A Next-Generation Cleaved, Soluble HIV-1 Env Trimer, BG505 SOSIP.664 gp140, Expresses Multiple Epitopes for Broadly Neutralizing but Not Non-Neutralizing Antibodies

If HIV's outer envelope constantly changes shape, then a vaccine targeting that envelope has to catch it in exactly the right configuration — the one that exposes the sites where neutralizing antibodies can actually do damage. Follow the chain one step further: if you could build a stable, soluble copy of that configuration in a bottle, you might finally have something worth injecting. That's what Sanders and colleagues set out to do, and BG505 SOSIP dot six hundred sixty-four is what they built. The central obstacle for an antibody-based HIV vaccine is the viral envelope glycoprotein, Env — a trimeric complex of three gp120 and three gp41 subunits held together by metastable, non-covalent interactions. It is the only target on the virus surface for neutralizing antibodies. Decades of work with the simplest Env forms, monomeric gp120 proteins, failed to induce broadly neutralizing antibodies, the kind capable of blocking diverse HIV strains. Part of the problem is architectural: when you remove the membrane anchor to make a soluble version, the trimer rapidly disintegrates into individual subunits. What the field needed was a homogeneous, stable, cleaved trimer that presents the epitopes broadly neutralizing antibodies recognize while hiding the sites that non-neutralizing antibodies target — the ones that generate immune responses that look impressive but do nothing useful against the virus. The BG505 strain was chosen deliberately. It comes from a subtype A transmitted-founder virus, meaning the sequence reflects what was actually transmitted between individuals rather than a virus that evolved inside a single host over years. Crucially, the BG505 gp120 monomer unusually binds PG9, one of the most potent broadly neutralizing antibodies known, making it a promising backbone for trimer work. From that starting point, Sanders and colleagues introduced a small set of precise engineering decisions. Two substitutions — A501C and T605C — create a disulfide bond, the so-called SOS link, physically tethering gp120 to the gp41 ectodomain so the cleaved trimer stays in one piece. A third change, I559P, breaks a helix in gp41 in a way that promotes trimerization and strengthens interactions among the three protomers. The natural furin cleavage site between gp120 and gp41 was replaced with six arginines to maximize cleavage efficiency, and the construct was co-expressed with furin itself. Finally, the protein was truncated at residue six hundred sixty-four — removing the membrane-proximal external region and the transmembrane domain — which reduced aggregation and improved solubility. The name encodes the history: SOSIP for the disulfide and proline, dot six hundred sixty-four for the truncation point. Expressed in HEK293T cells and purified by affinity chromatography followed by size-exclusion chromatography, the resulting protein behaved as a trimer. A dominant peak eluted at one hundred forty-four milliliters on the column, with a smaller monomer peak at one hundred sixty-four milliliters. Under reducing conditions on a gel, more than ninety-five percent of the material converted to separate gp120 and gp41 bands — confirming efficient cleavage. Those are the kinds of numbers that tell a biochemist the construct is doing what it was designed to do. Then came the stability data, and here the results get genuinely surprising. Differential scanning calorimetry, which measures how much heat a protein absorbs as it unfolds, showed that the trimer's thermal transition begins fourteen point four degrees Celsius higher than that of the gp120 monomer alone. The engineered trimer is more thermally stable than one of its own components in isolation. That's not what you would naively predict, and it suggests the interprotomer contacts introduced by the engineering are adding substantial thermodynamic stabilization to the whole assembly. Negative-stain electron microscopy confirmed the structural picture. The purified trimers are homogeneous — the particles look alike, not like a mixture of misfolded junk — and their three-dimensional reconstruction, calculated from over fifteen thousand particles at roughly twenty-four angstroms resolution, is nearly identical to reconstructions of native virion-associated spikes. For a listener who can't see the images: think of a squat mushroom shape, wide at the top where the three gp120 heads splay outward, narrowing at the stalk of gp41. The soluble trimer matches that shape. That structural fidelity is not cosmetic. It matters because the quaternary architecture of the spike — the way the three protomers pack against each other — creates or occludes epitopes. Get the shape wrong and you expose the wrong sites. Sanders and colleagues then ran the test that matters most for vaccine design: do antibodies that neutralize the live virus bind this soluble trimer, and do antibodies that can't neutralize the virus leave it alone? They used capture enzyme-linked immunosorbent assay, surface plasmon resonance, isothermal titration calorimetry, and negative-stain electron microscopy — four complementary methods, each with different sensitivity to avidity artifacts and conformational flexibility. The broadly neutralizing antibodies passed with high marks. Every major epitope class was represented: CD4-binding site antibodies, the N332-glycan-dependent V3 cluster including PGT121 through PGT128, the outer-domain glycan cluster including 2G12 and PGT135, and — most demanding of all — the quaternary epitopes that only form when three protomers assemble correctly. CH01, PG9, PG16, and PGT145 all bound strongly, and size-exclusion fractionation showed PG16 and PGT145 binding almost exclusively to trimer-containing fractions. That last result is direct evidence of genuine quaternary recognition. Isothermal titration calorimetry put numbers on the affinities: PGT128 Fab bound with a dissociation constant of five point seven nanomolar, PG9 Fab at eleven nanomolar, 2G12 immunoglobulin G at sixteen nanomolar, and PGT121 Fab at one hundred fifty-one nanomolar. These are tight, specific interactions. The non-neutralizing antibodies told the opposite story. CD4-binding-site non-neutralizers — b6, F105, 15e, F91 — showed little or no reactivity with the trimers even though they bind gp120 monomers readily. CD4-induced epitope antibodies like 17b and 412d required the addition of soluble CD4 to bind at all, because CD4 triggers a conformational shift that opens those sites. Non-neutralizing antibodies to the gp41 outer domain bound isolated gp41 subunits but not the intact trimer. There was one exception worth examining carefully. Three V3-directed non-neutralizing antibodies — 19b, 14e, and 39F — bound efficiently in enzyme-linked immunosorbent assay but showed minimal reactivity by surface plasmon resonance and very limited binding in the electron microscopy occupancy data. In electron microscopy, ninety-four percent of particles incubated with Fab b6 showed no bound Fab at all; for F240, ninety-eight percent were unbound. For 19b, sixty-one percent were unbound. Sanders and colleagues propose that the enzyme-linked immunosorbent assay capture format, which anchors the trimer through a C-terminal tag, may induce local unfolding that transiently exposes buried V3 epitopes — an avidity artifact of the plate rather than true epitope access. The surface plasmon resonance and electron microscopy results, which better reflect behavior in solution, support that interpretation. The distinction matters practically: surface plasmon resonance and electron microscopy are harder to fool than enzyme-linked immunosorbent assay. When Sanders and colleagues plotted trimer binding against virus neutralization across the full antibody panel, the correlation was strong — r equals zero point six five with a p-value below zero point zero zero zero one. Remove the 2G12 and V3 non-neutralizing antibodies from the calculation and it rises to r equals zero point eight eight. That is an unusually tight correspondence between a soluble protein and the functional antigenicity of a live virus, and it is the key validation that this trimer is doing what it was designed to do. What does this give the field? Two things. First, a structural tool. The trimers are already enabling detailed epitope mapping for antibodies like PG9, PGT122, and PGT135, and they serve as substrates for higher-resolution structural work. Second, a candidate immunogen — though Sanders and colleagues are careful about what they claim here. Immunogenicity studies in rabbits, guinea pigs, and macaques were underway at the time of publication, and the paper states plainly that whether the favorable antigenic profile translates into the induction of broadly neutralizing antibodies will be determined experimentally. The glycan shield that covers the trimer surface presents both a challenge and a potential tool: those glycans are part of what occludes non-neutralizing epitopes, and selectively adding glycans to further mask V3 is one direction the field is exploring, though any such changes risk creating new unintended epitopes. What BG505 SOSIP dot six hundred sixty-four represents is a molecularly honest baseline. For the first time, the field had a soluble, cleaved, stable Env trimer that looks like the native spike, behaves like the native spike antigenically, and can be produced in quantity. That is the prerequisite for everything that comes next — rational immunogen design, structural studies of broadly neutralizing antibody-Env complexes, and eventually the question of whether any of it actually works in an animal and then a human. The chain of reasoning that opened this conversation — catch the envelope in the right shape, put it in a bottle — Sanders and colleagues showed it can be done. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.