An antibody that prevents serpin polymerisation acts by inducing a novel allosteric behaviour

α₁-Antitrypsin (α₁-AT) is the most abundant circulating protease inhibitor and a member of the serpin superfamily. It is predominantly expressed by hepatocytes in the liver and acts to prevent excessive proteolytic damage by the enzyme neutrophil elastase, particularly in the lung [1]. In common with other inhibitory serpins, its native conformation is ‘metastable’, acting essentially as a kinetically stabilised folding intermediate. This is the basis of its mechanism of action [2]: following cleavage of a solvent-exposed ‘reactive centre loop’ (RCL) by the target protease, this loop inserts as an additional sixth strand into the central β-sheet A [3] (indicated in Supplementary Figure S1). In the process, the protease is translocated from one pole of the serpin to the other, culminating in distortion of its catalytic triad. The result is the formation of an essentially irreversible covalent complex [4].

Mutations are known that facilitate escape from the metastable conformation to achieve the thermodynamically preferred 6-stranded β-sheet A form of α₁-AT. This is primarily associated with the formation of functionally inactive ordered aggregates known as polymers [5]. α₁-AT is synthesised in the endoplasmic reticulum (ER) as the first step in the secretory pathway; however, polymerisation of mutant protein largely prevents further transit, resulting in accumulation in this organelle as inclusions [6]. Despite the resulting distension of the ER and changes in its luminal viscosity, α₁-AT polymers do not trigger the unfolded protein response (UPR) [7–9]. Of many mutations in the SERPINA1 gene identified in patients presenting with plasma deficiencies [10], the Z mutation (Glu342Lys) is the most common, being present in 4% of the Northern European Caucasian population [11]. In homozygotes, it leads to the retention [6] or degradation [12] of 85–90% of the synthesised α₁-AT [13]; the hyperaccumulation of protein is associated with liver cirrhosis [14], and the deficiency in circulating inhibitor predisposes an individual to emphysema [5]. Plasma replacement therapy is the only routine clinical treatment available. However, due in large part to high cost and concerns over efficacy [15], it is not universally offered.

The rational design of new therapies for α₁-AT deficiency requires an understanding of the pathway through which α₁-AT self-associates into ordered polymers. Many in vitro studies have found that induction of polymerisation using denaturant or heat proceeds via a polymerisation-prone intermediate ensemble [16–21], in a process that can be summarised as follows:

where M represents the native monomer, M* is the activated monomeric intermediate with some characteristics of the polymer [22] and P represents the terminal, hyperstable polymer. However, it is clear that polymerisation is also a diverse process. Notably, polymers formed in the presence of denaturant lack a cryptic epitope that distinguishes them from those found ex vivo in patient material [23,24]. This epitope is, in contrast, expressed by polymers induced at elevated temperatures [23]. Differences are also manifest in the character of the intermediate state, which is compact when formed by heating, and expanded and molten globule-like in the presence of denaturant [25–27]. It is unsurprising then that varied models of the α₁-AT polymer have been proposed, based on biophysical [28,29] and crystallographic [30] data. These models differ in the nature of the domain swap that forms the basis of the polymer chain, but one common feature predicted by all extant models is an expanded 6-stranded β-sheet A, with RCL residues accommodated in equivalent positions to the canonical cleaved, and therefore hyperstable, conformation (Supplementary Figure S1). This reflects the ability to block polymerisation using peptide mimetics of the RCL and the pronounced stability in denaturants [6,31]. However, the feasibility of a polymer model does not itself predicate its pathological relevance: the majority of synthesised Z α₁-AT does not polymerise but is degraded by the ER-associated degradation pathway [32], indicating that distinct populations are present in a physiological setting.

The close relationship between conformational stability and function renders the serpin scaffold amenable to fine-tuning through intermolecular interactions with a variety of ligands, such as hormones, oligosaccharides and peptides. As a result, it has been possible to identify monoclonal antibodies with non-native-binding activities, providing tools for modulation of serpin activity and stability in novel ways [33–38]. Manipulation of the polymerisation pathway represents one therapeutic strategy for these diseases of accumulation and deficiency. For example, there is a clear relationship between the kinetic stability of the native state and the tendency to polymerise [39]. However, previous attempts at using RCL peptide analogues [40–42] and small-molecule compounds [43] have resulted in molecules that have abrogated α₁-AT inhibitory activity.

We recently described a monoclonal antibody (mAb_4B12) that blocks polymerisation of Z α₁-AT, both when induced by heat in vitro and also during expression in a cellular model of disease, while retaining most inhibitory activity [44]. Here we have characterised its mechanism of action in detail. This was not a consequence of conformational selectivity: mAb_4B12 displayed similar affinity for both native and inserted conformations, and blocked polymerisation of many α₁-AT deficiency variants. Using a sparse cysteine-scanning protocol, the mAb_4B12 epitope was localised to the vicinity of residues Glu32, His43 and Leu306 on helices A and I, with the side chain of His43 being an obligate component of the binding site. This is a region not typically associated with conformational modulation, and in apparent contradiction to a core tenet of models of serpin conformational change, when bound α₁-AT exhibited an enhanced ability to incorporate RCL mimetic peptides. mAb_4B12 was also able to destabilise the covalent complex with a model protease, trypsin, indicating local conformational change propagated to the base of the molecule from the binding site. These observations suggest that the antibody is behaving as an allosteric antagonist of polymerisation, and permit us to draw conclusions regarding the importance of the timing of β-sheet A opening to the polymerisation mechanism.

General reagents were obtained from Sigma-Aldrich or Alfa Aesar, and Escherichia coli growth media from Formedium, unless otherwise specified.

Single-point mutations were designed using PrimerX (http://bioinformatics.org/primerx/) and introduced in pQE-30-based plasmids (Qiagen) containing wild-type (‘M’ allele) α₁-AT [45] on a well-characterised Cys232Ser background (α₁-AT_C232S), whose behaviour is essentially indistinguishable from the Cys232 parent [39,45–49]. Site-directed mutagenesis was undertaken using the Quikchange site-directed mutagenesis kit according to the manufacturer's protocols (Agilent). Expression plasmids were transformed into XL-1 blue cells (Novagen), and recombinant proteins were expressed and purified as described previously [49,50], with the addition of 50 mM β-mercaptoethanol to single-cysteine mutants before buffer exchange into 20 mM Tris, 100 mM NaCl (pH 7.4) and storage at −80°C.

Assays were performed as recently described [39]. Briefly, plates were coated overnight with antigen-purified rabbit polyclonal anti-α₁-AT antibody in phosphate-buffered saline (PBS), and they were washed (using 0.9% w/v sodium chloride and 0.05% v/v Tween 20), blocked with blocking buffer (PBS, 0.25% w/v bovine serum albumin, 0.05% v/v Tween 20 and 0.025% w/v sodium azide) for 2 h and washed again. Plates were incubated for 2 h at 4°C with 0.2 µg ml⁻¹ of each antigen, followed by washing and incubation with one-third serial dilutions of 2 µg ml⁻¹ mAb_4B12 in PBS. The wells were then washed before incubating with rabbit anti-mouse HRP antibody (Abcam). The reaction was developed in the dark with TMB substrate solution (Sigma-Aldrich) and stopped with 1 M H₂SO₄, and an endpoint measurement of HRP activity at 450 nm was made using a SpectraMax M2e plate reader (Molecular Devices). All steps were performed at room temperature.

Fluorophore-based thermal stability assays were performed as detailed [39,49], using a 5-fold final concentration of SYPRO Orange dye solution (Life Technologies) in PBS, with a protein concentration of 0.1 mg ml⁻¹ in 20 μl. Samples were heated from 25 to 95°C at a rate of 1°C min⁻¹ on an Applied Biosystems 7500HT real-time PCR instrument, with fluorescence in the ROX range recorded. The descending arm of the curves following the transition was truncated once the intensity dropped 20% below the maximum. The midpoint of denaturation (T_m) was determined from the optimum fit of an equation describing two-state unfolding [39,51,52]:

where F_T represents the fluorescence intensity at temperature T, T_m is the transition midpoint temperature, F_N and F_I are the fluorescence intensities of the native and intermediate states, respectively, and m_N and m_I reflect their temperature dependence around T_m. Thermodynamic parameters were not derived from these curves due to a lack of reversibility of unfolding; instead, T_m is used as a relative measure of stability under the given experimental conditions.

Bis–Tris non-denaturing PAGE using 3–12% w/v acrylamide gels (Life Technologies) was used to evaluate the state of α₁-antitrypsin polymerisation and antibody complex formation. The latter causes a decrease in α₁-AT migration due to both the increased size of the antibody–antigen complex and an elevated net isoelectric point. Gel-based polymerisation endpoint experiments were performed as detailed [49]. Proteins were loaded at 2–4 μg per lane for staining with Coomassie blue and at 0.05–2 μg for fluorescence detection using a UV transilluminator.

As described recently [22], α₁-AT_C232C (wild type) was labelled at the endogenous Cys232 site by incubation with a 10-fold molar excess of Alexa Fluor 488 maleimide or Alexa Fluor 594 maleimide (Life Technologies) overnight at 4°C. Following quenching with 5 mM cysteine, a 1 ml HiTrap Q sepharose column (GE Healthcare) was used to remove unconjugated label. The resulting protein was diluted to 0.1 mg ml⁻¹ in PBS in the presence and absence of a 2-fold molar excess of antibody with a volume of 10 μl. Polymerisation was reported by an increase in Förster resonance energy transfer (FRET) between the donor (λ_ex = 470 nm) and acceptor dye (λ_em = 605 ± 15 nm) upon heating in a Realplex⁴ quantitative PCR instrument (Eppendorf). The ascending and initial descending components of the FRET signal were well described by an empirically determined single or double sigmoidal function in Prism (Graphpad):

where F_t is the fluorescence at time t, H is the dynamic range of the curve at t_0.5,2, L is the baseline, m_h is the slope of the decay in maximal fluorescence over time, w is the fraction contribution to the signal by the curve with a half-time of t_0.5,1 and (1 − w) is the fraction contribution to the signal by the curve with a half-time of t_0.5,2. This equation permitted numerical calculation of the overall time to half-maximal fluorescence intensity using Octave (GNU Software Foundation).

Methoxypolyethylene glycol maleimide 5 kDa (PEG5K; Sigma-Aldrich) was conjugated to an introduced cysteine residue on α₁-AT by incubating at a 10-fold molar excess to 0.5 mg ml⁻¹ α₁-AT overnight at room temperature. The reaction was quenched by the addition of a 10-fold molar excess of l-cysteine with respect to PEG5K; successful conjugation was assessed by a molecular weight shift under SDS–PAGE. For mAb_4B12-binding assays, mAb_4B12 and PEGylated α₁-AT and non-PEGylated counterparts were incubated at a molar ratio of 2:1 for 30 min at room temperature in PBS. For protection assays, protein variants were first bound to mAb_4B12 for 30 min, subsequently conjugated with PEG5K maleimide for 5 min to 3 h, and the extent of labelling was determined from a change in mobility of the antigen–antibody complex by SDS–PAGE. Steric hindrance by the PEG5K molecule of mAb_4B12 binding was assessed by electrophoretic mobility using bis–Tris non-denaturing PAGE (Life Technologies) and ELISA assays.

A free radical probe — to be used as a reporter of binding or conformational change at a specific site on α₁-AT — was introduced onto single-cysteine mutants by incubation with 3-(2-iodoacetamido)-proxyl at a 20-fold excess in PBS overnight at 25°C. Following this, the reaction was quenched with 10 mM cysteine, and free label was removed by repurification using a 1 ml HiTrap Q sepharose column (GE Healthcare) as described previously [49]. Probe behaviour was assessed using continuous-wave electron paramagnetic resonance (CW-EPR) at room temperature on a Bruker eScan spectrometer at a protein concentration of 50–100 µM in a glass capillary, with an incident microwave power of 1 mW, a modulation amplitude of 0.1 mT, a scan range of 150 Gauss and spectra averaged over 500 scans (20 s per scan). First-derivative EPR spectra were normalised to the same number of spins using the double integral.

Inhibitory activity of α₁-AT variants was measured and calculated as detailed previously [49,53]. Bovine trypsin (Sigma-Aldrich) was titrated using p′-guanidinobenzoate HCl [54]. The SI (stoichiometry of inhibition) values of wild type and mutants of α₁-AT were determined by incubation of the protein for 15 min at room temperature with 0.1 μM bovine trypsin in 20 μl of protease assay buffer [20 mM Tris, 100 mM NaCl, 0.1% (w/v) PEG 8000, 10 mM CaCl₂, pH 8.0]. An aliquot of 180 μl of 200 μM Nα-benzoyl-l-arginine 4-nitroanilide substrate (Sigma-Aldrich) was added and the rate of increase in absorbance at 405 nm was recorded for 5 min using a SpectraMax M2e plate reader (Molecular Devices). Linear regression was used to extrapolate the amount of inhibitor required to abrogate enzyme activity. The association rate constant of inhibitor with enzyme (k_ass) was measured by reaction progress curves under pseudo-first-order conditions for 4 h at 25°C with a final concentration of 50–250 nM inhibitor, 1 mM substrate and 5 nM bovine trypsin.

First-order dissociation rates were determined by continuously monitoring protease dissociation [55] using a chromogenic reporter substrate. α₁-AT–trypsin complexes were formed by incubating 0.5 μM α₁-AT with 0.5 μM bovine trypsin in assay buffer [20 mM Tris, 100 mM NaCl, 0.1% (w/v) PEG 8000, 10 mM CaCl₂, pH 8.0] at room temperature for 1 h. Complexes were then diluted in the same assay buffer containing 1 mM Nα-benzoyl-l-arginine 4-nitroanilide hydrochloride substrate (Sigma-Aldrich), to a final volume of 200 μl. Complex dissociation was monitored by following the increase in absorbance at 405 nm using a SpectraMax M2e plate reader (Molecular Devices) at 25°C for 4 h. Proteinase-only controls were included to convert the rate of change in absorbance to active proteinase concentration.

It has previously been noted that a model peptide corresponding with the RCL of anti-thrombin, when incorporated into α₁-AT, causes an increase in intrinsic protein fluorescence [41]. Wild-type α₁-AT at 0.5 mg ml⁻¹ in PBS was correspondingly incubated with a 50-fold molar concentration of an 11-mer (Ac-SEAAASTAVVI-NH2) and 4-mer (Ac-FLAA-NH2) peptide, at 37°C, in a 50 µl volume. Peptide incorporation was monitored for 24–48 h with excitation at 290 nm and emission at 330 nm, on a SpectraMax Gemini plate reader (Molecular Devices) using a half-well plate with a UV-transparent base (µClear, Greiner). Samples were degassed prior to the experiment and overlaid with VaporLock (Qiagen). Peptide incorporation was confirmed at the end of the experiment by non-denaturing PAGE.

M and Z α₁-AT were denatured for a minimum of 4 h at room temperature at a concentration of 10 mg ml⁻¹ in 6 M guanidine hydrochloride. Aliquots of 20 µl were placed in the lid of a microcentrifuge tube containing 1 ml of Tris-buffered saline (TBS) with or without 0.6 mg ml⁻¹ mAb_4B12. Refolding was initiated by a rapid flip of the tube followed by several gentle inversions. Aliquots of 2.5 µl were removed at various time points between 0 and 240 min, diluted with 60 µl of TBS, and a 1.25-fold molar excess of trypsin. After a 5 min incubation, Suc-AAPV-pNa substrate at a final concentration of 400 µM was added and the rate of increase in absorbance at 405 nm was recorded. For each refolded sample, the percentage of inhibitory activity was calculated with reference to an identical non-denatured sample from the rates of substrate turnover (V):

Our recent work described the 4B12 monoclonal antibody as a suppressor of α₁-AT polymerisation. This involved the use of the monomeric form of Z α₁-AT as the antigen to generate a hybridoma library, and its use in a primary screen that accordingly selected antibodies able to recognise this form of the protein [44]. A polymerisation-enhancing antibody, identified using a similar protocol, was found to exhibit marked conformational selectivity [22]. It is therefore possible that mAb_4B12 too favours one conformation over others. To test this, the affinity of mAb_4B12 for native, cleaved, polymeric and 11-mer RCL peptide-inserted forms of α₁-AT was determined by ELISA and found to be indistinguishable (Figure 1A). Thus, mAb_4B12 recognises an epitope that is present on — and is essentially unvaried between — native, loop-inserted and polymeric forms of α₁-AT.

The Z (Glu342Lys) mutation is situated in a region termed the ‘breach’: the initial site of interaction between the inserting RCL and β-sheet A during the inhibition of a target protease. However, mutations that increase the tendency to polymerise are also found elsewhere in α₁-AT. To determine whether mAb_4B12 activity extends beyond the breach region, the antibody was assessed for its ability to prevent heat-induced polymerisation of the mild deficiency variants I (Arg39Cys), S (Glu264Val) and Queen's (Lys154Asn). These mutations localise to helix A, helix G and helix F, respectively (Supplementary Figure S1, top). An endpoint experiment was performed, in which variants were either preincubated with a 2-fold molar excess of mAb_4B12 and then heated at 55°C for 4 h, or heated at 55°C prior to mAb_4B12 addition. The well-characterised Cys232Ser mutant [39,46–49] (α₁-AT_C232S) was used as the control, and the result was evaluated by non-denaturing PAGE, in which polymerised α₁-AT and mAb_4B12-bound α₁-AT appear higher on the gel (Figure 1B).

Without preincubation with the antibody, there was a loss of the monomer form upon heating for all variants. In contrast, mAb_4B12 inhibited the formation of a higher-order polymeric population for α₁-AT_C232S, and the S and Queen's variants (Figure 1B). This effect was not evident for the I variant, despite possessing similar characteristics to S α₁-AT [56]. The appearance of a band corresponding with a non-antibody-bound I variant suggests that this mutant interfered with complex formation. While the use of fixed experimental conditions does not exclude the possibility of further differences between mutants, it could nevertheless be concluded from these data that mAb_4B12 can mitigate the consequences of mutations outside of the breach region.

mAb_4B12 was previously found to reduce — but not to eliminate — inhibitory activity against human neutrophil elastase. This suggested that its binding site was situated in a region of the molecule only indirectly associated with the inhibitory mechanism [44]. In comparison, antibodies interacting with β-sheet A of protease nexin-1 were found to almost completely abolish inhibitory activity through direct interference with the RCL insertion mechanism [37]. Variable effects on inhibitory activity have been seen in studies with serpin mutants, dependent on the identity of the target protease [45,57–59]. To investigate whether this applied to mAb_4B12 binding, the effect on the inhibition of trypsin by α₁-AT was assessed. In contrast with elastase, the efficiency of the interaction, as represented by the SI, was almost identical in the presence and absence of antibody (1.17 ± 0.02 and 1.06 ± 0.03, respectively; SE of the linear regression, n = 4; Supplementary Figure S2A). Such contrasting behaviour has been shown in other studies to result from changes in the rate of RCL insertion during inhibition [58] or in the rate of deacylation due to variation in the interaction between a protease, its RCL tether and the region of the serpin in which it comes to rest [4,45,59]. Trypsin was subsequently selected as a tool with which to evaluate the effect of mAb_4B12 on resistance to heat-induced inactivation.

The ability of mAb_4B12 to preserve α₁-AT in a monomeric form, observed in Figure 1B, was confirmed by an experiment in which fluorescently labelled α₁-AT_WT was incubated for 12 h over a range of temperatures. Analysis by non-denaturing PAGE revealed several temperatures that had induced marked sequestration into higher-order polymers in its absence, but not in its presence (Figure 1C). However, antibodies have been found that induce another serpin, PAI-1, to adopt an inactive yet monomeric latent conformation [33,34]. In the case of α₁-AT, the latent conformation is a by-product of the polymerisation pathway that can be induced by heating [60] and augmented by the presence of buffer components that disfavour intermolecular interactions [61]. If the antibody was exerting an analogous effect — permitting conformational change while sterically preventing the association of single subunits — a comparable loss of activity might be expected to occur. α₁-AT_WT was heated for 16 h at 55°C, and residual activity was determined by titration with trypsin. There was a ∼10% loss of inhibitory activity in the presence of mAb_4B12 with respect to an unheated control, and a 75% decrease in its absence (Table 1 and Supplementary Figure S2B). Thus, binding substantially maintains α₁-AT in the native state, rather than permitting the adoption of a monomeric yet inactive form.

mAb_4B12 was found previously to be an effective inhibitor of polymerisation in vitro at 45°C and in mammalian cells at 37°C [44]. α₁-AT labelled with Alexa Fluor-488 and Alexa Fluor-594 fluorophores can be used as a tool to follow polymerisation as a relative increase in FRET efficiency [39]. α₁-AT_C232S (at 0.1 mg ml⁻¹) was heated at temperatures between 50 and 64°C with, or without, a 2-fold molar concentration of mAb_4B12. The resulting progress curves showed an antibody-mediated delay in the temperature at which polymerisation ensued (Supplementary Figure S2C). When the natural logarithm of half-times of polymerisation from the progress curves was plotted against the inverse absolute temperature, a linear relationship was apparent (Figure 1D). Interpolation at the mid-range temperature of 55°C revealed a decrease in the rate of polymerisation by 16.5-fold (±2.5 SEM, n = 4) due to mAb_4B12 binding. The slopes of the lines permitted calculation of the apparent free energy barrier to polymer formation under the conditions of the experiment. At 349 ± 8 kJ mol⁻¹ (SE of the linear regression, n = 4), this barrier was found to be modestly elevated in the presence of the antibody with respect to α₁-AT_C232S alone (281 ± 4 kJ mol⁻¹; SE of the linear regression, n = 4). As a result, assuming relative temperature insensitivity of these values within the supraphysiological temperature range considered, extrapolation of these relationships to 37°C gives a more pronounced (∼70-fold) decrease in the rate of polymerisation.

Thermal stability experiments, in the presence of the environment-sensitive probe SYPRO Orange, report the transition from the α₁-AT native state to a polymerisation-prone intermediate conformation along the polymerisation pathway [39]. Consequently, stabilisation of the native state will manifest as an increased midpoint of denaturation (T_m), which in a two-protein system can be deconvoluted by fitting a double two-state unfolding equation [22] (Supplementary Figure S2D). α₁-AT_C232S was incubated in the presence and absence of the antibody, with temperature increasing from 25 to 95°C at a rate of 1°C min⁻¹. These experiments revealed a +10.3 ± 0.8°C (SEM, n = 6) enhancement in the stability of the mAb-bound α₁-AT component (Figure 1E, left) against progression through the polymerisation-prone intermediate.

α₁-AT shows a predictable decrease in the rate of heat-induced polymerisation with an increase in native-state stability [39]. Therefore, temperature can be used as a probe of mechanism: deviations from this trend indicate effects on other characteristics of the polymerisation pathway [22,39,62]. A comparison of polymerisation half-life interpolated at 55°C (Figure 1D) and stability determined by thermal unfolding (Figure 1E, left) is shown in Figure 1E (right). This analysis reveals a rate of polymer formation that is broadly consistent with that expected from the enhancement of native-state stability. Taking into account the presence of the epitope in both monomer and polymer, these collective data suggest that mAb_4B12 exerts its effects primarily through changes in native-state stability. There is additionally a modest interference with the polymerisation mechanism, which is predicted to be more pronounced at lower temperatures.

The suppression of polymerisation in the context of an ability to bind both monomer and polymer suggests that mAb_4B12 acts through modulation of a dynamic process. To glean further details of the underlying mechanism, we sought to locate the site of interaction. This initially involved a sparse screen of single-cysteine substitutions distributed around α₁-AT. Ten candidate surface-exposed positions not involved in side chain-mediated interactions were selected (Figure 2A) with reference to the crystal structure of the native state [63]. Disruption of a putative binding site in the vicinity of each point mutation was then achieved by conjugation of a methoxypolyethylene glycol 5000 (PEG5K) maleimide moiety to the cysteine side chain. The ability of each PEGylated derivative to be bound by mAb_4B12 was assessed by sandwich ELISA with respect to the non-PEGylated form. Due to the steric interference imposed by this bulky moiety, disruption manifested as a reduction in maximal binding, rather than a nuanced shift in affinity. This tranche of mutants identified α₁-AT_E32C-PEG5K as exhibiting reduced binding activity with respect to α₁-AT_E32C (Figure 2A).

A thiol protection experiment was also performed, with each mutant pre-complexed to antibody followed by PEGylation. This was done to verify that the PEG5K group was not indirectly causing disruption of the epitope in these mutants. The complexes were then resolved by non-denaturing PAGE (Figure 2B, top). All mutants excluding α₁-AT_L353C and α₁-AT_E32C exhibited an upwards mobility shift characteristic of PEGylation. Parallel interference/protection experiments for these two mutants provided an inconclusive result for α₁-AT_L353C due to the lack of PEGylation in all lanes (Figure 2B, bottom). In contrast, pre-PEGylation of α₁-AT_E32C prevented a mobility shift diagnostic of mAb_4B12 binding, and conversely pre-binding of antibody again blocked PEGylation of this variant. Notably, this position is in proximity to the Arg39Cys (I) mutation that appeared to exhibit reduced binding to mAb_4B12 (Figure 1B) and lies within a region structurally conserved between native and loop-inserted conformations [44].

Based on this preliminary observation, a finer-grained cysteine scan was performed of nine residues in the vicinity of Glu32 — on helices A, C, G, H and I — with sites again selected based on the accessibility and lack of significant side chain contacts. Position 36 failed to conjugate with the PEG5K moiety and was excluded from the experiment. An ELISA-based assessment of the maximal binding of the variants by mAb_4B12 demonstrated perturbation of the interaction upon PEGylation at 306 (Figure 2C). Strikingly, the unconjugated α₁-AT_H43C mutant also demonstrated a lack of binding activity with respect to the α₁-AT_C232S control. Overall, these results (summarised in Supplementary Figure S3 and Figure 3B) support the proximity of Glu32 and Leu306 to the epitope, and imply a critical role for the histidine side chain at position 43 in binding.

The bulky PEG5K moiety had been chosen to increase the likelihood of identifying a binding site despite sparse sampling at the α₁-AT surface. In contrast, site-directed spin labelling can be used as a sensitive and direct reporter of protein–protein interactions, with minimal perturbation of the interface due to small probe size [64]. Mutants corresponding with Glu32, Leu306 and His43, and two flanking sites from the PEGylation experiments, Glu266 and Asn81, were labelled with a 3-(2-iodoacetamido)-proxyl-free radical label, and CW-EPR spectra were collected in the presence and absence of mAb_4B12 (Figure 3A). One of the key determinants of the morphology of a CW-EPR spectrum is spin probe mobility, which relates to spectral width. Here, interaction with the antibody resulted in the appearance of a distinct species, reflected by a marked increase in spectral broadness and reduction in intensity, for both α₁-AT_E32C and α₁-AT_L306C. Additional effects were seen at position 266, at which site PEG5K conjugation had not prevented binding (Figure 2C), therefore likely delineating one edge of the interface. The other tested sites did not show significant peak broadening (Figure 3A). One surprising result is the lack of change reported by the label at residue 36, given its situation within the triangle delineated by 32-43-306. However, we note that an antibody–antigen interface is neither uniform nor completely planar; it may be that occlusion by the antibody does not significantly alter its behaviour. Overall, these results confirm the involvement of positions 32 and 306 at the α₁-AT/mAb_4B12 interface. The combined results from the PEGylation and CW-EPR experiments are summarised in Figure 3B.

The stark increase in thermal stability of antibody-bound α₁-AT (Figure 1E, left) was used as a further tool to evaluate the effect of mutations on binding and the polymerisation suppression mechanism. First, the role of the His43 side chain in the interface was verified. In contrast with the α₁-AT_C232S control, the α₁-AT_H43C mutant failed to exhibit a comparable increase in T_m in the presence of mAb_4B12, consistent with the lack of interaction between these components (Table 2 and Supplementary Figure S4A). To confirm that this was due to the loss of a histidine — rather than the acquisition of a cysteine side chain — an additional α₁-AT_H43A mutant was generated. This variant displayed an identical behaviour with that of α₁-AT_H43C. Therefore, the histidine side chain at position 43 is an obligate component of the epitope.

It has been observed that constraints imposed by introduced disulphides generally result in global thermal stabilisation of α₁-AT with a consequent increase in resistance to polymerisation [39,65]. One possibility is that mAb_4B12 binding could be mimicking this effect. Of the two structural elements within the epitope — helices A and I — the latter is considered the more labile element [27]. Indeed, it has been proposed to be denatured in a polymerisation-prone unfolding intermediate [66,67], and according to one model of polymerisation, it is expected to be extricated from the molecule and fully denatured in the antitrypsin polymer [29]. To test the possibility that the antibody merely acts by preventing displacement of helix I from its native position, two disulphides were engineered into two separate molecules: α₁-AT_79-306, tethering position 306 at the C-terminus of helix I to residue 79 on helix C; and α₁-AT_298-333 that tethers the N-terminus of helix I to β-strand 5A (Figure 4, right panels). In the resulting mutants, the disulphides were present in the majority (∼90 and ∼60%, respectively) of molecules, as determined by a difference in SDS–PAGE migration between reduced and non-reduced samples.

Thermal denaturation experiments were undertaken; despite the mixed population of α₁-AT_298-333, the thermal stability curve was well described by a single two-state unfolding equation. In contrast with typical observations with disulphides [39], α₁-AT_79-306 and α₁-AT_298-333 did not demonstrate an increase in global native stability with respect to the α₁-AT_C232S control, instead exhibiting a reduction in T_m of −1.2 ± 0.6 and −3.5 ± 0.7°C, respectively (Table 3 and Supplementary Figure S4B; SEM, n = 5–6). Therefore, any difference observed in the polymerisation propensity would be ascribable to specific effects of a disulphide on the polymerisation pathway, rather than a global stabilisation effect [39]. This was investigated by performing endpoint polymerisation experiments [49], with monomeric protein heated for 4 h at temperatures ranging between 45 and 63°C, and samples were analysed by non-denaturing PAGE. The resulting profiles were found to be almost the same for the control α₁-AT_C232S protein and the disulphide mutants (Figure 4, left panels). It was noted that there was no persistence of a fraction of the monomer for α₁-AT_298-333 that might be attributable to the disulphide-bonded population. These data indicate that constraining helix I at either terminus does not stabilise the native conformation against polymerisation, strongly suggesting that this is not the basis for the activity of mAb_4B12.

The effect of introduced disulphides on the ability of mAb_4B12 to confer enhanced thermal stability was assessed. An additional mutant that prevents sheet opening at the top of β-sheet A, α₁-AT_191-339 [39], was also considered. This (>90% disulphide-bonded) mutant exhibited an increase in T_m to 71.2 ± 0.7°C (SD, n = 5) when bound to mAb_4B12 (Table 3 and Supplementary Figure S4B). In contrast, the presence of the 79–306 and 298–333 disulphides yielded a T_m increase for the mAb_4B12 complex attenuated by −6.9 and −4.5°C with respect to the antibody-bound α₁-AT_C232S control. This tentatively suggested that these constraining disulphides antagonised the ability of mAb_4B12 to stabilise the native state. In support of this, treatment with DTT did indeed cause α₁-AT_298-333 + mAb_4B12 to manifest increased stability to within −0.1°C of the α₁-AT_C232S + mAb_4B12 control.

In contrast, reduction with DTT led to a milder +1.9°C increase in the T_m of α₁-AT_79-306 + mAb_4B12, such that it remained −4.4°C lower than the control and the reduced α₁-AT_298-333 + mAb_4B12 sample. This suggested that the Gly79Cys/Leu306Cys double mutation itself contributed substantially to the limited mAb_4B12 stabilisation. As 306 is situated within the epitope, it is possible that this variant is bound with less affinity by the antibody. To check this, an ELISA was performed (Supplementary Figure S4C). Under the conditions of the ELISA, comparable EC₅₀ values were found for α₁-AT_C232S (10.8 ± 0.4 ng ml⁻¹), α₁-AT_298-333 (9.1 ± 0.7 ng ml⁻¹) and α₁-AT_191-339 (12.2 ± 0.9 ng ml⁻¹), making these variants directly comparable. In contrast, while the antibody did recognise α₁-AT_79-306, there was a modest ∼1.6-fold increase in the EC₅₀ of the interaction to 17.1 ± 1.8 ng ml⁻¹ (±SEM of the regression, n = 3). Despite rendering interpretation of the effect of α₁-AT_79-306 on mAb_4B12-induced stabilisation non-trivial, this result highlights the contribution of residues in this region to the mAb_4B12 epitope and is consistent with the limited recovery of T_m upon reduction in the 79–306 disulphide.

Taken together, the results in Table 3 and Figure 4 indicate that a reduction in the mobility of the helix I termini does not adequately account for the observed resistance to polymerisation. Rather, the mechanism of stabilisation by mAb_4B12 is antagonised by a disulphide constraint at the helix I N-terminus, but not by one in the breach region at the top of β-sheet A.

During sparse screening to delineate the epitope, some α₁-AT cysteine mutants yielded intermediate results in the protection assay when compared with α₁-AT_E32C (Figure 2B, top). Subsequent parallel binding interference/protection assays (as performed in Figure 2B, bottom) identified two sites, position 301 on helix I, and residue 296 at the junction between helix I and strand 6A, that exhibited a distinct behaviour (Figure 5A). These positions could be fully PEGylated in the absence of mAb_4B12, and the PEG5K group did not prevent binding of mAb_4B12, but when the antibody was bound first, these mutants exhibited a reduced ability to accept the PEG5K maleimide. This reduction in susceptibility to PEG5K conjugation was not intrinsic to the presence of antibody, as cysteines introduced at different locations (including residue 155, on helix F) showed a full mobility shift (Figure 5A).

In light of the observations with a tethered helix I N-terminus, the previously noted mAb_4B12 induced reduction in inhibitory activity against neutrophil elastase [44] and the effect of helix I mutations on inhibitory activity [67], these derivatisation results raise the possibility that antibody binding mediates β-sheet A behaviour through structural changes in this region. To assess this possibility directly, spin labels were used as sensitive reporters of differences (primarily mobility and polarity) in the local environment. In addition to 296 and 301, three other positions along the RCL-β-sheet A axis were chosen (Figure 5B). Upon antibody complex formation, for α₁-AT_T296C, α₁-AT_S301C and α₁-AT_V333C, there was a change reported in the environment of the spin probe, resulting in the adoption of an altered mobility regime. The effect was more marginal for α₁-AT_T339C, consistent with the lack of interference of α₁-AT_191-339 with antibody-mediated stabilisation, and was absent for α₁-AT_I360C. These results collectively suggest propagation of conformational change to the lower region of β-sheet A.

The ability of peptide mimetics of the RCL to antagonise polymerisation represented key early evidence of the involvement of the RCL in polymer formation [6,31]. While a peptide derived from one serpin can anneal with another, different rates of incorporation have been observed between the M and Z α₁-AT variants, interpreted to be the result of altered β-sheet A dynamics [41]. To explore the possibility that mAb_4B12 binding precipitated a comparable behaviour in this region, the rate of incorporation of two model peptides, an 11-mer and 4-mer, was determined from progress curves reporting intrinsic tryptophan fluorescence [41,68] (Supplementary Figure S5A). These data were satisfactorily fit by a single-phase exponential equation, and surprisingly showed a 2.5- and 1.8-fold increase in the rate of peptide incorporation in the presence of antibody with respect to the non-antibody-bound control (Figure 6A). Thus, contrary to expectations, antibody binding appears to increase β-sheet A liability.

An altered rate of association between a serpin and a protease during inhibition can provide a measure of changes in RCL/β-sheet A behaviour. This is exemplified by anti-thrombin, whose partially inserted RCL is expelled upon interaction with heparin, altering its inhibitory kinetics towards proteases in the coagulation cascade [69,70]. Similarly, it has been proposed that the Z variant of α₁-AT may also exhibit partial insertion in association with its polymerisation-prone nature [41], a change also predicated by the C-terminal polymerisation mechanism [30].

To understand whether such effects are seen in response to antibody binding, progress curves for the inhibition of trypsin under pseudo-first-order conditions were recorded. However, these showed an atypical morphology, suggestive of a significant degree of serpin–enzyme complex dissociation [55]. To explore this possibility further, complex stability experiments were performed: α₁-AT and trypsin were combined at a high concentration, and rapidly diluted to low concentrations in the presence and absence of antibody, with regain of enzyme activity followed as a function of time (Supplementary Figure S5B). The recovery of protease activity was indeed found to be stimulated in the presence of mAb_4B12, with a dissociation rate (k_diss) increased by 4.8-fold and 2.9-fold for plasma and recombinant α₁-AT, respectively (Table 4; indicated by a shift to the right in Figure 6B). This was not observed in the absence of antibody binding, using an IgG1 isotype control, or for the binding-compromised α₁-AT_H43C mutant. In addition, the Fab fragment derived from mAb_4B12 showed a comparable effect on complex dissociation, rendering it unlikely that the antibody bulk was exerting a steric influence on the final resting position of the protease. The minimal effect on SI with trypsin indicates that the progression from encounter to inhibited complex occurs with sufficient efficiency; therefore, accelerated complex dissociation is consistent with perturbation at the base of β-sheet A in the region where the inhibited protease comes to rest.

Examples exist of monoclonal antibodies that assist in folding of an antigen to a native, active conformation [71], and it is possible that this is the basis of the activity of the single-chain 4B12 analogue in cells [44]. To test this possibility, wild-type (M) and Z α₁-AT were denatured at 10 mg ml⁻¹ in 6 M guanidine and snap-refolded with a 50-fold dilution in the presence and absence of a 2-fold concentration of mAb_4B12. The inhibitory activity at various time points was determined by incubating for 5 min with a 1.25-fold molar excess of bovine trypsin. For each sample, residual activity was calculated as a proportion of that of an identical but non-denatured control. While it was not possible to derive accurate refolding rates from these data, M and Z variants with mAb_4B12, and M without, showed a trend of increasing inhibitory activity over time; in contrast Z alone did not (Supplementary Figure S5C). Correspondingly, the resulting steady-state levels of activity for the M samples (−mAb_4B12: 55 ± 2%, +mAb_4B12: 55 ± 4%; SEM of the fit, n = 3) and Z + mAb_4B12 (60 ± 5%) were similar, and higher than that of Z-mAb_4B12 (42 ± 1%; Supplementary Figure S5D).

Mutations in the SERPINA1 gene result in attenuated secretion of α₁-AT by hepatocytes, due to the formation of polymers in the ER [72] and proteasomal degradation of misfolded protein [32]. These polymers have been shown to consist of hyperstable, folded protein in an unbranched arrangement [73] and form through self-association of a polymerisation-prone monomeric species [21,22]. However, there is debate as to whether polymers arise from a near-native conformation in vivo or whether the polymerisation-prone species is a substantially unfolded intermediate on the folding pathway [29,30,44,60]. In this regard, it is notable that a single-chain variant of a monoclonal antibody that prevents polymerisation from native material in vitro is also effective during co-expression of a polymerisation-prone variant of α₁-AT in a cell model of disease [44]. Here, we have sought to characterise this antibody further to understand its mechanism of action, and the implications for the polymerisation pathway that it antagonises.

Despite the use of monomeric Z α₁-AT in the process that identified mAb_4B12, this antibody was found to suppress polymerisation for deficiency variants with mutations at different locations within the molecule (Figure 1B). Based on evidence that these variants assert their effects on the molecule via a common mechanism [26,56], this indicates that the action of the antibody is mechanism- (rather than variant-) centric. By ELISA, mAb_4B12 was demonstrated to lack conformational selectivity (Figure 1A), precluding action as a conformational ‘sink’ that would stabilise, for example, the native state over others. This is suggestive of a site of binding which is relatively invariant between the 5-stranded native and 6-stranded inserted conformations of the protein, and correspondingly that it is modulation of dynamic changes that form the likely basis of its mode of action. Consistent with this, the epitope, identified by indirect and direct measures of binding (Figures 2 and 3), was found to span helices A and I. These elements are present in a region of the protein that shows high similarity between crystal structures of the native and cleaved conformations [44]. Helix I itself, however, has been proposed to undergo plastic deformation during insertion of a cleaved RCL [74], and mutations within it have been shown to affect inhibitory activity [67]. It is notable then that mAb_4B12 activity was antagonised upon tethering of this element by an introduced disulphide linking positions 298 and 333 (Table 3). In addition, antibody binding was associated with a procession of indirect conformational effects along the helix I–β-sheet A axis (Figure 5A,B), with a concomitant increase in thermal transition temperature of ∼10°C (Figure 1E, left). These effects manifested an altered behaviour in β-sheet A, reflected by a protease-dependent perturbation of inhibitory efficiency [44], increased ability to accommodate exogenous RCL mimetic peptides (Figure 6A), perturbation of the α₁-AT–trypsin complex (Figure 6B) and altered strand 5A dynamics (Figure 5B).

It is noteworthy that the mAb_4B12 epitope is adjacent to the binding site of a monoclonal antibody that stabilises the serpin PAI-1 against inactivation, which includes the helix I-strand 5A loop [36], and one that confers substrate-like behaviour through direct interference with a covalently bound protease [35]. Antibodies that decrease PAI-1 stability have also been found that interact with regions associated with conformational transitions: the C-sheet upon strand 1C displacement [33] and helix D [34]. Others that suppress inhibitory activity have been noted to interact directly with components of the inhibitory apparatus, localising to the vicinity of helix F of PAI-1 — which is known to move during RCL insertion [38] — and the lower portion of β-sheet A that accepts the incoming RCL during inhibition by protease nexin-1 [37]. Each of these epitopes is distinct from the one presented here, and collectively they span a significant proportion of the serpin fold.

Proposed models of polymerisation differ in the degree of structural perturbation they predict, such as unravelling of the C-terminus [30] or unfolding of helix I and strand 5A [29]. Release of strand 1C [39,46] and deformation of helix F [65,75,76] are additional changes that have been shown to occur during polymerisation. All models, however, anticipate adoption of an RCL-inserted state for β-sheet A, be it a consequence of intermolecular or intramolecular interactions (Figure 1A), and it is frequently expected or implied that the breach region [74] of β-sheet A plays a significant role [6,30,41,46,77]. For example, in anti-thrombin, heparin causes this sheet to close, expelling the partially inserted RCL from the ‘breach’ [69] and increasing the stability of the molecule by ∼7°C [78]. However, there does not appear to be significant involvement of the breach in the mAb_4B12 mechanism (Table 3 and Figure 5B). Similarly, the binding of vitronectin by PAI-1 in the cleft that partitions helix E from helix F stabilises the molecule against inactivation by inducing a shift in β-sheet A towards the ‘closed’ state, evidenced by a reduced rate at which extrinsic peptide is incorporated [79]. This highlights a distinguishing feature of mAb_4B12-mediated stabilisation: paradoxically, binding renders β-sheet A more receptive to peptide incorporation (Figure 6A).

Furthermore, along the sheet lies the shutter domain, involved in the regulation of its opening and identified from the clustering of mutations that compromise molecular activity or stability [10,73,80]. The site of influence of mAb_4B12, in the distal portion of β-sheet A, flanks this region. It is proposed that mAb_4B12 acts by inducing a shift to a ‘quasi-inserted’ conformation in the lower region of β-sheet A, near the shutter, with minimal disruption of the sheet in the breach region. The native α₁-AT fold is peppered with unfavourable interactions whose resolution imbues the inserted state with a pronounced increase in thermodynamic stability [2]. Accordingly, it is proposed that the incomplete shift to an ‘open’ conformation, out-of-step with other changes that occur during polymerisation, results in a partial stabilisation of the protein by resolution of some of these unfavourable interactions (Figure 6C). Furthermore, it is suggested that this stabilisation is exerted primarily on the native state (Figure 1E), conferring enhanced kinetic stability; however, the proposed structural change would be expected to remain compatible with the polymerisation mechanism, due to the ability of the antibody to bind the polymer form.

Some observations presented permit further conclusions to be drawn with regard to the status of helix I in polymers induced in vitro and in cells. As shown here, mAb_4B12 recognises polymer and monomer equally, with the C-terminus of helix I forming a component of its epitope (Figures 1A, 2 and 3), supported by the decreased affinity for a 79–306 double mutant (Supplementary Figure S4C). In addition, no stabilisation was observed against thermal transition or polymerisation when helix I was tethered using introduced disulphides (Figure 4). These data indicate that this structural element is likely to be substantially intact in the polymer. This is of particular relevance to the ‘β-hairpin model’, which extends implications from a crystal structure of an anti-thrombin dimer to α₁-AT, based on limited proteolysis experiments and an introduced disulphide [29]. One of its core tenets is that helix I is substantially unwound. Therefore, while expectations of the relevance of this conformation to α₁-AT have been curtailed [23,30,39], at the very least these data indicate that a re-appraisal of the model, in which an intact helix I is countenanced, is required.

It has not yet been established whether polymerisation predominantly occurs from a near-native state or during folding. Many lines of evidence favour the former: there is a well described absence of the UPR, indicating in turn an absence of an unfolded protein load [7,8,32]; expression in the absence of ER chaperones results in the appearance of atypical small circular oligomeric species [81]; there is a lack of the misfolded latent conformation in ex vivo tissue [60] and the molten globule-like unfolding intermediate [27] results in a polymer that is distinct immunologically [23], and most likely therefore structurally [25]. It is noteworthy that co-expression of a single-chain variant (scFv) of mAb_4B12 results in a reduction in the intracellular polymerisation of Z α₁-AT by 60% [44]. The ability of mAb_4B12 to also suppress heat-induced polymerisation from a properly formed native conformation in vitro is suggestive of processes common to both environments. The discontinuous nature of the mAb_4B12 epitope identified here provides evidence of a monomeric antigen that is substantially folded when it is bound. The observation that mAb_4B12 can improve the yield of active Z α₁-AT to a level comparable with the wild-type variant during refolding in vitro is supportive of a folding pathway that manifests its mutation-induced differences only once this structured, near-native state is achieved; therefore, this is the form that mAb_4B12 most probably exerts its effects on in vivo.

The primary site of α₁-AT polymer formation within the ER of hepatocytes renders antibody-based molecules unlikely therapeutic reagents for α₁-AT deficiency. Nevertheless, use of mAb_4B12 as a tool has identified a region that can allosterically regulate β-sheet A dynamics. Small molecules that assert allosteric effects have been approved for use in the clinic [82,83]; one of the key desirable properties they can exhibit is greater specificity than those targeted to highly conserved elements such as active sites [84]. The data presented here suggest a means — not directly involving the RCL or β-sheet A — by which polymerisation can be suppressed. Notably, while replication of an antibody–antigen interface is beyond the ability of small molecules, a relatively small component of the mAb_4B12 epitope is associated with conformational liability. Taken together, the results presented here and our previous data obtained using a single-chain variant of mAb_4B12 [44] suggest that polymerisation proceeds in vivo from a near-native conformation, and that the state of the molecule prior to the ‘polymer decision point’ is sufficiently well structured that it represents a tractable target for small-molecule intervention.

α₁-AT, α₁-antitrypsin; CW-EPR, continuous-wave electron paramagnetic resonance; ER, endoplasmic reticulum; FRET, Förster resonance energy transfer; k_ass, second-order association rate constant; k_diss, first-order dissociation rate constant; mAb, monoclonal antibody; PBS, phosphate-buffered saline; PEG5K, methoxypolyethylene glycol maleimide 5 kDa; RCL, reactive centre (or site) loop; scFv, single-chain antibody fragment; SI, stoichiometry of inhibition; TBS, Tris-buffered saline; T_m, midpoint of thermal denaturation; UPR, unfolded protein response.

N.M.-S., J.A.I. and D.A.L. conceived the study. N.M.-S., J.A.I., A.M.J., E.S., C.W.M.K. and A.O. designed experiments. N.M.-S., J.A.I., A.M.J., M.L., S.V.F., A.S.N., J.L.W., E.S. and A.O. performed experiments. J.A.I., N.M.-S., D.A.L., A.M.J., M.L., S.V.F., A.S.N., E.S., J.L.W., C.W.M.K. and A.O. analysed/interpreted data. N.H.-C., I.H., E.M. and J.P. generated reagents. J.A.I., N.M.-S., A.M.J. and D.A.L. wrote the manuscript, which was edited and approved by all authors.

N.M.-S. was supported by a Marie Curie IEF Fellowship. A.M.J. and S.V.F. are recipients of EPSRC/GlaxoSmithKline CASE Studentships. A.S.N. is a Pomona Scholar. E.M. is supported by the Sapienza University of Rome (Scientific Research calls 2014 and 2015). I.H. is an eALTA Fellow. This work was funded in part by a grant from the Alpha-1 Foundation to J.A.I. D.A.L. is funded by the Medical Research Council (UK) (G0901786) and the National Institute for Health Research/University College London Hospitals NHS Foundation Trust Biomedical Research Centre.

The Authors declare that there are no competing interests associated with the manuscript.