A Ginzburg-Landau model for the expansion of a dodecahedral viral capsid E. Zappaa, G. Indelicatob, A. Albanoc, P. Cermellic aDepartment of Mathematics, University of York, UK 3 bYork Centre for Complex Systems Analysis, Department of Mathematics, University of 1 0 York, UK 2 cDipartimento di Matematica, Universit`a di Torino, Italy n a J 7 1 Abstract ] We propose a Ginzburg-Landau model for the expansion of a dodecahedral h p viral capsid during infection or maturation. The capsid is described as a - dodecahedron whose faces, meant to model rigid capsomers, are free to move h t independent of each other, and has therefore twelve degrees of freedom. We a m assumethattheenergyofthesystemisafunctionofthetwelvevariableswith [ icosahedral symmetry. Using techniques of the theory of invariants, we ex- pand the energy as the sum of invariant polynomials up to fourth order, and 1 v classify its minima in dependence of the coefficients of the Ginzburg-Landau 6 expansion. Possible conformational changes of the capsid correspond to sym- 3 0 metry breaking of the equilibrium closed form. The results suggest that the 4 only generic transition from the closed state leads to icosahedral expanded . 1 form. Our approach does not allow to study the expansion pathway, which 0 3 is likely to be non-icosahedral. 1 : Keywords: v i 2010 MSC: 92-XX, 20C40 X r a 1. Introduction Most viruses are made of a protein shell, the capsid, built of identical protein units, that encapsidates and hence protects the nucleic acid (RNA Email addresses: [email protected] (E. Zappa), [email protected] (G. Indelicato), [email protected] (A. Albano), [email protected] (P. Cermelli) Preprint submitted to International Journal of Nonlinear Mechanics or DNA). Although viruses exhibit a wide diversity of shape a large number of them display icosahedral symmetry and this is irrespective of the number and the chemical composition of the protein subunits, the capsomers, that constitute the capsid. The basic principles to account for the icosahedral arrangement of the protein in a capsid were outlined by Caspar and Klug in the quasi-equivalence theory [1], and this continues to be a fundamental framework in virology. During their life-cycle viruses undergo structural transitions. These phe- nomena can be triggered by a change of the environment, such as a variation of temperature or pH. The occurence of a transitions induces a radial expan- sion of the capsid and a rearrangement of the capsomers and consequently the opening of pores on the capsid, so that the genetic material is exposed and eventually released in the host cell [2, 3, 4, 5]. In this work we focus on icosahedral viruses whose expansion can be mod- elledthroughtheindependentmotionoftwelvepentagonalblocks. Thesecan be viral particles whose capsomers are pentamers (group of five proteins) or more complex viruses whose protein subunit are arranged to form pentagonal blocks [5]. In general, the energy of the capsid should account both for the cohesive forces between capsomers and for an internal pressure that tends to expand the capsid. However, instead of making specific choices, we adopt here the Ginzburg-Landau approach, and assume that the energy of the capsid is a function of the twelve independent degrees of freedom of the capsomers, and its explicit expression is given in terms of invariant polynomials of the icosahedral group. The coefficients of the expansion are the basic control pa- rametersofourmodeland, inturn, theyshouldberelatedtotheenvironment of the capsid. We relax the symmetry conditions and classify the minima of the energy according to their symmetry, in dependence of the parameters. The appear- ance of new minima corresponds to conformational changes of the capsid, with symmetry possibly lower than icosahedral. We restrict to minima of the energy corresponding to one of the three maximal subgroups of the icosahe- dral group: the tetrahedral group, and the dihedral groups of order 6 and 10. The analysis shows that only four of the nine expansion parameters are relevant in this model: the parameter space is subdivided into regions, in which minima have a given symmetry. The main result of our work is that the only generic transition from a closed configuration with icosahedral sym- 2 metry turns out to be to an icosahedral expanded state. Minima with lower symmetry are not accessible from the closed reference state because they involve the cooperative change of more than one control parameters. Hence, our study shows that a phenomenological model based on the Ginzburg-Landau expansion is able to describe one of the main features of theexpansionofaviralcapsid: eventhoughthetransitionpathwayisunlikely to be icosahedral (cf. [6, 7]), the final equilibrium state is still icosahedral (cf. [7]). Our work complements other approaches to the study of conformational changes of viral capsids, either based on coarse-graining and phenomenolog- ical interaction potentials between the capsomers and relying on domain- decomposition techniques [8, 7], or on the geometrical description of the fine features of the capsid via libraries of point sets with icosahedral symmetry [9, 10, 11, 6], normal mode analysis of the atomic ensemble of the capsid [12] or, finally, using the continuum theory of thin shells [13, 14]. 2. Basics of the Ginzburg-Landau approach Let us consider a system described by state variables (x ,...,x ) = x ∈ 1 n Rn and a symmetry group G acting on Rn. The action of G provides a representation ρ : G −→ GL(R,n). We associate to the system a free energy E : Rn ×A −→ R (1) (x,α) (cid:55)−→ E(x,α), where A ⊆ Rm is an open subset, α ∈ A are parameters affecting the system (i.e., temperature, pH, etc.), and E ∈ C2(Rn×A). We require the energy to be invariant with respect to ρ, i.e. E(ρ(g)x,α) = E(x,α) ∀g ∈ G, ∀x ∈ Rn ∀α ∈ A. (2) The minima of E(·,α) with respect to x correspond to the stable phases of the system. Let us introduce the following definition Definition 2.1. Let ρ : G −→ GL(Rn) be a representation of a finite group G. The isotropy subgroup of x ∈ Rn is Σ = {g ∈ G : ρ(g)x = x}. x 3 Minima of E(·,α) can be classified according to their isotropy group. In particular, if x is a minimum of E(·,α) such that 0 Σ = G, x0 we say that the system is in a high-symmetry phase. In general, we are interested in studying the local minima of the energy as α varies. The system undergoes a phase transition when the number and the symmetry of the minima changes as α varies [15]. Notice that the level sets of E(·,α) are invariant under ρ. Therefore, if there exists a minimum x such that 0 Σ = H, H < G, x0 all points of the orbit ρ(G)x also are minima. If G is finite, the orbit ρ(G)x 0 0 is also finite and has |G/H| elements. Hence, low symmetry phases occur in different variants, while high symmetry phases are in general unique. Since we are interested in icosahedral viruses we consider the icosahedral group I which consists of all the rotations which leave a regular icosahedron invariant. It has order 60, and it is generated by the elements g (twofold 2 rotation) and g (fivefold rotation) such that (g )2 = (g )5 = (g g )3 = e, the 5 2 5 2 5 identity element. It is isomorphic to the alternating group of order 5, A . 5 Moreover I acts naturally on the twelve faces of a dodecahedron by per- mutation. In this way, we obtain a function σ : I −→ S , where S is the 12 12 symmetric group of order 12, such that σ(g ) = (1,6)(2,5)(3,9)(4,10)(7,12)(8,11), 2 σ(g ) = (1,2,3,4,5)(7,8,9,10,11). 5 It is easy to see that σ is an homomorphism, since σ(g )2 = σ(g )5 = (σ(g )σ(g ))3 = id . 2 5 2 5 S12 σ is called a permutation representation of I, being an homomorphism be- tween a group and a permutation group. The symmetric group S acts nat- 12 urally on R12 by permuting the indices of a vector. This action induces a rep- resentation ρ : I −→ GL(R,12) such that ρ(g )2 = ρ(g )5 = (ρ(g )ρ(g ))3 = 2 5 2 5 I , with I the identity matrix of order 12. 12 12 4 3. Formulation of the model We now formulate a model for the expansion of the capsid in the frame- work of the Ginzburg-Landau theory. We focus here on viruses with capsid with icosahedral symmetry, whose expansion can be described via the independent motion of twelve pentagonal blocks. Therefore, before expansion, the capsid can be modelled as a regular dodecahedron (Figure 1). We associate to each capsomer, labeled by i = 1,...,12, a translation parameter x along the axis of the pentagonal face, i with x ∈ R (cf. Figure 2). The translation variables x have meaning only i i when non negative, but we allow them to assume negative values in order to simplify the analysis of the minima of the energy. However, we will only accept as physically meaningful the minima whose components are all non negative. The translation parameters x ,i = 1,...,12, represent the state variables i of the system, i.e., x = (x ,...,x ). Note that 0 = (0,...,0) represents the 1 12 5 4 6 1 3 2 Figure 1: A dodecahedron: each pentagonal face is labelled by a number from 1 to 12. The face opposite to face i is labelled by i+6, for i=1,...,6. closed configuration of the capsid, which has by definition full icosahedral symmetry, i.e., its isotropy subgroup is the whole group I. We require the energy E to be a C2 function E : R12 × A → R, with A ⊂ Rm, such that: (i) it is invariant with respect to the representation ρ of I on R12, i.e., E(ρ(g)x,α) = E(x,α) ∀g ∈ I, ∀α ∈ A, ∀x ∈ R12. (ii) 0 = (0,...,0) is a critical point of E, i.e. ∇ E(0,α) = 0 ∀α ∈ A. x 5 (iii) lim E(x,α) = +∞ ∀α ∈ A. |x|→+∞ As a first approach we also require the energy to be polynomial in the vari- ables x . This assumption relies on the fact that, if the energy is sufficently i smooth, it can be expanded as a Taylor polynomial in a neighbourhood of 0, which still has the property of invariance with respect to ρ. Our goals are: • to find explicit expressions of the energy E as the sum of icosahedrally invariant polynomials; • to find the minima of E in dependence of the parameters α. When 0, which is always an extremum due to hypoyhesis (ii), is a minimum, the capsid is in the closed configuration. We interpret the appearance of new minima of the energy as conformational changes of the capsid associated to its expansion, driven by variations of the parameters α, that in turn are related to the environment of the capsid. x i x j Figure 2: Basic variables of the model. 4. Explicit form of the energy In this section we describe a method to find an explicit form of the energy. As we pointed out in the previous section, we assume it to be a polynomial invariant under the action of I. Therefore, we first study the ring of invariant polynomials and its structure. Complete proofs for the assertions below can be found, for example, in [16] or [17]. 6 Let V be a vector space over a field K of characteristic zero (in what follows, we will be interested mainly in K = R) with dimV = n. We fix a basis A of V and let {x , ..., x } be the dual basis of V∗. 1 n Let G be a finite group and ρ : G −→ GL(V) a representation of G. The action of G on V induces the dual action on V∗ and hence an action on the symmetric algebra Sym(V∗) ∼= K[x ,...,x ]. We denote by K[x]G = 1 n K[x ,...,x ]G the subring of polynomials invariant under the action of G. 1 n K[x] is a graded ring, i.e., +∞ (cid:77) K[x] = K[x] k k=0 where K[x] is the vector space of the homogeneous polynomials of degree k. k Since the action of G preserves the grading, K[x]G is a graded subring and we can write +∞ (cid:77) K[x]G = K[x]G k k=0 where K[x]G = K[x ,...,x ]G is the set of all the invariant polynomials of k 1 n k degree k. One way to find invariant polynomials is to take the average over the orbits of the action. More precisely, let us introduce the so-called Reynolds operator R : K[x] −→ K[x]G G defined by 1 (cid:88) R (f)(x) = (g ·f)(x). G |G| g∈G It is clear that R is a K-linear map that is the identity on the sub- G ring K[x]G (in particular, it is surjective). Hilbert showed, as a consequence of his famous Basis Theorem, that Theorem 4.1 (Hilbert). The invariant ring K[x]G of a finite group G is finitely generated. LaterE.Notherprovedthatthedegreeofthegeneratorsisboundedabove by the order of the group: 7 Theorem 4.2 (E. Noether). Let |G| be the order of the group G. Then K[x]G is generated, as a K-algebra, by (finitely many) polynomials of degree less or equal than |G|. Hence there exist p ,...,p ∈ K[x]G such that K[x]G = K[p ,...,p ], 1 t 1 t i.e., every polynomial in K[x]G can be written as a polynomial in the p i and degp ≤ |G| for i = 1,...,t. Moreover, we can find the generators by i applying the Reynolds operator to a basis of the space of polynomials of degree less or equal than |G|, for example to the monomials. This gives the following Algorithm 4.1. In order to find a vector space basis of K[x]G, apply the k Reynolds operator to all monomials in K[x] . This yields a generating set of k K[x]G as a vector space. By linear algebra, a basis can be extracted from it. k To find a generating set for the full invariant ring, apply the previous step for k = 1,2,...,|G|. This algorithm has been performed successfully using the computer al- gebra software SINGULAR [18]. However, in the case where the order of G is large and the representation of G has a high degree, computations can be difficult, due to the high dimension of K[x] . One way to determine some in- k variant polynomialswith quiteefficientcomputations isto usethe irreducible representations of the group, as we are now going to discuss. Let ρ : G −→ GL(V) be a representation of V, and let A be a basis of V, with dimV = n. Using Maschke’s theorem, we know that there exist irre- ducible representations (defined over the algebraic closure of K) ρ : G −→ i GL(V ), i = 1...r, with V = V ⊕...⊕V , such that ρ = ρ ⊕...⊕ρ . Given i 1 r 1 r a basis B of V , we have i i [ρ (g)] ... 0 1 B1 [ρ(g)]B = ... ... ... 0 ... [ρ (g)] r Br where B = B ∪ ... ∪ B and [ρ(g)] is the matrix of ρ(g) in the basis B. 1 r B Moreover, if P is the matrix representing the change of basis from A to B, we have [ρ(g)] = P−1[ρ(g)] P ∀g ∈ G. B A We denote by x(cid:48) = (η(1),...,η(1) ,η(2),...,η(2) ,...,η(r),...η(r) ) and x = (x ,...,x ) the co1ordinatedsimoVf1 a1vector indimRVn2 in the1 basesdBimVarnd A, 1 n 8 respectively, so that x = Px(cid:48). We consider the two rings of invariants K[x]G and K[x(cid:48)]G. More precisely, p ∈ K[x]G if p([ρ(g)] x) = p(x) ∀g ∈ G A while f ∈ K[x(cid:48)]G if f([ρ(g)] x(cid:48)) = f(x(cid:48)) ∀g ∈ G. B The two rings are clearly isomorphic, and an isomorphism is the following ϕ : K[x(cid:48)]G −→ K[x]G f(x(cid:48)) (cid:55)−→ f¯(x) = f(P−1x). Inthisway, itispossibletofindpolynomialsoftheringK[x]G byfirstfinding polynomials of the ring K[x(cid:48)]G and then using the function ϕ. Let B = {v(i)} , j = 1...dimV . We have, for every vector x(cid:48) ∈ Rn, i j j i since V = V ⊕...⊕V 1 r x(cid:48) = x(cid:48) +...+x(cid:48) 1 r with d(cid:88)imVi x(cid:48) = η(i)v(i) ≡ (η(i),...,η(i) ). i j j 1 dimVi j=1 Since [ρ] = [ρ ] ⊕...⊕[ρ ] , invariance of f can be rewritten as B 1 B1 r Br f([ρ ] x(cid:48) +...+[ρ ] x(cid:48)) = f(x(cid:48) +...+x(cid:48)). 1 B1 1 r Br r 1 r Therefore, we have r (cid:77) K[x(cid:48)]G ⊆ K[x(cid:48)]G. i i=1 In other words, if p(i)(x(cid:48)) = p(i)(η(i),...,η(i) ) ∈ K[x(cid:48)]G, i.e. i 1 dimVi i p(i)([ρ (g)] x(cid:48)) = p(i)(x(cid:48)) ∀g ∈ G, i Bi i i then the polynomial r (cid:88) p(i)(x(cid:48)) i i=1 9 is an invariant polynomial in K[x(cid:48)]G. This is a sufficient condition for the invariance which is useful for computations. Let us consider K[x(cid:48)]G, k > 0. Using Algorithm 4.1, it is possible to find i k a basis {p(i)} of K[x(cid:48)]G and write kj j i k (cid:88) p(i)(x(cid:48)) = c p(i)(x(cid:48)), c ∈ K. k i j kj i j j Since dimV < dimV the computations are easier (sometimes in a consider- i able way). We then consider the polynomial (cid:32) (cid:33) d r (cid:88) (cid:88) f(x(cid:48)) = p(i)(x(cid:48)) k i k=1 i=1 and finally using the function ϕ we find ϕ(f) = f¯(x) = f(P−1x) ∈ K[x]G. We organize the above results in the following algorithm Algorithm 4.2. Let V be a vector space with dimV = n, and A a basis of V. Let ρ : G −→ GL(V) be a representation of a finite group G, and K[x ,...,x ]G = K[x]G the ring of the polynomials invariant under the ac- 1 n tion of G. Perform the following steps: 1. find the decomposition ρ ⊕ ... ⊕ ρ using the character table of the 1 r group G and the formula (see [19]) χ = χ +...+χ ; (3) ρ ρ1 ρr 2. find explicitly the decomposition V = V ⊕...⊕V finding a basis B for 1 r i each V using the projection operator (see [19]) P : V −→ V defined by i i i (cid:88) P = χ (g)ρ(g); i ρi g∈G 3. find [ρ ] , i = 1,...,r and, letting B = B ∪...∪B , find the matrix i Bi 1 r P of the change of basis from A to B such that [ρ] = P−1[ρ] P; B A 4. having fixed the degree d > 0 and denoting by x(cid:48) the coordinates of a vector in the basis B, find the polynomials p(i)(x(cid:48)) forming a basis of kj i K[x(cid:48)]G, using Algorithm 4.1, for i = 1,...r and k = 1,...,d; i k 10