Development of a small molecule that corrects misfolding and increases secretion of Z α1-antitrypsin

Severe α1-antitrypsin deficiency results from the Z allele (Glu342Lys) that causes the accumulation of homopolymers of mutant α1-antitrypsin within the endoplasmic reticulum of hepatocytes in association with liver disease. We have used a DNA-encoded chemical library to undertake a high throughput screen to identify small molecules that bind to, and stabilise Z α1-antitrypsin. The lead compound blocks Z α1-antitrypsin polymerisation in vitro, reduces intracellular polymerisation and increases the secretion of Z α1-antitrypsin three-fold in mammalian cells including an iPSC model of disease. Crystallographic and biophysical analyses demonstrate that GSK716 and related molecules bind to a cryptic binding pocket, negate the local effects of the Z mutation and stabilise the bound state against progression along the polymerization pathway. Oral dosing of transgenic mice at 100 mg/kg three times a day for 20 days increased the secretion of Z α1-antitrypsin into the plasma by 7-fold. There was no observable clearance of hepatic inclusions with respect to controls. This study provides proof-of-principle that ‘mutation ameliorating’ small molecules are a viable approach to treat protein conformational diseases.


Introduction
Alpha-1 antitrypsin deficiency affects 1 in 2000 people of Northern European descent, leading to liver and lung disease (1). Ninety-five percent of severe deficiency results from the 'Z' allele (Glu342Lys) that perturbs the folding of a1-antitrypsin resulting in the secretion of only 15% of the mature protein. The remaining protein is retained within the cell by persistent binding to molecular chaperones (2) and then either degraded via the ERAD-proteasome pathway (3)(4)(5) or folded into ordered polymers that may be cleared by autophagy (6) or accumulate within the endoplasmic reticulum (ER) of hepatocytes (7). The accumulation of polymers causes neonatal hepatitis, cirrhosis and hepatocellular carcinoma, and can sensitise the liver to damage from environmental insults such as alcohol, fat or viral hepatitis (8,9). The consequent deficiency of a1-antitrypsin within the circulation results in insufficient protection of the lungs from neutrophil elastase, leading to early onset emphysema (1).
The Z mutation lies at the head of strand 5 of b-sheet A of a1-antitrypsin. It perturbs the local environment, allowing population of an unstable intermediate that we have termed M* (10) in which b-sheet A opens and the upper part of helix F unwinds (11,12). Polymerisation from this state involves insertion of the RCL into b-sheet A. In vivo, this may be intermolecular resulting in a loop-sheet dimer which extends to form longer polymers (13), or intramolecular with a domain-swap of the C-terminal region providing the inter-subunit linkage (14). The resulting polymer is deposited within hepatocytes.
The aim of our work was to develop a small molecule corrector of Z a1-antitrypsin folding that was able to block the formation of polymers within the endoplasmic reticulum of hepatocytes and that was suitable for oral dosing as a potential treatment for a1-antitrypsin deficiency. To achieve this we needed to overcome a number of challenges: (i) the drug target is a highly mobile folding intermediate located in the endoplasmic reticulum; (ii) disparity in the size of the interface between a small molecule and the large protein-protein interaction that it is designed to block; (iii) oral dosing greatly restricts suitable chemical space; (iv) as a non-classical drug target, small molecule binders may well not be well-represented in compound screening libraries; (v) the relatively high concentration of circulating monomeric Z a1-antitrypsin sink for compound, restricting its access to the target in the hepatocyte and requiring high total blood concentrations of drug to achieve sufficient free drug concentration and target engagement in the liver.

Results
Identification of GSK716 through Encoded Library Technology screening, structure guided drug design and cellular profiling Z a1-antitrypsin is a conformationally dynamic molecule (7,15) that polymerises from a near-native conformation late in the folding pathway (16,17) and therefore represents a non-classical target for drug discovery. A cell-free assay approach to hit finding was undertaken so as not to miss compounds that bind a1-antitrypsin and block polymerisation but lack the molecular properties to cross cell membranes. This comprised of: (i) an Encoded Library Technology (ELT) screen (18) of a library comprising nominal diversity of 2×10 12 unique components to identify binders to Z a1antitrypsin and (ii) a high throughput screen (HTS) of the GSK compound collection (~1.7 million compounds) for small molecules that could block polymerisation of Z a1antitrypsin. In both screening approaches glycosylated Z a1-antitrypsin, purified from the plasma of Z a1-antitrypsin homozygotes (19), was used since this represents the disease-relevant human pathophysiological drug target that populates an intermediate on the polymerisation pathway (15,20). ELT selections were performed by incubating Z a1-antitrypsin with DNA-encoded compound libraries for 1 hour at 4°C and 37°C for 3 rounds of selection with subsequent capture of Z a1-antitrypsin using a1-antitrypsin Select Resin (GE Healthcare). A variation on this protocol using pre-immobilised Z a1-antitrypsin was also used for library selections. In the HTS assay, polymerisation of purified Z a1-antitrypsin was induced by incubation at 37°C for 72 hours in the presence of test compounds, with end-point quantification of polymers performed using the polymer-specific monoclonal antibody, 2C1 (21) in a TR-FRET-based immunoassay. A number of small molecules that could block polymerisation of Z a1antitrypsin were obtained through the HTS but none progressed beyond the early lead optimisation stage. However, a single lead series of chiral hydroxy-carboxamides (GSK425) was identified from the ELT screen that also demonstrated functional activity at blocking polymerisation in the TR-FRET immunoassay (pIC50 6.5) (Fig. 1a,   b).
Optimisation of this initial hit followed a structure-based design approach, exploiting knowledge from iterative crystal structures of small molecule ligands complexed with a1-antitrypsin. The central hydroxy carboxamide and propyl chain were found to be critical for binding to Z a1-antitrypsin and hence further medicinal chemistry development focussed on modification of the phenyl and indole heterocycle. This resulted in an ~100-fold increase in potency and the discovery of the 2-oxindole GSK716 (pIC50 8.3) (Fig. 1a, b).

GSK716 is a potent inhibitor of polymerisation in vitro and in cell models of disease
GSK716 binds to Z a1-antitrypsin with a high affinity mean pKD 8.5 ± 0.12 (n = 18) as determined by a competition binding assay with a fluorescently labelled derivative (Fig.   1c). The binding demonstrates selectivity with a 50-fold lower affinity for plasmapurified wild-type M a1-antitrypsin at mean pKD 6.8 ± 0.18 (n = 10) (Fig. 1c). The shape of the curves and native mass spectrometry (not shown) are consistent with a single high-affinity compound binding site. No binding of the fluorescent derivative to polymers of Z a1-antitrypsin was observed, indicating conformational selectivity for the monomeric protein (Fig. 1d). The rate of interaction of the compound with the target was monitored through changes in intrinsic tryptophan fluorescence (10); this property was used to determine the second-order association rate constants for GSK716 binding to Z (4.1 x 10 4 M -1 s -1 ) and M a1-antitrypsin (2.1 x 10 2 M -1 s -1 ) (Figs. 1e, f).
From the association rate constants and the affinity values, first-order dissociation rate constants were calculated and found to be of the same order of magnitude for Z (6.1 x 10 -5 s -1 ) and M a1-antitrypsin (1.6 x 10 -5 s -1 ). Therefore, the selectivity of the compound for Z over M a1-antitrypsin is dominated by the difference in the rate of association rather than dissociation.
The ability of GSK716 to block Z a1-antitrypsin polymerisation in the ER during folding was assessed by adding GSK716 to CHO-TET-ON-Z-A1AT cells (8) with simultaneous induction of Z a1-antitrypsin expression using doxycycline. In comparison with controls, GSK716 completely blocked the intracellular formation of Z a1-antitrypsin polymers, as measured by staining with the 2C1 anti-Z a1-antitrypsin polymer monoclonal antibody (pIC50 = 6.3) (Figs. 2a, b). It also increased the secretion of Z a1-antitrypsin approximately 3-fold compared to vehicle control (mean pEC50 6.2 ± 0.23; n = 74) (Fig. 2b). Similar potency between the effects on secretion and polymerisation was observed throughout members of the lead series supporting the hypothesis that these effects are caused by the same pharmacological mode of action. GSK716 had a similar effect on the secretion and polymerisation of constitutively expressed Z a1-antitrypsin in iPSC-derived human hepatocytes with the ZZ a1-antitrypsin genotype (22). It inhibited polymerisation and increased secretion with a mean pIC50 of 6.4 ± 0.45 (n = 16) and mean pEC50 of 6.5 ± 0.37 (n = 14), respectively, inducing an approximately 3-fold increase of secreted levels of Z a1antitrypsin (Figs 2c, d). GSK716 treatment reduced the levels of intracellular Z a1antitrypsin polymer compared with cells assessed before compound addition (Fig. 2c), demonstrating that polymers can be cleared over the time course of the experiment, and that accumulation of polymers is reversible in ZZ-iPSC-hepatocytes.
The pre-treatment of CHO cells induced to express Z a1-antitrypsin with GSK716 significantly reduced the formation of soluble and insoluble polymers (

GSK716 binds to a novel cryptic binding site
A high-resolution crystal structure of a1-antitrypsin complexed with the lead compound GSK716 was generated by soaking compound into apo a1-antitrypsin crystals ( Table   1). The structure reveals that interaction with the compound induces the formation of a cryptic binding site that is not evident in apo structures, at the top of β-sheet-A behind strand 5. This region is referred to as the 'breach' as it is the point at which the reactive centre loop first inserts during protease inhibition (23), and includes the site of the Z (Glu342Lys) mutation (Fig. 3a). The structure reveals that the 2-oxindole ring of GSK716 stacks with the side chain of Trp194 whilst the carbonyl group forms a hydrogen bond with the mainchain Trp194 (Fig. 3b). Trp194 adopted a new position due to rearrangement of residues Gly192 to Thr203 consistent with the change in Correspondingly, in a constant-temperature experiment, oligomers were generated at higher temperatures in the presence of the compound than in its absence when visualised by non-denaturing PAGE (Fig. 4B). Native state stability can also be probed by equilibrium unfolding using chemical denaturants, where a peak in bis-ANS fluorescence corresponds with a maximally populated unfolding intermediate (25).
The profiles in Fig in which the compound stabilises β-sheet A against conformational change that mediates both inhibitory activity and pathological misfolding.

Characterisation of drug-like properties of GSK716
GSK716 selectivity and PK properties were profiled in order to investigate the suitability of GSK716 for progression into in vivo studies and the potential for taking it forward as a clinical candidate for testing in humans. Since GSK716 results in loss of inhibitory activity of a1-antitrypsin, the effect of the compound was assessed on other closely-related serpins. GSK716 did not effect the inhibitory activity of antithrombin, neuroserpin and a1-antichymotrypsin towards their cognate proteases (Fig. 5C).
Furthermore, there were no off target effects in a panel of assays considered predictive of known safety liabilities that precluded further development of GSK716 (Suppl. Table   1). GSK716 exhibited low metabolic clearance in human hepatocytes (0.31 ml/min/g tissue), with moderate to high clearance in mouse hepatocytes (4.56 ml/min/g tissue).

Since
It exhibited weak time dependent inhibition of CYP3A4 resulting in a 1.59 fold shift in IC50. Taken together, oral bioavailability is predicted to be high in human, with measured F in rat (48%) and dog (71%) at ≤3mg/kg being consistent with 100% absorption and losses via first pass metabolism only. Mean exposure of GSK716 in blood in the male CD-1 mouse increased with dose following single PO administration at 10, 30 or 100mg/kg (mean dose-normalised Cmax 58±112, 113±27 and 113±27; DNAUCinf 202±101, 294±47 and 403±246, respectively).

GSK716 increases secretion of Z a 1-antitrypsin in a transgenic mouse model of Z a 1-antitrypsin deficiency
GSK716 was evaluated in a transgenic mouse model with an engineered random insertion of the human Z a1-antitrypsin gene (6). The PK-PD relationship of GSK716 was explored by dosing Z a1-antitrypsin transgenic animals with 10, 30 or 100 mg/kg GSK716 three times a day. Blood and liver were harvested on day 6 at 3hr (~Cmax) and 8hr (Cmin) after the dose for the measurement of total and free drug in both tissues.
Blood was also harvested for the measurement of monomeric Z a1-antitrypsin in plasma. Total concentrations of GSK716 were determined by LC-MS/MS and the free drug in both tissues was determined using equilibrium dialysis to determine free fraction in the samples, subsequently used to derive unbound concentrations. Blood concentrations demonstrated that the Cmin levels of free drug were at or above 300nM, the cellular secretion assay EC50, for the majority of the dosing period following 100mg/kg dosing, whereas 30mg/kg and 10mg/kg doses resulted in free drug levels in blood significantly below the cellular EC50 concentrations for a large part of the dosing period. Both free and total drug concentrations of GSK716 at the targeted site of action in the liver were equivalent to those in blood ( Table 2).
A significant fraction of the total Z a1-antitrypsin in the circulation is in the polymeric conformation (27). There are no antibodies that are specific for monomeric Z a1antitrypsin and so to directly determine its concentration, a deconvolution method was developed based on immunoassays with antibodies for either total or polymeric a1antitrypsin, and calibration curves with purified monomeric and polymeric Z a1antitrypsin. Monomeric Z a1-antitrypsin was measured in plasma samples following 6 days of dosing and levels were normalised to each animals' predose control levels to account for the natural variation of Z a1-antitrypsin between animals. Administration of 100mg/kg GSK716 resulted in a mean 7-fold increase in circulating monomeric Z a1-antitrypsin levels demonstrating robust target engagement in the liver (Fig. 6A).
Interestingly, 30mg/kg and 10mg/kg groups also gave significant, dose-dependent increases in circulating Z a1-antitrypsin despite free concentrations being below the cellular EC50 for secretion for much or all of the dosing period. Total drug levels and changes in Z a1-antitrypsin following 3 days of dosing were indistinguishable to those following 6 days of dosing. There was no effect on circulating serum albumin after 5 days of dosing which provides evidence that GSK716 is specific for Z a1-antitrypsin.
Moreover the effects are not mediated by metabolites of GSK716 as the major metabolites have much reduced or no binding to a1-antitrypsin.  Fig. 1). There was no difference observed in total liver polymer load when assessed by manual or quantitative scoring (Fig. 6b). There was no significant fibrosis in any of the liver sections. Intrahepatic a1-antitrypsin was also assessed by ELISA. The vast majority (95-100%) of a1-antitrypsin within the liver is polymer with the monomer typically being below the level of detection. Treatment with GSK716 increased the monomer measured in liver homogenate by approx. 4-fold in keeping with the changes seen in blood. in Z a1-antitrypsin in transgenic mouse liver following 12-33 weeks of treatment, albeit without reports of data at earlier timepoints (29). It is likely that GSK716 will need to be dosed to transgenic Z a1-antitrypsin mice for significantly longer than 20 days to demonstrate an effect on total liver polymer levels. It remains to be seen whether the intra-hepatic polymer needs to be cleared from the liver in order to have some functional benefit or whether the accumulated polymer inclusions are inert and abrogation of polymer production is sufficient to restore the functioning of the ER and hence the health of the cells. Moreover treatment with GSK716 may be sufficient to protect against the two-hit process whereby the Z a1-antitrypsin polymers sensitise the liver to a secondary insult such as alcohol, drug or liver fat (8) (9).

Discussion
Polymer formation and inhibitory activity are inextricably linked in the serpin mechanism (30) and thus small molecules that block polymerisation may have the unwanted effect of also blocking inhibitory activity. Bound GSK716 inhibits the serpin activity of a1-antitrypsin and so would not be expected to increase protease inhibitory activity during the dosing period. However, the half-life of monomeric Z-a1-antitrypsin in human is 6 days whereas drug would be expected to be cleared with a half life of a few hours after dosing raising the possibility of a pulsatile dosing regimen that would lead to increased, active serpin. The slow development of the lung disease in individuals with a1-antitrypsin deficiency over many decades suggests that acute effects associated with inhibition of serpin activity are unlikely.
There is increasing recognition that heterozygosity for wildtype M and mutant Z a1antitrypsin alleles predisposes to liver disease (9). Our small molecule approach to block polymer formation has an advantage over siRNA therapies to 'knock down' a1antitrypsin production in treating the MZ a1-antitrypsin heterozygote. GSK716 has a 100-fold greater affinity for Z than M a1-antitrypsin and so may be dosed at a level that prevents polymerisation of Z a1-antitrypsin without reducing the inhibitory effect of the wildtype M protein.
In summary we report the first small molecule drug-like correctors of Z a1-antitrypsin folding obtained via optimisation of hits from an Encoded Library Technology screen (18,31,32) that are suitable for oral delivery, correct folding in human patient iPSCderived hepatocytes and increase circulating Z a1-antitrypsin levels in a transgenic mouse model of a1-antitrypsin deficiency.

Materials and Methods
Alpha1-antitrypsin was purified from the plasma of M (wildtype) and Z a1-antitrypsin homozygotes and recombinant Cys232Ser a1-antitrypsin was expressed and purified as detailed previously (19,33,34).

DNA-encoded library technology (ELT) screen
An encoded library technology (ELT) screen with a nominal diversity of 2×10 12 unique components was used to identify small molecules that bind monomeric Z a1-antitrypsin at 4°C, 22°C, and 37°C. Affinity selections were performed as described previously (31).

In vitro assay of Z a 1-antitrypsin polymerisation
An antibody-based time-resolved fluorescence resonance energy transfer (TR-FRET) assay was developed to monitor the polymerisation of 5nM Z a1-antitrypsin following incubation with varying concentrations of compounds at 37°C for 72 hours. This assay used the 2C1 monoclonal antibody (1.25nM) that is specific to pathological polymers

Compound association experiments
Kinetic parameters of GSK716 binding to M, Z, S (Glu264Val) and Baghdad (Ala336Pro) (35) a1-antitrypsin were measured by detecting intrinsic tryptophan fluorescence of the protein (excitation at 280 nm and detection of emission at 320 nm) on a stopped flow apparatus (Applied Photophysics) (10,36). A competition assay for binding to M and Z a1-antitrypsin and Z a1-antitrypsin polymers (19,33,34), based on an Alexa488-labelled analogue of GSK716 (A488-GSK716), was used to determine the binding affinity of test compounds.

Thermal stability, unfolding and assessment of protease inhibition
The native state stability of a1-antitrypsin on addition of compounds was investigated by thermal denaturation in the presence of a 5X concentration of SYPRO Orange dye solution (Life Technologies) (37). Resistance to heat-induced polymerisation was determined using an end-point constant-temperature assay. Equilibrium unfolding was evaluated with a Bis-ANS dye (10) and rapid refolding following denaturation in 6M urea was assessed by non-denaturing PAGE. The inhibitory activity of a1antitrypsin was measured by titration against the model protease bovine achymotrypsin. The activity of antithrombin, neuroserpin and a1-antichymotrypsin were assessed against human thrombin, bovine trypsin and bovine a-chymotrypsin respectively.

Crystallisation and structure determination
Crystallisation of recombinant Cys232Ser a1-antitrypsin was carried out in 2-well MRC crystallisation plates with a Mosquito robot (TTP Labtech) using 100nl protein solution and 100nl well solution. Crystals grew from 22% w/v PEG1500, 0.2M MES pH6.0 and were soaked for 24hrs with 25mM compound (5% v/v DMSO). X-ray diffraction data were collected at Diamond on beamline IO3. Structure solution was carried out by molecular replacement using PHASER (38). The model used to solve this structure was a related complex (data not shown) which had been solved by molecular replacement using PDB entry 2QUG as a starting model. Building was carried out using Coot (39) and refinement with REFMAC (40).

In vitro pharmacokinetics
ChromLogD was measured as previously described (41). Permeability in MDR1-MDCK cells with pgp inhibitor, CYP3A4 TDI fold IC50 shift and hepatocyte clearance were performed as detailed in the supplementary material.

The Paper Explained
Problem Intracellular protein aggregation can result in 'gain of function' cell toxicity.
It has proved challenging to develop small molecules that can stabilise intracellular mutant proteins, prevent self-aggregation and so ameliorate disease. Severe a1antitrypsin deficiency results largely from the Z allele (Glu342Lys) that causes the accumulation of homopolymers of mutant a1-antitrypsin within the endoplasmic reticulum of hepatocytes in association with liver disease.

Results
We have undertaken a medicinal chemistry campaign to develop an orally bioavailable small molecule that binds to intra-endoplasmic reticulum mutant Z a1antitrypsin, corrects the folding defect and increases secretion in a transgenic model of disease.
Impact This study reports the successful targeting of an aggregation-prone mutant in order to prevent the intracellular polymerisation and accumulation of a1-antitrypsin that underlies a1-antitrypsin deficiency. It provides proof-of-principle that the identification of 'mutation ameliorating' small molecules is a viable approach to treat protein conformational diseases.          (Table 2). (a) 100mg/kg GSK716 resulted in a mean 7-fold increase in circulating monomeric Z a1-antitrypsin levels demonstrating robust target engagement in the liver. The observed steady state total and free compound levels of GSK716 in the transgenic Z a1-antitrypsin mouse were predicted well by the in silico PK model built on: (i) in vitro metabolic clearance data, (ii) plasma protein binding data, (iii) in vivo PK data from wild type mice and (iv) a term comprising a 5µM circulating sink for drug with an affinity of 1.5 nM, representing the Z a1-antitrypsin within blood. The target free drug concentration was selected based on the observed potency in the in vitro secretion assays in which the total drug approximates to the free drug in the assay. Interestingly, 30mg/kg and 10mg/kg groups also gave significant, dose-dependent increases in circulating Z a1-antitrypsin despite free concentrations being below the cellular EC50 for secretion for much or all of the dosing period.

Figures and Tables
Significance at p<0.05 by a student's t test is denoted by *: pairwise: each animal is compared with itself, treated vs pre-treatment; between groups: compound treated vs vehicle. (b) Representative example images of 2C1 stained livers from 20 day vehicle (i and ii) and 100mg/kg TID GSK716 (iii and iv) treated animals. The asterisks indicate regions of hepatocytes that are negative on 2C1 immunostaining. Quantification of 2C1 stained area in livers from vehicle and 100mg/kg TID GSK716 treated animals at day 15 (v) and 21 (vi), shown as total area and low-only or high and mid-only intensity stained areas. There were no significant differences between groups by KW ANOVA with Tukey's post-hoc analysis, data are the median +/-interquartile ranges.