Targeted degradation of α-Synuclein using an evolved botulinum toxin protease

Significance

The ability to evolve proteases that selectively cleave a desired protein in vivo could provide access to useful therapeutic agents. This would be especially powerful when targeting intrinsically disordered proteins (IDPs), a hard-to-drug class of proteins involved in many human diseases including cancer and neurodegenerative diseases. Here, we demonstrate the stepwise evolution of clinically used botulinum protease to proteolyze the IDP α-Synuclein which forms plaques in the brains of patients suffering from Parkinson’s disease.

Abstract

There is considerable interest in the targeted degradation of proteins implicated in human disease. The use of sequence-specific proteases for this purpose is severely limited by the difficulty in engineering the numerous enzyme–substrate interactions required to yield highly selective proteases while maintaining catalytic activity. Herein, we report a strategy to evolve a protease for the programmed degradation of α-Synuclein, a presynaptic protein closely linked to Parkinson’s disease. Our structure-guided evolution campaign uses the protease from botulinum neurotoxin and showcases the stepwise change of specificity from its native substrate SNAP25 to the selective degradation of α-Synuclein. The protease’s selectivity is further demonstrated in human cells where near complete degradation of overexpressed human α-Synuclein is observed with no significant effects on cell proliferation. This stepwise strategy may serve as a general approach to evolve highly selective proteases targeting dysregulated proteins.

The ability to evolve proteases that selectively and efficiently cleave a desired protein in vivo would provide novel therapeutic agents to degrade proteins contributing to human disease. Indeed, the protease from botulinum neurotoxin A (BoNT/A), by virtue of its exquisite sequence specificity for its native target SNAP25, is used clinically for local deactivation of peripheral neurons and can undergo repeated administration without loss in efficacy due to antidrug antibodies (1, 2). Structurally, like many other bacterial toxins (3), BoNT/A consists of a “light chain” catalytic protease domain tethered to a “heavy chain” that is responsible for endocytosis by acting as a receptor-binding and translocation domain. After uptake into lysosomes, the disulfide bond connecting the two subunits is cleaved and the light chain protease is released into the cytosol. The heavy chain thus provides an efficient means for the protease to cross cellular membranes with selective tropism for a specific cell type and proteolyze an intracellular target. The target of the BoNT/A light chain, SNAP25, is an intracellular, intrinsically disordered protein (IDP). SNAP25 is an integral part of SNARE assemblies required for neurotransmitter exocytosis (4) and selective proteolysis of SNAP25 by BoNT/A leads to synaptic deactivation. Unfortunately, other such highly selective proteases that can be used to degrade targets of clinical interest are lacking.

The extended >30 amino acid substrate binding site of the BoNT/A Zn²⁺ metalloprotease and distinct catalytic and substrate binding sites (5) make it an ideal starting point for evolving proteases to other targets of interest. SNAP25 binds to BoNT/A through numerous polar and nonpolar interactions, both C- and N-terminal to the substrate cleavage site. These interactions are mediated by unstructured loops and β-sheets of the protease, that position the substrate in a favorable conformation for hydrolysis in close proximity to the active site Zn²⁺ ion. Previous work on evolving clinically used BoNT/A (6–8) or other BoNT serotypes (9, 10) has demonstrated the possibility of partially changing substrate selectivity. However, the complex interplay of evolving selective substrate binding with concomitant retention of enzymatic activity and protein stability has made the evolution of altered substrate selectivity in BoNT a challenging task, particularly for nonhomologous targets. Consequently, previous efforts focused on evolving specificity toward highly homologous SNAP23 (6–8), used BoNT serotypes with less complex surface interactions than BoNT/A and SNAP25 (10), or both (9). Herein, we describe a structure-guided, stepwise strategy to evolve the enzymatic activity of the clinically used BoNT/A protease for the selective degradation of proteins that bear little homology to its native substrate.

Results

α-Synuclein as a Target.

As a proof of concept, we chose α-Synuclein (α-Syn) as target due to its close association with Parkinson’s disease (PD) and its lack of any significant homology to SNAP25. In PD, loss of dopaminergic neurons is accompanied by the presence of Lewy bodies, abnormal plaques consisting mostly of precipitated α-Syn (11)—a 14 kDa protein that is highly expressed in neurons. Certain point mutations in the protein’s N-terminal region significantly increase the risk of developing PD and some are associated with early disease onset (12). Gene duplications (13) and triplications (14) also lead to early-onset autosomal dominant PD, indicating a correlation of dysregulated α-Syn expression levels to the occurrence of PD. Furthermore, it has been observed that neuronal subpopulations with high relative α-Syn levels are more susceptible to Lewy body formation (15). Deletions of the gene cause minimal developmental changes in mouse studies (16, 17). Taken together, these data suggest that lowering intracellular Synuclein levels or degrading aggregates represents a promising strategy for improving disease outcome.

α-Syn is an IDP, a class of proteins, which lack a folded structure in the absence of an interacting partner. Indeed, IDPs or proteins with intrinsically disordered regions have been identified as key targets in human disease such as c-myc (18, 19), p53 (20), and EWS-FLI1 (21) in oncology and β-amyloid, TDP43 (22), and TMEM106B (23) in neurodegenerative diseases. The first step in designing a protease to degrade α-Syn was to identify a target cleavage site, which ideally corresponds to a conformationally flexible loop or disordered region of the protein required for activity. To this end, we aligned the amino acid sequence from SNAP25 cleaved by BoNT/A (24) with the central “nonamyloid core” region (NAC) (25) of α-Syn (Fig. 1A), a region that is considered responsible for the aggregation properties of the synucleins. Cleavage within this region would likely prevent the formation of aggregates, although it is unclear whether this region would be susceptible to cleavage in existing aggregates. We identified a potential α-Syn cleavage site between residues Q79 and K80 that corresponds to SNAP25’s Q197/R198 amide bond hydrolytically cleaved by wild-type BoNT/A. Additionally, α-Syn residues T72, Q79, G86, and S87 align with SNAP25 residues T190, Q197, G204, and S205, respectively. Attempting to proteolyze α-Syn at this site using an evolved BoNT/A therefore likely requires changing the recognition sequence of BoNT/A for many of the key residues in the putative 30 amino acid substrate binding site. However, even a single projected mutation in SNAP25 (R198K) has been shown to decrease the proteolytic activity of wild-type BoNT/A below detectable levels (26), indicating a significant effort is required to alter the selectivity of this protease to target α-Syn.

Fig. 1.

(A) Alignment of the C-terminus of SNAP25 and part of α-Synuclein’s NAC region highlighting homologous sites with teal boxes and the cleavage sites of WT-BoNT/A on SNAP25 by a red triangle. (B) In the presence of WT-BoNT/A the SNAP25-based linker connecting T7LZ to T7RNAP is proteolytically cleaved, producing active T7 RNA polymerase. (C) Proteolytic activity leading to sfGFP production is demonstrated experimentally when protease expression is induced with arabinose, resulting in strongly increased production of sfGFP. When activity of the protease pBAD promoter is inhibited with glucose, only low levels of sfGFP production are observed. In the absence of a competent protease, the tethered T7 lysozyme induces abortive transcription by T7 RNAP, leading to only basal sfGFP production by Escherichia coli, independent of whether glucose or arabinose is added. (D) Partial amino acid sequences of different linkers tethering T7 lysozyme to T7 RNA polymerase that were employed in this study.

Screening and Mutagenesis Strategy.

To evolve the sequence specificity of BoNT/A, we required means to efficiently screen large numbers of diverse protease clones. To this end, we constructed the plasmid SS-SNAP harboring a fusion of a mutant lysozyme from coliphage T7 linked to T7 RNA polymerase (Fig. 1B) that has been previously employed in PACE evolution campaigns (10, 27). The linker connecting the two proteins initially consisted of a peptide sequence containing 20 amino acids from SNAP25 surrounding the cleavage site, which likely plays the dominant role in determining the BoNT/A substrate Michaelis constant (K_M). Additionally, our screening system harbors an sfGFP gene under control of a T7 promoter. In this construct, T7 lysozyme (T7LZ) acts as the natural inhibitor of T7 RNA polymerase (T7RNAP). Accordingly, with T7LZ translationally fused to T7RNAP only minimal background sfGFP production is observed by fluorescence measurement (Fig. 1 B and C). When Escherichia coli DH10B (naturally araD139) are cotransformed with the plasmid SS-SNAP and a second plasmid harboring BoNT/A protease under control of the pBAD promoter, we measured trace production of sfGFP in the presence of 1 wt.% glucose (Fig. 1C), comparable to the fluorescence observed in the absence of the protease plasmid. A strongly increased fluorescence signal was observed when BoNT/A expression was induced in the presence of 0.01 wt.% arabinose, indicating a correlation between proteolytic activity and sfGFP production. Consequently, cells expressing proteolytically active BoNT clones can be readily isolated based on their relative sfGFP expression levels using this system and fluorescence-activated cell sorting (FACS). Additionally, cells producing sfGFP independent of proteolytic activity, e.g., through mutations in the T7LZ or T7RNAP, can be distinguished due to their unchanged sfGFP levels when inducing or inhibiting protease expression. By varying the linker sequence connecting T7LZ to T7RNAP, protease activity against various targets can be probed. Moreover, nonspecific proteases are likely to be toxic if they act on the essential bacterial proteome, providing a simple negative selection.

We next sought to identify key residues in BoNT/A involved in protease–substrate interaction by examination of the cocrystal structure of an inactive BoNT/A mutant (E224Q, Y366F) bound to a SNAP25 peptide fragment (5). We identified three important regions mediating interactions between SNAP25 and BoNT/A proximal to the Zn²⁺ catalytic site (site-saturation mutagenesis 1–3; SSM1–3, Fig. 2 A and C) and four more remote sites that were judged to be of secondary importance (SSM4a–d, Fig. 2A). The design of our first (Fig. 2B) targeted mutation library SSM1 was based on the hypothesis that the high selectivity of BoNT/A for residues proximal to the cleavage site between Q197 and R198 is mediated in part by an unstructured loop flanking Y366, because the Y366F mutation severely impairs its catalytic activity (5). Additionally, we postulated that the interaction of BoNT/A D370 located in this loop is responsible for the observed substrate selectivity at SNAP25 sites R198 and A199 (26). SSM1 is an NNK library derived from WT-BoNT/A with five randomized amino acids flanking Y366 but conserving the tyrosine itself (Fig. 2A). For the corresponding substrate, we constructed a selection plasmid bearing a quadruple-mutant SNAP25 target sequence (R198K/A199T/T200V/L203A, SNAP25 numbering, Fig. 1 A and D) as a first step in transforming the wildtype BoNT/A cleavage site sequentially to that of α-Syn.

Fig. 2.

(A) Co-Crystal structure of deactivated double-mutant BoNT/A-LC (E224Q/Y366F) and SNAP25 fragment (yellow) (PDB: 1XTG) with regions highlighted in color that mediate substrate binding and cleavage. (B) Diversification and screening scheme for evolving substrate selectivity of botulinum protease with the substrates used for each respective step. Clones chosen for subsequent rounds are highlighted in black. (C) Mutations observed in selected clones from site-saturation mutagenesis. An asterisk marks sequences that were carried on for further mutagenesis. (D) Mutations observed in selected BoNT/A clones isolated over the course of this study. Gray boxes highlight amino acids mutated from the WT sequence. Red letters highlight residues mutated a second time during the evolution campaign. ^apreceded by a 32 amino acid insert between BoNT/A amino acids S254 and T255 with the sequence DSEVSFEELRTFVAINPNRVLQSNTNAYYEMS.

Subsite Mutation and Screening.

Bacteria transformed with this protease library SSM1 and quadruple-mutant screening plasmid SS1 (Figs. 1D and 2B) were then sorted by FACS. Throughout this study, ~300 million cells were sorted per round and ~40 000 cells with the highest sfGFP levels were collected. Sequencing of green colonies from this initial selection revealed a strong mutational bias, with BoNT/A mutations F369L and D370P being frequently observed in selected clones (Fig. 2C, Tier 1 mutations). To further investigate the evolved proteolytic activity of these proteases, we selected clone 5 from this round (“Protease 1”, Fig. 2D) and observed a significant signal increase in our fluorescence assay using SS1 (Fig. 3A). Additionally, when Protease 1 was coexpressed with a Flag-tagged T7LZ-T7RNAP harboring the R198K/A199T/T200V/L203A quadruple mutant SNAP25 linker sequence (FLAG-SS1), the substrate could be clearly detected by Western blot even when inhibiting protease expression (Fig. 3B). When protease expression was induced, complete disappearance of the substrate band was observed, indicating complete proteolytic cleavage of the SNAP25 quadruple mutant sequence.

Fig. 3.

(A) Fluorescence assay in E. coli as a measure for proteolytic activity of respective BoNT/A mutant against indicated screening system. Fluorescence data from cells without protease is shown for comparison. (B) Western blot showing activity of protease mutants on Flag-tagged screening system substrates with induced (Top) or uninduced (Bottom) protease expression. Proteases selected from site-saturation mutagenesis libraries are active on only 1 to 2 homologous substrates without prior negative selection.

We chose Protease 1 (cf. Fig. 2C, clone 5 from SSM1) to continue the next round of our stepwise evolution (Fig. 2B) campaign and focused the design of SSM2 on the BoNT region recognizing the C-terminus of SNAP25 comprising residues K201, M202, L203, and G204 (Figs. 1A and 2A). SNAP25 residue K201 C-terminal to the cleavage site binds to a BoNT β-sheet near E257, which in turn interacts with BoNT’s K244, located in a second, antiparallel β-strand. This round of mutagenesis focused on generating clones that cleave α-Syn E83, reversing the substrate polarity of SNAP 25, which has a lysine at the corresponding position (Fig. 1A). Consequently, we prepared protease library SSM2 from Protease 1 with randomized BoNT/A residues K244 and E257 along with 4 proximal sites F243, V245, G255, and L256 (Fig. 2). These BoNT residues are all part of the same β-sheet binding the substrate C-terminal region. We constructed the corresponding screening substrate, SS2, additionally harboring mutations K201E and M202G compared to SS1 (Fig. 1D). Cells harboring this sextuple mutant SS2 and second-tier protease library SSM2 were again sorted for the highest levels of sfGFP after inducing protease expression. Sequencing of active clones revealed a strong bias at select sites (Fig. 2C, Tier 2 mutations). Notably, position E257 was mutated from glutamate to lysine or arginine, an observation in line with our expectation to reverse the polarity of SS2 at this site as compared to SS-SNAP. BoNT mutation K244Y was also prominently featured, perhaps interacting with the newly introduced cationic residue at position 257 within the BoNT β-sheet structure. The observation of aromatic or branched aliphatic amino acids overrepresented at BoNT positions 243, 245, 255, and 256 may also contribute to β-sheet stability. A strong increase in fluorescence was observed when expressing the protease clone 8 from this selection (“Protease 2”, Fig. 2D) in the presence of SS2 (Fig. 3A). We further validated the proteolytic activity by Western blot analysis, using Flag-tagged plasmid FLAG-SS2 (Fig. 3B).

With an evolved protease readily cleaving SNAP25 sextuple mutant SS2 in hand, we turned to the N-terminal region of the substrate encompassing SNAP25 residues T190–Q197 as a next step (Fig. 2B). This substrate region binds to another β-strand in BoNT/A, itself deeply nested in the protease core as part of an 8-strand β-sheet motif. BoNT/A residues C165, K166, S167, F168 Q184, and R231 appear involved in interactions with SNAP25 in this region, and we accordingly prepared library SSM3 containing proteases randomized at these positions (Fig. 2A). For screening, we devised and created the corresponding substrate SS3 (Fig. 1D), which has an additional six substrate residues that bind to this BoNT/A β-strand altered relative to SS2: R191G, I192V, D193T, E194A, A195V, and N196A (SNAP25#, Fig. 1A). Overall, the substrate sequence in SS3 differs at 12 sites from the original SNAP25 sequence and already closely resembles that of α-Syn (Fig. 1A). Despite this large change from the previous substrate sequence SS2, we were able to isolate active protease clones against SS3 when sorting cells expressing this round 3 library by FACS. Strong mutational biases were again observed, with R231F being the only functional mutation isolated at this position. Clones with conserved C165 appeared to require an additional cysteine at BoNT position 184, presumably forming a new disulfide bond. K166R was another frequently isolated mutation, with lysine at this site being notably absent. Otherwise, native amino acid identities were frequently observed, perhaps indicating the importance of this deeply embedded core region of the enzyme. We investigated clone 1 from this selection (“Protease 3”, Fig. 2D) in more detail. In our fluorescence assay using SS3, an increase in fluorescence was observed in the presence of this Protease 3 (Fig. 3A). However, the increase was significantly smaller than that in previous iterations. We also constructed Flag-tagged plasmid Flag-SS3, and cleavage of this mutant substrate was visualized by Western blot, which showed significant but incomplete degradation of Flag-SS3 by Protease 3 (Fig. 3B).

Before optimizing for increased proteolytic activity, we first sought to investigate whether selectivity is retained on a larger α-Syn fragment. Consequently, we increased the substrate length with screening system SS-aSyn containing a 60 amino acid linker sequence derived exclusively from α-Syn connecting T7LZ and T7RNAP (Fig. 1D). The increased substrate size also served the dual purpose of identifying possible remote binding sites for increased substrate affinity and minimizing the possibility of selecting mutants that bind to either T7LZ or T7RNAP. Notably, we observed that the longer linker sequence leads to slightly less efficient inhibition of T7RNAP activity and higher background fluorescence in the absence of protease (Fig. 3A). Despite Protease 3 cleaving highly homologous SS3, no significant activity of this protease was detected on SS-aSyn either by fluorescence assay or by Western blot (Fig. 3). We thus attempted to further increase both activity and selectivity of Protease 3 by mutating distal SNAP25 binding sites in the protease. Consequently, four libraries were separately constructed from Protease 3, each randomizing residues in separate unstructured loops at BoNT sites K23, I24, P25, A27, G28, and Q29 (SSM4a, Fig. 2 A and B), G169, H170, E171, V172, N174, and T176 (SSM4c), and V304, G305, T306, T307, A308, and Y312 (SSM4d), as well as in β-sheet residues D131, T132, C134, N136, Y144, and S146 (SSM4b). From these four libraries, protease clones active against SS-aSyn were identified by FACS. The mutated regions isolated in each round of selection were then permutated. First, a separate PCR for each of the pooled active protease libraries SSM1-4 was used to produce linear gene fragments; pooling of the resulting PCR products and reassembly using Gibson assembly yielded plasmids encoding full-length proteases with interchanged gene cassettes. We again tested the proteolytic activity of Protease 4 containing permutated sequences from previous selections (Fig. 2D) using SS-aSyn and observed a significant increase in fluorescence in the presence of protease (Fig. 3A). Importantly no fluorescence increase was detected in the presence of SS-SNAP, SS1 or SS2 (Fig. 3A). However, further analysis using our fluorescence assay showed remaining activity against SS3. A Western Blot revealed that more, but still only partial proteolysis of substrate SS-aSyn was achieved (Fig. 3B) and confirmed that Protease 4 retained significant activity on previous substrate SS3, suggesting that selectivity also required additional improvement.

Random Mutagenesis and Recombination.

To further optimize the catalytic activity and specificity of our evolved proteases for α-Syn, we conducted random, error-prone mutagenesis using Protease 4 as the template, reasoning mutations distal to the binding site might optimize the geometry of key interactions. Clones were sorted for activity, and a library of 400 active clones was then subjected to NeXT DNA shuffling (28) along with clones from previous screens. Cells were again challenged against SS-aSyn and sorted for high fluorescence intensity. The active protease clones were then counterscreened for minimal activity against SNAP25 using SS-SNAP and this complete protocol was repeated to yield Protease 5. Interestingly, Q363R reverted to its original wild-type amino acid (Fig. 2D), and we also observed additional mutations E368L and A372L in this unstructured loop. In addition, the β-sheet evolved in SSM2 contained the adjacent mutation V242A and proximal mutations N248D and S254R. All point mutations identified during random mutagenesis appeared to mediate interactions with the C-terminal region of the substrate except for T109M. Additionally, some sites mutated in previous rounds showed variability, suggesting that these regions play a minor role in controlling overall activity and selectivity.

Protease 5 displayed significantly increased activity in our fluorescence-based activity assay against SS-aSyn, and a strong increase in sfGFP production was observed even for low protease expression conditions in the presence of glucose (Fig. 4A). To evaluate its selectivity, we cotransformed cells with Protease 5 plasmid and separately one of our four screening plasmids SS-SNAP, SS1, SS2, or SS3. As a control, we also collected data using cells containing no protease, but only the respective screening plasmid. In the absence of protease, cells showed only marginal fluorescence in the presence of arabinose. In the presence of our evolved protease, none of the earlier screening systems showed significant proteolytic activity as judged by our fluorescence assay either in the presence of glucose or arabinose (Fig. 4A). Selection system SS-aSyn bearing the 60 residue α-Syn linker was the only substrate exhibiting a significant increase in fluorescence in the presence of Protease 5 over the no protease control, independent of whether expression was induced, highlighting the activity and fidelity of this evolved protease. We further confirmed these data by Western Blot, where induction of protease expression led to complete disappearance of the substrate, and partial proteolysis was observed even when protease expression was inhibited (Fig. 4A). The high selectivity of this protease for aSyn may result in part from degradation of bacterial proteins by lower fidelity proteases with broader substrate specificity, resulting in slower growth rates and subsequently exponential underrepresentation during selection.

Fig. 4.

(A) Fluorescence assay using E. coli in the absence and presence of Protease 5 as an indicator for its activity and selectivity against substrate screening systems used in this study. Substrate cleavage was also evaluated by Western Blot on Flag-tagged substrates with F-actin serving as a loading control. Only the SS-aSyn substrate was significantly degraded by the final Protease 5. (B) Heatmap visualizing the selectivity trajectory of our evolution campaign by plotting the fluorescence increase in the presence of respective substrate screening systems and proteases in E. coli. White boxes highlight the substrate screening system the respective protease was selected against. (C) Western blot showing substrate degradation by WT-BoNT (Flag-SNAP substrate) or evolved Protease 5 (Flag-SS-aSyn substrate). Cells were grown for 3 h total and arabinose was added at the indicated time prior to harvest. The high-contrast view shows traces of SS-aSyn substrate when no arabinose was added (0 min timepoint). (D) Western blot of HEK293T cells transfected with the indicated ratios of CMV plasmids encoding either full-length human α-Syn or Protease 5. (E) Viability of HEK239T measured 24 h after being transfected with CMV plasmids encoding the indicated proteins. Viability of cells transfected with SNAP25 plasmid was normalized to 100% viability.

Proteolytic Activity of Evolved Protease 5.

The time course of substrate cleavage in bacteria was assessed by transforming bacteria with SS-SNAP and WT-BoNT protease or SS-aSyn and evolved Protease 5. Saturated cultures were diluted into multiple aliquots and grown for a total of 3 h at 37 °C. At specific timepoints, protease expression was induced with arabinose. For WT-BoNT, partial SS-SNAP25 substrate cleavage was observed when protease expression was induced for 30 min (Fig. 4C). In contrast, no substrate band could be observed at 30 min with evolved Protease 5. Furthermore, only trace amounts of substrate could be observed at high exposure even when no arabinose was added at all (timepoint 0 min), indicating highly efficient degradation by our evolved protease at low uninduced protease expression levels. This is in accordance with our previous finding that notable SS-aSyn substrate degradation was observed by Western blot when using Protease 5, even under conditions that repress protease expression with glucose (Fig. 4A). The presence of bands for the substrate Flag-SS-aSyn in the absence of protease plasmid additionally showcases that Protease 5 is responsible for degradation of this construct. We then purified Protease 5 with an N-terminal GB1 solubility tag and subsequently incubated it with a substrate consisting of full-length α-Synuclein with N-terminal YFP and C-terminal CFP fusions. Analysis by LCMS revealed that this substrate was cleaved between α-Synuclein residues T92 and G93 (SI Appendix, Fig. S1). These data further confirm that only two fragments are produced by evolved Protease 5. It proved difficult to directly quantify the specific activities for the evolved protease by FRET with this system due to aggregation of the various Synuclein containing fusion proteins. Both variation of the fluorescent proteins or Synuclein-derived linker did not yield any monomeric substrate amenable to in vitro measurements. These difficulties are in line with previous reports highlighting the complex and disputed aggregation status of α-Synuclein (29, 30). A short peptide consisting of 17 amino acids derived from Synuclein amino acids G68–G84 for kinetic measurements proved to be insoluble.

To begin to investigate whether our findings extend to the degradation of full-length, native Synuclein in mammalian cells, we constructed two CMV vectors expressing Protease 5 and human α-Syn. When HEK293T cells were transfected with only the α-Syn expression plasmid, a strong α-Syn band was observed by Western blot using antibody Syn211 (Fig. 4C). Under these imaging conditions, none of the endogenous α-Syn produced by HEK293T was detectable, underpinning the high level of α-Syn overexpression achieved by transient transfection. In contrast, cotransfection of both protease and Syn plasmids (1:5 ratio, respectively) led to significant degradation of α-Syn when probed after 24 h. Cotransfected cells were investigated for viability to determine any detrimental effects of either proteolytic fragments or off-target protease activity, but no significant decrease in viability was observed (Fig. 4). Further experiments will be required to define what dosage of Protease 5 is required to degrade α-Syn and/or its aggregates at therapeutically relevant levels.

Discussion

We report the first complete reprogramming of the BoNT/A protease specificity for the targeted degradation of α-Syn while retaining the high catalytic activity of the native BoNT/A starting enzyme. This dramatic change in selectivity was achieved in a rational, stepwise manner which can form the basis for future BoNT/A evolution campaigns. Over the course of these experiments, 45 different sites were mutated, producing a protease that displays high activity and sequence specificity on a new target protein with minimal homology to the native substrate. Our novel protease largely depletes high levels of full-length α-Synuclein overexpressed in mammalian cells, a protein playing a largely unknown role in the human nervous system but of utmost importance in Parkinson’s disease and other synucleinopathies. We also show that our structure-guided, stepwise approach infers minimal promiscuity to the resulting proteases, in contrast to prior reports of protease evolution (9, 27).

This work suggests that it will be possible to redirect the clinical utility of BoNT/A to the degradation of other non-SNARE proteins by a similar directed evolution strategy. The level of homology between SNAP25 and α-Synuclein can be found in virtually any IDP, highlighting the adaptability of our approach to other IDP targets. It is also likely that conformationally flexible loops or other disordered regions in proteins can be identified computationally and targeted, but this awaits experimental validation. However, for therapeutic use, another requirement is that the protease be targeted to the cell type of interest. Significant efforts are currently focused on modifying the targeting and cell entry of bacterial toxins. Re-engineering of the heavy chain cargo has been demonstrated and may allow for evolved protease to be delivered to new target cells (31–35) The conjugation of evolved proteases with bacterial heavy chains also leads to functional assemblies (10). Furthermore, AAVs have been shown to be viable tools for the delivery of some protein therapeutics and may be used for protease delivery, as well (36–38). As a next step, we are currently attempting to generalize the strategy described herein to degrade the IDP C-myc and C-terminal tail of K-Ras, and are integrating it with our recently described orthogonal T7-based continuous bacterial evolution system to further facilitate the process.

Materials and Methods

Site-Saturation Library Assembly.

The respective protease plasmid was amplified with primers containing degenerate NNK codons at the indicated sites using Q5 DNA Polymerase (New England Biolabs). PCR fragments were assembled using Gibson assembly (39). Assembly mixtures were purified using QIAquick PCR Purification Kit (Qiagen). Electrocompetent NEB 10-beta cells (New England Biolabs) were mixed with precooled, purified DNA, and electroporated and recovered using the manufacturer’s protocol. Pooled cells from multiple electroporations were then plated on LB agar and grown at 37 °C overnight. Small aliquots were diluted and plated on separate plates to estimate library diversity. The next day, cells were resuspended in PBS, pelleted and plasmid DNA was isolated using ZymoPURE II Plasmid Midiprep Kit (Zymo Research).

Error-Prone PCR.

Gene fragments were amplified using Taq polymerase rendered more error-prone by addition of 0.3 mM MnCl₂ and an uneven mix of dNTPs. In short, primers (each 0.2 µM), 400 ng plasmid DNA template, 4 µL Taq DNA polymerase, “error-prone” dNTPs (350 µM dATP, 400 µM dCTP, 200 µM dGTP, 1,350 µM dTTP), MgCl₂ (2.95 mM), and molecular biology grade water to 388µL were mixed and MnCl₂ (12 µL, 0.3 mM) was added. The mixture was distributed into a PCR strip with 8 wells containing each 50 µL each. Molarities indicate final concentrations. A PCR protocol consisted of the following: Initial Denaturation: 95 °C for 1:00 min, cycle 34×: 95 °C for 30 s, 60 °C for 45 s, 68 °C 1:30 then one final amplification at 68 °C for 5:00 min. PCR mixtures were then pooled into 200 µL batches, purified using QIAquick PCR Purification Kit (Qiagen) and assembled into circular plasmids using Gibson assembly. Plasmid backbones for Gibson assembly were amplified under nonerror-prone conditions using Q5 DNA polymerase. Isolation of library plasmids was executed as described above for site-saturation libraries.

Random DNA Shuffling.

Mutagenized clones were pooled and 100 fmol were amplified using Taq polymerase in ThermoPol Buffer following the manufacturer’s protocol with 30% of dTTP substituted with dUTP. PCR products were isolated using QIAquick PCR Purification Kit (Qiagen) and digested using 2 units USER enzyme (NEB) for 50µL reactions in 1× rCutSmart (NEB) for 2 h at 37 °C. Complete digestion was validated by agarose gel. To the digested mixtures was added 1 µL DpnI and incubated for 15 min at 37 °C. DNA fragments were isolated using the QIAex II Kit (Qiagen). Digested fragments were reassembled into full-length genes as previously reported (40) and cloned into linear plasmids using Gibson assembly and electrocompetent NEB 10-beta cells (New England Biolabs).

Protease Activity Selection by FACS and Fluorescence Measurements.

NEB 10-beta cells (New England Biolabs) competent cells containing respective selection plasmids SS-SNAP, SS1, SS2, SS3, or SS-aSyn were made electrocompetent (41), mixed with the respective protease plasmid library, and transformed by electroporation. Recovered cells were plated on LB agar containing maintenance antibiotics and 0.01 wt.% arabinose and grown overnight at 37 °C. Cells were harvested in PBS and sorted by fluorescence-activated cells sorting (FACS) using a Sony MA 900 cell sorter until >300million events were reached and ~40,000 cells with the highest sfGFP levels had been collected. Collected cells were plated on LB agar containing maintenance antibiotics and 0.01 wt.% arabinose and grown overnight at 37 °C. The next morning, single green colonies were picked into a 96-well plate containing 150 μL/well 2xYT media with maintenance antibiotics. Cultures were grown to saturation and diluted 30-fold into 96-well plates containing 2xYT media with maintenance antibiotics and either 1 wt.% glucose or 0.01 wt.% arabinose and shaken for 6 h at 37 °C in a Logphase 600 microbiology reader (BioTek) at 800 rpm in a black-walled, clear bottom 96-well plate. Cells were pelleted and resuspended in PBS, and fluorescence was measured at 470 nm excitation/510 nm emission using a SpectraMax iD3 plate reader (Molecular Devices). Clones with higher fluorescence in the presence of glucose were excluded as false positives and the remaining clones were ranked and selected according to their relative fluorescence in the presence of arabinose.

Western Blots.

NEB 10-beta cells containing the respective FLAG-tagged selection system plasmid and protease plasmid were grown overnight. Then cultures were diluted 30-fold into 96-well plates containing 2xYT media with maintenance antibiotics and either 1% glucose or 0.01% arabinose and shaken for 6 h at 37 °C in a Logphase 600 microbiology reader (BioTek) at 800 rpm in a black-walled, clear bottom 96-well plate. Then, 50 μL of culture were pelleted and resuspended in 50 μL 1× NuPage LDS sample buffer (Life Technologies). Proteins were separated by electrophoresis on Bolt 4 to 12% Bis-Tris Plus WedgeWell gels (Invitrogen) at 200 V for 30 min using submerged in MOPS SDS Running Buffer (Invitrogen). Proteins were blotted onto 0.45 um PVDF membrane (Immobilon) at 20 V for 60 min using Bolt Transfer Buffer (Invitrogen) containing 10% MeOH. Membranes were blocked for 1 h using 5% nonfat dry milk in TBST (1× TBS+0.1% Tween80 (Millipore). Membranes were then incubated overnight with the primary antibody and then washed 6× with TBST over 30 min. Membranes were incubated with secondary antibody conjugated to horseradish peroxidase for 1 h at room temperature and washed 12× with TBST over 1 h followed by a final wash with 1× TBS. Horseradish peroxidase conjugates were visualized using ProSignal Dura ECL Reagent (Prometheus) on a ChemiDoc Touch Imaging System (Bio-Rad).

Western Blots visualizing native α-Synuclein were conducted with an adjusted method to avoid detachment of α-Synuclein from the PVDF membrane after transfer. In short, after transfer, the membrane was incubated in PBS containing 0.4% paraformaldehyde for 30 min, then blocked for 1 h with 5% nonfat dry milk in TBST. Afterward, the membrane was incubated with antibody Syn211 (Invitrogen) in 5% nonfat dry milk in TBST overnight, followed by washing 6× with TBST over 30 min, incubation with HRP conjugated secondary goat anti-mouse HRP conjugated antibody (Invitrogen) in 5% nonfat dry milk in TBST for 1 h and washing with TBST 12× over 1 h. After a final wash with 1× TBS, horseradish peroxidase conjugates were visualized using ProSignal Dura ECL Reagent (Prometheus) on a ChemiDoc Touch Imaging System (Bio-Rad).

Tissue Culture.

HEK293T cells were cultured as adherent cells in tissue culture treated petri dishes in a humidified incubator in a 5% CO2 atmosphere at 37 °C using Dulbecco’s modified Eagle’s medium with 4.5 g/L glucose, l-glutamine, and sodium pyruvate (Corning) supplemented with 10% FBS (Gibco). Cells were passaged 1:10 at ~80% confluency using TrypLE Express (Gibco) to detach cells. For Western Blots 0.2 × 10⁶ cells were seeded into clear 6-well dishes, allowed to attach and transfected using FugeneHD (Promega) following the manufacturers protocol. In short, 2 μg DNA was topped up to 92 μL total volume using OptiMEM reduced serum medium (Gibco). Then, 8 μL FugeneHD reagent was slowly added without touching the plastic walls, pipetted up and down to mix the reagents, and incubated for 10 min and room temperature. The complete mixture was then added dropwise to the culture dish and incubated for further 24 h. For viability assays, 0.01 × 10⁶ cells were seeded into a white-walled 96-well plate (Corning Costar) and transfected with 10 μL of FugeneHD transfection mixture prepared as above and again incubated for 24 h afterward. Measurements were recorded in triplicates.

Protein Expression.

BL21(DE3) chemically competent cells (New England Biolabs) were transformed with a pET21 vector encoding evolved Protease 5 preceded by an N-terminal His-tag and a GB1 solubility tag. The substrate was expressed using the same cell line with a pET21 plasmid encoding a fusion of an N-terminal His-tag, YFP, full-length human α-Synuclein, and CFP. To express either protein, a single colony was picked and used to inoculate an overnight culture of 2xYT with maintenance antibiotic. The next day, the culture was diluted 1:100 into a 4 L baffled flask containing 1 L of 2xYT with maintenance antibiotic, grown to OD = 1, and then cooled on ice for 1 h. IPTG was added to a final concentration of 0.2 mM and the flask was shaken at 200 rpm at 18 °C overnight. Cells were pelleted, resuspended in 100 mL ice-cold lysis buffer containing cOmplete EDTA free protease inhibitor (Roche), 300 mM NaCl, and 20 mM HEPES and lysed by sonication using a Q700 (Qsonica) using the following parameters: Amplitude 15, 1 s on time, 2 s off time, and for a total processing time of 15 min. The sample was kept on ice during the sonication. Cell debris was removed by centrifugation at 29,000 rpm, the supernatant was further cleared by filtration, 5 M imidazole (neutralized by HCl, pH = 7) was added to a final concentration of 10 mM, and a pre-equilibrated Ni-NTA agarose resin (Qiagen) column was then incubated with this cell lysate for 1 h at 4 °C. The lysate was allowed to exit the column by gravity flow, the remaining resin was washed with lysis buffer containing 25 mM imidazole followed by elution of the desired protein using lysis buffer containing 250 mM imidazole. The imidazole was removed from the eluted protein using a PD10 column (Cytiva).

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article SI Appendix.