How the Serpin α₁-Proteinase Inhibitor Folds

Background: Functional serpins uniquely adopt a metastable conformation by an unknown folding pathway.

Results: Ability of constituent α₁-proteinase inhibitor (α₁PI) peptides to associate reveals the order of folding.

Conclusion: Metastability results from the inability of sheet A strand 4 to efficiently insert before completion of C-terminal sheet B.

Significance: Pathway helps explains the nature of the polymerogenic intermediate of the Z variant of α₁PI.

Abstract

Serpins are remarkable and unique proteins in being able to spontaneously fold into a metastable conformation without the aid of a chaperone or prodomain. This metastable conformation is essential for inhibition of proteinases, so that massive serpin conformational change, driven by the favorable energetics of relaxation of the metastable conformation to the more stable one, can kinetically trap the proteinase-serpin acylenzyme intermediate. Failure to direct folding to the metastable conformation would lead to inactive, latent serpin. How serpins fold into such a metastable state is unknown. Using the ability of component peptides from the serpin α₁PI to associate, we have now elucidated the pathway by which this serpin efficiently folds into its metastable state. In addition we have established the likely structure of the polymerogenic intermediate of the Z variant of α₁PI.

Introduction

Serpins, of which α₁-proteinase inhibitor (α₁PI,² also frequently called α₁-antitrypsin) is a very abundant and well studied example, are large protein proteinase inhibitors that possess a common core fold composed of three β-sheets and eight or nine α-helices (see Fig. 1A). They can adopt two major conformations, native and latent. In the native state, the reactive center loop (RCL), which contains the primary proteinase recognition site, is exposed and poised to interact with its target. Using a nomenclature in which strands of a given β-sheet X are represented as s#X, the exposed RCL is linked between s5A on the N-terminal end and s1C on the C-terminal end (see Fig. 1A). In the more stable latent state the RCL is inserted into β-sheet A as strand 4 (s4A), and s1C is absent from β-sheet C and exposed (1) (see Figs. 1A and B). In a unique departure from the protein-folding paradigm, the native, functional state of serpins is metastable, whereas the latent state is the most stable, although inactive, state (2). Although quantitative data on the difference in ΔG between these states is lacking, it is clearly manifested by a large difference in denaturation temperature between the two states. For the serpin PAI-1 the difference in T_m for unfolding of the two conformations is about 17 °C (3), whereas for α₁PI it is of similar, although less well defined, magnitude (4). Surprisingly, neither chaperone nor prodomain is required for efficient folding to the metastable state. Consequently, serpins without disulfides can be denatured from either native or latent states and refolded to the metastable conformation.

FIGURE 1.

Serpin conformations. Native α₁PI (panel A) and latent α₁PI (panel B) show the key residues and secondary structural elements mentioned in the text. The color scheme for s5A (green), s4A(RCL) (yellow), the remainder of sheet A (red), and s1C/s4B/s5B (cyan) is maintained in the folding pathway shown in Fig. 8.

Although the question of how proteins fold is of wide general interest, what makes the question of serpin folding in particular a fascinating one is that such metastability is an absolute requirement for the serpin inhibition mechanism. Thus serpins act as suicide substrate inhibitors of serine and cysteine proteinases by using a massive conformational change to translocate the proteinase >70 Å (5) and kinetically trap the acyl-enzyme intermediate by distortion of the proteinase active site (6, 7). In this conformational change the exposed RCL (now cleaved by the attacking but still attached proteinase) inserts into β-sheet A in the same way as in formation of the latent conformation. The energy required for these changes derives from the metastability of the native conformation, which is thus crucial for function (8, 9). As yet, however, the pathway that directs folding to the metastable rather than latent conformation (i.e. what determines that the RCL is exposed rather than immediately incorporated into β-sheet A as its central fourth strand) has not been elucidated.

A separate, but important aspect of serpin folding is that there is a major negative consequence of metastability in that it can lead to pathology when mutations affect the folding pathway. Such is the case for the Z variant of α₁PI, where intracellular polymerization leads to liver disease, and emphysema results from reduction in circulating levels of functional α₁PI (10), which is the principal inhibitor of neutrophil elastase. Here we have used folding of constituent peptide fragments to identify for the first time the pathway by which the serpin α₁PI, folds into the metastable state. By implication we consider that this is likely to be true for all serpins. In addition we have established the likely structure of the polymerogenic folding intermediate of the Z variant, which has implications for how polymerization is initiated.

EXPERIMENTAL PROCEDURES

Protein Expression and Purification

Full-length His₆-tagged α₁PI multi-8, cloned in pQE-30 (11, 12), was expressed as a soluble protein at 25 °C and purified on Ni²⁺-Sepharose (GE Healthcare) using the manufacturer's protocol. The eluted α₁PI was dialyzed against 20 mm Tris-HCl, pH 8.0, 50 mm NaCl overnight and further purified by ion exchange chromatography on a Q-Sepharose HP column with a gradient of 50–1000 mm NaCl in 20 mm Tris-HCl, pH 8.0. The A350C and Z (E342K) variants were generated using the QuikChange protocol (Stratagene) on the α₁PI multi-8 background. The A350C variant was expressed and purified as for WT α₁PI. The Z variant was found exclusively in inclusion bodies and needed to be solubilized in 6 m urea before purification. Further purification was from the denatured state using Ni²⁺-Sepharose using the manufacturer's protocol followed by ion-exchange chromatography on Q-Sepharose in 6 m urea.

DNA encoding the fragments 1–323, 324–394, 1–344, and Z 1–344 were generated from the α₁PI multi-8 plasmid by PCR with Expand (Roche Applied Science) using primers designed to introduce cysteines at positions 323, 324, or 344 (although the latter cysteine was always subsequently blocked with iodoacetamide (IAA)). The 1–323, 1–344, and Z 1–344 fragments, including an N-terminal His₆ tag, were cloned in pQE-30, and the 324–394 fragment was cloned in pQE-60 (both Qiagen). All fragments were expressed in SG13009 cells (Qiagen) in 2YT medium and isolated from inclusion bodies. All buffers for purification of the fragments were degassed, and β-mercaptoethanol was added (0.1% volume) fresh. The 1–323, 1–344, and Z 1–344 fragments were purified by Co²⁺-affinity chromatography (TALON, Clontech), equilibrated in and loaded using 50 mm NaP, pH 7.4, 300 mm NaCl, and 6 m urea, and eluted with 50 mm NaOAc, pH 5.0, 300 mm NaCl, 6 m urea. The eluted fragment was further purified by ion exchange chromatography on Q-Sepharose HP (20 mm Tris-HCl, pH 8.0, 0–1000 mm NaCl, 6 m urea). The 324–394 fragment was dissolved in 20 mm Tris-HCl, pH 8.0, 6 m urea and passed through a Q-Sepharose FF cartridge (GE Healthcare). The flow-through containing the fragment was dialyzed overnight against 20 mm MES, pH 6.0, 6 m urea. Final purification was on a SP-Sepharose HP column (GE Healthcare) eluted with a 0–500 mm NaCl gradient in 20 mm MES, pH 6.0, 6 m urea. For NMR experiments, uniformly ¹⁵N-labeled α₁PI, 1–323, and 324–394 were expressed in minimal medium containing 1 g/liter ¹⁵NH₄Cl.

C26 (residues 369–394) was synthesized (Genscript) with an extra C-terminal Cys residue for labeling. C36 (residues 359–394 with an extra C-terminal Cys residue and an N-terminal His tag) was cloned by PCR from α₁PI multi-8, and the fragment was inserted in pQE30 (Qiagen), modified to contain a TEV cleavage site. C36 was expressed in SG13009 cells (Qiagen) in 2YT medium and grown to A₆₀₀ = 0.6–1.0 at 37 °C, and cells were harvested 4–5 h after induction with 1 mm isopropyl 1-thio-β-d-galactopyranoside. C36 was purified from inclusion bodies under denaturing conditions by Ni²⁺-nitrilotriacetic acid chromatography followed by Q-HP ion exchange chromatography in 6 m urea, 20 mm Tris-HCl, pH 8.0, eluting with a 0–1 m NaCl gradient.

For fluorescence studies, including native PAGE and quenching experiments, single-cysteine-containing species (C26, C36, and A350C variant of α₁PI) were labeled with 5-iodoacetamidofluorescein using a 10-fold molar excess of fluorophore over peptide after incubation of the protein with DTT at a 1:1 ratio. Fluorescent fractions of labeled C26 and C36 were purified by reverse phase HPLC and checked by MS. This gave peptides with 1:1 fluorescein to peptide stoichiometries. To obtain the fluorescently labeled fragment 1–354 with the label at Cys-350, the iodoacetamidofluorescein-labeled A350C α₁PI was cleaved at position 354 in the RCL with papain (1:100, enzyme:α₁PI) molar ratio for1 h at 37 °C. Papain activity was inhibited with IAA, and cleaved α₁PI was denatured by adding guanidine HCl to a final concentration of 8 m and incubating the protein at 42 °C for an hour. After dialysis against 20 mm Tris-HCl, pH 8.0, 6 m urea, the peptides were separated by Q-Sepharose HP chromatography as described for the 1–323 fragment. This gave a fragment, 1–354, with 0.84:1 fluorescein:protein label at Cys-350. For quenching studies on the species obtained by associating C36 with Cys-350-labeled 1–354, the folded two-chain species was purified from the folding mixture before use to ensure that a homogeneous species was used in the experiment.

Folding Procedure

Small scale protein refolding assays were performed by placing a small droplet of denatured α₁PI fragment(s) in 6 m guanidine HCl at the bottom of a microcentrifuge tube and quickly diluting the protein with PBS. Final α₁PI concentration was 210 nm (1–323, 1–344, and full-length fragments) or 630 nm (C26 and C36). For peptide binding studies, the mixture was left on ice overnight. Refolding experiments involving full-length α₁PI and α₁PI-Z were performed at 37 °C. Large scale refolding was performed by slowly dripping denatured α₁PI (1–323, 1–323 + 324–394, 1–354, or 1–354 + 359–394) into cold PBS with rapid stirring. After incubation overnight, the protein was concentrated using a Ni²⁺-nitrilotriacetic acid cartridge (Qiagen). Two-chain molecules were further purified on a Q-Sepharose HP column eluted with a 50–1000 mm NaCl gradient in 20 mm Tris-HCl, pH 8.0.

CD and NMR Spectroscopy

CD measurements were performed on a Jasco J-710 in 50 mm sodium phosphate, pH 7.4, at 23 °C. The proteins were diluted to a final concentration of 2 μm, and spectra were recorded between 300 and 190 nm with 5 samplings in 2-mm quartz cells. Spectra are shown normalized to equal concentrations. NMR ¹H, ¹⁵N heteronuclear single quantum correlation spectra were recorded on a 900 MHz Bruker US² equipped with a cryoprobe. Spectra were recorded in 20 mm sodium phosphate, pH 7.4, 50 mm NaCl supplemented with 10% (vol) D₂O. Sample concentrations were between 100 μm (for ¹⁵N WT α₁PI) and 20 μm (for ¹⁵N 1–323).

Thermal Denaturation

To measure the thermal stability of the product obtained from associating C36 with 1–354, the tryptophan fluorescence was followed at 335 nm with excitation at 280 nm. Intrinsic fluorescence was recorded as each sample (1 μm) was heated in 5 °C increments from 30–80 °C.

Fluorescence Spectroscopy

The environment of the reactive center loop was assessed by measuring the quenchability of a fluorescein label at the Pro-9 (A350C) position. Fluorescein fluorescence at 515 nm was recorded at varying KI concentrations (0–200 mm) with excitation at 494 nm.

Kinetic Assays

50 μl of α₁PI samples were withdrawn from the refolding mixture at the times indicated and mixed with 1 mol eq of HNE. After incubation at RT for 2 min, the samples were mixed with 1 ml of 200 μm Ala-Ala-Pro-Val-p-nitroanaline (Sigma) in 100 mm Tris-HCl, pH 8.0, 1 mm EDTA, and 0.1% PEG 8000 and assayed by changes in absorbance at 405 nm for 180 s in a Shimadzu UV-2101PC spectrophotometer at 25 °C. Fractional activity was plotted relative to HNE activity in the absence of inhibitor.

Gel Scanning

Gels were scanned using a Bio-Rad Geldoc XR+ scanner with automatic exposure setting. Gel bands were quantified using the supplied Image Lab software using automatic background subtraction.

Size Exclusion Chromatography of Fluorescein-labeled Species

1–344 or Z 1–344 (273 pmol in 4 or 5 μl, respectively, of 6 m guanidine hydrochloride) was placed in a microcentrifuge tube and renatured by rapid dilution with 1 ml of PBS at room temp. After 10 min, 750 μl of the solution was transferred to a tube containing 3 μl of 66 μm iodoacetamidofluorescein-labeled C36. After 5 min of incubation 200 μl was analyzed on a Superdex 75 column (10 × 300 mm) run at 0.35 ml/min in 20 mm sodium phosphate buffer, pH 7.4, containing 250 mm NaCl and 0.1% PEG 3350. Elution was monitored by fluorescein fluorescence at 515 nm, with excitation at 494 nm. The fraction of total fluorescence eluting with the peak corresponding to the position of native α₁PI (run separately) is considered to represent C36 non-covalent complex with 1–344 or Z 1–344.

RESULTS AND DISCUSSION

C-terminal Region Directs Metastable Folding

Because residues that change secondary structure between the native and latent conformational states of serpins all lie in the extreme C-terminal portion (see Fig. 1), it seemed to us likely that the elements that direct folding to the metastable state would reside there. We therefore examined the ability of a polypeptide representing part of the α₁PI C-terminal region to associate with the much longer chain that lies N-terminal to it to form a functional proteinase inhibitor.

The choice of break point was after residue 323. Residues 323 and 324 are both aspartates in an exposed loop at the bottom of β-sheet A, immediately preceding s5A (Fig. 1A), and so are unlikely to be critical for folding. Both residues were changed to cysteine, and separate polypeptides 1–323 and 324–394 were expressed and purified. Refolding of the 1–323 fragment by the same protocol as used for refolding of single-chain α₁PI followed by the addition of the 323–394 fragment gave two species on SDS-PAGE run under non-reducing conditions. However, only a single, two-chain species was observed when reducing agent was added (data not shown), suggesting that the two species differed only by a disulfide between Cys-323 and Cys-324 in one form but not the other. Each of these species was then made in pure form for subsequent characterization. The disulfide-containing form could be obtained as the sole product by allowing oxidation of the two cysteines to proceed for a further day (Fig. 2A). The two-chain, non-disulfide-containing form was obtained as a single species by blocking the cysteines with IAA before refolding the two chains (Fig. 2A).

FIGURE 2.

SDS-PAGE of α₁PI under non-reducing (panel A) and reducing (panel B) conditions, with (+) or without (−) HNE. WT, wild-type α₁PI; Cov, α₁PI with disulfide linkage between separate chains of 1–323 and 324–394; TC, two-chain α₁PI formed from separate, non disulfide-linked, IAA-blocked, 1–323 and 324–394 chains. HNE⁺ represents HNE covalently linked to 324–358 present in covalent complex of HNE with α₁PI. M, molecular mass markers.

To show that both oxidized single-chain and IAA-treated two-chain species were correctly folded, we carried out inhibition studies with human neutrophil elastase (HNE) and porcine pancreatic elastase. Both α₁PI species formed covalent 1:1 SDS-stable proteinase complexes, which are a hallmark for a functional serpin inhibiting by the suicide substrate inhibition mechanism (shown for the HNE reaction in Fig. 2) and gave second order rate constants and stoichiometries of inhibition comparable to wild-type α₁PI (Table 1). Because the serpin inhibition mechanism has an absolute requirement for the correctly folded metastable conformation, these findings indicate that the two-chain serpin had the normal metastable fold. Significantly, however, the existence of a two-chain folded species suggested that folding of the smaller fragment onto the larger one did not require a covalent link between them. More importantly, this suggested that there must be a preferred folding pathway that results in association of the smaller C-terminal chain with the prefolded larger N-terminal chain.

TABLE 1.

Inhibition parameters for α₁PI species

Species	SI (HNE)	SI (PPE)	kapp PPE	kapp × SI
			m−1s−1	m−1s−1
WT α1PI	1.0	1.7	1.5 × 105	2.7 × 105
1–323:324–394 disulfide linked	1.1	1.9	1.6 × 105	3.0 × 105
1–323:324–394 two-chain IAA	1.5	2.4	1.3 × 105	3.0 × 105

Confirmation of the correct fold of the two-chain species was provided by several additional approaches. CD spectroscopy gave a far UV spectrum indistinguishable from that of wild-type α₁PI (Fig. 3A), whereas the spectrum of 1–323 had much lower ellipticity, with secondary structure estimated as ∼22% α-helix and 27% β-sheet compared with values of 31 and 32%, respectively, for wild-type α₁PI (Fig. 3). Two-dimensional NMR of two-chain α₁PI, with the ¹⁵N label in either the 323-residue N-terminal chain or the 71-residue C-terminal chain, gave a heteronuclear single quantum correlation spectrum that was well dispersed and had resonances at positions representing subsets of the spectrum of wild-type single chain uniformly ¹⁵N-labeled α₁PI so that superpositioning of spectra of the two two-chain α₁PIs, with the label alternately in the N- or C-terminal region, gave a composite spectrum (Fig. 4B) the same as for uniformly labeled single chain α₁PI (Fig. 4A). The spectrum of 1–323 alone was more like that of a molten globule, with poor dispersion in the ¹H dimension (Fig. 4C), although one that, from the CD spectrum, must contain much of the final β-sheet content. These findings are consistent with our hypothesis that the element(s) directing the metastable fold must lie C-terminal to residue 323, as a break at that position did not alter the ability of the two parts to fold together correctly.

FIGURE 3.

Evidence for folding of two-chain α₁PI. Shown are CD spectra of WT α₁PI (black), two-chain α₁PI formed from 1–323 + 324–394 (orange), disulfide-linked “two-chain” α₁PI formed from 1–323 + 324–394 (plum) and fragment 1–323 (cyan).

FIGURE 4.

NMR evidence for correct folding of two-chain α₁PI. Shown are two-dimensional ¹H,¹⁵N HSQC NMR spectra of uniformly labeled wild-type α₁PI (panel A), overlay of spectra of 1–323 + 324–394 two-chain α₁PI with the label in the 1–323 fragment (black) and of 1–323 + 324–394 two-chain α₁PI with the label in the 324–394 fragment (red) (panel B) and uniformly labeled fragment 1–323 (panel C).

Folding Pathway

s5A Inserts First

The above results suggest that 1–323 adopts much of its critical secondary structure first and only subsequently associates with the C terminus, which comprises, in order, the secondary structure elements s5A, s4A, s1C, s4B, and s5B. We then examined which of these elements associates next. To accomplish this we attempted to fold a peptide comprising the residues that form strands s1C, s4B, and s5B (C36, labeled with fluorescein at the C terminus) onto prefolded 1–323. If s1C/s4B/s5B can bind efficiently without the presence of a completed β-sheet A, the elimination of s4A and s5A from the C-terminal peptide, i.e. by using C36 (359–394) rather than the 324–394 used above, should not prevent it from associating with 1–323. We found, however, that these chains were minimally capable of associating (Fig. 5) under identical conditions to those under which the longer C-terminal peptide (324–394) readily associated with 1–323 to give functional serpin (see above). However, when a longer fragment that contained s5A (1–344) was used with C36 there was strong association of the two polypeptides, detected by fluorescence on native PAGE (Fig. 5). These findings clearly indicate that s5A must associate with the remainder of the serpin before s1C, s4B, and s5B can efficiently associate. This finding is contrary to the recent suggestion that was based on a structure of dimeric antithrombin (13). In that structure the dimer was formed by a swap of strands s4A and s5A between monomers. It was suggested there that formation of the dimer occurred from an intermediate on the normal folding pathway, which implied that s5A must be the last element of the metastable conformation to form as, by microscopic reversibility, it would be the element most easily removed. Our present results and more recent results from the same group (14) are not consistent with this.

FIGURE 5.

Strong association of fluorescein-labeled C36 with 1–344 detected by fluorescence. The gel is overexposed to show the minimal association of fluorescein-labeled C26 with 1–344 or of 1–323 with either C26 or C36.

s1C inserts Next Followed by s4B and s5B

We next sought to determine whether there was further sequential association within the remaining C-terminal elements. We used the same approach of examining the ability of fluorescently labeled C-terminal peptides to associate with 1–344 but used the smaller 26-residue fragment (C26) comprising only s4B and s5B but lacking s1C. Somewhat surprisingly, we found only poor association under the same conditions that C36 associated well (Fig. 5). An estimate of the relative amounts of C-terminal peptide associated with 1–344 gave a preference for C36 containing s1C over C26 of about 8–10:1, as judged from the relative intensities of the fluorescent bands of associated complex. This might result from the location of s1C within α₁PI. s1C is the outermost strand of β-sheet C and ends close to where the long s4B/s5B β-hairpin must insert into the hydrophobic center of β-sheet B to complete the sheet. Association of s1C before insertion of the s4B/s5B hairpin might optimally position the hairpin for insertion. However, it should be noted that the discrimination we observe between C26 and C36 is not absolute. This supports the role of s1C as promoting association without being an absolute requirement. This may, however, explain the many examples of lowered levels of secreted serpin associated with mutations in s1C (15). As a final experiment, which reinforced the above conclusion of the need for s5A insertion before C-terminal association, we examined the ability of C26 to associate with 1–323. As expected, there was negligible binding (Fig. 5).

s4A Inserts Last

The above folding pathway of s5A insertion before association of the unit s1C/s4B/s5B leaves the RCL (s4A) exposed. If it were to insert into β-sheet A efficiently before the C-terminal unit could associate, the resulting serpin would be in the noninhibitory latent conformation (Fig. 1B). Because the active metastable conformation is the normal end point, this implies that insertion of s4A into sheet A requires the presence of s1C/s4B/s5B in the structure so that at the time that s4A might favorably associate with sheet A, it is already constrained at both ends (by s5A and s1C) and is thus prevented from doing so. To test this, we examined the ability of 359–394 (C36) to associate with a longer N-terminal fragment that contained not only s5A but most of s4A (1–354) (see Fig. 1A). These two chains spontaneously associated to give a protein with a CD spectrum identical to that of cleaved α₁PI (i.e. a species with s4A inserted into sheet A) and distinct from that of native α₁PI (Fig. 6). The species also migrated on native PAGE the same as for cleaved α₁PI and had greatly enhanced thermal stability compared with native α₁PI (data not shown), suggesting that it had adopted the cleaved conformation, in which s4A inserted in sheet A to give a six-stranded sheet. In contrast, the CD spectrum of 1–354 on its own was similar to that of 1–323 and to that of 1–323 + 359–394 (see Figs. 3 and 6), suggesting that insertion of the RCL (s4A) into β-sheet A has not occurred and that it must require the prior association of s1C, s4B, and s5B. This is the critical step that ensures that folding is directed to the metastable state and not to the more stable latent state, as in the normal folding of the full-length α₁PI, association of s1C/s4B/s5B completes the native metastable structure with s4A (the RCL) exposed.

FIGURE 6.

CD evidence that s4A inserts only after s1C/s4B/s5B. Two-chain α₁PI formed from 1–354 + 359–394 (red) has a spectrum that resembles that of cleaved α₁PI (green) rather than of native α₁PI (black). In contrast, 1–354 (blue) alone gives a spectrum similar to that of 1–323 + 359–394 (cyan), which neither associate nor contain s4A. The spectrum of 359–394 (purple) adds little to the overall spectrum.

Separate evidence that insertion of s4A into β-sheet A can only occur after s1C/s4B/s5B (359–394) has associated came from a fluorescein reporter attached to the RCL (s4A) at position 350 (P9). From x-ray structures of wild-type α₁PI it is known that this is a highly solvent-exposed position (16, 17). Insertion of the RCL into β-sheet A with concomitant long-distance movement of the P9 residue as it becomes part of β-sheet A results in a large change in environment for the P9 side chain, as has been found in studies on complex formation with proteinases (18). Accessibility of the fluorophore was probed by KI quenchability. In labeled native α₁PI, the fluorescein at P9 was readily quenched by KI, with a K_Q of 6.0 m⁻¹ (Fig. 7). In labeled, folded 1–354 the fluorescein was even more accessible, with a K_Q of 7.1 m⁻¹ (Fig. 7), consistent with an RCL that was even more exposed than native α₁PI (as might be expected from the absence of the constraint that is imposed from attachment to s1C in full-length native α₁PI). However, after the C-terminal fragment C36 was added to labeled 1–354 (only residues P1-P4 were missing) and the resulting two-chain labeled species was purified to obtain a homogeneous protein, the fluorophore was very much less accessible to KI, with a K_Q of only 4.2 m⁻¹, suggesting a much more hindered environment, as expected if strand s4A had now inserted into β-sheet A (note that the P9 residue would be on the outer face of the sheet and so still be quenchable by KI). Consistent with this species having s4A inserted, the quenching constant for fluorescein at the P9 position of cleaved α₁PI was nearly identical (4.0 m⁻¹). It should be noted that this experiment, in which a cleaved-like state (s4A inserted) that must proceed through a more native-like state first (s4A exposed) is generated, is analogous to one carried out by others on the RCL-cleaved form of the serpin ovalbumin (19). In that study it was shown that the two chains of RCL-cleaved ovalbumin, after denaturation and separation, could re-associate to form the same cleaved conformation, but that this occurred via a native-like species, i.e. that the last step was insertion of s4A.

FIGURE 7.

Fluorescence evidence that s4A only associates after s1C/s4B/s5B. A Stern-Volmer plot of KI quenching of fluorescein attached to position 350 in the RCL is shown. Shown are native α₁PI (black), 1–354 (blue), cleaved α₁PI (red), and 1–354 after the addition of 359–394 (green).

Complete Pathway for α₁PI

Taken together, the above results suggest the following pathway for the folding of α₁PI into the metastable state (Fig. 8). First, the N-terminal portion up to residue 323 adopts a conformation with much secondary structure already present, as shown by the CD spectrum, although with conformational flexibility, which is likely to result from incomplete β-sheets A and B and which is manifested in an NMR spectrum that reflects conformational flux. The next segment to associate is the secondary structural element that occurs immediately after in the primary structure, namely s5A, rather than s1C/s4B/s5B (Fig. 8ii). That s1C/s4B/s5B cannot efficiently associate with the large folded N-terminal region until s5A has bound (Fig. 8iv) may well be due to the extensive hydrophobic contact interface between β-sheets A and B. Thus, if β-sheet A has not been completed by incorporation of s5A, β-sheet B would not be able to make the many hydrophobic interactions with the underside of sheet A that are presumably critical for stability. It is significant that many of the residues in this region are conserved among serpins whether inhibitory or not (20) and so suggests that what is true for the folding of α₁PI (M8 form) and ovalbumin is likely to be more generally true for all serpins.

FIGURE 8.

Folding pathway of α₁PI. i, molten globule-like conformation of 1–323 forms first; ii, incorporation of s5A (327–342) into β-sheet A; iii, full insertion of s5A (green) now allows incorporation of s4B and s5B (cyan) behind s5A and of s1C (cyan). This leaves the RCL (s4A, yellow) exposed, as the RCL can only insert into β-sheet A after s4B, s5B, and s1C have associated. RCL can insert into β-sheet A either after extraction of s1C or RCL cleavage to give either latent-like (iv) or cleaved structures (v), respectively. The requirement that s4B and s5B must associate before s4A can associate with sheet A ensures that state ii cannot proceed directly to the latent-like state (iv). Color scheme: 1–323 (gray), s5A (324–344) (green), RCL (s5A) (yellow), and s1C/s4B/s5B (359–394) (cyan).

At this point, with s5A inserted, the native 5-strand conformation of β-sheet A has been completed, whereas the polypeptide from residue 345 to the C terminus is still not associated (Fig. 8ii). This includes the RCL (s4A) and strands 1C, s4B, and s5B. Of these, it is the latter group that next associates. If s1C is absent, the hydrophobic hairpin of s4B/s5B binds much less well, suggesting either that prior association of s1C positions the hairpin correctly to fit into the opening in sheet B or that a concerted process involving all elements occurs. The resulting structure is then that of the correctly folded native, metastable form of the serpin (Fig. 8iii). Thus, the key step in ensuring that the metastable conformation is adopted is that s4A cannot associate efficiently until s1C/s4B/s5B has done so. As with the requirement for s5A to associate before s1C/s4B/s5B can do so, this is likely to result from the intimate contacts between sheets B and A, which are thus likely to help to stabilize the expanded 6-strand β-sheet that forms in cleaved α₁PI (Fig. 8v). In a timely publication from another group, support for this folding pathway is provided by an x-ray structure of a trimeric form of α₁PI in which the elements s1C/s4B/s5B from one monomer associate with and complete the structure of a second monomer (14) in such a way that none of these units remains exposed. This is evidence of a folding intermediate in which these elements are not yet associated with the folded N-terminal region but in which s5A is associated.

Folding of Z Variant Slowed

The above folding pathway has implications for the folding and polymerogenicity of Z-variant α₁PI. This variant involves mutation of Glu-342 to Lys at the top of s5A (Fig. 8iii), which results in a normal folding pathway intermediate having a lengthened half-life with the consequence that, although the intermediate can still successfully proceed to functional metastable serpin, it can also lead to polymers from abnormal buildup of the concentration of intermediate (21). In homozygotes, such polymers cause liver disease when formed within the hepatocyte (22). Based on our proposed folding pathway, the Z mutation might be expected to make it harder for s5A to fully insert, as another positively charged residue (Lys-290) is adjacent to Lys-342 at the top of s6A.

To examine this we incubated fluorescein-labeled C36 for 5 min with peptides 1–344 or 1–344 containing the Z mutation and then quantitated the amount of C36 that associated with the 1–344 species by separating peptide from protein by size exclusion chromatography. The first-eluting fluorescein-containing peak, which was confirmed to also contain 1–344 by SDS-PAGE (not shown), was more than twice as intense for the WT 1–344 than for the Z-variant form (integrated intensities of 11.5 ± 2.3% and 5.2 ± 2.2% of total peptide, respectively, from four separate runs on each species) (Fig. 9). Non-associated fluorescein-labeled C36 was well separated and eluted much later (Fig. 9). In addition, a recent study using single tryptophan variants of α₁PI demonstrated a structural difference between wild-type and Z variant α₁PIs in the vicinity of the mutation at the top of the A sheet, suggestive of greater solvent accessibility in this region (23), perhaps resulting from a widened opening at the top of the sheet.

FIGURE 9.

Ability of fluorescein-labeled C36 peptide to associate with WT 1–344 or Z-variant 1–344. Shown are FPLC Superdex 75 elution profiles, monitored by fluorescein fluorescence, of fluorescein-labeled C36 incubated for 5 min on its own (A), with WT 1–344 peptide (B), or with Z-mutation-containing 1–344 (C). The elution position of folded α₁PI, run separately, is indicated by an arrow. The slightly earlier elution of the non-covalent complex compared with native α₁PI is likely due to the extra His tag and linker in the complex.

To further characterize this incompletely folded intermediate, we followed the folding of Z α₁PI by both native gel and activity measurements. Wild-type α₁PI folded rapidly and gave a single band on native PAGE within the first time point (5 min) (Fig. 10A). Separate kinetic measurements of the attainment of activity as an elastase inhibitor and, thus, of correct folding were consistent with this (Fig. 10B). In contrast, Z α₁PI was much slower to achieve maximum activity (Fig. 10B), which correlated with the initial formation of a slower-moving band on native PAGE that slowly converted to a faster-moving, presumably native species (Fig. 10A). Fitting of the kinetic data to a simple exponential decay process gave half-lives of 0.7 min for wild-type and 4.2 min for the Z variant. Taken together with our above finding of the poorer association of C36 with the 1–344 containing the Z mutation, this suggests that the Z-variant intermediate contains an incompletely inserted s5A, as we have shown above that prior association of s5A is necessary before C36 can efficiently associate. This is exactly as predicted from our folding pathway both in terms of the accumulation of a not yet active intermediate (s1C/s4B/s5B not associated) and the final attainment of similar folding efficiency as wild type rather than the accumulation of inactive latent conformation (s4A cannot associate until s1C/s4B/s5B has done so). In addition, it explains an otherwise odd finding in a recent study on the accessibility of α₁PI residues in an unfolding intermediate that may correspond to the present folding intermediate. There, residue 381, otherwise buried and at the end of s5B, became fully accessible in the unfolding intermediate (24).

FIGURE 10.

Kinetics of folding of WT and Z variant α₁PI followed by native PAGE (panel A) and activity measurements against HNE (panel B). Sufficient α₁PI was used in panel B to give complete inhibition of the HNE if refolding were 100% successful and the stoichiometry of inhibition (number of mol of serpin required to inactivate 1 mol of proteinase) in each case were ∼1. For both, the folding efficiency was ∼75%, because for Z α₁PI the SI is closer to 2 (25).

It has been shown that formation of polymers of the Z variant occur from a retarded normal folding intermediate (21). Our above findings of a rapidly formed, but not yet active intermediate for the Z variant, is consistent with it being the same polymerogenic intermediate. In light of our proposed folding pathway, such an intermediate is expected to have the structure depicted in Fig. 8ii although with the very C-terminal residue(s) of s5A not yet inserted as a result of the Z mutation (and hence accounting for its longer half-life). The implication with regard to the nature of Z-polymers formed in the hepatocyte during folding is that they are initiated by association of the exposed s1C/s4B/s5B of one monomer with the incomplete β-sheet B of another. Significantly this is very similar to what has recently been found in the x-ray structure of a heat-induced trimeric form of α₁PI (14).