New insight into serpin polymerization and aggregation

Abstract

We recently solved the crystallographic structure of a dimeric form of the serpin antithrombin which has fundamentally changed the way we think about serpin polymerization. Like for other diseases that have protein deposition as a hallmark, the serpinopathies are associated with discrete inter-protomer linkage followed by subsequent association into larger fibrils and aggregates. Polymerization of the serpins is an off-pathway event that occurs during folding in the endoplasmic reticulum. Our structure reveals the nature of the polymerogenic folding intermediate, the reason that the inter-protomer linkage is hyperstable, and suggests a mechanism of lateral association of polymers into soluble fibrils and insoluble aggregates. While the basis of cellular toxicity is still unclear, novel therapeutic approaches targeting the folding intermediate or the lateral association event are now conceivable.

The serpins are serine protease inhibitors that utilize a unique and well characterized β-sheet expansion event as a necessary part of their mechanism.1–3 This conformational/topological change from a five-stranded (N in Fig. 1) to a six-stranded β-sheet A (L in Fig. 1) results in a doubling of the protein's stability, and is driven by a free-energy term of around −32 kcal/mol.4 Thus, the native form of serpins is metastable in defiance of the Anfinsen principle,5 requiring a folding pathway that kinetically traps the five-stranded state. This unusual protein folding requirement allows for an off-pathway event known as polymerization, where one protomer completes the A sheet of another.6 For secreted serpins, this occurs in the endoplasmic reticulum (ER) where polymers are seen to accumulate as insoluble proteinaceous inclusions. Polymerization has been described for several serpins and is always associated with a loss of functional levels due to accumulation within the cell, and occasionally it is associated with the death of the secretory cells through a poorly understood mechanism.7–9 The best examples of the latter are provided by the Z variant of α₁-antitrypsin (α₁AT) leading to liver disease and the Syracuse variant of neuroserpin that causes early onset dementia.

We recently solved a crystal structure of a self-terminating (closed) serpin dimer that revealed a large domain swap including the fourth and fifth strands of β-sheet A.10 We then modelled an open polymerization competent polymer based on the structure (P in Fig. 1) that explained their facile propagation, the hyperstability and flexibility of the inter-protomer linkage, and also suggested a structure for the polymerogenic folding intermediate (M* in Fig. 1). We proposed that the final step in folding to the native state is the insertion of strand 5 into β-sheet A and the association of the coiled linker domain to the ‘bottom’ of the molecule. This event would leave the fourth strand (the reactive centre loop) accessible to serve as bait for proteolytic attack, necessary for the functioning of the serpin mechanism. While many details are yet to be confirmed, the position of certain polymerogenic mutations on and underlying strand 5A supports the proposal.10

One unexpected implication of our model is the requisite unfolding and exposure of helix I and the following coiled region in linear serpin polymers. Exposure of this ‘linker region’ was verified in linear polymers of serpins antithrombin and α₁AT through limited proteolysis and fluorescence studies, and explains the observation that polymers are hydrophobic and exhibit an increased propensity towards aggregation. Unglycosylated serpins typically aggregate when polymerized in vitro (with heat or low concentrations of chaotrophes), even at vanishingly low concentrations, whereas high concentrations are required to observe aggregation of glycosylated serpin polymers. We hypothesized that aggregation/precipitation occurs via lateral association of linear polymers, either through specific β-strand linkage or non-specific hydrophobic interactions involving the linker region. Sequence analysis of helix I suggest that it is a ‘frustrated’ β-strand11 for several serpins including α₁AT, supporting the idea that aggregates of serpin polymers form through an extended β-sheet mechanism akin to other ‘conformational diseases.’12

While there are clear parallels between the ‘serpinopathies’ and conformational diseases such as Alzheimer, Huntington and the prion encephalopathies (e.g., ordered intermolecular linkage, β-sheet expansion, cell death, dementia, accumulation of insoluble aggregates, domain-swapping),12,13 the detailed molecular mechanism revealed by our crystal structure is unique to the serpins. Domain swaps in other proteins are generally characterized by normal activity and stability, and may not play a role in the secondary association event that leads to the toxic species.14,15 For serpins the domain swap leads to hyperstability and the exposure of hydrophobic regions not seen in the monomeric state. Another key difference is the manner of cellular toxicity and the nature of the toxic species. It is becoming clear that for Alzheimer, Huntington and other conformational diseases the toxic fragments are likely to be the soluble (proto)-fibrils, not the insoluble aggregates (inclusions).16,17 Serpin polymerization generally leads to disease through loss of secretion of the active species, and only in two special cases is it through gain-of-function cellular toxicity, and although the toxic mechanisms are incompletely resolved, they appear to involve the insoluble aggregates.

The most common cause of cirrhosis among children is the homozygous Z mutation in α₁AT.18 Antitrypsin is expressed at high levels by hepatocytes (1.3–3.5 g/l in blood plasma) and its expression can increase in response to infection and other stimuli.19 However, only about one-third of the homozygous carriers ever manifest liver disease18 and it never occurs in carriers of a single Z allele, indicating that hepatocytes are generally well equipped to deal with the mutant protein. Soluble Z α₁AT in the ER binds to chaperones and is subsequently targeted to the ERAD pathway for clearance by the proteosome, and insoluble polymers and aggregates are thought to activate autophagy for degradation in the lysosomes.20,21 In such a model, accumulation of polymers and cellular toxicity only occur when the proteosomal and autophagic pathways have been saturated by the high level of expression of mutant α₁AT in the liver.21 In contrast, neuroserpin is expressed at low levels in neurons and mutation leads to polymerization and dementia in an autosomal dominant fashion.22,23 However, cell death and disease are still associated with accumulation of inclusion bodies within the ER, and the toxic mechanisms are likely to be similar.9

In summary, we have elucidated a novel mechanism of serpin polymerization that involves a hyperstable domain swap of a folding intermediate. Formation of linear polymers exposes hydrophobic regions that mediate lateral polymer association and eventually leads to intra ER accretion and cellular toxicity. Our proposal suggests new avenues for the rational design of compounds to combat the diseases caused by serpin polymerization, either through targeting the folding intermediate or the lateral association of soluble polymers.

Figures and Tables

Figure 1 Serpin folding and polymerization. The pathway of serpin folding proceeds from the unfolded state (U) to the native state (N) via a stable intermediate (M*). The native conformation is the only active state, and is composed of a five-stranded A sheet (red) and a 20 residue reactive centre loop (RCL, yellow). Serpin inhibitory function requires the native conformation to be a kinetically trapped metastable state. Completion of sheet A by incorporation of the RCL as strand 4, to form the latent (L) state, results in the doubling of the serpin's thermodynamic stability (the six strands are labelled on L). Folding and unfolding of native serpins is known to proceed via a stable intermediate denoted M*, which also corresponds to the polymerogenic form.24–26 The key feature of the M* state is that strand 5 is not yet incorporated into sheet A, and can thus insert in an intermolecular fashion to form off-pathway polymers (P, each protomer of the pentamer is in a different colour). The polymers have complete A sheets and are thus hyperstable. As a consequence of polymerization, the linker region (cyan), containing helix I, remains unfolded. We hypothesize that the hydrophobic linker (indicated by the oval) is responsible for the lateral association of polymers into insoluble aggregates.