Threonine Aldolase-Catalyzed Enantioselective α-Alkylation of Amino Acids Through Unconventional Photoinduced Radical Initiation

Abstract

Visible light driven pyridoxal radical biocatalysis has emerged as a promising strategy for the stereoselective synthesis of valuable non-canonical amino acids (ncAAs). Previously, the use of well-tailored photoredox catalysts represented the key to enable efficient pyridoxal phosphate (PLP) enzyme-catalyzed radical reactions. Here, we report a PLP-dependent threonine aldolase-catalyzed asymmetric α-C–H alkylation of abundant amino acids using Katritzky pyridinium salts as the alkylating agents. The use of engineered threonine aldolases allowed for this redox-neutral radical alkylation to proceed efficiently, giving rise to challenging α-tri- and tetrasubstituted ncAA products in a protecting-group-free fashion with excellent enantiocontrol. Mechanistically, this enantioselective α-alkylation capitalizes on the unique reactivity of the persistent enzymatic quinonoid intermediate derived from the PLP cofactor and the amino acid substrate to allow for novel radical C–C coupling. Surprisingly, this photobiocatalytic process does not require the use of well-established photoredox catalysts and operates through an unconventional photoinduced radical generation involving PLP-derived aldimine. The ability to develop photobiocatalytic reactions without relying on classic photocatalysts or photoenzymes opens up new avenues for advancing stereoselective intermolecular radical reactions which are not known in either organic chemistry or enzymology.

Graphical Abstract

Over the past several years, using visible light to unveil novel enzymatic activities, photobiocatalysis has emerged as an appealing strategy for discovering synthetically useful radical reactions,^1–3 including those that are not known in either organic chemistry or enzymology.^4–6 By leveraging the intimate substrate-protein interactions within the enzyme’s active site, photobiocatalysis illuminates a new avenue to exert stereocontrol over free radical-mediated transformations, a challenging task eluding small-molecule catalysts.^7–9 By capitalizing on the excited-state cofactor redox properties of NAD(P)H^{1–2, 10–12}- and flavin^{1–4, 13–36}-dependent enzymes, including ketoreductases (KREDs),^{1–2, 10–11} ene reductases (EREDs)^{1–2, 4, 13–35} imine reductases (IREDs)¹² and fatty acid photodecarboxylases (FAPs),³⁶ a range of photoenzymatic radical transformations not encountered in the biological world were developed. Recently, by exploiting the synergy between well-established photoredox catalysts and biocatalysts, cooperative photobiocatalysis has allowed the repurposing of a wider range of enzymes,^{5–6, 37–44} including pyridoxal phosphate (PLP)^5–6 and thiamine pyrophosphate (TPP)^43–44-dependent enzymes lacking a photoredox cofactor, to catalyze stereoselective free radical reactions. To further advance the field of new-to-nature photobiocatalysis, new strategies to generate reactive radical intermediates without relying on coenzyme photochemistry or well-established photosensitizers are highly desirable.

Our laboratory has been developing new strategies for stereoselective radical biocatalysis.^45–47 In 2023, enabled by synergistic photobiocatalysis, we reported pyridoxal radical biocatalysis, where we used photoredox catalysis to unlock radical activities of PLP-dependent tryptophan synthases for stereoselective β-dehydroxylative C–C coupling.⁵ Very recently, we expanded this radical pyridoxal enzymology to PLP-dependent threonine aldolases^48–55 to allow for the stereoselective α-functionalizations of simple and abundant amino acids using organoboron reagents under oxidative conditions.⁶ To simplify this PLP enzyme-catalyzed radical α-C–C coupling not previously known in organic chemistry or biochemistry, we set out to devise redox-neutral conditions for radical coupling by employing easily accessible electrophilic radical precursors. If successfully implemented, this redox-neutral radical α-C–H functionalization process would eliminate the use of stoichiometric quantities of oxidants, allowing a broader range of enantioenriched non-canonical amino acids (ncAAs) to be prepared with enhanced efficiency without protection group manipulations. In light of ncAA’s essential role^56–58 as building blocks in clinically significant peptide therapeutics,⁵⁸ bioactive natural products,⁵⁹ and functional unnatural proteins⁶⁰ including proteins for photobiocatalysis,^61–62 this biocatalytic method for enantioselective amino acid synthesis will find use in an array of research areas.

Our postulated catalytic cycle is depicted in Scheme 1. Starting from the internal aldimine form of a judiciously selected threonine aldolase I,^48–54 transimination with the amino acid substrate 1 would afford an external aldimine II. Protein-enabled α-deprotonation⁴¹ of II would furnish a persistent quinonoid intermediate III housed in the enzyme pocket. If a simple and efficient radical generation protocol could be devised, externally formed transient radical IV derived from easily available radical precursors such as Katritzky pyridinium salts^63–64 would enter the enzyme active site and add to the α-position of III in a stereocontrolled fashion, leading to a nitrogen-centered radical V. Further electron transfer/proton transfer (ET/PT) of III would then generate a new external aldimine VI, which upon transimination with the conserved lysine residue would afford the ncAA product 3 and regenerate the internal aldimine, thereby completing the catalytic cycle. Importantly, elegant contemporaneous work from the Hyster lab also exploited this proposal and furnished invaluable methods for amino acid synthesis.⁶⁵

Scheme 1.

PLP-Dependent Threonine Aldolase-Catalyzed Redox-Neutral Enantioselective α-Alkylation of Amino Acids. TmTA active site structure was made using PDB # 1LW5.

At the outset of this study, using glycine (1a) as the amino acid substrate, we examined electrophilic radical precursors that could be activated upon single electron reduction in the presence of Thermotoga maritima threonine aldolase (TmTA),⁴¹ the optimal PLP biocatalyst identified in our previous work.⁶ After extensive optimization, the use of redox-active N-hydroxyphthalimide ester (2a’) did not provide the desired α-C–C coupling product phenylalanine (3a) in part due to amidation of 2a’ with the amino moiety of glycine (1a) (Table 1, entry 1). The employment of Katritzky pyridinium salts (2), a class of widely used radical precursors that could be conveniently prepared from abundant amines,^63–64 furnished 3a in 5% yield and >99:1 e.r. (entry 2). Evaluation of our TmTA variant library resulting from prior studies⁶ revealed that TmTA E88T provided the highest yield of 3a in this reaction (entry 3, 14% yield, >99:1 e.r.). TmTA E88T was thus used as the biocatalyst for further optimization.

Table 1.

Discovery and Optimization of Photobiocatalytic Enantioselective α-Benzylation of Glycine Using Katritzky Salts.

entry	TmTA variant	added catalyst	yield of 3a	e.r. of 3a
1a	TmTA	-	0%	-
2	TmTA	-	5%	>99:1
3	TmTA E88T	-	14%	>99:1
4	TmTA E88T	2 mol% [Ru(bpy)3]Cl2	74%	>99:1
5	TmTA E88T	2 mol% fac-Ir(ppy)3	34%	>99:1
6	TmTA E88T	10 mol% Eosin Y	44%	99:1
7	TmTA E88T	10 mol% Fluorescein	18%	>99:1
8	TmTA E88T	10 mol% Rose bengal	34%	98:2
9	TmTA E88T	10 mol% Rhodamine B	49%	>99:1
10	TmTA E88T	10 mol% Rhodamine 6G	62%	>99:1
11	TmTA E88T	10 mol% 4CzIPN	7%	>99:1
12	TmTA E88T	5 mol% PLP	82% (80%b, 83%c)	>99:1
13			(43%)d	(96:4)d
14	TmTA E88T (1 mol%)	5 mol% PLP	88%	>99:1
15	TmTA W86N	5 mol% PLP	67%	>99:1
16	TmTA E88T H83F	5 mol% PLP	45%	98:2

To further enhance the efficiency of this photobiocatalytic C–C coupling, we evaluated a range of photocatalysts^66–69 to facilitate the redox events involved in this process. The use of transition-metal-based photosensitizers⁶⁶ such as Ru(bpy)₃Cl₂ and fac-Ir(ppy)₃ afforded 3a with improved yield (entry 4, 74% yield, >99:1 e.r., and entry 5, 34% yield, >99:1 e.r., respectively). The use of organic photocatalysts,⁶⁷ including eosin Y (entry 6, 44% yield, 99:1 e.r.), fluorescein (entry 7, 18% yield, >99:1 e.r.), rose bengal (entry 8, 34% yield, 98:2 e.r.), rhodamine B (entry 9, 49% yield, >99:1 e.r.), rhodamine 6G (entry 10, 62% yield, >99:1 e.r.), also resulted in varying degrees of C–C coupling activity enhancement. Interestingly, the use of 4CzIPN led to a slight decrease in the yield of 3a (entry 11, 7% yield, see Table S1 in the SI for further details on photocatalyst effects).

Surprisingly, optimal results were obtained when an additional 5 mol% of free PLP was introduced with no other added photoredox catalysts (entry 12, 82% yield, >99:1 e.r.). The use of Ches and Tris buffers in lieu of KPi buffer led to similar results. Performing the reaction at room temperature resulted in lower yield and ee (entry 13). Importantly, in the presence of 5 mol% PLP, the inclusion of additional exogenous photocatalysts uniformly lowered the yield of 3a (Table S2), suggesting a new radical initiation mechanism is operative. Increasing the loading of TmTA E88T from 0.5 mol% to 1 mol% led to a small increase of yield (entry 14, 88% yield, >99:1 e.r.). Other TmTA variants such as TmTA W86N (entry 15) and TmTA E88T H83F (entry 16) we previously engineered⁶ were found to be less effective. The pH of the reaction medium had a significant impact on this threonine aldolase-catalyzed process. Basic conditions were essential for optimal biocatalytic efficiency (Table S3). Control experiments omitting either TmTA E88T or the visible light source (440 nm) led to no product 3a formation, confirming the photobiocatalytic nature of this transformation (Table S4). Irradiation at alternative wavelengths (e.g., 390 nm, 467 nm and 525 nm) provided inferior yields of 3a (Table S4).

With the optimized photobiocatalytic protocol in hand, we examined the substrate scope of Katritzky salts (2) for the asymmetric α-radical alkylation of glycine (1a, Table 2). It was found that methyl substituents at the para- (3b), meta- (3c), and ortho- (3d) positions of the N-benzylpyridinium salt 2 were compatible with this radical C–C coupling, leading to amino acid products with excellent enantiocontrol (>99:1 e.r.). Electron-donating benzyl groups, including a 4-methoxy (3e) and a 4-methylthio (3f)-substituted benzyl group, were also tolerated under our conditions. Halogen substituents such as a 4-fluoro (3g), a 4-chloro (3h) and a 4-bromo (3i), were also compatible. Additionally, Katritzky salts possessing an electron withdrawing group, such as a trifluoromethoxy (3j), a trifluoromethyl (3k) a carboxylic acid (3l) and a cyano (3m) group, were successfully converted into the corresponding unnatural phenylalanines in good yield and excellent enantioselectivity. We note that these electron-deficient substrates proved much lower yields in our recently developed oxidative C–C coupling with organoboron reagents,⁶ thus showcasing the complementarity of this redox-neutral C–C coupling to our recently developed protocol. Furthermore, sensitive functional groups, including a free carboxylic acid (3l) and a cyano (3m) substituent, were readily accommodated under the mild photobiocatalytic coupling conditions. Substrates with a relatively bulky para-substituent on the benzyl moiety, including a phenyl (3n) and a tert-butyl (3o), were transformed by the PLP enzyme TmTA E88T without additional engineering, highlighting its high levels of substrate promiscuity. Importantly, heterocyclic substrates such as a pyridine (3p) and a thiophene (3q)-containing substrate also underwent smoothy biotransformations. Finally, without additional optimization, unstabilized alkyl radical (3r) could also be engaged by the PLP enzyme under these photobiocatalytic conditions, highlighting the potential of the system to accommodate a wider range of radical intermediates.

Table 2.

Substrate Scope of TmTA E88T-Catalyzed Asymmetric Synthesis of α-Trisubstituted ncAAs

Reaction conditions: 2 (3.0 mM), 1a (30 mM), PLP (0.15 mM), 0.5 mol% TmTA E88T, hv (440 nm), 200 mM KPi buffer, pH 8.5, DMSO (3% v/v), 50 °C, 12 h.

^a1 mol%TmTA E88T was used.

We next investigated the photobiocatalytic transformation of α-branched amino acids using alanine (1b) as the model substrate (Table 3). The synthesis of such α-tetrasubstituted amino acids using powerful native biocatalysis based on transaminases⁷⁰ or ammonia lyases⁷¹ is unknown, underscoring the challenge of this biotransformation. To our satisfaction, without further engineering, TmTA E88T was found to be highly efficient in the enantioselective radical α-benzylation of 1b with Katritzky salt 1a, providing the corresponding α-tetrasubstituted ncAA 4a in excellent yield and > 99:1 e.r. (Table 3, entry 1–3). Interestingly, in contrast to the vast majority of PLP enzymes, TmTA E88T was capable of converting both enantiomeric forms of alanine (i.e., D-alanine and L-alanine, entries 1 and 2), producing the same major enantiomeric product with excellent enantiocontrol. It was found that D-alanine (D-1b, entry 1) could be transformed more efficiently than L-alanine (L-1b, entry 2),⁴¹ but both D-1b and L-1b displayed the same level of enantioselectivity and preference for 4a formation. In accord with this finding, the use of racemic alanine ((rac)-1b) afforded 4a in 82% yield and >99:1 e.r. (entry 3). Together, these results highlighted the ability of threonine aldolase TmTA E88T to allow for the enantioconvergent transformations of α-branched amino acid substrates, a process that remained uncommon among PLP-dependent enzymes. UV-vis spectroscopic studies with TmTA E88T showed that this threonine aldolase mutant formed persistent quinonoid intermediates with both D-alanine (D-1b) and L-alanine (L-1b). Under otherwise identical conditions, a more prominent quinonoid absorption signal (500 nm) was observed with D-alanine than L-alanine, which was consistent with our photobiocatalysis data. Other TmTA variants were found to be less effective than TmTA E88T (see Table S5 in the SI for details). Interestingly, in its native aldolase activity, TmTA E88T did not catalyze the addition of either D- or L-alanine to benzaldehyde (see Table S6 in the SI for details), thereby underscoring the potential of photobiocatalysis to unveil otherwise illusive catalytic activities towards challenging substrates.

Table 3.

Studies on TmTA E88T-Catalyzed Asymmetric α-Benzylation of Alanine^a

We next briefly surveyed the scope of enantioselective benzylation of α-branched amino acid substrates 1 in their racemic form (Table 4). Similar to the transformations of glycine (1a), with alanine (1b), substituted Katritzky salts bearing an ortho- (4b), meta- (4c) and para-methyl (4d) and para-trifluoromethyl (4e) group were effectively transformed to afford the α-tetrasubstituted ncAAs with excellent yields and enantioselectivities. Other larger α-branched amino acid substrates, such as the α-ethyl amino acid 1c, also underwent efficient transformation with high levels of enantioselectivity (4f). Interestingly, when serine (1d) was employed, both the α-hydroxymethyl substituted ncAA (4g) and phenylalanine (3a) formed, different product distributions with regard to 4g and 3a was observed with L-serine (L-1e) and D-serine (D-1e). Starting from D-1e, 4g and 3a formed in 51% yield, 78.5:21.5 e.r. and 10% yield, >99:1 e.r., respectively. With L-1e, 4g formed in only 3% yield and 57% 3a was observed with >99:1 e.r.. The formation of 3a from serine likely involves a retroaldol reaction catalyzed by TmTA E88T, furnishing the quinonoid intermediate upon the release of formaldehyde (CH₂O). This enzymatic quinonoid subsequently underwent radical α-benzylation to provide 3a.

Table 4.

Substrate Scope of TmTA E88T-Catalyzed Asymmetric Synthesis of α-Tetrasubstituted ncAAs^a

To gain insights into the mechanism of this photobiocatalytic enantioselective amino acid α-alkylation, we used cyclopropyl containing radical clock substrates, including cyclopropyl Katrizky salt 2s and α-cyclopropyl amino acid 1e, to probe the intermediacy of transient radical species in this photobiocatalytic C–C bond formation (Scheme 2). When 2s was employed, ring-opening homoallyl coupling product 3s’ formed exclusively in 7% yield and >99:1 e.r. (Scheme 2a). No cyclopropyl containing product 3s was observed, suggesting the intermediacy of radical species and rapid ring opening in the conversion of 2s. In contrast, when the α-cyclopropyl amino acid 1e was applied, no ring-opening product 4h’ was observed, and the desired cyclopropyl-containing ncAA 4h formed in 13% yield and >99:1 e.r. (Scheme 2b). This result suggested that there is likely no radical formation at the α-position of the cyclopropyl amino acid substrate 1f via electron transfer during photobiocatalysis.

Scheme 2.

Photobiocatalytic Transformations of Radical Clock Substrates

To probe the radical nature of this transformation, we next carried out the standard photobiocatalytic reactions with TmTA E88T in the presence of 1 equiv of 2,2,6,6-tetramethyl-1-piperidinyloxy (TEMPO, Table 5). In the presence of TEMPO, the formation of amino acid product 3a was completely inhibited, and TEMPO trapping product 5a formed in 86% yield (Table 5, entry 1). In the absence of the threonine aldolase TmTA E88T, benzyl radical trapping product 5a still formed in 86% yield under these conditions (entry 2). By contrast, in the absence of light, Katritzky salt 2a was fully recovered (99%) and no TEMPO trapping product 5a was observed (entry 3). Together, these data suggested an unusual enzyme-independent, photoinduced radical initiation mechanism.

Table 5.

TEMPO Trapping Studies in the Presence and Absence of TmTA E88T

entry	variation from standard	conv. of 2a	yield of 3a	yield of 5a
1	-	88%	0%	86%
2	no TmTA E88T	90%	0%	86%
3	no hυ (440 nm)	1%	0%	0%
4	no TmTA E88T, no PLP	12%	0%	10%
5	no TmTA E88T, no 1a	21%	0%	19%
6	no TmTA E88T, H2O instead	21%	0%	20%
	of KPi buffer (pH 8.5)		0%
7	no TmTA E88T, no 1a, H2O instead of KPi buffer (pH 8.5)	1%	0%	0%

To further understand this visible light-promoted radical initiation without involving the PLP enzyme, we performed additional TEMPO trapping studies in the absence of the biocatalyst TmTA E88T. When PLP was omitted, the yield of TEMPO-trapping product 5a was reduced dramatically (10% yield, Table 5, entry 4). Similarly, when glycine (1a) was omitted, the yield of TEMPO-trapping product 5a dropped to 19% (entry 5). These results showed that although glycine (1a) and PLP on their own can trigger benzyl radical formation to some degree in the presence of Katritzky salt 2a under blue light irradiation, a substantially more potent species formed from PLP and glycine is responsible for the efficient radical initiation under these photochemical conditions.

We further employed EPR spectroscopy and spin trapping reagent 5,5-dimethyl-1-pyrroline-N-oxide (DMPO) to investigate radical generation under catalytically relevant conditions (Figure 1, see the SI for experimental details). In this study, a range of reaction mixtures were subjected to visible light irradiation at 440 nm in the presence of DMPO. Strong spin-trapped EPR signals were observed only for the reaction mixture containing DMPO, PLP, glycine (1a) and Katritzky salt 2a. Consistent with our TEMPO trapping data (vide supra), this indicates that the aldimine species formed from PLP and glycine (1a) is highly effective for benzyl radical formation. Only a small amount of spin-trapped product was observed when PLP and Katritzky salt 2a were illuminated at 440 nm in the absence of glycine (1a), suggesting much less efficient radical initiation with free PLP. The hyperfine splitting pattern of the major DMPO-spin trapped species, simulated with g = 2.00715, a_iso ¹H = 58.7 MHz (or 2.09 mT) and a_iso ¹⁴N = 40.6 MHz (or 1.45 mT), is consistent with a carbon-centered radical species being trapped (red trace).^72–73 A minor species with g = 2.00738, a_iso ¹⁴N = 44.2 MHz (or 1.57 mT) and without ¹H splitting was also detected, suggesting the formation of a minor product without a β-H (yellow trace).

Figure 1.

Room temperature X-band CW EPR spectra of DMPO-spin trapped reaction mixtures before and after blue light irradiation. Red trace: simulation of the spin trapped benzyl radical species; yellow trace: simulation for the minor species.

Time-resolved luminescence decay study using time-correlated single photon counting (TCSPC) revealed that the dominating excited-state decay pathway (90% relative amplitude) of the quinonoid state of our engineered threonine aldolase TmTA E88T is less than 0.2 ns (see the SI for details). This level of lifetime is generally considered too short for efficient bimolecular diffusional quenching of excited state species.⁷⁴ As the sterically encumbered nature of the N-benzyl-2,4,6-triphenylpyridinium salt 2a precludes its pre-association with the quinonoid species in the enzyme active site, a photoenzyme mechanism involving photoinduced electron transfer between the excited-state enzymatic quinonoid with Katritzky salt 2a is less likely to occur under the current conditions. We hypothesized that the aldimine species derived from the PLP cofactor and the amino acid might be responsible for photoinduced radical initiation. As can be seen from Figure 2, the steady-state emission spectra of the aldimine derived from PLP and glycine (1a) or alanine (1b) exhibits a broad peak with hypsochromic shift compared to that of PLP alone (see SI for the spectra). At pH = 9.0, the decay of the PLP excited states is faster than the instrument detection limit of 25 ps (Table 6). In contrast, significantly longer-lived emissive excited states were observed with PLP aldimines derived from glycine or alanine at both pH = 7.5 and 9.0 (Table 6 and Figure 2, right), which is consistent with previous studies.^75–76 The relatively long-lived excited states of PLP aldimine or their photoproducts are kinetically feasible for the photoinduced radical initiation from Katritzky salts. Furthermore, PLP aldimine exhibited an increased molar absorptivity at λ = 440 nm compared to free PLP (Figure 2, left), indicating the former is a more potent photosensitizer under blue light irradiation.

Figure 2.

Luminescence decay of PLP and PLP aldimine derived from 1a and 1b.

Table 6.

Fitting of PLP and aldimine luminescence lifetimes (pH = 9)

Photosensitizer	decay lifetime τ/ns (rel. ampl. /%)
PLPa	< 0.01
PLP + glycine (1a)b	0.167 (55)	6.53 (45)
PLP + alanine (1b)b	0.159 (45)	6.15 (55)

^aluminescence at λ = 545 nm.

^bluminescence at λ = 500 nm.

Based on these results, we propose that the aldimine species formed spontaneously from PLP and glycine (1a) plays a critical role in the photoinduced formation of benzyl radical from Katritzky salt 2a (Scheme 3). Under blue light irradiation, this PLP aldimine species is much more efficient in benzyl radical formation with pyridinium 2 compared to its constituent PLP or glycine. This unconventional PLP aldimine-enabled free radical initiation represents a departure from classic radical generation mechanisms previously exploited in new-to-nature photobiocatalysis, opening up new possibilities for radical biocatalytic reaction development. Once formed, this benzyl radical would then diffuse into the enzyme active site to engage the quinonoid species for α-C–C bond formation. Our kinetic isotope effect study with glycine and glycine-d₂ revealed a k_H/k_D value of 2.2 (see the SI for details), suggesting the α-deprotonation might be involved in the rate-determining step of this photobiocatalytic redox-neutral C–C coupling.

Scheme 3.

Proposed Photoinduced Benzyl Radical Initiation from 2a and PLP Aldimine.

In summary, we developed a visible light-driven, threonine aldolase-catalyzed enantioselective α-alkylation of simple amino acids using Katritzky salts as the radical precursors. A range of amino acid substrates, including glycine and other unprotected α-branched amino acids, were readily accommodated under these photobiocatalytic conditions, giving rise to α-tri- and tetrasubstituted amino acid products with excellent enantiocontrol without using protecting groups. This threonine aldolase-catalyzed redox-neutral C–C coupling leverages the unique reactivity of PLP-bound quinonoid intermediate, further advancing the general concept of pyridoxal radical biocatalysis⁵ for the discovery of intermolecular asymmetric radical reactions which are not known to either chemistry or biology. Furthermore, the novel PLP aldimine-enabled photoinduced radical generation will inspire the development of other radical-based synthetic methodology and biocatalytic processes.