Viral Capsid Self-Assembly

Why viral capsid self-assembly matters

• It is the cheapest way to wrap a large volume in a small amount of code. A poliovirus genome is about 7,400 nucleotides — barely enough for a dozen proteins. Encoding one 33 kDa subunit and reusing it 60 times (the principle of genetic economy, Crick & Watson 1956) is the only way to build a 30 nm container around that genome without the gene for the container being bigger than the genome itself.
• It is a vaccine platform worth billions. The HPV vaccines Gardasil and Cervarix are nothing but the L1 coat protein self-assembled into empty virus-like particles (VLPs) — no DNA inside, no infectivity, but the same surface the immune system sees on a real virion. The hepatitis B surface-antigen vaccine works the same way. VLP vaccines have been given to hundreds of millions of people.
• It is the engine of gene therapy. Adeno-associated virus (AAV) capsids self-assemble around a therapeutic gene and deliver it to human cells; AAV vectors carry the approved therapies for spinal muscular atrophy (Zolgensma) and inherited retinal disease (Luxturna).
• It is a brand-new antiviral target. Capsid assembly modulators are a 2020s drug class that don't block an enzyme — they poison the geometry. HIV's lenacapavir (approved 2022) binds the capsid protein and forces it to mis-assemble; hepatitis B core-protein modulators do the same. Break the lattice and the virus can't build a working shell.
• It is a model system for all of structural self-organization. The same physics — many identical parts, weak reversible bonds, an energy landscape with one deep minimum — explains protein folding, microtubule assembly, and designed protein nanocages. Capsids are the textbook case because their symmetry makes the geometry exactly solvable.
• It defined an era of structural biology. Tomato bushy stunt virus (Harrison, 1978) and the HK97 bacteriophage chainmail capsid (Wikoff & Johnson, 2000) were landmark atomic structures. Today single-particle cryo-EM resolves entire 960-subunit capsids to better than 3 Å, turning the abstract triangulation rules into visible atoms.

How a capsid assembles itself

Start with a soup of freshly translated coat-protein subunits in the host cytoplasm. Each subunit is small — typically 200–400 amino acids — and folds into a compact domain, very often the eight-stranded jelly-roll β-barrel that recurs across unrelated virus families (picornaviruses, plant viruses, parvoviruses, adenovirus). The barrel presents one flat-ish face and edges studded with complementary chemical patches: hydrophobic ridges, charged residues for salt bridges, and backbone donors and acceptors for hydrogen bonds.

Assembly proceeds through a nucleation-and-growth pathway, like crystallization. First a small unstable nucleus forms — often a trimer or a pentamer of subunits, or a coat-protein dimer bound to a packaging signal on the genome. Below a critical nucleus the cluster is more likely to fall apart than grow; above it, adding the next subunit is downhill, and the shell sweeps to completion cooperatively. Because no single interface is strong (often just 3–8 kcal/mol of free energy per contact), wrongly placed subunits dissociate and re-bind. This error correction by reversibility is the whole reason a process with no proofreading enzyme reliably yields one perfect closed shell rather than a heap of misfolded junk.

The closure is forced by geometry. To make a finite closed shell from triangular units you must include exactly 12 pentamers — 12 points where five subunits meet instead of six — a hard constraint from Euler's theorem (the same reason a soccer ball has exactly 12 pentagonal panels). The protein therefore has to adopt two slightly different conformations: a flatter one at the hexamers and a more curved one at the pentamers. Many capsid proteins do this with a flexible "molecular switch" — a movable N-terminal or C-terminal arm, or a wedge of ordered RNA — that toggles the local curvature. That switch is what lets quasi-equivalence work: chemically identical subunits sit in geometrically non-identical seats by bending a hinge.

Two big classes of virus deviate from pure spontaneous folding. Tailed dsDNA bacteriophages and herpesviruses first build an empty procapsid around internal scaffolding proteins, then eject the scaffold, then run a powerful ATP-driven terminase motor at a unique portal vertex to pump a single naked genome into the rigid shell against pressures of tens of atmospheres. Helical viruses like tobacco mosaic virus skip the icosahedron entirely: subunits stack in a helix directly along the RNA, one base of structure laid down per turn, so the genome's length sets the rod's length exactly.

The players and conditions

• Coat / capsid protein (CP). The one repeated subunit. Usually one gene product; some viruses use two or three different proteins (poliovirus uses VP1–VP4 cut from one precursor). The jelly-roll β-barrel fold is the most common architecture.
• The genome. ssRNA, dsRNA, ssDNA, or dsDNA. In many RNA viruses the genome is not cargo loaded last — it is a structural scaffold that nucleates and templates assembly through packaging signals.
• Packaging / nucleation signal. A specific sequence and fold — TMV's origin-of-assembly loop, MS2's operator stem-loop, the φ-sites of HIV, the pac and cos sites of DNA phages — that the CP recognizes to start assembly on the right nucleic acid.
• Scaffolding proteins (large viruses only). Transient internal chaperones that template the procapsid shape, then leave or are proteolyzed before DNA enters.
• The portal and terminase (tailed phages, herpesviruses). A 12-subunit ring at one special vertex plus an ATPase motor that translocates DNA.
• Solution conditions. pH, ionic strength, and divalent cations (Mg²⁺, Ca²⁺) tune the interface strength. TMV protein forms disks at pH 7 and the helical lock at pH ~5–6; many capsids assemble only below a threshold pH or above a threshold protein concentration.

Caspar–Klug geometry: the allowed shells

In 1962 Donald Caspar and Aaron Klug explained why capsid sizes are quantized. An icosahedral shell is built from 60 × T subunits, where the triangulation number T counts the quasi-equivalent positions per asymmetric unit. T is not free: it must equal h² + hk + k² for non-negative integers h, k. Each capsid has exactly 12 pentamers and 10 × (T − 1) hexamers.

T number	(h, k)	Subunits (60×T)	Pentamers	Hexamers	Example virus	Diameter
T = 1	(1,0)	60	12	0	Parvovirus, STNV, AAV	~18–26 nm
pseudo T = 3	—	60 (3 distinct CPs)	12	—	Poliovirus, rhinovirus	~30 nm
T = 3	(1,1)	180	12	20	Cowpea mosaic, nodavirus	~28–30 nm
T = 4	(2,0)	240	12	30	Hepatitis B core, Sindbis	~32–45 nm
T = 7	(2,1)	420	12	60	Papillomavirus, HK97 phage	~55–60 nm
T = 13	(3,1)	780	12	120	Rotavirus, bluetongue inner	~70–80 nm
T = 16	(4,0)	960	12	150	Herpesvirus capsid	~125 nm

Note the special cases. Pseudo-T = 3 picornaviruses look like a T = 3 shell but use three different proteins (VP1, VP2, VP3) in the three positions instead of one protein in three conformations — same lattice, more genes. Giant viruses such as Mimivirus push to T = 1141 and beyond, and HIV breaks icosahedral symmetry entirely: its core is a fullerene cone of about 250 capsid (CA) hexamers closed by exactly 12 pentamers — 7 at the wide end, 5 at the narrow end — which is why it's a cone rather than a sphere.

Quantified figures: sizes, forces, and timescales

• Subunit size: coat proteins are commonly 200–400 residues, ~25–60 kDa. TMV CP is 158 residues (17.5 kDa); HBV core is 183 residues.
• Copy number: 60 (T=1) to 960 (T=16) for icosahedral capsids; ~2,130 identical subunits in a single TMV helical rod.
• Shell dimensions: 18 nm (parvovirus) to 125 nm (herpesvirus) across; capsid walls are typically 2–5 nm thick. TMV is a hollow rod 300 nm long, 18 nm wide, with a 4 nm central channel.
• Interface energy: roughly 3–8 kcal/mol of free energy per subunit–subunit contact — deliberately weak so errors self-correct; total shell stabilization is the sum over hundreds of such contacts.
• Speed: in vitro reassembly of TMV or simple RNA virus capsids completes in seconds to a few minutes once nucleated; HBV capsids assemble on the order of seconds at physiological protein concentration.
• Packaging pressure (DNA phages): the φ29 and T4 terminase motors pump DNA against internal pressures of ~50–60 atmospheres, and the φ29 motor is one of the strongest known, stalling near 50–60 piconewtons of force.
• Packaging density: dsDNA inside a phage head reaches ~500 mg/mL — near the density of crystalline DNA — which is why so much pressure is needed to load it and why it ejects so fast on infection.
• Symmetry: the icosahedral point group has 60-fold rotational symmetry (6 five-fold, 10 three-fold, 15 two-fold axes) — the maximum for a finite assembly of asymmetric subunits.

Self-assembly vs motor-driven assembly

Property	Spontaneous self-assembly (simple capsids)	Scaffold + motor assembly (large dsDNA viruses)
Examples	TMV, STNV, MS2, poliovirus, HBV, AAV	T4, λ, φ29 phages; herpesvirus
Energy source	None — thermodynamic (entropy + weak bonds)	ATP-burning terminase motor
Genome timing	Co-assembled; often nucleates the shell	Empty procapsid first, genome pumped in last
Scaffolding proteins	Usually none	Required, then removed/proteolyzed
Genome state inside	Loosely condensed with the protein	Tightly spooled, ~50–60 atm pressure
In-vitro reconstitution	Yes — RNA + protein in a tube (TMV, 1955)	No — needs portal, motor, and ATP
Error correction	Reversible binding; mis-assembly falls apart	Portal geometry + headful sensing
Typical T / size	T = 1–4, 18–45 nm	T = 7–16+, 55–125 nm

Real organisms, diseases, and named examples

• Tobacco mosaic virus (TMV). The original self-assembly proof: Fraenkel-Conrat and Williams (1955) mixed purified TMV RNA with purified coat protein and recovered infectious rod-shaped virus, showing the assembly instructions live in the protein, not in any enzyme. Klug's later work on TMV and capsids won the 1982 Nobel Prize in Chemistry.
• Poliovirus and rhinovirus (picornaviruses). Pseudo-T = 3 shells of VP1–VP4. Pleconaril and related "WIN compounds" wedge into a hydrophobic pocket under the canyon floor and rigidify the capsid so it can't uncoat — capsid-targeted antivirals that predate the modern assembly-modulator class.
• Human papillomavirus (HPV). The L1 protein self-assembles into 72-pentamer T = 7 VLPs; expressed in yeast or insect cells with no viral DNA, these empty shells are Gardasil and Cervarix.
• Hepatitis B virus (HBV). The core antigen self-assembles into T = 3 and T = 4 capsids; core-protein allosteric modulators (CpAMs) speed or mis-direct assembly and are in trials as functional-cure antivirals.
• HIV-1. The Gag-derived capsid (CA) protein builds a non-icosahedral fullerene cone; lenacapavir (Sunlenca, 2022) binds CA and disrupts both assembly and uncoating, a long-acting twice-yearly drug.
• Bacteriophage HK97. Its T = 7 capsid forms covalent isopeptide "chainmail" rings that catenate the subunits like molecular knights' armor — a unique stabilization strategy discovered from its 2000 crystal structure.
• Adeno-associated virus (AAV). A T = 1 parvovirus capsid (VP1/VP2/VP3) self-assembled around a therapeutic transgene — the workhorse of approved gene therapies.

Common misconceptions and pitfalls

• "The subunits are glued by strong, permanent bonds." No — the interfaces are deliberately weak (a few kcal/mol each). Weakness is the feature: it lets mis-placed subunits unbind and the shell reach its one correct minimum. Strong irreversible bonds would freeze in errors.
• "Capsids are spheres." They look round at low resolution but are icosahedra (or helices, or HIV's cone). The 12 pentamers and the discrete T-number sizes are the giveaway that the geometry is faceted, not spherical — you cannot tile a true sphere with identical units.
• "Any number of subunits can form a shell." Only 60 × T with T = h² + hk + k² gives a stable icosahedral closure. You can't build a stable 100-subunit or 300-subunit icosahedral capsid; the math forbids it.
• "Quasi-equivalent subunits are different proteins." Usually they are the same protein in slightly different conformations, switched by a flexible arm or a bound RNA wedge. (Pseudo-T = 3 picornaviruses are the exception that uses genuinely different proteins.)
• "Assembly needs ATP and enzymes." Simple capsids need neither — they are spontaneous. Only large dsDNA viruses use an ATP motor, and even then only to pump the genome, not to build the protein shell.
• "The genome is loaded into a finished sphere like filling a balloon." True for tailed phages and herpesviruses, but for most RNA viruses the genome co-assembles and actively nucleates and templates the shell — protein and nucleic acid build each other simultaneously.
• "A virus-like particle is infectious." A VLP has the same protein shell but no genome inside, so it cannot replicate — which is exactly why empty self-assembled capsids make safe vaccines.