Biomass‐Derived Diformylxylose as a Renewable Solvent for Biocatalysis Applications

TL;DR

This paper shows that diformylxylose (DFX), a green solvent derived from biomass, can effectively support enzyme reactions, especially for ketoreductases and lipases, outperforming traditional solvents.

Contribution

The study introduces DFX as a renewable solvent that enhances enzyme performance and substrate solubility in biocatalysis.

Findings

KRED TeSADH W110A achieved full conversion in DFX at high substrate concentration, outperforming DMSO and DMF.

Lipase CalB reached 95% conversion in DFX at high substrate loading.

DFX supports high enzyme activity for KREDs and lipases but shows limited compatibility with transaminases and imine reductases.

Abstract

Developing sustainable biocatalytic processes requires alternative solvents that support enzyme activity while reducing environmental impact. This study explores the potential to use diformylxylose (DFX), a xylose‐derived green solvent, as a cosolvent in enzymatic reactions, and compares its application to reaction outcomes in conventional solvents such as dimethyl sulfoxide (DMSO) and dimethylformamide (DMF). A comprehensive enzyme panel, including ketoreductases (KREDs), lipases as well as transaminases (TAs) and imine reductases (IREDs) was tested for activity and stability in DFX. In the green solvent, the selected KREDs and the immobilized lipase CalB retained high or even superior catalytic activity compared to conventional media, while the selected biocatalysts from other enzyme classes such as TAs, and IREDs exhibited limited compatibility under the tested conditions underscoring the enzyme‐specific nature of solvent effects. Notably, the KRED TeSADH W110A achieved full conversion when asymmetrically reducing phenyl‐ring‐containing ketones at 300 mM substrate concentration in DFX, significantly outperforming reaction conditions with DMSO and DMF (∼40% conversion). Lipase CalB also exhibited remarkable activity, reaching 95% conversion at 300 mM 4‐nitrophenyl butyrate loading. The findings highlight DFX as a promising alternative solvent for biocatalysis applications, particularly for KRED‐ and lipase‐mediated reactions.

It feels greener in diformylxylose (DFX)! The renewable solvent dissolves aromatic ketones at remarkable concentrations and sustains the performance of robust enzymes such as ketoreductases (KREDs) and lipases. By combining superior substrate dissolution with improved enzyme performance, DFX offers a green medium for selective, high‐load biotransformations.© 2026 WILEY‐VCH GmbH

Figures

Performance of TeSADH W110A in selected solvent systems on a set of ten ketone substrates. Reaction conditions: substrate (100 and 120 mM in case of substate 1 and 4), NADP+ (0.2 mM), cell‐free TeSADH containing extract (25 mg/mL), Tris‐HCl buffer (50 mM, pH 8.0), 2‐propanol (10% v/v) as cosubstrate, and organic solvents (DFX, DMSO, DMF) at 35% (v/v) in the reaction medium. All reactions were carried out at 50°C for 24 h. Conversions were determined via GC or HPLC analysis. Darker shades indicate higher conversions. Data are the mean values ± s.d. of at least two independent replicates.

(a) Effect of solvent composition on enzymatic conversions. Observed conversions when substrate 1 was reduced by TeSADH W110A in different solvent mixtures. Reaction conditions: substrate 1 (various concentrations), solvents tested DFX, DMSO, DMF at 50% (v/v) in reaction medium containing NADP+ (0.2 mM), cell‐free extract of E. coli TeSADH W110A (25 mg/mL), Tris‐HCl buffer (50 mM, pH 8.0), and 2‐propanol (10% v/v) at 50°C, 24 h. (b) Effect of DFX concentration on the conversions of substrate 1 by E. coli TeSADH W110A. Reaction conditions: substrate 1 (30–400 mM), NADP+ (0.2 mM), cell‐free extract of E.coli TeSADH W110A (25 mg/mL), Tris‐HCl buffer (50 mM, pH 8.0), 2‐propanol (10% v/v), and DFX as cosolvent at varying concentrations (18%, 35%, and 50% v/v). All reactions were carried out at 50°C for 20 h. Data are the mean values ± s.d. of at least two independent replicates.

(a) Time‐course of the bioreduction of substrate 11 catalyzed by cell‐free extract of E.coli ChKRED20 in the selected solvent systems. Reaction conditions: substrate 11 (200 mM), cofactor NAD+ (0.2 mM), cell‐free extract (25 mg/mL), Tris‐HCl buffer (50 mM, pH 8.0), 2‐propanol (10% v/v) as cosubstrate, and tested solvents (DFX, DMSO, DMF) at 35% (v/v) in the reaction medium. (b) Time‐course of the bioreduction of substrate 1 catalyzed by cell‐free extract of E.coli TeSADH_W110A in the selected solvent systems. Reaction conditions: substrate 1 (300 mM), cofactor NADP+ (0.2 mM), cell‐free extract (25 mg/mL), Tris‐HCl buffer (50 mM, pH 8.0), 2‐propanol (10% v/v) as cosubstrate, and tested solvents (DFX, DMSO, DMF) at 35% (v/v) in the reaction medium.Reactions were performed at 50°C for up to 48 h. Tris‐HCl buffer without additional solvent served as the control. Conversion rates were monitored over time.

Comparison of the structural stability of KRED_36 (originating from our in‐house KRED library [21]) and TeSADH_W110A in 18% (v/v) DFX solvent using MD simulations. (a) In DFX, the radius of gyration (RoG) of KRED_36 shows a consistent 0.2 nm increase across all five replicates, indicating potential enzyme unfolding. (b) The changes of the RoG observed for TeSADH_W110A have a smaller increase of 0.1 nm, suggesting greater protein stability under these conditions. (c) EvoEF2 total energy analysis of simulation output conformations, revealing lower total energy for TeSADH_W110A (−7251.3 ± 69.6 kcal/mol) compared to KRED_36 (−6408.9 ± 51.7 kcal/mol), which is a pronounced difference considering their similar structural fold.

Tables

TABLE 1. Observed conversions (%) and ee values (%) for selected enzymes in DFX in comparison to conventional solvent system.

Entry	Enzyme	Substrate, concentration, mM	Reaction conditions	Product
				Conversion, %	ee, %
1	TeSADH W110A		DFX	>99 ± 0.2	>99 (S)
			DMSO	38.3 ± 0.3	>99 (S)
			DMF	28 ± 0.8	>99 (S)
			Buffer	20 ± 0.4	>99 (S)
2			DFX	91 ± 0.2	93 (S)
			DMSO	75 ± 0.3	75 (S)
			DMF	67 ± 0.7	59 (S)
			Buffer	91.7 ± 0.4	68 (S)
3	LKADH		DFX	86.2 ± 0.8	>99 (S)
			DMSO	74.9 ± 0.2	99 (S)
			DMF	73.9 ± 0.3	99 (S)
			Buffer	75 ± 0.2	>99 (S)
4	ChKRED20		DFX	77 ± 0.4	99 (R)
			DMSO	75 ± 1.1	97 (R)
			DMF	76.5 ± 0.9	96 (R)
			Buffer	73.4 ± 0.9	99 (R)
5	PpADH		DFX	98.2 ± 0.4	n.d
			DMSO	98 ± 0.3
			DMF	31.7 ± 0.2
			Buffer	46.6 ± 0.2
6	Lipase (CalB)		DFX	95 ± 0.2	n.a
			DMSO	39.4 ± 2.1
			DMF	40.8 ± 2.2
			Buffer	43.8 ± 0.9

Full text

Introduction

1

Biocatalysis enables highly selective and efficient chemical transformations under mild conditions often allowing to improve process metrics, including the process mass index [1, 2, 3, 4]. However, industrial adoption of biocatalytic steps can be hampered by poor substrate solubility in aqueous systems and a concomitant low enzyme stability in organic solvents. While enzyme engineering can be used to address the biocatalysts’ stability limitation in reactions containing dimethyl sulfoxide (DMSO) and dimethylformamide (DMF), the solvents’ limited biodegradability and hepatotoxicity, respectively, pose environmental and operational challenges [5, 6].

In response to these concerns, efforts have been directed toward developing alternative solvents including deep eutectic solvents or those derived from renewable sources such as biomass‐derived cyrene, CPME or 2‐MeTHF [7, 8, 9, 10, 11, 12]. In this context, sugar‐derived solvents have garnered attention for their low toxicity and biodegradability, while also offering tunable polarity and hydrogen‐bonding properties [13]. In detail, key physicochemical parameters can be modulated through structural modifications or by designing solvent mixtures with different sugar‐based components. Such flexibility enables the fine‐tuning of solvent properties to better meet the requirements of the targeted process [9].

A particularly promising example is diformylxylose (DFX), a polar aprotic solvent obtained in one step from biomass‐derived xylose, first introduced and characterized in depth by Komarova et al. [14]. DFX possesses physicochemical properties such as a high π* value (0.92) and strong hydrogen bond acceptor character (*β *= 0.82) that are remarkably close to those of industrial solvents like DMSO and DMF, while offering the added advantages of being bio‐based, inherently biodegradable due to the sugar‐derived backbone, nonmutagenic, and relatively thermally stable [15]. Prior work established DFX as a viable alternative for polar aprotic applications in chemical synthesis, successfully demonstrating its performance in alkylation, hydrogenation, and Heck cross‐coupling reactions [14]. These reactions not only required high solvent polarity but also challenged solvent resilience under catalytic and thermal stress. Furthermore, it was recently shown that DFX can be produced catalytically by using immobilized heteropoly acids for the acetalization of xylose [16] and that the compound can be further upcycled to yield, e.g., furfural [17].

Considering these beneficial properties of DFX, the current study seeks to build a bridge between the green solvent's performance in chemical syntheses and its potential benefit in biocatalytic reactions. Apart from a single, orthogonal study on the use of structurally analogous xylose acetal solvents such as dipropylxylose (DPX) in enzymatic polycondensation with lipase B (CalB) [18], DFX compatibility with enzymes has not yet been investigated—even though biocatalysis represents a key potential application area considering its increasing importance in sustainable synthesis [1]. Given the structural similarities between DFX and biocatalysis‐compatible solvents like DMSO, and considering DFX's favorable Kamlet‐Taft parameters, we hypothesized that DFX could support enzymatic reactions, particularly for hydrophobic substrates and under elevated temperatures. Here, we explore DFX's compatibility with multiple enzyme families including ketoreductases (KREDs), lipases, imine reductases (IREDs) and transaminases (TAs) and benchmark DFX's biocatalytic performance against conventional solvents (aqueous systems, DMSO, DMF), focusing on conversion, substrate scope, and stereoselectivity.

Keto‐reductions and selective ester cleavages are among the most established biocatalytic reactions in industrial synthesis [19, 20] due to the corresponding enzyme's broad substrate scope, operational robustness, and proven scalability [2, 21]. KRED libraries enable access to enantioselective alcohols from diverse ketone substrates, while lipases are frequently employed for selective esterification and transesterification reactions under aqueous or organic conditions [22, 23]. Thus, special emphasis was placed on investigating the DFX performance of TeSADH W110A from Thermoanaerobacter ethanolicus [24] and ChKRED20 from Chryseobacterium sp. CA49 [25], two well‐characterized KREDs, as well as the activity of immobilized lipase CalB [23], a widely used hydrolase.

As a first step in our assessment of using DFX as a cosolvent in biocatalysis reactions, we selected a representative set of thermostable KREDs, IREDs, TAs, and lipases. Enzymes were either ordered as lyophilized powders (e.g., LKADH, ATA008; purchased from Enzymicals AG, Germany), expressed recombinantly via synthetic genes (e.g., ChKRED20 [25] and PpADH [26]; ordered from TWIST Biosciences) or derived from our previously established in‐house KRED library (e.g., KRED36 [21]). All genes were cloned into a pET28b(+) vector and transformed into E. coli BL21(DE3) for protein production. Crude cell‐free extracts or lyophilized whole cells were used directly in the biocatalytic reactions. For cofactor regeneration in NAD(P)H‐dependent reactions (e.g., KREDs, IREDs), an isopropanol‐based recycling system was employed. Product formation was monitored using gas chromatography with flame ionization detection (GC–FID) or liquid chromatography coupled to mass spectrometry (LC–MS), depending on substrate properties.

The most performant enzyme in this study was TeSADH W110A [27], a thermostable KRED that showed exceptional conversion efficiencies in DFX. In the carbohydrate‐based solvent, TeSADH W110A exhibited a broad acceptance of aromatic, aliphatic, and ether‐linked ketones (substrates 1 – 10; Figure 1). For example, when exploring activity of TeSADH W110A on 1‐phenoxypropan‐2‐one (1), DFX supported excellent activity (>99%) and selectivity (>99% ee) (Table 1, Entry 1). Notably, prior studies using aqueous‐organic systems such as buffer/[bmim]BF_4_] reported high conversion (∼96%) and enantioselectivity (∼97% ee) for 1, but only at ≤100 mM substrate concentrations [27]. In contrast, DFX helped the enzyme maintain excellent activity and selectivity (>99%) even at 300 mM substrate load, demonstrating its capacity for high substrate load biocatalysis reactions. In addition, we observed that reactions in DFX not only achieved higher conversion but also consistently improved enantioselectivity, yielding (S)‐alcohols with excellent stereochemical purity. For instance, substrate 2 (a phenyl‐ring‐containing ketone) was almost fully converted in DFX, even at elevated substrate concentration of 300 mM, yielding the expected (S)‐alcohol with an enantiomeric excess of 91%, whereas conversions in DMSO and DMF plateaued at ∼75% and ∼67%, respectively, and yielded e.e. values below 75% (Table 1, Entry 2).

Performance of TeSADH W110A in selected solvent systems on a set of ten ketone substrates. Reaction conditions: substrate (100 and 120 mM in case of substate 1 and 4), NADP+ (0.2 mM), cell‐free TeSADH containing extract (25 mg/mL), Tris‐HCl buffer (50 mM, pH 8.0), 2‐propanol (10% v/v) as cosubstrate, and organic solvents (DFX, DMSO, DMF) at 35% (v/v) in the reaction medium. All reactions were carried out at 50°C for 24 h. Conversions were determined via GC or HPLC analysis. Darker shades indicate higher conversions. Data are the mean values ± s.d. of at least two independent replicates.

As a next step, we investigated the operational temperature window of the biocatalytic reaction by conducting temperature‐dependence experiments between 40°C and 70°C under otherwise identical high‐load conditions. This temperature range was chosen to ensure complete liquefaction of DFX, which is solid at room temperature (T m = 48°C) [14]. These experiments revealed solvent‐specific temperature‐response profiles. Reactions carried out in DFX showed an operational optimum at 50°C for the asymmetric reduction of substrate 1 by TeSADH W110A, while reactions carried out in DMSO exhibited a local, albeit lower, optimum at 60°C and those in DMF displayed a pronounced loss of activity at elevated temperatures (Figure S12).

Building on these findings, we set out to further explore optimal reaction conditions for DFX‐based biotransformations (Figure 2a) and to learn more about the reasons of the TeSADH variant's improved performance. Toward this goal, we varied the DFX concentration between 18% and 50% (v/v) and tested it with substrate 1 concentrations ranging from 30 to 400 mM (Figure 2b); when applying 50% DFX, TeSADH W110A achieved the highest conversion across all tested concentrations of 1. Remarkably, quantitative conversions were maintained at substrate loadings of up to 120 mM, even at reduced DFX concentrations (18% and 35%), before a decline in yields was observed (Figure 2b). We speculate that DFX's high polarity and low volatility not only enhanced substrate solubility but also stabilized enzyme conformation as indicated by protein–solvent molecular dynamics (MD) simulations (vide infra).

(a) Effect of solvent composition on enzymatic conversions. Observed conversions when substrate 1 was reduced by TeSADH W110A in different solvent mixtures. Reaction conditions: substrate 1 (various concentrations), solvents tested DFX, DMSO, DMF at 50% (v/v) in reaction medium containing NADP+ (0.2 mM), cell‐free extract of E. coli TeSADH W110A (25 mg/mL), Tris‐HCl buffer (50 mM, pH 8.0), and 2‐propanol (10% v/v) at 50°C, 24 h. (b) Effect of DFX concentration on the conversions of substrate 1 by E. coli TeSADH W110A. Reaction conditions: substrate 1 (30–400 mM), NADP+ (0.2 mM), cell‐free extract of E.coli TeSADH W110A (25 mg/mL), Tris‐HCl buffer (50 mM, pH 8.0), 2‐propanol (10% v/v), and DFX as cosolvent at varying concentrations (18%, 35%, and 50% v/v). All reactions were carried out at 50°C for 20 h. Data are the mean values ± s.d. of at least two independent replicates.

To benchmark the use of DFX against a broader panel of solvents, we evaluated the TeSADH W110A–catalyzed reduction of substrate 1 in the biomass‐derived solvent Cyrene, CPME, and 2‐MeTHF, as well as in glycerol and MeCN. Under intensified reaction conditions (300 mM substrate 1, 50% (v/v) solvent), DFX afforded the highest conversion (>99% ± 0.2%), outperforming all alternative solvents, with CPME emerging as a close second (89% ± 0.8% conversion) (Figure S13).

Going forward, we turned to ChKRED20, a thermostable KRED known for its ability to reduce structurally diverse prochiral ketones with high stereoselectivity, particularly favoring the production of (S)‐alcohols even under elevated temperature and solvent stress [25]. In our study, ChKRED20 demonstrated comparable robustness in DFX. Using a loading of 500 mM substrate 11, a substituted aryl ketone, ChKRED20 achieved more than 75% conversion in the presence of 35% DFX outperforming all other tested solvents under these conditions (Table 1; Entry 4; Figure S5). Testing ChKRED20 on substrate 1 showed a similar picture (Figure S6). Notably, with 1 (1‐phenoxypropan‐2‐one) and 11 (3,5‐bis(trifluoromethyl)acetophenone), time‐course analyses (Figure 3a) revealed that DFX consistently supported the highest conversion across the entire time range, reaching >95% conversion after 48 h. This highlights not only the enzyme's robustness but also DFX's ability to sustain prolonged biocatalytic activity without enzyme deactivation, even in the presence of hydrophobic substrates. A similar trend was observed for TeSADH W110A with substrate 1, where reactions performed in DFX consistently showed higher conversion over time compared to DMSO, DMF, and buffer‐only conditions (Figure 3b).

(a) Time‐course of the bioreduction of substrate 11 catalyzed by cell‐free extract of E.coli ChKRED20 in the selected solvent systems. Reaction conditions: substrate 11 (200 mM), cofactor NAD+ (0.2 mM), cell‐free extract (25 mg/mL), Tris‐HCl buffer (50 mM, pH 8.0), 2‐propanol (10% v/v) as cosubstrate, and tested solvents (DFX, DMSO, DMF) at 35% (v/v) in the reaction medium. (b) Time‐course of the bioreduction of substrate 1 catalyzed by cell‐free extract of E.coli TeSADH_W110A in the selected solvent systems. Reaction conditions: substrate 1 (300 mM), cofactor NADP+ (0.2 mM), cell‐free extract (25 mg/mL), Tris‐HCl buffer (50 mM, pH 8.0), 2‐propanol (10% v/v) as cosubstrate, and tested solvents (DFX, DMSO, DMF) at 35% (v/v) in the reaction medium.Reactions were performed at 50°C for up to 48 h. Tris‐HCl buffer without additional solvent served as the control. Conversion rates were monitored over time.

These findings emphasize that biocatalytic performance in DFX is determined by the combined effects of enzyme selectivity, substrate properties, and enzyme‐solvent interactions, including solvent effects on enzyme stability (vide infra). For instance, while TeSADH W110A achieved full conversion of 1 and 2 even at 300 mM using 50% DFX (Figures S1 and S3), ChKRED20 proved more effective for substrate 11 (Figure S5), underscoring enzyme‐specific preferences. Substrates bearing ether or aliphatic moieties showed variable conversions, reflecting the interplay between hydrophobicity, polarity, and binding dynamics. Although descriptors such as logP and TPSA (Table S2) offer some guidance, they did not consistently predict reactivity, indicating that substrate solubility, as well as enzyme–substrate and enzyme–solvent compatibility are key determinants for performance in DFX‐based bioreductions.

To probe generalizability, we evaluated the activity of additional ADHs, namely LKADH (provided by Enzymicals) and PpADH from Paracoccus pantotrophus [26] in DFX (Table 1, Entries 3 and 5; Figures S4 and S7). Both enzymes showed excellent conversion (>85%) in DFX above 40°C when applied on bulky ketoesters (e.g., methyl 2‐oxo‐2‐phenylacetate 12). These results underscore DFX's compatibility with structurally diverse oxidoreductases, belonging to different families such as classical short‐chain dehydrogenases/reductases (SDRs; ChKRED20, PpADH, LKADH) and medium‐chain dehydrogenases/reductases (MDRs; TeSADH) [28, 29].

Beyond the oxidoreductases, we tested immobilized lipase CalB as a representative from the hydrolase family. Using a standard ester hydrolysis assay with 300 mM substrate 14, CalB achieved a notable conversion of ∼95% (Table 1, Entry 6, Figure S8). In contrast, the buffer‐based reaction resulted in 43%, while using DMSO and DMF led to 39% and ∼41% product formation, respectively. The observed high conversions bode well for DFX use in selective esterification reactions, especially involving fatty acids and triglycerides as substrates [30]. Specifically, lipases such as CalB catalyze the transesterification of triglycerides with short‐chain alcohols (e.g., methanol or ethanol), yielding fatty acid methyl esters (FAMEs), which are key components of biodiesel [31]. The compatibility of DFX with CalB therefore suggests the possibility for implementing potentially more sustainable biodiesel processes using xylose‐derived solvents.

An analysis of the best‐performing substrates in DFX revealed several common features. Substrates such as phenoxy‐2‐propanone 1, 4‐phenyl‐2‐butanone 2, 4‐fluoroacetophenone 9, 3,5‐bis(trifluoromethyl)acetophenone 11, and methyl 2‐oxo‐2‐phenylacetate 12 all contain aromatic scaffolds and display moderate hydrophobicity. These features likely enhance solubility in the polar aprotic DFX medium at 50°C aligning with prior findings that hydrophobicity and planarity can positively impact biocatalytic efficiency in low‐water or organic media [32, 33]. While high conversions were obtained with KREDs and immobilized lipase CalB, other enzyme classes such as TAs (e.g., Sp*‐*ATA [34]) and a limited set of IREDs did not yield measurable product formation in DFX under the tested conditions despite their reported thermostability (Figures S9 and S10). These results demonstrate that while DFX can be a valuable solvent for biotransformations catalyzed by thermostable enzymes, it is not universally applicable, and enzymatic performance must be evaluated on a case‐by‐case basis.

To understand the enzyme‐dependent tolerance of DFX, we explored the thermal stability of two well‐performing enzymes (TeSADH_W110A and ChKRED20) with two negative examples that did not tolerate DFX as solvent (KRED_36 and TA Sp*‐ATA), (Figures S9 and S11) using nano differential scanning fluorimetry (nanoDSF) (Figure S12). For these experiments, the enzymes were prepared as purified proteins (PP; 1.6 mg/mL) and tested across different solvent conditions (buffer only, as well as 18% and 30% (v/v) DFX, DMSO, and DMF). Across all conditions, both TeSADH and ChKRED maintained high thermal stabilities and were characterized by T m values > 70°C (Figure S12A and S12B), aligning well with their high catalytic performance in DFX. In contrast, TA Sp‐ATA and KRED_36 and PpADH showed considerably decreased thermal stabilities at both DFX concentrations, with T m values dropping below 55°C (Figure S14C–E). Considering that all bioreactions were run at 50°C, enzyme denaturation is a plausible explanation for the low observed performance of Sp‐*ATA and KRED_36.

The varying enzyme stability in DFX was further explored by performing MD simulations of the structurally homologous (RMSD = 2.271 Å) MDR‐type KREDs TeSADH_W110A and KRED_36 [21]. To enable direct comparison of the conformational dynamics, simulations of both enzymes were performed under identical conditions: 18% (v/v) DFX, 10% (v/v) isopropanol, setting the simulation temperature to 50°C. All ionizable residues were protonated according to their expected states at pH 8.0. Analyzing the resulting trajectories revealed that the radius of gyration (RoG) of KRED_36, a common measure of enzyme unfolding, consistently increased by 0.2 nm across all replicates ‐ twice the change observed for TeSADH_W110A (0.1 nm) (Figure 4a,b). This suggests destabilization or partial denaturation of KRED_36, likely caused by its reduced thermal stability at this DFX concentration. Further analysis of the final simulation frames using EvoEF2 [35] showed a total energy of −6408.9 ± 51.7 kcal/mol for KRED_36, compared to −7251.3 ± 69.6 kcal/mol for TeSADH_W110A, which is a pronounced difference considering their similar structural fold (Figure 4c).

Comparison of the structural stability of KRED_36 (originating from our in‐house KRED library [21]) and TeSADH_W110A in 18% (v/v) DFX solvent using MD simulations. (a) In DFX, the radius of gyration (RoG) of KRED_36 shows a consistent 0.2 nm increase across all five replicates, indicating potential enzyme unfolding. (b) The changes of the RoG observed for TeSADH_W110A have a smaller increase of 0.1 nm, suggesting greater protein stability under these conditions. (c) EvoEF2 total energy analysis of simulation output conformations, revealing lower total energy for TeSADH_W110A (−7251.3 ± 69.6 kcal/mol) compared to KRED_36 (−6408.9 ± 51.7 kcal/mol), which is a pronounced difference considering their similar structural fold.

Conclusions

2

The experimental and computational analyses collectively highlight that enzyme compatibility with DFX is governed not only by solvent characteristics but also by factors such as protein flexibility, thermostability, and susceptibility to solvent‐induced unfolding. Notably, we observed that the tested enzyme's thermostability varied depending on the applied solvent system, with differences as large as ∼20°C. As such, we recommend including simple thermal stability assays in DFX as an early step in enzyme selection when choosing to use this bioderived solvent. Taken together, this work identifies DFX as a robust biomass‐derived alternative to conventional polar aprotic solvents in biocatalysis, consistently delivering the highest conversions across the tested solvent set and defining a practical niche for high‐load enzymatic transformations.

By evaluating the performance of several well‐established biocatalyst families (KRED, Lipases, IREDs, TAs) [1, 2, 21, 36, 37] in DFX, this work positions the xylose‐derived solvent as an alternative to conventional solvents in selected biocatalysis applications, with potential relevance in pharmaceutical and fine chemical manufacturing and support the broader implementation of DFX in sustainable biocatalytic workflows. Notably, we found that DFX represents a green alternative to conventional cosolvents such as DMSO and DMF by reducing the carbon footprint of the overall enzymatic reactions; past life cycle modeling showed that the footprint of DFX can be as low = −0.4 kg CO_2, eq,/kg solvent, i.e., carbon negative at factory gate, when produced from corn cobs using green reagents, versus 3.1 and 1.4 kg CO_2, eq,/kg solvent for DMF and DMSO, respectively [15]. Preliminary studies also showed that DFX was “inherently biodegradable” according to OECD standards [15], nonmutagenic according to the AMES test [14] and did not inhibit the growth of algae or bioluminescent bacteria [38]. The xylose‐derived solvent also boosted conversion and selectivity of the targeted reactions in some cases (e.g., TeSADH W110A and ChKRED20 with 1 and 11, respectively). To exploit these beneficial effects further, future studies should explore DFX's compatibility with additional enzyme classes known to possess high thermostabilities such as peroxidases or squalene‐hopene cyclases. Furthermore, enzyme engineering and design may be profitably employed to make more protein scaffolds amendable to DFX‐containing bioreactions.

Experimental Section

3

General Materials and Methods

3.1

All chemicals were purchased from commercial suppliers (Sigma‐Aldrich, VWR, Carl Roth) and used without further purification unless otherwise specified. DMSO and DMF were reagent grade (≥99.9%) and used as received. DFX was synthesized in‐house from D‐xylose using a previously reported procedure [14]. Briefly, 109 g of D‐xylose and 50 g of paraformaldehyde were added to 200 ml of 2‐Me‐THF. 12 g of H_2_SO_4_ (95−97 wt%) were added dropwise while stirring at 400 rpm. The reaction mixture was then heated at 80°C for 3 h while stirring. The resulting solution was then cooled to room temperature. To isolate DFX, the reaction mixture was neutralized with a saturated NaOH solution, filtered, and concentrated in vacuo using a rotary evaporator with a bath temperature of 45°C. The final residue was crystallized by cooling to room temperature and was filtered while washing with EtOH to remove impurities and byproducts. The resulting DFX product was a white crystalline solid of ≥98% purity with a reaction yield of 81% and an isolated yield of 74%. As DFX is a solid at room temperature, it was preweighed and melted at 50°C–55°C until fully liquified. The required volume was calculated based on desired final concentration (e.g., 18%, 35%, or 50% v/v) and added to prewarmed reaction mixtures under stirring to ensure homogeneous distribution. All DFX‐containing reactions were conducted at ≥50°C (unless otherwise noted) to maintain the solvent in its liquid state.

Analytical‐grade solvents were used for extraction and chromatography. GC–MS analyses were performed on an Agilent 8890 GC system equipped with an Agilent 5977B GC/MSD and an automatic liquid sampler. GC–FID analyses were performed on an Agilent 7890B instrument. HPLC and chiral HPLC analyses were carried out on an Agilent 1260 systems using appropriate columns, including Chiralpak OD for enantiomeric separations and a Poroshell 120 EC‐C18 column (2.7 µm, 3 × 50 mm, Agilent Technologies) for achiral analyses.

Enzyme Expression and Preparation

3.2

Members of several enzyme families were evaluated with regard to their compatibility with various solvent systems, including the xylose‐derived solvent DFX. Tested KREDs included the engineered variant TeSADH W110A from Thermoanaerobacter ethanolicus JW 200, ChKRED20 from Chryseobacterium sp. CA49, and PpADH from Pseudomonas putida (all ordered as synthetic genes from TWIST) and LKADH from Lactobacillus kefiri (the latter obtained as a commercial lyophilized formulation from Enzymicals). In addition, we evaluated IREDs p‐IR13, p‐IR355, p‐IR361, and p‐IR23 from the Prozomix IRED gene library (maintained in‐house as glycerol stock) as well as transminases Sp‐ATA (ordered as a synthetic gene) [34] and ATA008 from Enzymicals (lyophilized formulation from commercial supplier). For hydrolase reactions, immobilized Candida antarctica lipase B (CalB, Novozyme 435, 2000 TBU/g) was employed directly as beads. Synthetic genes were cloned into vector pET28b+ with an N‐terminal His‐tag and the encoded enzymes were produced in E. coli BL21(DE3). In detail, plasmids encoding the desired enzymes were transformed into competent cells and grown overnight in LB medium. 5 ml overnight culture was inoculated into 400 ml ZYM‐5052 autoinduction medium supplemented with corresponding antibiotics (50 mg/ml) and incubated at 20°C for 16–20 h.

Cells were harvested by centrifugation (3400 g, 10 min, 4°C), washed with buffers optimized for each enzyme (see Supporting Information), and resuspended in the same buffer. Lysis was performed by ultrasonication (70% duty cycle, output 7–8, total time 2 min) on ice. Lysates were clarified by centrifugation (16,000 g, 20 min, 4°C), and the supernatants were either used directly in biotransformations or stored at –20°C until usage. Unless otherwise noted, total protein concentration was adjusted to 25 mg/mL.

General Reaction Conditions

3.3

Biotransformations were carried out in 2 mL GC vials in a total reaction volume of 200 µL. Reactions were incubated at 50°C (unless otherwise noted) with orbital shaking at 800 rpm for 20–48 h.

The standard reaction mixture contained the substrate (10–1000 mM), 0.2 mM NAD^+^ or NADP^+^ as appropriate, 5 or 10% v/v 2‐propanol (as cosubstrate), 50 mM Tris‐HCl buffer (pH 7.5 or 8.0 depending on enzyme), and 25 mg/mL cell‐free extract (or 5 mg/mL for lyophilized enzymes). Cosolvents (DFX, DMSO, DMF) were added at 18%, 35%, or 50% (v/v), and pre‐equilibrated to 50°C before mixing. Lipase reactions were carried out with substrate 13 or 14 (10–300 mM) and 2 mg CalB beads per vial.

For the time‐course study with ChKRED20, reactions were sampled at 0, 1, 2, 4, 6, 8, 20, 24, and 48 h. Conversions were determined by GC after quenching. All reactions were performed at 35% solvent concentration and 50°C.

Reaction Work‐up and Analytical Methods

3.4

For the product extraction, 1 mL of ethyl acetate was added directly to each biotransformation vial. The mixtures were vortexed and centrifuged (14,000 rpm, 10 min), and the organic phase was transferred to GC vials for analysis.

Most biotransformations (substrates 1–12) were analyzed by chiral GC–FID using a MEGA‐DEX AS‐Beta column (30 m × 0.25 mm × 0.25 µm, MEGA S.r.l., Italy) with an FID detector, using nitrogen (N_2_) as the carrier gas. The injector and detector temperatures were set to 250°C. Oven programs were optimized per substrate (see Supporting Information). Enantiomeric excess (ee) was determined by comparing retention times with racemic standards or commercial references.

For substrates 5, 13, and 14, conversion was determined by LC–MS using a Poroshell 120 EC‐C18 column (2.7 µm, 3 × 50 mm, Agilent Technologies). The mobile phase consisted of water and acetonitrile (each containing 0.1% formic acid), with gradient elution. Detection was performed using DAD and ESI–MS. These compounds were analyzed by LC–MS due to their polarity or thermal lability. All reactions were analyzed in at least duplicate.

MD Simulations

3.5

Tetrameric homology models of the structurally homologous MDRs TeSADH_W110A [24] and the previously described KRED_36 [21] were built using the SWISS‐MODEL homology modeling web server [39, 40, 41, 42, 43]. The enzyme structures were superimposed to align their coordinate centers, and simulation boxes were solvated with 18% (v/v) DFX and 10% (v/v) isopropanol using the Packmol package. Small molecule parameterization and MD simulations at 50°C and were carried out as previously described [44]. All ionizable residues were protonated according to their expected states at pH 8.0. Changes in the radius of gyration across all five simulation replicates, as well as the total energy of the output frames, were calculated using our recently published analysis scripts, based primarily on MDTraj [45] and EvoEF2 [35], respectively [46].

Supporting Information

Additional supporting information can be found online in the Supporting information section. Supporting Fig. S1: Conversion of phenoxy‐2‐propanone (1) catalyzed by TeSADH W110A in various solvent systems at 300 mM substrate loading. Reactions were carried out using crude lysate (25 mg/mL) of E. coli‐expressed TeSADH_W110A in 50 mM Tris‐HCl buffer (pH 8.0) containing 0.2 mM NADP^+^ and 10% (v/v) isopropanol as cosubstrate. Performance was compared across four solvent systems: 50% (v/v) DFX, 50% (v/v) DMSO, 50% (v/v) DMF, and aqueous buffer (Tris‐HCl). Incubations were performed at 50°C for 20 h. Data are the mean values ± s.d. of at least two independent replicates. Supporting Fig. S2: Conversion of 4‐phenylbutan‐2‐one 2 catalyzed by TeSADH W110A in various solvent systems at increasing substrate concentrations (60–300 mM). Reactions were carried out using crude lysate (25 mg/mL) of E. coli‐expressed TeSADH_W110A in 50 mM Tris‐HCl buffer (pH 8.0) containing 0.2 mM NADP^+^, 10% (v/v) isopropanol as cosubstrate performance in four solvent systems: 35% DFX, 35% DMSO, 35% DMF (v/v), and aqueous buffer (Tris‐HCl). Incubations were performed at 50°C for 20 h at increasing substrate concentrations (60–300 mM). Data are the mean values ± s.d. of at least two independent replicates. Supporting Fig. S3: Conversion of 4‐phenylbutan‐2‐one 2 catalyzed by TeSADH W110A in various solvent systems at 300 mM substrate loading. Reactions were carried out using crude lysate (25 mg/mL) of E. coli‐expressed TeSADH_W110A in 50 mM Tris‐HCl buffer (pH 8.0) containing 0.2 mM NADP^+^ and 10% (v/v) isopropanol as cosubstrate. Performance was compared across four solvent systems: 50% (v/v) DFX, 50% (v/v) DMSO, 50% (v/v) DMF, and aqueous buffer (Tris‐HCl). Incubations were performed at 50°C for 20 h. Data are the mean values ± s.d. of at least two independent replicates. Supporting Fig. S4: Conversion of 4‐fluoroacetophenone 9 by LKADH in various solvent systems. Reactions were performed at 50°C for 20 h in 50 mM Tris‐HCl buffer (pH 7.5), using 10% (v/v) isopropanol and 5 mg/mL lyophilized E. coli/LKADH. Solvent systems included aqueous buffer, DMSO, DMF, and DFX (tested at 35% v/v). Data are the mean values ± s.d. of at least two independent replicates. Supporting Fig. S5: Conversion of 3,5‐bis(trifluoromethyl)acetophenone 11 catalyzed by ChKRED20 in different solvent systems. Substrate 11 (50–500 mM), ChKRED20 cell‐free lysate (25 mg/mL), 0.2 mM NAD^+^, 10% (v/v) isopropanol, 50 mM Tris‐HCl buffer (pH 8.0), 35% cosolvent (DFX, DMSO, or DMF), 50°C, 24 h. Enantiomeric excess >99% (S). Data are the mean values ± s.d. of at least two independent replicates. Supporting Fig. S6: Conversion of 1‐phenoxypropan‐2‐one 1 catalyzed by ChKRED20 across increasing substrate concentrations in Tris buffer, DMSO (35%) and DFX (35%). 1‐phenoxypropan‐2‐one (50 mM – 2 M), ChKRED20 cell‐free lysate, NADH with 10% (v/v) isopropanol, 50 mM Tris‐HCl buffer (pH 7.5), 35% cosolvent (DFX or DMSO), 50°C, 20 h. Conversions measured by GC. Data are the mean values ± s.d. of at least two independent replicates. Supporting Fig. S7: PpADH‐catalyzed conversion of methyl benzoylformate 12 (25–400 mM) in various solvents. Reaction conditions: 25–400 mM substrate, 35% (v/v) cosolvent (DFX, DMSO, DMF), 5% isopropanol, 25 mg/mL E. coli/ PpADH, 50 mM Tris‐HCl (pH 7.5), 40°C, 20 h. Data are the mean values ± s.d. of at least two independent replicates. Supporting Fig. S8: Conversion of Lipase B (CalB) in four different solvent systems (DFX, DMSO, DMF, aqueous buffer). All reactions were carried out at 50°C for 5 h using 2 mg of immobilized CalB. Conversions were determined by LC–MS under the same gradient and detection conditions for all samples. Data are the mean values ± s.d. of at least two independent replicates. Supporting Fig. S9: Transaminase‐catalyzed amination of selected ketones in different solvent systems (Tris‐Cl buffer, DMSO and DFX). Reaction conditions: 10 g/L ketone, 3.3 g/L biocatalyst, 35% (v/v) cosolvent, 2 mM PLP, 2 M isopropylamine·HCl, 0.1 M Tris–Cl buffer (pH 7.5), 0.2 mL reaction volume, 50°C, 800 rpm, 20 h. Data are the mean values ± s.d. of at least two independent replicates. Supporting Fig. S10: Conversion of four selected IREDs in four different solvent systems (DFX, DMSO, DMF, aqueous buffer).The reactions were carried out at 40°C for 18 h in 100 mM Tris‐HCl buffer (pH 7.4), with 4 mM NADPH, 20 mM glucose, 0.02 mg/mL GDH, and 50% (v/v) clarified lysate containing the IRED. Substrate concentrations 6 were 5 mM ketone and 50 mM amine. Reactions were analyzed by LC–MS. Data are the mean values ± s.d. of at least two independent replicates. Supporting Fig. S11: Comparative performance of TeSADH W110A and KRED_36 in DFX, DMSO, and aqueous buffer. Conversion of a model ketone (substrate 1) at 20 mM and 40 mM substrate loadings using TeSADH W110A and KRED_36 in three different solvent systems: DFX (18% v/v), DMSO (18% v/v), and aqueous buffer (pH 8.0, 50 mM Tris‐HCl). Reactions were carried out at 50°C for 20 h using NADP^+^ (0.2 mM) and 10% (v/v) isopropanol for in situ regeneration. Cell‐free extract (25 mg/mL) from E. coli expressing KREDs was used as biocatalyst. Data are the mean values ± s.d. of at least two independent replicates. Supporting Fig. S12: Temperature dependence of enzymatic conversions in selected solvents at high substrate load. Conversions for the reduction of substrate 1 (300 mM) by TeSADH W110A were determined at 40°C–70°C in DFX, DMSO, DMF (35% v/v), and buffer. Reaction conditions: NADP^+^ (0.2 mM), cell‐free extract of E. coli TeSADH W110A (25 mg mL^−1^), Tris–HCl (50 mM, pH 8.0), 2‐propanol (10% v/v), 24 h. Data are the mean values ± s.d. of at least two independent replicates. Supporting Fig . S13: Effect of solvent type on the reduction of substrate 1 catalyzed by TeSADH W110A at high substrate loads. Reaction conditions: substrate 1 (300 mM), organic solvent (50 % v/v; DFX, DMSO, DMF, glycerol, Cyrene, MeCN, 2‐MeTHF, CPME, or buffer), NADP^+^ (0.2 mM), cell‐free extract of E. coli expressing TeSADH W110A (25 mg mL^−1^), Tris–HCl buffer (50 mM, pH 8.0), and 2‐propanol (10 % v/v) as cosubstrate. Reactions were performed at 50°C for 24 h. CPME = cyclopentyl methyl ether, 2‐MeTHF = 2‐methyltetrahydrofuran. Data are the mean values ± s.d. of at least two independent replicates. Supporting Fig. S14: Curve plots of the first derivative of the 350 nm/330 nm fluorescence intensity ratio (thermal unfolding) versus temperature for selected enzymes under various solvent conditions. A) Purified TeSADH_W110A (1.6 mg/mL in 50 mM Tris‐HCl, pH 8) in buffer, 18% and 30% (v/v) DFX, DMSO, and DMF; B) Purified ChKRED20 (1.6 mg/mL total protein) under identical conditions; C) Purified Sp‐ATA (1.6 mg/mL in 2 mM PLP, 0.1 M HEPES pH 7). D) Purified KRED_36 (1.6 mg/mL in 50 mM Tris‐HCl, pH 8) in buffer, 18% and 30% (v/v) DFX, DMSO, and DMF. E) Crude lysate of PpADH (2.0 mg/mL total protein in 50 mM Tris‐HCl, pH 7.5) in buffer, 18% and 30% (v/v) DFX, DMSO, and DMF; All measurements were performed using nanoDSF (Prometheus Panta) between 25°C and 95°C in duplicate. Supporting Fig. S15: External calibration curves recorded by GC–FID for the indicated substrate and product standards under identical analytical conditions as used for the reaction samples. Calibration standards were prepared in a representative solvent composition reflecting the solvent ratio used in the biotransformations. Supporting Fig. S16: External calibration curves recorded by LC‐MS for the indicated substrate and product standards under identical analytical conditions as used for the reaction samples. Calibration standards were prepared in a representative solvent composition reflecting the solvent ratio used in the biotransformations. Supporting Table S1: List of primers used for site‐directed mutagenesis. Supporting Table S2: Calculated physicochemical properties of selected ketone substrates used in this study. LogP, molecular weight (MW), topological polar surface area (TPSA) was calculated using PubChem database. These descriptors were analyzed to explore correlations between substrate structure and observed solvent‐dependent performance in biocatalytic reductions.

Funding

This study was supported by NCCR Catalysis (Grant 225147).

Conflicts of Interest

J.S.L. is part owner and board member of Bloom Biorenewables Ltd, a start‐up company that is commercializing the aldehyde functionalization chemistry of biomass‐derived molecules. J.S.L. and A.O.K are inventors on an international patent (WO/2022/223480) on the use of xylose‐substituted molecules as solvents.

Supporting information

Supplementary Material