Enzymatic Upgrading of Biomass‐Derived Aldoses to Rare Deoxy Ketoses Catalyzed by Transketolase Variants

Abstract

A sustainable, convenient, scalable, one‐step method for the two‐carbon chain elongation of cheap and biomass‐derived pentoses (l‐arabinose, and 2‐deoxy‐d‐ribose) and hexose l‐rhamnose was developed to produce C_n+2 deoxy ketoses (C‐7 and C‐8) using transketolase, an enzyme catalyzing the quasi‐irreversible transfer of a ketol group from an α‐keto acid to an aldehyde. Deoxygenated ketoses ‐ commonly obtained by chemical synthesis ‐ were afforded through a suitable combination of both nucleophile and electrophile substrates in the presence of rationally designed TK variants. Pyruvate as nucleophile with pentose l‐arabinose (C‐5) as electrophile gave 1‐deoxy‐L‐gluco‐heptulose (C‐7), while ß‐hydroxypyruvate (HPA) as nucleophile with acceptors 2‐deoxy‐d‐ribose (C‐5) and 6‐deoxy‐l‐mannose (l‐rhamnose) (C‐6) led to formation of 4‐deoxy‐d‐altro‐heptulose (C‐7) and 8‐deoxy‐l‐glycero‐l‐galacto‐octulose (C‐8), respectively. These three deoxy ketoses were easily obtained with efficient TK variants under mild conditions with complete or high substrate conversions, good to excellent yields and high diastereoselectivities. This strategy offers interesting prospects to study the biological activities of these three rare and valuable deoxy ketoses on various cellular targets.

A sustainable, convenient, scalable, one‐step method for the two‐carbon chain elongation of cheap, biomass‐derived pentoses (l‐arabinose, and 2‐deoxy‐d‐ribose) and a hexose (l‐rhamnose) was developed to produce C_n+2 deoxy ketoses (C‐7 and C‐8) using transketolase, an enzyme catalyzing the irreversible transfer of a ketol group from an α‐ketoacid to an aldose.

Introduction

Much research has been devoted to the efficient valorization of biomass. Given that carbohydrates represent most of the biomass‐derived organic compounds, great attention is currently being paid to transformations of mono‐, oligo‐, and polysaccharides into valuable compounds.[ ¹ , ² , ³ , ⁴ , ⁵ ] Among monosaccharides, d‐glucose is by far the main building‐block of plant biomass; neverthless, from a synthetic point of view it would be interesting to value several other natural sugars highly abundant in biomass such as l‐arabinose, the second most abundant pentose isolated from hemicellulose and pectin. In addition, we targeted two naturally occuring deoxy monosaccharides including 6‐deoxy‐D‐mannose (l‐rhamnose), abundant in the pectin fraction of plant cell wall, and 2‐deoxy‐d‐ribose, ubiquitous in all living cells.

Enzymatic isomerization and epimerization[ ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ ] are well‐known carbon‐efficient ways to produce aldoses and ketoses from abundant biomass‐derived monosaccharides, but only a few studies have reported their use for the enzymatic synthesis of non‐natural deoxy ketoses. The most notable work in this area is that of Izumori's group: they converted l‐rhamnose into a series of 1‐ and 6‐deoxy ketoses using d‐tagatose‐3‐epimerase ^[10] and different sugar isomerases. ^[11] Common ways to obtain such compounds are total chemical synthesis or regioselective deoxygenation of natural monosaccharides, requiring multistep processes owing to the need for group protection. ^[12] One alternative biocatalytic route is carboligation, which directly generates deoxygenated sugars by the choice of suitable enzymes and substrates. This strategy obviates the protection of alcohol functions and uses mild conditions. Among carboligases, aldolases[ ¹³ , ¹⁴ , ¹⁵ , ¹⁶ ] such as dihydroxyacetone phosphate (DHAP)‐dependent aldolase, fructose 6‐phosphate aldolase (FSA) from E. coli, and transaldolase B from E. coli (TalB) have been largely used for asymmetric synthesis of stereochemically pure deoxy sugars from simple precursors by three‐carbon chain elongation. However, aldolase‐catalyzed reactions most often require phosphorylated substrates and additional enzymes to release free ketoses,[ ¹⁴ , ¹⁵ , ¹⁶ ] making them unsuitable for the synthesis of non‐phosphorylated long carbon‐chain rare ketoses differently configurated on C‐3 and C‐4.In the pentose phosphate pathway, the carboligase enzyme transketolase (TK) catalyzes the upgrading of aldoses phosphate into ketoses phosphate by stereoselective C−C bond formation in the presence of thiamine diphosphate (ThDP) and Mg²⁺ as cofactors (Scheme 1a).[ ¹⁷ , ¹⁸ , ¹⁹ ]

Scheme 1.

Natural and non‐natural TK_gst reactions: a) wild type TK_{gst ‐}catalyzed reversible reaction from physiological phosphorylated aldoses and ketoses; b) TK_gst variant‐catalyzed irreversible reaction from α‐ketoacids (pyruvate or hydroxypyruvate as nucleophiles) and biosourced (deoxy) aldoses as electrophiles for the synthesis of chiral 1‐or 4‐deoxyhetuloses and 8‐deoxy octulose.

To seek commercial viability in large‐scale processes with non‐phosphorylated substrates, the TK reaction was designed to employ ß‐hydroxypyruvate (HPA) as a donor, making the reaction quasi‐irreversible, and it was applied to a wide range of non‐phosphorylated aliphatic and aromatic aldehyde acceptors. TK variants exhibiting enhanced activity toward these substrates have been successfully developed.[ ¹⁷ , ¹⁸ , ¹⁹ ] In parallel, recent studies showed that aliphatic α–ketoacids can be recognized as donor substrates by dedicated TK variants. ^[20] In particular, the thermostable TK from Geobacillus stearothermophilus (TK_gst) ^[21] was engineered to accept pyruvate, 2‐oxobutyrate, and 3‐methyl‐2‐oxobutyrate as donors in place of HPA (Scheme 1b).[ ²² , ²³ , ²⁴ , ²⁵ ]

This work describes the upgrading of biomass‐derived pentoses (l‐arabinose and 2‐deoxy‐d‐ribose) and a hexose (l‐rhamnose) used as electrophiles to produce C_n+2 monodeoxy ketoses (C‐7 and C‐8) using TK_gst as biocatalyst in the presence of two possible α‐ketoacids as nucleophiles: HPA or pyruvate. The TK‐catalyzed reaction between pyruvate and l‐arabinose (5‐C) 3 was expected to give 1‐deoxy‐L‐gluco‐heptulose (C‐7) 7, while the reaction between HPA with 2‐deoxy‐ribose (C‐5) 1 and 6‐deoxy‐l‐mannose (l‐rhamnose, C‐6) 2 would yield 4‐deoxy‐d‐altro‐heptulose (C‐7) 4 and 8‐deoxy‐l‐glycero‐l‐galacto‐octulose (C‐8) 5, respectively (Scheme 1b). These long carbon‐chain monodeoxy ketoses offer interesting prospects to study their potential applications on various cellular targets. Indeed, some studies reported the biological properties of other deoxy heptuloses and deoxy octuloses. Deoxy heptuloses were described as inhibitors of glucose metabolism and in vitro insulin secretion, ^[26] intermediates in the production of pancrastatin, a new‐generation anticancer drug, ^[27] and antimetabolites exhibiting potential herbicide properties.[ ²⁸ , ²⁹ ] Deoxy octuloses such as 3,7‐anhydro‐1‐deoxy‐d‐glycero‐d‐gulo‐2‐octulose were isolated from Brassica rapa and show ability to prevent oxidative stress and neurocytotoxicity.[ ³⁰ , ³¹ ] Recent studies have shown the essential role of octuloses in the metabolism of parasites such as Trypanosoma and various Leishmania species, opening the way to potential octulose analog‐based inhibitors to treat the diseases caused by these parasites. ^[32]

Results and Discussion

Design of TK_gst Variants

To improve TK_gst activities toward the different donor and acceptor couples, we designed variants based on a combination of our previous results reported in the literature. We had previously shown that positions R521, S385 and H462 play a key role in the enhancement of TK_gst activity toward long carbon‐chain non‐phosphorylated aldoses (C‐5, C‐6) used as acceptors, being these residues are directly involved in the interaction with the phosphate group of physiological acceptors as e. g. D‐erythrose‐4‐phosphate substrates (Figure 1a).

Figure 1.

Model of wild‐type TK_gst based on the X‐ray crystal structure of TK_ban (PDB entry 3 M49) with a) natural acceptor substrate D‐erythrose‐4‐phosphate (E4P) and b) hydroxypyruvate (HPA) the common donor used for biocatalytic synthesis. The model was built using Modeler 9.14 and Chimera.

We previously showed that R521Y/H462N gave the greatest improvement toward d‐ribose and its (3S)‐epimer d‐xylose in the presence of HPA, with a 3.5 fold and 3.3 fold increase in activity, respectively, compared to wild‐type. ^[23] Based on these results, to improve TK_gst activity toward 2‐deoxy‐D‐ribose 1 or L‐rhamnose (6‐deoxy‐Lmannose) 2, we investigated to create a library by site saturation mutagenesis (SSM) using NDT codons on these both positions with additionnal S385 position (R521X/S385X/H462X). A total of more than 5000 clones were screened using a semi‐solid phase assay described earlier. ^[33] Of the six best hits discovered after the screening of l‐rhamnose 2 as acceptor, the triple mutant S385D/H462I/R521H was highlighted four times. However, with 2‐deoxy‐d‐ribose 1 as acceptor substrate, we failed to obtain any hit from the screening of this library. Concomitantly, we tested a few TK_gst variants isolated from our own TK libraries, and which were active on polyhydroxylated aldoses. We found that the earlier reported variant L382F/F435Y ^[23] displayed a marked activity toward 2‐deoxy‐d‐ribose.

When using pyruvate as nucleophile in place of the common donor HPA (figure 1b) with l‐arabinose 3 as acceptor, we tested the previously described H102L/L118I/H474S(G) variant, which exhibits increased activity toward aliphatic α‐ketoacid in the presence of aldoses (C2‐C4). ^[25]

Screening of TK_gst Variants Towards Targeted Donor and Acceptor Substrates

TK_gst variants were expressed in E. coli BL21 (DE3) pLysS strain and purified by Ni²⁺ chelating affinity column chromatography. The reactions catalyzed by TK_gst variants were first first run at 50 °C and pH 7.0 in phosphate buffer (50 mM) with both substrates at 50 micromole scale, in the presence of 2 mg TK_gst variant in a total volume of 1 mL. The reactions were monitored by in situ ¹H NMR spectroscopy using TSP‐d₄ as an internal standard, and the product in situ yields (%) were calculated after a reaction time of 24 h (Figure 2).

Figure 2.

Product in situ yields determined by in situ ¹H NMR using TSP‐d₄ as an internal standard after 24 h reaction time using both donors a) HPA or b) pyruvate and different (deoxy) aldoses as acceptors, 2‐deoxy‐d‐ribose 1 ▪, 6‐deoxy‐l‐mannose (l‐rhamnose) 2 ▪, l‐arabinose 3 ▪. The reactions were performed at 50 °C and pH 7.0 with both substrates at 50 micromole scale, in the presence of 2 mg TK_gst variant in a total volume of 1 mL.

The wild‐type TK_gst catalyzed only a slight conversion of HPA as nucleophile (figure 2a) coupled with 2‐deoxy‐d‐ribose 1 as electrophile and none with 6‐deoxy‐l‐mannose (l‐rhamnose) 2. TK_gst variant S385D/H462I/R521H appeared as the best candidate to increase the product in situ yields with 6‐deoxy‐l‐mannose (l‐rhamnose) 2 (8.5‐fold), while these mutations were inefficient toward 2‐deoxy‐d‐ribose 1. The mutations S385D/H462I/R521H are structurally distant from the methyl group of 6‐deoxy‐L‐mannose (C‐6) 2 and cannot directly influence the substrate stabilization. However, the presence of Ile in place of His 462 creates a hydrophobic niche with Phe35, Val467 and the methyl group of Thr461, which could be favorable for the interaction with the methyl group of 6‐deoxy‐l‐mannose 2. This triple variant S385D/H462I/R521H was not suited to increase the conversion of 2‐deoxy‐D‐ribose 1 having a shorter carbon chain, while TK_gst variant L382F/F435Y gave a three‐fold in situ yield increase toward 1 compared to that obtained with wild type TK_gst. We can explain this result by additional stacking interactions with Tyr 435 improving the positioning of the substrate and by more interactions with hydroxyl groups of 1 and the solvent due to the presence of the phenolic hydroxyl group upon the polar replacement Phe435Tyr.

When using pyruvate as donor (Figure 2b), H102/L118I/H474S was revealed as the best mutant to convert l‐arabinose 3, as evidenced by a four‐fold increase in product analytical yield compared to the same reaction catalyzed by wild‐type TK_gst. It appears that the presence of Ile in place of Leu118 in the triple variant H102/L118I/H474S is required, as shown by the decreased conversion with H102L/H474G I. Ile 118 modifies the position of the ThDP C‐2 thiazolium ring, decreasing the distance from the carbonyl group of l‐arabinose 3, and thus favoring the reaction. In addition, H102L and H474S mutations are required to promote pyruvate binding as previously reported. ^[22] The replacement of His for Ser in position 474 is essential to preserve an hydrogen bond donating capacity to the ketol carbonyl moiety and the replacement of His102 by a non‐polar Leu residue improves the binding of a hydrophobic alkyl chain of the substrate.

Synthesis of Deoxy Ketoses Catalyzed by TK_gst Variants

The syntheses were performed with either HPA, lithium salt (55 mg, 0.5 mmol) or pyruvate, sodium salt (55 mg, 0.5 mmol) as the nucleophilic substrates and 2‐deoxy‐d‐ribose 1 (67 mg, 0.5 mmol), 6‐deoxy‐L‐mannose 2 (91 mg, 0.5 mmol), l‐arabinose 3 (75 mg, 0.5 mmol) or L‐mannose 4 (90 mg, 0.5 mmol) 1 or as the electrophiles (Table 1). The reaction proceeded in a final volume of 10 ml phosphate buffer (50 mM) at pH 7 in a 25 mL round‐bottom flask, under argon stream, in the presence of 3 mg (40 μM) per milliliter of TK_gst variant (L382F/F435Y, H102L/L118I/H474S or S385D/H462I/R521H). The reaction mixture was heated at 50 °C in a sand bath and stirred at 100 rpm using a magnetic stirrer equipped with temperature sensor. The substrates and products were quantified by in situ ¹H NMR analysis of aliquots (540 μL) taken from the reaction mixtures at defined intervals and using TSPd₄ at 5 mM as the internal standard, allowing the measurement of the final conversion levels. The products were purified by column chromatography and characterized by ¹H, and ¹³C NMR 1‐D NMR and homonuclear ¹H‐¹H and heteronuclear ¹H‐¹³C 2‐D NMR and HRMS.

Table 1.

Synthesis of deoxy ketoses at 50 °C using TK_gst variants in the presence of hydroxypyruvate (HPA) or pyruvate as nucleophiles and aldoses as electrophiles.

α‐ketoacids[a] Nucleophiles	Aldoses[a] Electrophiles	Deoxy ketoses Products[b]	TKgst variant[a]	Reaction Time (h)	In situ Yield (%)[c]	Isolated yield (%)	d.e. (%)[c]
HPA			L382F/F435Y	48	50	20	95
	2‐deoxy‐d‐ribose 1	4‐deoxy‐d‐altro‐heptulose 4
			S385D/H462I/R521H/	24	98	80	95
	6‐deoxy‐l‐mannose 2	8‐deoxy‐L‐glycero‐l‐galacto‐octulose 5
			/S385D/H462I/R521H	24	100	68	95
	l‐arabinose 3	l ‐gluco‐heptulose 7
			L382F/F435Y	24	97	77	95
	l‐mannose 4	L‐glycero‐l‐galacto‐octulose 8
Pyruvate			H102L/L118I/H474S	48	95	68	95
	l‐arabinose 3	1‐deoxy‐l ‐gluco‐heptulose 6

[a] The syntheses were performed at 0.5 millimole scale in the presence of 3 mg TK_gst variant per mL in a final volume of 10 mL at 50 °C and pH 7. [b] Determined by ¹H NMR and ¹³C NMR (SI). [c] Determined by in situ ¹H NMR using TSP‐d₄ as an internal standard and calculated on the basis of in situ product formation (SI).

For the syntheses performed with HPA as nucleophile, wild type TK_gst is inefficient toward deoxy aldoses 1 and 2 according to the results previously obtained at analytical scale, we selected the TK_gst variant L382F/F435Y with 2‐deoxy‐d‐ribose 1 while we preferred R521H/S385D/H462I with 6‐deoxy‐l‐mannose (l‐rhamnose) 2. The expected 8‐deoxy‐l‐glycero‐l‐galacto‐octulose 5 was obtained with complete conversion of substrates and higher yield (98 %) and shorter reaction time (24 h) compared to 4‐deoxy‐d‐altro‐heptulose 4 (20 % isolated yield and 48 h reaction time), in line with the previous analytical results. We would like to note that the enzymatic synthesis of neither valuable deoxy ketose 4 or 5 had been reported hitherto. In addition, these TK_gst variants can be applied for the synthesis of other valuable C‐7 and C‐8 ketoses with excellent yields and diastereoselectivities (SI). R521H/S385D/H462I led to L‐gluco‐heptulose 7 from HPA and L‐arabinose 3 (while wild‐type TK_gst gave no product) and L382F/F435Y allowed the synthesis of l‐glycero‐l‐galacto‐octulose 8 from HPA and L‐mannose 4. Interestingly, the purification of L‐gluco‐heptulose 7 was conveniently achieved by affinity chromatography on Dowex Ca²⁺ ion‐exchange resin in water, thus avoiding the use of silica gel chromatography and organic solvents.

Pyruvate as nucleophile and l‐arabinose 3 as electrophile in the presence of TK_gst variant H102L/L118I/H474S gave 1‐deoxy‐l ‐gluco‐heptulose 6 with complete conversion of substrates after 48 h and 68 % of isolated yield. However, longer carbon chain 4 and deoxy aldoses 1, 2 used previously with HPA did not lead to the corresponding products.

In all cases, a single stereoisomer was observed by in situ NMR analysis of the reaction mixture, showing the high stereoselectivity of TK_gst variants. The (S)‐configuration of the newly formed stereocenter was proved by comparison of NMR spectra with those of the same stereoisomers of products 4, 5 and 6.

Conclusions

We have developed a sustainable, convenient, scalable, one‐step method for the two‐carbon chain elongation of cheap and biomass‐derived pentoses (2‐deoxy‐D‐ribose, l‐arabinose) and hexose 6‐deoxy‐l‐mannose (l‐rhamnose) to produce the corresponding C_n+2 deoxy ketoses using efficient TK_gst variants in the presence of two α‐ketoacids (HPA or pyruvate). The engineering of TK_gst based on the active site analysis and on the combination of suitable mutations in key positions greatly improved wild‐type TK_gst activity towards both nucleophiles (HPA or pyruvate) and electrophile substrates (3 to 8.5 fold). The most efficient variant selected enabled the synthesis of three rare and valuable deoxy ketoses, of which one never reported before, with good to excellent yields (20 %–80 %) and high diastereoselectivities.

To enhance the sustainability of this strategy, the α‐ketoacids used as nucleophiles (HPA and pyruvate) can be easily obtained in situ from the corresponding amino acids (d‐serine or d‐alanine respectively) with d‐amino acid oxidase (AAO) coupled with TK_gst in a one‐pot process, so avoiding eco‐incompatible and costly synthesis or purchase of the α‐ketoacids, while controlling their instability. ^[25]

This environmentally friendly procedure offers an efficient alternative to the conventional chemical synthesis of deoxy analogs of natural ketoses. It opens interesting perspectives for testing their biological activity on different biological targets as suggested in recent literature.[ ²⁷ , ²⁸ , ²⁹ , ³⁰ , ³¹ , ³² ]

Conflict of Interests

The authors declare no conflict of interest.

Supporting information

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supporting Information

Arbia G., Joly M., Nauton L., Leogrande C., Tittmann K., Charmantray F., Hecquet L., ChemSusChem 2025, 18, e202401834. 10.1002/cssc.202401834

Data Availability Statement

The data that support the findings of this study are available in the supplementary material of this article.