Structure-activity relationships in catalytic amyloids

Abstract

Short peptide-based assemblies combined with metal ions are simple yet effective platforms for catalysis. This study investigates the catalytic properties of de novo amyloid forming peptides, how their catalytic properties change with a range of hydrolysis substrates and the relationship between catalysis and structural characteristics. We have shown that even slight modifications in peptide sequence have a drastic impact on catalytic efficiency and substrate selectivity. We have also established k_uncat benchmarks for a 4-methylumbelliferyl ester that is increasingly reported as a model substrate in protein design studies. Our studies have yielded catalytic amyloids which display levels of specific activity on par with those reported for the best de novo globular enzymes. This work expands the understanding of catalytic amyloid function and explores design principles for developing minimal peptide scaffolds with tunable esterase-like activity.

Graphical abstract

Catalytic amyloid assembly mediates dual-substrate hydrolysis, yielding designs with activity on par with globular proteins. Studying multiple substrates with an existing catalytic system reveals structure-activity relationships that can be applied to future designs.

1. Introduction

Metalloenzymes are fundamental in catalyzing physiologically critical reactions.[1–4] Over billions of years, diverse metalloenzymes have evolved which display excellent selectivity and activity. Yet, creating robust and functional artificial metalloenzymes for both natural and unnatural reactions remains a largely unsolved task, despite recent progress in the field.[5–12] A common approach to designing metalloproteins relies on a theozyme, a theoretically derived 3D representation of a transition state interacting with side chain residues, which is then subsequently computationally introduced into a protein scaffold and further optimized.[13, 14] This method has had some notable successes but still lacks the high levels of activity found in natural enzymes.[14, 15]

Alternatively, one can try to recreate Nature’s evolutionary path to creating highly efficient metalloproteins. When considering ancestral metalloenzymes it is likely that the original catalysts were far simpler than the complex three-dimensional enzymes we see today.[16–18] It is postulated that early enzymes were abiotically synthesized short peptides that relied on self-assembly to attain productive three-dimensional folds, and this hypothesis is supported by the growing number of examples of catalytically active supramolecular peptide assemblies.[19–21] With the wealth of cryoEM data that is now available, there is a growing recognition that even very short peptides can form highly complex 3D structures that go way beyond the classic cross-beta arrangement.[22–26]

Substrate recognition is key to enzyme design, and so as the field progresses, we start to consider less activated and more chemically complex substrate molecules.[27, 28] Studying multiple substrates provides many advantages: it can illuminate the relative selectivity of the studied catalyst and yields physicochemical insights about the mechanism of the reaction. This allows designers to build primitive substrate recognition sites, or to develop patterns of behavior specific to different peptide sequences. Previous design efforts have focused on how we can optimize designs to a known substrate, but catalytic amyloids are showing themselves to be increasingly diverse and complex, and so we were interested in studying the reactivity of alternate substrates with what should be a relatively promiscuous catalyst (Fig.1).[29–42] In natural systems, generally one observes a bias for a particular substrate, yet catalytic amyloids offer a platform to study metalloprotein behavior with a less specialized system.

Fig. 1.

Catalytic amyloids can hydrolyze coumarin esters upon self-assembly.

Hydrolytic catalysts are a subject of great interest, given their history as a model reaction in enzyme design, facilitated by a range of designs and potential substrates.[15, 43–50] Metallohydrolases are necessary for life, and they can be designed to catalyze industrially valuable reactions, like the breakdown of environmentally hazardous substances.[5, 21, 47, 51–60] We reasoned that simpler catalytic systems would display promiscuous activities as their less specialized structures would allow interaction, binding and hydrolysis of a broader range of substrates. There have been previous minimalist hydrolase models, both with and without co-factors, but these have tended to focus on para-nitrophenyl substrates, and we were interested to expand the range of characterized substrates to better inform future design challenges.[21, 43, 47, 61, 62] We have previously reported a series of zinc-binding amyloid fibrils which catalyze the hydrolysis of para-nitrophenyl acetate, which display levels of activity on par with that of carbonic anhydrase on a per-weight basis, can be mixed into heterofibrillar mixtures, and even preserve their activity when constrained as macrocycles.[19, 30, 37, 41, 63–66] 4-methylumbelliferyl acetate (4MU-Ac) has previously been explored as a substrate for de novo designed serine hydrolases, and other coumarin derived molecules have been studied as metalloprotease substrates.[46, 57, 67, 68] Coumarins are also an important secondary metabolites, controlling signaling and stress response, with 4-methylumbelliferone being found to enhance drought tolerance in M. sieversii.[69]

Ancestral enzymes are believed to have been extremely promiscuous, an evolutionary feature that allowed them broader function until such time as they became specialized or evolved.[70, 71] Given that our catalysts are extremely minimalist, we thought it reasonable that there would be a broad range of substrates hydrolyzed by a wide range of designs, and within this system we can narrow down the design rules for effective hydrolytic catalysis. Thus, we explored the hydrolysis behavior of a range of catalytic amyloids in the hydrolysis of 4-methylumbelliferyl esters and directly investigated the relationship between k_uncat of the given substrate and how that relates to other substrates. We report new findings showing the sequence dependence of designed amyloids and identifying how structure-activity relationships are related across different substrates.

2. Experimental

2.1. Materials and Instrumentation

All reagents were purchased from Ambeed, Fisher Scientific, VWR, Sigma Aldrich, and Chem-Impex and used without additional purification. All synthetic transformations were carried out under a nitrogen atmosphere. Peptides were synthesized using a Gyros PurePep Chorus peptide synthesizer, using room temperature fluorenylmethyloxycarbonyl (Fmoc) solid phase synthesis methods adapted from previous methods.[72] Synthesized peptides were purified using a linear gradient (solvent A: 0.1% TFA in MiliQ-water; solvent B: 0.1% TFA, 10% MilliQ-water, 90% acetonitrile) using the Shimadzu Prominence UFLC, which includes a Phenomenex C4 preparative column (21.2 mm × 250 mm). The purity of peptides was determined using a 1260 Agilent Infinity II HPLC using GL Sciences (Japan) C-18 column (4.6 mm × 150mm). Mass spectra were recorded using a Bruker Microflex LRF MALDI-TOF instrument. High resolution mass spectra were recorded on a Thermo Q-Exactive Oribtrap mass spectrometer. ¹H and ¹³C NMR spectra were measured on a Bruker 400 MHz spectrometer. Chemical shifts are reported in parts per million (ppm) and are calibrated using residual undeuterated solvent as an internal standard (CDCl₃: 7.26 ppm and 77.0 ppm for 1H and 13C NMR, respectively). UV-Vis and kinetic measurements were performed on Agilent Cary 60 UV-Vis spectrophotometer equipped with a multicell attachment using a 1 cm quartz cuvette. CD spectra were recorded on an Applied Photophysics Chirascan VX spectrophotometer. Fluorescence spectra were recorded using a BioTek Eon microplate reader using black Fluotrac 600 96-well plates, operating in the steady state mode at 25°C. Transmission electron micrographs were visualized using a ThermoFisher Spectra 300C Transmission Electron Microscope (operating in 80 kV mode) and the micrographs were captured with a Ceta-D camera.

2.2. Substrate synthesis

4-methylumbelliferyl acetate (4MU-Ac)

To a flame-dried one-neck round bottom flask equipped with a stirrer bar, 7-hydroxy-4-methylcoumarin (0.34 g, 1.9 mmol) was dissolved in DMF (21 mL). This was followed by neat DIPEA (N, N-diisopropylethylamine, 0.35 mL, 2 mmol) to form a yellow solution. Neat acetic anhydride (0.5 mL, 5.3 mmol) was added to the solution dropwise via syringe, which resulted in a colorless solution. This was left to stir at room temperature for 20 minutes and then was precipitated into cold methyl tert-butyl ether. The white solid was collected via centrifugation, dried, and used without further purification. ¹HNMR (CDCl₃, 400 MHz) δ 7.64–7.62 (1H, d), 7.14–7.13 (1H, d), 7.11–7.08 (1H, m), 6.29 (1H, s), 2.45 (3H, s), 2.36 (3H, s); ¹³CNMR (CDCl₃,100 MHz) - δ 168.64, 160.62, 154.12, 153.06, 151.90, 125.13, 118.11, 117.75, 114.04, 110.59, 21.13, 18.32; HRMS (ESI-TOF):- m/z [(M+H)⁺] = 219.0652, found=219.0632.

(±)-4-methylumbelliferyl 2-phenylpropionate (4MU-2PhPro)

In a flame-dried one-neck round bottom flask equipped with a stirrer bar, 2-phenylpropionic acid (0.7 mL, 5 mmol) was dissolved in DMF (5 mL). To this solution was added neat DIPEA (1.75 mL, 10 mmol) and a solution of HCTU (2-(6-chloro-1H-benzotriazol-1-yl)-1,1,3,3-tetramethylaminium hexafluorophosphate) (2.5 g, 5 mmol) dissolved in DMF (12 mL). This mixture was stirred for 10 minutes at room temperature, before addition of 4-methylumbelliferone (0.88 g, 6 mmol) dissolved in DMF (5 mL). This mixture was stirred at room temperature for 7 hours, before diluting the reaction mixture with water (300 mL). The resulting white suspension was washed twice with DCM (100 mL) and the organic layer removed under vacuum to yield the crude solid. This was purified on a Combi Flash Rf 220i using a 0–40 % ethyl acetate and hexane solvent system. ¹HNMR (CDCl₃, 400 MHz) δ 7.48 (1H, s), 7.31–7.17 (5H, m), 6.95–6.93 (1H, m), 6.90–6.87 (1H, m), 6.16 (1H, s), 3.94–3.89 (1H, q), 2.30 (3H, s), 1.56–1.54 (3H, d);¹³CNMR (CDCl₃,100 MHz) δ 172.61, 160.51, 154.13, 153.03, 152.12, 139.66, 128.66, 127.85, 127.56, 125.23, 117.90, 114.35, 110.46, 45.67, 18.73, 18.44; HRMS (ESI-TOF):- m/z [(M+H)⁺] = 309.1122, found= 309.1121.

2.3. Peptide Synthesis

Rink amide resin (0.65 mmol/g) was swollen for 20 minutes before synthesis. Each synthetic step consisted of: 1) Double Fmoc deprotection using 20% 4-methyl piperidine in DMF 2) coupling of target amino acid using a mixture of DIPEA and HCTU for 20 minutes 3) washing twice in DMF 4) N-terminal acetylation for 5 minutes with a capping solution (DIPEA: acetic anhydride: DMF; 2:1:100) to react with any unreacted N-termini. After synthesis, the resin was washed with dichloromethane and dried under a stream of nitrogen. The peptide was cleaved from the resin using a mixture of triisopropylsilane, water and trifluoracetic acid (2.5:2.5:95 v/v) at room temperature for 2 hours resulted in the simultaneous cleavage of peptide from resin and side chain deprotection. The crude peptides were precipitated using cold methyl tert-butyl ether and purified by reverse-phase preparative HPLC. The target mass and purity of the final peptides was confirmed using MALDI-TOF mass spectrometry and reverse-phase analytical HPLC.

2.4. Peptide stock solutions

Stock solutions of peptides were prepared from lyophilized powder by dissolving them in DMSO to a concentration of 1 mM (determined by weight). A 25 μL aliquot of 1 mM peptide solution was mixed with 950 μL of the working buffer (25 mM Tris buffer, pH 8, 1 mM Zn²⁺), and this solution was incubated at room temperature for 24 hours before use in kinetic experiments. For circular dichroism experiments, peptide stocks were prepared in 10 mM HCl, with concentrations determined using calculated extinction coefficients for absorbance at 214 nm, as described previously.[73]

2.5. Circular Dichroism (CD) spectroscopy

CD spectra were recorded between 260–200 nm using a quartz cuvette with a 1 mm path length at 25°C. The data was collected from four scans (4 s averaging time) for each spectrum and corrected for the absorbance of the buffer. The measurements were performed using peptide solutions (10 mM Tris, pH 8) in either the presence or absence of zinc (1 mM). Care was taken to ensure that the sample absorbance never exceeded 1.5 at all wavelengths for reliable ellipticity values. All CD measurements were taken after 24 hours of assembly. Mean residue ellipticity (MRE, deg*cm² dmol⁻¹) values were calculated using the following equation, where $\theta$ is ellipticity (mdeg), L is pathlength (cm), C is peptide concentration ($\mathrm{M}$) and $\mathrm{N}$ is number of residues.

$$\mathrm{M}\mathrm{R}\mathrm{E}=\theta /\mathit{10}\mathrm{*}C\mathrm{*}L\mathrm{*}N$$

2.6. Thioflavin T fluorescence assay

Stock solutions of thioflavin T (1 mM) were prepared in water. ThT and peptide stocks were mixed with 25 mM Tris buffer (pH 8) containing 1 mM Zn²⁺ to give a working solution with 100 μM peptide and 50 μM ThT. Λ_ex = 440 nm Λ_em = 460–650 nm. All fluorescence measurements were taken after 24 hours of fibril assembly.

2.7. pNPA kinetic assays

p-Nitrophenol acetate solutions in buffer were freshly prepared by dilution of a 0.1 M acetonitrile stock to a final concentration of 195 μM, 375 μM, 570 μM and 750 μM in 25 mM Tris, 1 mM Zn, pH 8.0.[73] Substrate solutions of variable concentrations (150 μL) were added to Zn(II)-peptide pH 8 solution (100 μM, 50 μL) or Zn(II) in buffer. The final acetonitrile content was 2% in all reaction mixtures. Upon mixing, absorbance of the product (ε_{405 nm} = 16600 M⁻¹ cm⁻¹) was monitored. Kinetic parameters were obtained by fitting the data to the Michaelis–Menten equation: ${\mathrm{v}}_0={\mathrm{k}}_{\text{cat}}{\left[\mathrm{E}\right]}_0{\left[\mathrm{S}\right]}_0/({\mathrm{K}}_{\mathrm{M}}+{\left[\mathrm{S}\right]}_0)$, where ${\mathrm{v}}_0$ was derived from the initial linear portion of the reading and the extinction coefficient of the product.[19] In cases where ${\mathrm{K}}_{\mathrm{M}}$ was too high to be reliably determined, the data was fit to a linear slope and ${\mathrm{k}}_{\mathrm{cat}}/{\mathrm{K}}_{\mathrm{M}}$ was estimated from slope/[E].

2.8. 4MU-Ac kinetic assays

The reaction was monitored at 377 nm, with the extinction coefficient (8,195 M⁻¹ cm⁻¹) of 7-hydroxy-4-methylcoumarin at pH 8, which we determined experimentally.(Fig. S1)[57] Substrate stock solutions of 4MU-Ac (3 mM, 6 mM, 12 mM, and 18 mM) were prepared in DMSO by weighing out the appropriate amount of powdered substrate. To form the catalyst solutions, 950 μL of 25 mM Tris pH 8 with 1 mM Zn²⁺ was mixed with 25 μL of 1 mM DMSO stock peptide or 25 μL of DMSO (for blank). Next, 25 μL of substrate stock solution (3 mM, 6 mM, 12 mM or 16 mM) was added, bringing the final peptide concentration to 25 μM, with a final DMSO concentration of 5%. The increase in product absorbance was monitored at 377 nm. The Michaelis-Menten kinetic equation was fitted to the initial measured rates of reaction, which were obtained by calculating the slope of the initial linear portion of the measured increase in absorbance.

$${\mathrm{V}}_0=\left({\mathrm{k}}_{\text{cat}}\times \left[\mathrm{E}\right]\times \left[\mathrm{S}\right]\right)/\left(\left({\mathrm{K}}_{\mathrm{M}}+\left[\mathrm{S}\right]\right)\right)$$

where ${\mathrm{k}}_{\text{cat}}$ is the turnover rate of the reaction, [$E$] is the concentration of peptide, ${\mathrm{K}}_{\mathrm{M}}$ is the Michaelis constant, and [$\mathrm{S}$] is the starting substrate concentration.

The buffer-catalyzed hydrolysis (k_uncat) of 4MU-Ac was evaluated at pH 8.0. We varied the buffer concentration (10 mM, 25 mM, 50 mM, 100 mM, and 200 mM) while maintaining a substrate concentration of 450 μM, with a final DMSO concentration of 5% in all cases. The substrate was mixed with the respective buffer solution containing 5% DMSO, and the reaction was monitored in a sealed 96-well plate. The pseudo-first-order exponential equation

$$\mathrm{Y}=\mathrm{a}[1–\mathrm{e}\mathrm{x}\mathrm{p}(–\mathrm{b}\mathrm{x}\left)\right]$$

was applied to the time-dependent absorbance data. Here, “$a$” denotes the maximum absorbance, while “$b$” represents the observed rate constant (k_uncat). For each buffer concentration, the apparent k_uncat was derived from the fitted b-values. The resulting rate constants displayed a linear correlation with buffer concentration. A straight line emerged when plotting k_uncat against buffer concentration, with the intercept indicating the rate of background hydrolysis.

2.9. 4MU-2PhPro Kinetic assays

The reaction was monitored at 377 nm, with the extinction coefficient (8,195 M⁻¹ cm⁻¹) of 7-hydroxy-4-methylcoumarin at pH 8, which we determined experimentally.(Fig. S1)[57] Substrate stock solutions of 4MU-2PhPro (15 mM) were prepared in MeCN by weighing out the appropriate amount of powdered substrate (the final concentration of acetonitrile in the reaction mixture was 2%). To achieve a final substrate concentration of 30–60 μM, 150 μL of a fresh substrate solution in 25 mM Tris 1mM Zn²⁺ (pH 8) was added to 50 μL of buffer (25 mM tris 1mM Zn²⁺, pH 8) or 50 μL of peptide (40 μM) solution in tris buffer, 1mM Zn²⁺ at pH 8 (to a resulting peptide concentration of 10 μM). The increase in product absorbance was monitored at 377 nm. The Michaelis-Menten kinetic equation was fitted to the initial measured rates of reaction, which were obtained by calculating the slope of the initial linear portion of the measured increase in absorbance.

2.10. Transmission Electron Microscopy (TEM)

Peptide samples were prepared in the same way as described above for kinetic experiments. 10 μL aliquots (100 μM peptide, 1 mM Zn²⁺) were adsorbed for 2–5 minutes on glow-discharged 200-mesh copper grids (Ted Pella). Grids were then stained using 2% (w/v) uranyl acetate, adding a 10 μL drop and allowing it to adsorb for 20 seconds. Excess uranyl acetate was then blotted using filter paper, and the grids were stored under vacuum until viewing.

3. Results

3.1. Synthesis and characterization of 4MU-Ac

4MU-Ac was synthesized as summarized in the experimental section using an optimized one-pot method derived from Ghosh and co-workers through the acetylation of 4-methylumbelliferone with acetic anhydride, which furnished the desired substrate quantitatively and in high purity.[74] The NMR and ESI-MS spectra are reported in Figures S2–S4.

We determined the k_uncat values for 4MU-Ac and pNPA at pH 8 (Fig. 2). The k_uncat values are critically important for relative comparison of activities of different substrates. To the best of our knowledge the k_uncat values of 4MU-Ac have not been explicitly determined in literature. To obtain these values we employed an extrapolation to zero buffer concentration, and we found the determined values to be around threefold less than the reported k_uncat for pNPA reported under similar conditions.

Figure 2.

Plots showing the relationship between buffer concentrations and the spontaneous reaction at pH 8: a) Tris buffer with pNPA; b) HEPES buffer with pNPA; c) Tris buffer with 4MU-Ac; d) HEPES buffer with 4MU-Ac; e) The determined k_uncat values for pNPA and 4MU-Ac at pH 8 using two different buffers.

3.2. Peptide design, characterization, morphology and assembly

We screened a previously established library of amyloid peptides in the presence of zinc, as well as diversified the library to identify the role of and fine tune the positions of various residues in the sequence.[19] Specifically, we expanded the peptide library to include Ac-IQIHIHI-NH₂, the reversed sequence of one of the most active previously reported amyloid peptides, and Ac-QIHIHI-NH₂, to test the importance of a unified hydrophobic backbone, as well as Ac-FHFHFQF-NH2, an isoleucine to phenylalanine substitution previously unexplored (Fig. 3a). The purity and identities of all peptides was confirmed with analytical HPLC and MALDI-TOF mass spectrometry (Figs. S8–10).[19] The CD spectrum of Ac-IQIHIHI-NH₂ in the presence of Zn²⁺ displays a pronounced negative peak at 210 nm, consistent with amyloid β-sheet structure (Fig. 3b) observed in other characterized zinc-amyloid species. Ac-QIHIHI-NH₂ in the presence of Zn²⁺displays shows a much less intense negative ellipticity, and the minima shifts towards 218 nm, reminiscent of a traditional β-sheet spectrum. Ac-FHFHFQF-NH₂ in the presence of Zn²⁺ displays a broad shoulder at 218 nm and a notable negative peak around 204 nm, suggesting that the peptide forms a well-organized structure with β-sheet attributes.

Fig. 3.

(a) The design pathways explored to alter the most active studied sequence. b) CD spectra of Ac-FHFHFQF-NH₂ (red), Ac-IQIHIHI-NH₂ (blue) and Ac-QIHIHI-NH₂ (green); 10 mM Tris pH 8, 1 mM Zn²⁺, 100 μM peptide. c). ThT fluorescence spectra of buffer (Black), Ac-FHFHFQF-NH₂ (red), Ac-IQIHIHI-NH₂ (blue) and Ac-QIHIHI-NH₂ (green); conditions: 25 mM Tris pH 8, 1 mM Zn²⁺, 100 μM peptide, 50 μM ThT. Negatively stained micrographs of d) Ac-FHFHFQF-NH₂, e) Ac-IQIHIHI-NH₂, and f) Ac-QIHIHI-NH_2. Conditions: 25 mM Tris pH 8, 1 mM Zn²⁺, 100 μM peptide,

We observed increased fluorescence of thioflavin T (ThT) in the presence of all peptides compared to the inherent fluorescence of (ThT), consistent of amyloid formation (Fig. 3c), however the intensity of the response varies between the species, with Ac-QIHIHI-NH₂ showing fluorescence only 50% higher than the blank signal. TEM micrographs reveal distinct fibrillar morphologies for Ac-IQIHIHI-NH₂, Ac-QIHIHI-NH₂ and Ac-FHFHFQF-NH₂ in the presence of Zn²⁺ despite very minor sequence variation (Fig 3d–f). Ac-IQIHIHI-NH₂ forms flat, sheetlike assemblies, while Ac-FHFHFQF-NH₂ features well-defined, thicker fibrils with a periodic twist. Even in the absence of a terminal hydrophobic residue, Ac-QIHIHI-NH₂ forms short, thin fibrils in the presence of zinc. We see a change in morphology in the absence of zinc (Fig S11), with the apo form showing a defined fibril morphology like Ac-IHIHIQI-NH₂, but surprisingly the addition of zinc yielded thinner and more flexible amyloid fibrils forming a meshed network.

3.3. Ester hydrolysis kinetics

We screened all amyloid designs for their ability to hydrolyze 4MU-Ac, as well as testing new designs for their pNPA hydrolysis activity (Fig. 4, Fig. S12). In the presence of zinc, all tested peptides exhibited considerably higher activity than the measured background rate (Table 1, Fig. 4a), and we found incubating peptides for 24 hours resulted in higher activity (Fig. S13). The peptide Ac-IHIHIQI-NH₂ possessed the highest measured hydrolysis activity of the fibrils tested in this study (k_cat = 2.0 ± 0.25 × 10⁻² s⁻¹, K_M = 0.6 ± 0.1 mM, k_cat/K_M= 32 ± 7.7 M⁻¹ s⁻¹). This is consistent Ac-IHIHIQI-NH₂ being a promiscuous hydrolase, as it also displayed the highest efficiency in hydrolyzing pNPA.[19] We observe k_cat values an order of magnitude greater than those reported by Lauko et. al. for the same substrate (k_cat = 0.12 × 10⁻² s⁻¹), although our enzyme relies on a zinc ion as a catalytic center rather than a serine-histidine pair.[67] We screened the most active peptides in two different buffers, Tris and HEPES pH 8, and found that the catalytic efficiency is several-fold slower in HEPES than it is in Tris (Fig. S14). The differences are largely due to decreased k_cat, even though Ac-IHIHIQI-NH₂ displays a lower Michaelis constant in HEPES buffer than Tris. We have previously seen that Tris can lower reactivity in copper-fibril catalyzed reactions but in this case the behavior is reversed.[29] This finding underscores the effects of buffer on assembly and metal coordination

Fig. 4.

a). Initial rates of 4MU-Ac hydrolysis catalyzed by different peptide assemblies (b). Determined k_cat values of different peptide assemblies for the hydrolysis of 4MU-Ac. (c). K_M values for different peptide assemblies for the hydrolysis of 4MU-Ac. Blue; Ac-IHIHIQI-NH₂, Black; Ac-FHFHFQF-NH₂, Green; Ac-VHVHVQV-NH₂, Red; Ac-LHLHLQL-NH₂. Reaction conditions: 25 mM Tris buffer, pH 8, 25 μM peptide, 1 mM Zn²⁺, 5% DMSO.

Table 1.

Kinetic parameters for 7-acetoloxy-4-methylcoumarin hydrolysis catalyzed by peptide assemblies in 25 mM Tris, pH 8, 1 mM Zn²⁺, 5% DMSO.

		4-MU Ac			pNPA
	Peptide	kcat (s−1)	KM (mM)	kcat /KM (M−1 s−1)	kcat (s−1)	KM (mM)	kcat /KM (M−1 s−1)
1	Ac-QIHIHI-NH2	0.008 ± 0.0007	0.93 ± 0.12	8.6 ± 1.4	0.005 ± 0.0003	0.16 ± 0.03	33.4 ± 7.6
2	Ac-IQIHIHI-NH2	0.003 ± 0.0001	0.36 ± 0.02	8.5 ± 0.6	0.012 ± 0.0002	0.17 ± 0.01	67 ± 4.4
3	Ac-FHFHFQF-NH2	0.007 ± 0.0004	0.33 ± 0.04	20.8 ± 2.7	0.02 ± 0.0008	0.38 ± 0.03	50 ± 5.0
4	Ac-LHLHLKL-NH2	n.d.	n.d.	3.3 ± 0.08	n.d. *	n.d.*	12 ± 2*
5	Ac-LHLHLRL-NH2	n.d.	n.d.	7.2 ± 0.1	0.03 ± 0.004*	1.8 ± 0.4*	18 ± 4*
6	Ac-LHLHLYL-NH2	n.d.	n.d.	8.6 ± 0.4	n.d.*	n.d.*	13 ± 5*
7	Ac-LHLHLQL-NH2	0.006 ± 0.0002	0.23 ± 0.02	26.4 ± 2.8	n.d.*	n.d.*	30 ± 3*
8	Ac-VHVHVQV-NH2	0.008 ± 0.0004	0.29 ± 0.03	27.3 ± 3.7	0.016 ± 0.002*	0.5 ± 0.1*	32 ± 2*
9	Ac-IHIHIQI-NH2	0.02 ± 0.0025	0.61 ± 0.12	32.0 ± 7.7	0.026 ± 0.0004*	0.5 ± 0.1*	62 ± 2*
10	Ac-IHIHIRI-NH2	0.01 ± 0.002	1.21 ± 0.38	8.8 ± 3.6	0.04 ± 0.008*	1.9 ± 0.5*	22 ± 8*
11	Ac-VHVHVRV-NH2	0.02 ± 0.002	1.19 ± 0.17	13.2 ± 2.4	0.04 ± 0.003*	1.5 ± 0.2*	26 ± 4*

^*Indicates values taken from previous report by Rufo et. al.[19]

^*n.d. indicates cases where K_M was too high to determine reliably, and so k_cat/K_M was determined from the linear fit of the data.

We investigated the effect of the residues in the hydrophobic core on the ability of the fibrils to catalyze 4MU-Ac hydrolysis. Despite the hypothetically identical coordination sphere of the metal ion, different hydrophobic cores have impact on the effect of turnover number (k_cat). The least active leucine core sequence shows a k_cat of 0.6 × 10⁻² s⁻¹ and replacing the leucines with phenylalanines or valines only yields a moderate increase to 0.7–0.8 × 10⁻² s⁻¹ (Table 1, Fig. 4b). However, the variant with an isoleucine core is almost three times as active as its structural isomer (Table 1, Fig. 4b), consistent with what we observed in previous studies of pNPA hydrolysis where the isoleucine core variants showed much higher k_cat values. Of note, there is also a large difference in the Michaelis constant of the peptides, where the much higher isoleucine core k_cat is accompanied by a larger Michaelis constant, while leucine, valine and phenylalanine core amyloids display similar values, K_M = 0.23–0.33 mM (Table 1, Fig. 4c). These values are sequence specific, as shuffling the sequence to form Ac-IQIHIHI-NH₂ results in a Michaelis constant similar to those observed for the unshuffled leucine, valine and phenylalanine cores. The k_cat concurrently decreases with the relative rearrangement of the metal binding His unit and the auxiliary hydrophilic residue (Gln). All leucine core peptides, regardless of 6-position residue identity, displayed lower catalytic efficiency, with Michaelis constants higher than the substrate range investigated (>450 μM). We also investigated the effect of residues flanking the His-His binding site. We have previously observed success with positively charged residues, but with this substrate we see that the introduction of charged residues, or even tyrosine residues result in a decrease of catalytic efficiency.

To exclude any possibility that the observed formation of 4-methylumbelliferone is a result of a stoichiometric reaction we performed a turnover experiment, where we observed product formation over a prolonged period to verify the regeneration of the enzyme and its subsequent further catalysis of the reaction.[75] The most active peptide, Ac-IHIHIQI-NH_2, showed at least 5 equivalents of product production in the presence of Zn²⁺ (Fig. S15).

Having established core and auxiliary residues which result in high levels of activity, we studied the effect of sequence on reactivity and found considerable variation when shuffling or truncating the lead sequence. When we study Ac-IHIHIQI-NH₂, we see high activity regardless of substrate and β-sheet character which has a strong negative peak at around 208 nm, but Ac-IQIHIHI-NH₂ displays much lower activity in the hydrolysis of 4MU-Ac, and its CD spectrum red shifts its minima to resemble a classical β-sheet. This lower activity suggests that the fine changes in the overall morphology of the β-sheet assembly lead to major changes in activity despite the preservation of overall metal coordination site. This trend does not persist in pNPA hydrolysis, as Ac-IQIHIHI-NH₂ is more active than Ac-IHIHIQI-NH₂, albeit the difference is small. In both cases, Ac-IHIHIQI-NH₂ displays a higher k_cat, but it is the difference in Michaelis constant and how the substrate interacts with the assembly that affects the overall catalytic efficiency, again consistent with substrate binding difference associated to morphology change. Ac-IHIHIQI-NH₂ and Ac-IQIHIHI-NH₂ were not suitable for cryoEM characterization and proved resistant to crystallization efforts. We can observe, using CD and negative stain TEM that even very small sequence changes influence morphology, and consequently the catalytic activity. We have previously shown that Ac-IHIHIQI-NH₂ CD spectrum does not change in the presence or absence of zinc, and a structural model of Ac-IHVHLQI-NH₂ derived from solid state NMR experiments can be used to infer structural changes within the amyloid.[19, 63] However, recent cryoEM models published on similar structures demonstrate a breadth of structural complexity even within amyloids formed from short peptide sequences.[76–78] These tertiary or even quaternary folds are akin to those found in natural proteins. Although we cannot at this time directly observe the structure of the most active fibrils, there is clear evidence that morphology and structure relate to the measured activity.

We took our approach one step further and shortened the isoleucine core. Historically in our amyloid designs, 4 residues in the core sequence have provided stable and active designs. However, given the results seen in our recent work on amyloid oxidative behavior, we studied the effect of N-terminal glutamine.[78] The peptide still assembles, and we can observe clear amyloid character in the TEM, ThT fluorescence, and CD studies (Figure 3, Table 1). This sequence does not show the high activity that we see in the full core variant. In this case the change in morphology has had a negative effect upon the reactivity.

Finally, we studied the behavior of Ac-IQIHIHI-NH₂ and Ac-IHIHIQI-NH₂ catalyzing the hydrolysis of 4-methylumbelliferyl 2-phenylpropionate (4MU-2PhPro) (Fig. S16), a substrate previously studied in early computational design efforts (Table 2).[46] When studying Ac-IHIHIQI-NH₂, we found rapid initial burst-phase activity (k₂/K_S = 89.3 ± 1.8 M⁻¹ s⁻¹) but observed very little catalysis after 1 turnover. These results were on par with those seen for the unevolved previously reported variants. The divide between the two peptides isomers was striking, with the flipped sequence almost an order of magnitude less efficient at catalyzing the reaction (k₂/K_S = 10.5 ± 0.8 M⁻¹ s⁻¹) but displayed the same single turnover behavior. In all cases the Michaelis constant was too high to determine experimentally, suggesting weak interactions between the enzyme and substrate. However, the observed pattern mirrors the behavior observed with 4MU-Ac, with a strong preference displayed for peptide 9.

Table 2.

Kinetic parameters for 4MU-Ac and 4MU-2PhPro hydrolysis catalyzed by different peptide assemblies in 25 mM Tris buffer, pH 8, 1 mM Zn²⁺. The buffer contained 5% DMSO for 4MU-Ac hydrolysis and 2% MeCN for 4MU-2PhPro hydrolysis.

Peptide		4-MUAc			4MU2-PhPro
		kcat (s−1)	KM (mM)	kcat /KM (M−1 s−1)	k2 (s−1)	KS (mM)	k2 /KS (M−1 s−1)
9	Ac-IHIHIQI-NH2	0.02 ± 0.002	0.61 ± 0.12	32.0 ± 7.7	n.d	n.d	89.3 ± 1.8
2	Ac-IQIHIHI-NH2	0.003 ± 0.0001	0.36 ± 0.02	8.5 ± 0.6	n.d	n.d	10.5 ± 0.8

^*n.d. indicates cases where K_M was too high to determine reliably, and so k_cat/K_M was determined from the linear fit of the data.

4. Discussion

Enzyme promiscuity is proposed as an evolutionary trait that allowed for the selection of different functions as needed.[79, 80] It is logical that simple, promiscuous enzymes that would catalyze multiple functions would have served as the earliest examples, as evolution of the enzymes, systems and organisms would refine and specialize the enzymes as needed.[81–83] Amyloids have been previously proposed to serve as early enzymes, and this work demonstrates their promiscuity.[17, 18, 23, 24, 84] The field of catalytic amyloids has grown sufficiently such that we have seen different patterns and designs of amyloids, notably with starkly different reactivities and selectivity.[38, 85] We can start to construct families of amyloids, classifying them in much the same way as traditional enzymes. Within the work described here, we see sequence specificity as well as the relationship between morphology and reactivity. They show enzyme-like activity, demonstrating multiple turnovers in catalyzing a reaction which they were not designed for. We see distinct K_M values for different designs, and for different substrates within the designs. The morphological changes serve to control reactivity and might serve as an early stand-in for the complex three-dimensional structure we see in modern, evolved enzymes.

All tested peptides catalyze the hydrolysis of 4MU-Ac in the presence of zinc, with Ac-IHIHIQI-NH₂ exhibiting the highest efficiency (k_cat/K_M = 32 ± 7.7 M⁻¹s⁻¹). There is a strong structural element in the measured activity, as varying factors such as the peptide backbone, sequence order and specific amino acids at position 6 influence catalytic efficiency. Overall, these results illustrate that modifying both sequence and structure allows for the rational design of catalytic amyloids, providing a flexible framework for developing minimal catalysts.

These fluorescent esters are only ~4 times less activated than para-nitrophenyl acetate, based on their respective k_uncat values, and amongst the recorded examples used in de novo protein design the differences in k_uncat are very small, ranging between approximately 1–5 times (Table S1). This different reactivity is reflected in the overall k_cat/K_M and k_cat values determined here and those in previously reported literature. We see k_cat values largely following the trend, ~4x lower than the k_cat reported for pNPA, and in some cases the determined Michaelis constants are not significantly different from those determined for pNPA either. Given that K_M is central to enzyme design and specificity, we observed a trend opposite to what we expected for the fibrils- even though 4MU-Ac is a larger, aromatic planar substrate, the K_M in most cases was lower or on par for that determined for para-nitrophenyl acetate, suggesting that these amyloids have a higher affinity for this substrate.

Enzyme design still suffers from low activity, as well as the time cost of designing and producing new enzymes with custom activity and the subsequent directed evolution to reach the desired level of activity.[6, 14, 82, 86–94] Directed evolution, when applied successfully, can yield extremely active enzyme designs, however it is not yet a routine process, and the reactions which have been engineered remain limited.[95, 96] Minimalist design routes rely on the minimum requirements to install enzymatic activity and have the advantage of yielding promiscuous enzyme designs as a result.[9, 97] We can apply this minimalist approach to catalytic amyloids also, where installing the active site in a range of backbones yields a variety of highly active catalysts.[19, 36, 37, 39, 98, 99] We have compiled a representative list of de novo protein designs which catalyze methylumbelliferyl esters, with different functionalization, and we see turnover numbers and catalytic efficiencies on par with the levels of activity reported in well-defined, folded enzyme designs (Table S2). Given the similarity in reported k_uncat values as described above, we can discuss the activity across enzymes characterized using similar substrates. Our most active design, zincated Ac-IHIHIQI-NH₂, had a turnover (k_cat) of 0.02 s⁻¹, K_M= 0.61 mM, with an overall efficiency (k_cat/K_M) of 32 M⁻¹ s⁻¹. MID1sc, a helical protein with a zinc co-factor designed using computational principles has a reported catalytic turnover (k_cat) of 0.011 s⁻¹, with an overall efficiency (k_cat/K_M) of 18 M⁻¹ s⁻¹.[57] Recent reports of in silico designed proteins using a RFdiffusion approach have produced serine hydrolase designs “super” and “win”, which showed preference for the butyl and acetyl esters of 4-methylumbelliferone respectively, and yielded designs with respective k_cat values of 0.0014 s⁻¹ and 0.0012 s⁻¹, and efficiencies of 22 M⁻¹ s⁻¹ and 410 M⁻¹ s⁻¹.[67] There are naturally occurring proteins with 4MU-Ac hydrolysis activity too, and Rsp3690 (k_cat = 0.65 s⁻¹, K_M = 0.24 mM, k_cat/K_M = 2700 M⁻¹ s⁻¹) is only 30-fold more active in catalytic turnover, with overall catalytic efficiency two orders of magnitude greater.[68] Given the low molecular weight of the active unit in the fibrils, their specific activity rivals that of the natural enzyme. Per weight calculations (Rsp3690= 39079.5 Da vs Ac-IHIHQI-NH₂ = 979.5 Da) show that the efficiency of the enzyme far exceeds its small size (4.15 × 10³ μmol min⁻¹ mg⁻¹ M⁻¹ vs 1.96 × 10³ μmol min⁻¹ mg⁻¹ M⁻¹).

In this work, we describe a minimalist set of peptides which serve as promiscuous catalysts for the hydrolysis of carboxy esters. All tested peptides catalyze the hydrolysis of 4MU-Ac in the presence of a zinc cofactor, with Ac-IHIHIQI-NH₂ exhibiting the highest efficiency (k_cat/K_M = 32 ± 7.7 M⁻¹ s⁻¹), resulting in levels of activity on par with the best examples of in silico designed enzymes. There is a strong structural element in the measured activity, as varying factors such as the hydrophobic core, sequence order and auxiliary residues at position 6 influence the morphology and catalytic efficiency of the resultant amyloid assemblies. Overall, these results illustrate that catalytic amyloids can act as promiscuous, highly active catalysts, providing a flexible framework for developing metal-responsive, minimalist proteins for a range of biotechnological applications.

Highlights:

• Zinc-binding amyloids promote a variety of hydrolysis reactions

• Substrate variety allows us to explore what drives structure-activity relationships

• Amyloids can promote reactions with efficiencies on par with de novo computationally designed enzymes

• Minimalist systems with asymmetric arrangements possess nuanced activities.