diverse protein component of Env, many of the potential N-glycosylation sites (PNGSs) are well conserved; their total number has also remained relatively constant despite years of viral evolution (5, 6). While the glycans are derived from the host cell's glycosylation machinery, their high density leads to a large abundance of underprocessed, oligomannose-type glycans (Man5–9GlcNAc2) (7,–11). The glycan shield is therefore surprisingly homogenous given the extent of glycosylation, and it is itself a target for bNAbs that recognize exclusively glycan or mixed glycan/protein epitopes (12,–25).
The Env spike is a trimer of heterodimers composed of gp41 and gp120 subunits, which are generated when a protease, usually furin, cleaves the proprotein gp160 glycoprotein (26,–31). Furin most probably cleaves gp160 in the trans-Golgi network (TGN), but it is able to cycle between the TGN, the endosomal compartments, and the cell surface and is also believed to be able to act in the early secretory pathway (27, 32, 33). Furin is an important factor in Env immunogen production, as it is now commonly coexpressed to generate fully cleaved recombinant soluble trimers that structurally mimic native Env (34, 35). In contrast, structural and antigenic analyses of many uncleaved Env constructs show that they adopt nonnative conformations, with gp120 subunits dangling from a gp41 core, unless additional modifications are made to overcome the defect (36,–39). Cleavage is therefore important for the structural integrity of the trimer, which in turn greatly influences how the trimer is glycosylated (39, 40).
Several studies on monomeric gp120 proteins have demonstrated the presence of a cluster of oligomannose-type glycans, the “intrinsic mannose patch” (IMP), which arises because the glycan shield creates steric constraints that restrict the actions of endoplasmic reticulum (ER) and Golgi apparatus α-mannosidases (7, 8, 11, 41,–43). The IMP is localized at the outer domain of gp120 and comprises the epitopes of numerous glycan-influenced bNAbs (15, 21, 23, 24, 44,–47). The IMP is a characteristic feature of gp120 that is conserved across all HIV-1 clades (7, 8, 48) and longitudinally during infection (49). Analyses of recombinant soluble, native-like trimers (40, 50), membrane-associated trimers (51), and virion-derived Env (7,–9) all show that their oligomannose-type glycan contents are even higher than those of gp120 monomers, implying that a “trimer-associated mannose patch” (TAMP) exists (52). However, the glycan sites forming any such TAMP have not previously been defined.
Cleaved, soluble, native-like Env trimers are currently one focus of structure-guided immunogen design programs. A detailed site-specific N-glycosylation analysis of the prototype of this class of trimers, BG505 SOSIP.664, revealed fine structural details of the glycan shield (35, 50, 53). The gp120 subunit of this trimer is dominated by underprocessed Man8–9GlcNAc2 glycans, which form a mosaic of densely organized oligomannose clusters on the outer domain and the trimer interface (40, 50). In contrast, glycosylation sites with a mixed processing state (i.e., sites containing both highly processed, complex, and oligomannose-type glycans) are found at the trimer apex and near its N and C termini (50). Compact folding seems to be the major driver behind the oligomannose-dominated profile of the glycan shield on native trimers. For example, uncleaved, nonnative constructs that adopt a much more open and irregular conformation have a significantly reduced content of oligomannose-type glycans, because the now more accessible structures are efficiently processed (39, 40). We refer here to these uncleaved, nonnative glycoproteins as pseudotrimers, to reflect their lack of the 3-fold symmetry that is a characteristic feature of the structures of natively folded trimers (54, 55).
Here we reveal the detailed effects of cleavage-induced, native-like trimerization on N-glycan processing. Our quantitative, site-specific analysis of a set of comparably expressed and purified monomeric (i.e., gp120), uncleaved pseudotrimeric, and fully cleaved trimeric envelope glycoproteins of the same genotype (i.e., BG505) identifies the location of the cleavage-dependent TAMP. Thus, the glycans that form the TAMP are found in regions near the trimer apex and the protomer interface. We show that compared to those in pseudotrimers, these regions are even more highly processed on monomeric gp120, implying that the oligomerization events impose additional regional restrictions on glycan processing. This finding is echoed across the panel of glycoproteins in the analysis of O-linked glycosylation at T499, near the trimer base, which is present at almost unmeasurable levels on native-like trimers. We are also able to classify the glycan processing states of all N-glycans present on the various constructs, thus extending our earlier site-specific analysis of the BG505 SOSIP.664 trimer (50). Our findings emphasize how the glycan shield is influenced by the quaternary structure of the Env spike and highlight the relevance of native-like structures to Env vaccine development.
Glycosylation processing on monomeric, uncleaved, and cleaved Env proteins.
To thoroughly investigate the influence of trimerization- and cleavage-induced structural integrity on Env glycosylation, we used comparable conditions to express and purify monomeric gp120, native-like SOSIP.664 trimers, and uncleaved, nonnative WT.SEKS gp140 pseudotrimers, all based on the BG505 genotype. The design and characteristics of these proteins have been described previously (35, 53, 54, 56,–59). The WT.SEKS construct differs from SOSIP.664 in that it lacks the stabilizing SOSIP mutations and contains an inactivated furin cleavage site (RRRRRR mutated to SEKS); the resulting uncleaved pseudotrimers have a nonnative conformation (36). The monomeric gp120 protein is isogenic with the gp120 subunit of the SOSIP.664 trimer. The designs of all three constructs are summarized in Fig. 1A. The three proteins were all produced in transiently transfected human embryonic kidney (HEK) 293F cells and purified using 2G12 affinity chromatography followed by size-exclusion chromatography (SEC).
Ion mobility mass spectrometry analysis of BG505 Env glycoproteins. Mobility-extracted singly charged negative-ion electrospray spectra are shown for N-linked glycans found on the following BG505 Env proteins: gp120 monomers (A), WT.SEKS pseudotrimers (B), and SOSIP.664 trimers (C). The inset in panel B shows an example of a DriftScope image derived from gp120 monomers, with singly charged ions encircled with a yellow oval. The peaks of the oligomannose series Man5–9GlcNAc2 are highlighted in green. A list of identified glycans is given in Table S1 in the supplemental material.
Quantitative site-specific N-glycosylation analysis.
We performed a quantitative, site-specific N-glycosylation analysis using in-line liquid chromatography-ESI MS (LC-ESI MS) to thoroughly investigate how the formation of native-like, cleavage-induced trimers influences the glycan shield. We previously validated this method (50). We used a panel of different proteases to create peptides and glycopeptides from the gp120, WT.SEKS, and SOSIP.664 proteins, enriched glycopeptide populations, and then quantified the relative abundances of individual glycans by assessing the ion intensities over all charge states. The quantified and summarized data for the three Env constructs show that the processing states of distinct N-glycosylation sites in the glycan shields of gp120 monomers and pseudotrimers are very similar to each other but also quite different from those for native-like SOSIP.664 trimers (Fig. 3).
Models of a fully glycosylated BG505 gp120 monomer and a SOSIP.664 trimer. Models of the glycosylated gp120 monomer (A) and the glycosylated SOSIP.664 trimer (B) were derived from one previously described elsewhere (50). The monomer is oriented as it would appear in situ as a subunit of the SOSIP.664 trimer. The glycans on the models are colored according to their oligomannose content, as derived in the present study. N-glycan sites for which quantitative results of sufficient quality could not be obtained were classified by qualitative analysis according to their susceptibility to endo H and PNGase F digestion (Table S6).
The site-specific analysis of BG505 gp120 revealed numerous positions where exclusively or almost exclusively oligomannose-type glycans are present. These sites define the gp120 IMP (7). Note that the glycan repertoires are highly similar for the corresponding locations in the gp120 subunits of the uncleaved WT.SEKS pseudotrimer and the SOSIP.664 trimer (Fig. 3A). However, whereas Man9GlcNAc2 moieties dominate the sites forming the IMP on the SOSIP.664 trimer (e.g., N234, N363, N339, and N392) (Fig. 3A), there is additional but modest trimming of these sites toward smaller oligomannose glycans on the gp120 monomer and the pseudotrimer. Other sites on the gp120 monomer that contain a significant population of complex-type glycans are less processed on the native-like trimer (e.g., N137, N156, N160, N197, N276, and N355). The overall dominance of oligomannose-type glycosylation, however, is conserved among the three Env proteins, which supports the definition of the IMP as an intrinsic feature of the gp120 protein, whatever its quaternary context (7). The N295 site seems to be protected comparably from glycan processing enzymes on all three constructs; it is the only site where the quaternary context and cleavage status do not increase glycan processing. The highly conserved “supersite of vulnerability” on the uncleaved WT.SEKS glycoprotein, i.e., N332, carries a small population of complex-type glycans, whereas this site is entirely oligomannose on both the cleaved SOSIP.664 trimer and the gp120 monomer. A likely explanation is that the conformationally heterogeneous pseudotrimers include a subpopulation on which processing enzymes can access and modify this antigenically important site at the heart of the IMP. The site-specific glycosylation states of the gp41 subunits of the BG505 WT.SEKS pseudotrimer and the BG505.664 trimer (Fig. 3B) are largely similar in that complex-type glycosylation dominates. However, the native-like trimer has an elevated population of oligomannose glycans at N625 and N637, indicating steric inaccessibility in those regions.
To display the site-specific glycosylation analysis and to facilitate structural interpretation, we mapped the distribution of the different glycan processing states across the BG505 gp120 monomer (Fig. 4A) and the SOSIP.664 trimer (Fig. 4B), using a model of a fully glycosylated trimer derived from a cryo-electron microscopy structure (50, 61). All the glycosylation sites are displayed, with the sites classified as mainly oligomannose, mainly complex, or mixed.
Formation of native-like trimers results in interprotomer control of glycan processing.
Based on the BG505 SOSIP.664 glycosylation map, we created two heat maps displaying the site-specific increase in oligomannose-type glycans on the structure of the cleaved trimer relative to the gp120 monomer (Fig. 5A) and the pseudotrimer (Fig. 5B). The cleavage-dependent formation of native-like trimers has a significant effect on the processing state of several glycans, including those at the trimer apex and the protomer interface. The increase in oligomannose-type glycans on the native trimer relative to the monomer or pseudotrimer can be viewed as a decrease in enzymatic processing consistent with the structural constraints that limit processing. For example, the glycan at N276 on SOSIP.664, near the CD4 binding site (CD4bs), has an ∼75% oligomannose content, whereas the corresponding value for the same site on the gp120 monomer is ∼5%; the glycan at N276 is the gp120 glycan that differs the most in this regard (Fig. 5A). Other glycans affected similarly but less profoundly are N156 (∼60% oligomannose on monomers versus 100% on trimers, hence a 40 percentage point (pp) increase in Fig. 4), N160 (∼60 pp increase), and N197 (∼50 pp increase); each of these sites is located at the trimer apex or protomer interface and thus is in a different quaternary context on the SOSIP.664 trimer compared to the gp120 monomer. Similarly, the outer domain N355 glycan is highly processed on monomeric gp120 but is a mixed glycan site (∼40 pp increase in oligomannose) on the SOSIP.664 trimer. In this case, the model suggests that the presence (trimer) or absence (monomer) of proximal gp41 glycans may be the key influence (Fig. 4A).