Introduction

Deoxysugars exhibit various interesting biological activities and often play important roles in many physiologically significant reactions, in particular for the interaction of anti-cancer drugs with DNA (Kirschning et al. 1997). For example, the compound 2-deoxy-l-ribose constitutes the central building block of l-nucleosides and their derivatives, several of which are highly valuable antiviral agents (Jung and Xu. 1997; Nakajima et al. 2004). During biosynthesis of deoxyhexoses, hexose-6-phosphate is transformed into nucleoside diphosphate (NDP) by a guanylyltransferase to produce NDP-sugar; the latter is then deoxygenated by a specific dehydratase (Trefzer et al. 1999; Rodríguez et al. 2002). Various enzymatic methods have been developed to produce deoxysugars in vitro. An enzymatic production route for dTDP-4-keto-6-deoxy-d-glucose, a key intermediate of various deoxysugars in antibiotics, starting from dTMP, acetyl phosphate, and glucose-1-phosphate has also been presented (Oh et al. 2003; Kang et al. 2006). 2-Deoxy-d-ribose 5-phosphate aldolase (DERA, EC 4.1.2.4) catalyzes the reversible aldol reaction between acetaldehyde and d-glyceraldehyde-3-phosphate (Gap-3P) to generate 2-deoxy-d-ribose 5-phosphate (DR5P). The most attractive feature of DERA is its ability to catalyze sequential aldol additions because both the substrates and the product are aldehydes. The first aldol product can be accepted by the enzyme as acceptor again (cross-aldol) (Brovetto et al. 2011). Therefore, this property of DERA has been applied to perform sequential aldol reactions from acetaldehyde and generate (3R, 5R)-2,4,6-trideoxyhexose, a chiral precursor for the side chain of statins in vitro (Gijsen and Wong 1995). DERA also catalyzes reversible condensation between acetaldehyde and other non-phosphorylated aldehydes and thus shows potential for use in de novo synthesis of many kinds of 2-deoxysugars.

Despite the versatility of DERA, some obstacles limit its large-scale application in organic synthesis. DERA has a strong preference for phosphorylated substrates and is rapidly and irreversibly inactivated at high aldehyde concentrations (Gijsen and Wong 1994). To increase its affinity to non-phosphorylated substrates and tolerance to higher aldehyde concentrations, DERA has been mutated via several random mutagenesis strategies, such as error-prone PCR and DNA shuffling (Dean et al. 2007; Jennewein et al. 2006). Some mutant DERA variants have been shown to exhibit higher catalytic efficiency than wild-type DERA. For example, the mutant DERA(S238D) from Escherichia coli has an increased affinity towards non-phosphorylated 2-deoxy-d-ribose compared with 2-deoxy-d-ribose 5-phosphate (DeSantis et al. 2003). The mutant DERA(F200I) also has a higher catalytic efficiency to chloroacetaldehyde and acetaldehyde compared with wild type. Additional deletion of an amino acid (F200I/ΔY259) further increases its resistance to higher concentrations of chloroacetaldehyde (Jennewein et al. 2006). However, no study on altering DERA to improve enzyme activity and resistance to other polyhydroxy aldehydes, such as l-glyceraldehyde, has been reported to date.

In this study, DERA from Klebsiella pneumoniae (KDERA) was mutated to improve its activity and substrate resistance towards non-phosphorylated polyhydroxy aldehydes. Subsequently, a whole-cell transformation strategy using resting cells of recombinant E. coli strain was adopted to further increase its resistance to higher substrate concentrations. This strategy was successfully applied to efficiently produce C5 and C6 2-deoxysugars from polyhydroxy aldehydes.

Materials and methods

Materials and culture conditions

Acetaldehyde, glycolaldehyde, d/l-glyceraldehyde, d-erythrose, 2-deoxy-d-ribose, and isopropyl-β-d-thiogalactopyranoside (IPTG) were purchased from Sigma-Aldrich (St. Louis, MO, USA). All restriction enzymes and DNA ligase were purchased from Novagen (Darmstadt, Germany). Ni-NTA affinity chromatography column was bought from QIAGEN. Yeast extract and tryptone were purchased from OXOID LID, and brain heart infusion was purchased from Becton, Dickinson and Company.

All bacterial strains and plasmids used in this study are listed in Table 1. The E. coli strain DH5α was used for plasmid construction and as the donor for gene deoC amplification. K. pneumoniae CGMCC 1.10612 was also used as the donor for gene deoC amplification. E. coli BL21(DE3) was used for enzyme preparation. The pET-21a(+) plasmid was used as the foundation vector for expression of deoC in E. coli BL21(DE3). Cell culture was routinely carried out in Luria-Bertani (LB) medium containing 10 g/L tryptone, 5 g/L yeast extract, and 10 g/L NaCl. When necessary, 100 μg/mL ampicillin was used.

Table 1 Strains and plasmids used in this study
Full size table

Homology modeling and molecular docking

Build Homology Models (MODELER) module in Discovery Studio (DS) 4.1 Client (Accelrys, Inc., USA) was used for predicting the 3D structure of DERA from K. pneumoniae CGMCC 1.10612. The crystal structures of DERAs of E. coli K12 (PDB accession no. 1P1X), E. coli (PDB accession no. 3NQ8), and E. coli (PDB accession no. 3NPV) were used as templates. Pairwise alignment of DERA and template sequences was done using the Load Structure and Alignment program. The structure of KDERA was generated following the method described above. The substrate 2-deoxy-D-ribose (DR) or 2-deoxy-D-ribose 5-phosphate (DR5P) was initially positioned in the active site. The position of the substrate was then manually adjusted to avoid steric conflict with the active site residues. The resulting protein structure harboring DR or DR5P in the active site was loaded onto the docking software and subjected to docking simulation using the DS CDOCKER default parameters. The 10 top-scored poses were visually inspected to identify correct substrate orientations for productive binding that match spatial requirements between the amino group of the substrate and DR or DR5P. Protein structure and ligand–protein interaction diagrams were prepared using PyMOL.

Construction of deoC expression plasmids and site-directed mutagenesis

The deoC gene () of EDERA was amplified from the genome of the E. coli strain DH5α (NCBI Gene ID, 948902) using EDERA-F/EDERA-R primer sets (Table S1). This resulting fragment was then digested with NdeI and HindIII and cloned into pET-21a(+) to obtain pEDERA. Similarly, the deoC gene of KDERA was amplified from the K. pneumoniae CGMCC 1.10612 genome (NCBI Gene ID, 5341458) using KDERA-F/KDERA-R primer sets and subcloned into pET-21a(+) to obtain pKDERA.

Mutations were introduced into the deoC gene using the QuikChange site-directed mutagenesis kit (Stratagene) with oligonucleotides listed in the Supporting Information Table S1 and the plasmid pKDERA as template. DNA sequencing analysis confirmed the presence of the expected mutations in the gene sequence. Thus, a set of expression plasmids named pKDERA1-12 was obtained.

Gene expression and protein purification

The KDERA expression plasmids were transformed into E. coli BL21 (DE3). All recombinant E. coli strains were propagated at 37 °C in 100 mL LB medium containing ampicillin (100 mg/L) up to an optical density (OD)600 of 0.6–0.8 and subsequently induced by 1 mM IPTG at 20 °C. After incubation for 20–24 h, cells were harvested, washed twice, and suspended in 50 mM triethanolamine (TEA) (pH 7.0) buffer. Cell suspension was lysed by sonication, and the cellular debris was removed by centrifugation at 12,000×g for 10 min. Clear supernatant was collected and loaded onto a Ni2+-NTA-agarose column pre-equilibrated with binding buffer (50 mM TEA buffer, 100 mM NaCl, 20 mM imidazole, pH 7.0). The retained proteins were recovered with elution buffer (50 mM TEA buffer, 100 mM NaCl, 300 mM imidazole, pH 7.0). The eluted fraction containing purified protein was dialyzed and stored at −20 °C. Protein concentrations were determined by the Bradford method using bovine serum albumin as a standard.

Analysis of enzyme activity

The KDERA activity was determined using a coupled reaction system (1 mL) which contained 50 mM TEA buffer (pH 7.0), 10 mM substrates (d-glyceraldehyde and acetaldehyde), and 10 μL (~0.1 mg) diluted KDERA. The reaction was carried out at 30 °C for 20 min. One unit of enzyme activity was defined as the amount of enzyme that catalyzed the formation of 1 μmol product in 1 min. Using the methods mentioned above, the enzyme activity with 50 and 200 mM substrates were also determined.

Aldol reaction with d-glyceraldehyde and acetaldehyde in vitro

The reaction mixture containing aldehydes (0.05, 0.1, 0.2, 0.3, 0.5, or 1.0 M), 50 mM TEA buffer (pH 7.0), and 1 mg purified DERA aldolase was transferred to a 1.5-mL Eppendorf tube and shaken at 30 °C and 200 rpm for 24 h. To prevent aldehyde volatilization, the tube was sealed with parafilm. A control experiment with only the aldehydes at identical concentration was simultaneously carried out under the same conditions. The resulting mixture was treated with 10 M NaOH, centrifuged (22,000 rpm, 20 min), and then analyzed by high-performance liquid chromatography (HPLC) using Welch Xtimate Sugar-Ca column (7.8 × 300 mm, 8 μm), with 0.1 mM EDTA-Ca2+ as mobile phase at a flow rate of 0.4 mL/min and column temperature of 60 °C. 1H NMR and 13 C NMR spectra were recorded on a Bruker DMX-600 NMR spectrometer at 400 MHz.

Whole-cell transformation

The recombinant E. coli strains were initially cultivated in 100 mL of LB medium to an OD600 of 0.6 and then induced with 1 mM IPTG for 20 h at 20 °C and 120 rpm. Subsequently, the cells were harvested by centrifugation (8,000×g, 10 min, 4 °C), washed twice, and suspended in TEA buffer (50 mM). The cells were transferred into 1.5-mL Eppendorf tube containing 1 M acetaldehyde and polyhydroxy aldehydes with an initial OD600 of approximately 20. The temperature was held at 30 °C and stirred at 120 rpm. Samples were collected every hour and analyzed by HPLC. New products were purified by HPLC, concentrated by lyophilization, and then structurally characterized by NMR spectroscopy.

To characterize the whole-cell transformation in a 50-mL flask, the reaction mixture (10 mL) containing resting cells (OD600 = 80), 3 M d-glyceraldehyde and acetaldehyde, and 200 mmol TEA buffer (pH = 7.0) was carried out at 30 °C and 200 rpm for 24 h. A control experiment with only the aldehydes at identical concentration was simultaneously carried out under the same conditions. Samples were collected every 2 h and analyzed by HPLC.

Results

Mutation of DERA for higher enzyme activity to polyhydroxy aldehyde

DERA accepts not only phosphorylated Gap-3P but also non-phosphorylated aldehydes and thus can be useful for de novo synthesis of 2-deoxysugars (Gijsen and Wong. 1995). Therefore, the deoC genes from K. pneumoniae and E. coli were first cloned and expressed in E. coli BL21(DE3). Sequence alignment showed that EDERA showed 92 % identity with KDERA. After one-step purification with N-terminal 6×His-tag, SDS-PAGE of the obtained proteins showed the desired molecular mass of EDERA (27.7 kDa) and KDERA (27.6 kDa) (Fig. S1). Afterwards, we chose d-glyceraldehyde as a substrate to test the catalytic efficiency of conversion of wild-type DERA to polyhydroxy aldehydes. The results indicated that the activity of wild-type EDERA and KDERA on d-glyceraldehyde substrate was 0.33 and 0.4 U/mg, respectively (Table 2). However, these activities were lower than those of Gap-3P (2.5 U/mg) (Horinouchi et al. 2003). The results further confirmed that KDERA had strong preference for phosphorylated aldehydes.

Table 2 The enzyme activity and substrate resistance data of 2-deoxy-d-ribose 5-phosphate aldolase (DERA) mutants
Full size table

A multi-site-directed mutagenesis strategy based on crystal structure was used in this study to increase the affinity of KDERA to polyhydroxy aldehydes. A homology model of wild-type KDERA was constructed using the crystal structure of EDERA (Protein Data Bank code 1GAD) as template. Previous studies found that maintenance of the hydrophilic nature of the binding pocket and substitution of negatively charged amino acids with neutral and positively charged amino acids helped increase the preference for non-phosphorylated substrates (Dean et al. 2007; DeSantis et al. 2003). Therefore, the amino acids S238, S239, G171, and L20 around the substrate binding pocket were chosen and mutated to generate the mutants KDERAK1 (S238D), KDERAK2 (S239D), KDERAK3 (G171S), and KDERAK4 (L20D), respectively, all of which have neutral or positively charged amino acids (Fig. 1). The enzyme activity of KDERAK1 and KDERAK2 on d-glyceraldehyde was 12.5 % higher than that of wild-type KDERA (Table 2). However, KDERAK3 and KDERAK4 showed no activity towards d-glyceraldehyde. The sites of both mutations are strictly conserved, and the mutations could have resulted in structural perturbation (DeSantis et al. 2003). No improvement in substrate resistance was found for the mutants KDERAK1 and KDERAK2.

Fig. 1
figure 1

a The molecular docking result between wide-type DERA from K. pneumonia (KDERA) and 2-deoxy-d-ribose 5-phosphate (DR5P). b The molecular docking result between mutant KDERAK12 (S238D/F200I/ΔY259) and 2-deoxy-d-ribose

Full size image

The docking results between KDERA and DR5P showed that the amino acid residues I165, T168, M185, and F200 formed a small hydrophobic cluster close to the peptide backbone of the amino acids K167 and K201, which are the active sites of KDERA (Fig. 1a). The mutant DERAM185V from E. coli showed a fivefold increase in enzyme activity towards chloroacetaldehyde (Jennewein et al. 2006), which indicated that hydrophobicity around the active sites is needed to be enhanced when polyhydroxy aldehydes were used. Therefore, the amino acids I165, T168, M185, and F200 around the active sites were chosen and mutated to obtain the mutants KDERAK5 (F200I), KDERAK6 (T168L), KDERAK7 (F165I), and KDERAK8 (M185V), respectively. Among them, KDERAK5 and KDERAK8 showed 2.4- and 1.8-fold improvements, respectively, in enzyme activities compared to wild-type KDERA. KDERAK7 had only 87.5 % of the activity of wild-type KDERA, while KDERAK6 was virtually inactive. Notably, all the mutants, except for KDERAK6, showed preferable substrate tolerance compared with wild-type KDERA when the reactions were performed with 200 mM d-glyceraldehyde and acetaldehyde. Although the performances of KDERAK5 and KDERAK8 were satisfactory, their performance in combination in the KDERAK9 (M185V/F200I) mutant was not (Table 2).

Improvement of tolerance of DERA towards polyhydroxy aldehydes

The activity of wild-type KDERA sharply dropped by 43 % under 200 mM d-glyceraldehyde and acetaldehyde (Table 2), which limited its large-scale synthetic applications. To improve the resistance of DERA to both d-glyceraldehyde and acetaldehyde, the two mutants KDERAK10 (F200I/ΔY259) and KDERAK11 (F200I/S258T/Y259T + C-terminal extension KTQLSCTKW) were created as described previously (Jennewein et al. 2006). Substrate tolerance of the mutant KDERAK10 was 37.3 % higher than that of wild-type KDERA (Table 2).

The effective mutation sites mentioned above were combined to generate the mutants KDERAK12 (S238D/F200I/ΔY259), KDERAK13 (S238D/F200I/S258T/Y259T + C-terminal extension KTQLSCTKW), and KDERAK14 (M185V/S238D/F200I/ΔY259). Among all mutants, KDERAK12 exhibited preferable 2-deoxy-d-ribose synthetic ability (1.26 U/mg) and still maintained 88 % of residual activity at 200 mM d-glyceraldehyde and acetaldehyde (Table 2).

Six gradients of both substrates (0.05, 0.1, 0.2, 0.3, 0.5, and 1.0 M) were used to further describe the substrate tolerance of KDERAK12. When the substrate concentration increased to 1 M from 0.05 M, the enzyme activities of KDERAK12 and 2-deoxy-d-ribose decreased by 63.1 and 6.13 %, respectively (Fig. 2).

Fig. 2
figure 2

Substrate tolerance analysis of KDERAK12(S238D/F200I/∆259). A series of d-glyceraldehyde and acetaldehyde concentrations (M), 0.05, 0.1, 0.2, 0.3, 0.5, and 1, were used. The KDERA enzyme activity was calculated under the given substrate concentration. One unit of enzyme activity was defined as the enzyme amount catalyzing the formation of 1 μmol product per min. Mean and standard deviation were calculated based on triplicate experiments

Full size image

Optimization of whole-cell transformation for production of 2-deoxy-d-ribose

Whole-cell transformation is a useful tool to produce pharmaceutical intermediates (19). Therefore, whole-cell transformation of resting cells of the BL21(pKDERA12) strain, which contained the expressing plasmid pKDERAK12, was applied to produce 2-deoxy-d-ribose with acetaldehyde and d-glyceraldehyde as substrates. We first tested the effect of different reaction temperatures (25, 30, 37, and 50 °C), reaction buffers (phosphate-buffered saline (PBS), 3-Morpholinopropanesulfonic acid buffer (MOPS), Tris–HCl, triethanolamine buffer (TEA), disodium hydrogen phosphate-citric acid buffer (PC), and pH values (6, 7, 8, and 9) on 2-deoxy-d-ribose production. The initial substrate concentration and cell density (OD600) were 0.5 M and 20, respectively. The results in Fig. 3 indicated that the optimal temperature was 30 °C while the optimal pH was between 6 and 7. The 2-deoxy-d-ribose yield was barely altered by the reaction buffers PBS, MOPS, TEA, and PC but significantly decreased when using Tris–HCl. Therefore, the reaction conditions of 30 °C, pH 7, and TEA buffer were chosen.

Fig. 3
figure 3

Optimization of whole-cell transformation conditions. Whole-cell transformation using resting cells of recombinant strain BL21(pKDERA12) was carried out with an initial cell density OD600 of 20. The 2-deoxy-d-ribose productivity and yield were calculated within 1 and 24 h, respectively. a The effect of temperature on 2-deoxy-d-ribose production. b The effect of pH on 2-deoxy-d-ribose production. c The effect of reaction buffer on 2-deoxy-d-ribose production. d The effect of substrate concentration on 2-deoxy-d-ribose production

Full size image

The substrate concentrations were further optimized by assigning substrates with concentrations of 0.2, 0.5, 1, 1.5, 2, and 3 M. The initial 2-deoxy-d-ribose productivity was 70.86 % at 1 M acetaldehyde and d-glyceraldehyde. Interestingly, the 2-deoxy-d-ribose yield further increased to 0.81 mol/mol d-glyceraldehyde compared with that in the in vitro experiment (0.41 mol/mol d-glyceraldehyde) under 1 M substrate (Fig. 3d).

Adjustment of cell density to further increase resistance to higher substrate concentration

The results in Fig. 3d indicated that initial 2-deoxy-d-ribose productivity and yield decreased to 45.89 and 91.66 % under 3 M acetaldehyde and d-glyceraldehyde, respectively. We speculated that a higher cell density could tolerate higher substrate concentration; therefore, initial cell density of the culture was further modified to OD600 of 20, 40, 60, 80, 100, and 120 at 3 M substrate concentration. Reaction equilibrium was obtained within 9 h when an initial cell density of 120 was used (Fig. 4a). No significant substrate inhibition occurred when OD600 was greater than 80. To further describe 2-deoxy-d-ribose synthetic ability at OD600 of 80 in detail, five gradients of both substrates (1, 2, 2.5, 3, and 3.5 M) were used. Within 16 h, reaction equilibrium was obtained under 3 M of both acetaldehyde and d-glyceraldehyde (Fig. 4b). However, the 2-deoxy-d-ribose productivity and yield significantly decreased when substrate concentration was increased to 3.5 M (Fig. 4c). Thus, these results suggested that the whole-cell transformation strategy could increase substrate resistance of KDERAK12 to 3–3.5 M when the OD600 was 80.

Fig. 4
figure 4

Effect of initial cell density OD600 on 2-deoxy-d-ribose production. a Time course of 2-deoxy-d-ribose production at 3 M acetaldehyde and d-glyceraldehyde. A series of cell density OD600, 20 (black square), 40 (white circle), 60 (black triangle), 80 (circle with right half black), 100 (black circle), and 120 (white pentagon), were shown. b Optimization of substrate concentration when cell density OD600 was 80. A series of acetaldehyde and d-glyceraldehyde concentrations (M), 1.0 (black square), 2.0 (black circle), 2.5 (black triangle), 3.0 (white diamond), and 3.5 (white circle), were shown. c The transformation condition was similar with b. The 2-deoxy-d-ribose concentration and yield were obtained at 16 h. d Time course of whole-cell transformation with cell density OD600 of 80 and substrate concentration at 3 M. The transformation was performed for about 24 h. Acetaldehyde (white circle), d-glyceraldehyde (black triangle), and 2-deoxy-d-ribose (black square) were shown

Full size image

The time course of production of 2-deoxy-d-ribose by the strain BL21(pKDERA12) was also characterized in a 50-mL flask. The initial acetaldehyde and d-glyceraldehyde concentration and cell density (OD600) were set at 3 M and 80, respectively. After whole-cell transformation for 16 h, 2.14 M (287.06 g/L) 2-deoxy-d-ribose was produced with a yield of 0.71 mol/mol d-glyceraldehyde (Fig. 4d). The average productivity was 0.13 mol/L·h (17.94 g/L·h).

Production of other 2-deoxysugars by changing the type of polyhydroxy aldehydes

DERAs have been reported to accept a wide variety of acceptor aldehydes and are used in the synthesis of various carbohydrate derivatives (Samland and Sprenger 2006). Therefore, we further tested the ability of KDERA to produce other polyhydroxy aldehydes (l-glyceraldehyde and d-erythrose) by whole-cell transformation. The initial cell density (OD600) and substrate concentration were 20 and 1 M, respectively. New products were identified and structurally characterized by NMR spectroscopy (Fig. S2). Our data indicated that the BL21(pKDERA12) strain could adopt d-erythrose to form 2-deoxy-d-altrose (Fig. 5). 2-Deoxy-l-xylose and 2-deoxy-l-ribose were simultaneously produced from l-glyceraldehyde. The ratio of the content of 2-deoxy-l-xylose to 2-deoxy-l-ribose in total aldol products was 3.5:1. At the end of transformation, the BL21(pKDERA12) strain produced 0.57 M (93.5 g/L) 2-deoxy-d-altrose with a yield of 0.57 mol/mol d-erythrose and 0.53 M (71.6 g/L) 2-deoxy-l-xylose and 0.15 M (20.3 g/L) 2-deoxy-l-ribose with a yield of 0.69 mol/mol l-glyceraldehyde.

Fig. 5
figure 5

HPLC analysis of new 2-deoxysugars from different polyhydroxy aldehydes. a Production of 2-deoxy-d-ribose from acetaldehyde and d-glyceraldehyde: (a) acetaldehyde; (b) d-glyceraldehyde; (c) standard substance 2-deoxy-d-ribose; (d) supernatant of transformation. b Production of 2-deoxy-l-ribose and 2-deoxy-l-xylose from acetaldehyde and l-glyceraldehyde: (a) acetaldehyde; (b) l-glyceraldehyde; (c) supernatant of transformation. c Production of 2-deoxy-d-altrose from acetaldehyde and d-erythrose: (a) acetaldehyde; (b) d-erythrose; (c) supernatant of transformation

Full size image

Discussion

DERA catalyzes the reversible aldol reaction of acetaldehyde and Gap-3P to yield 2-deoxyribose-5-phosphate. This enzyme uses aldehyde as a natural donor and accepts a wide variety of aldehydes as acceptor substrates (Brovetto et al. 2011). However, DERA has low preference for phosphorylated aldehydes (DeSantis et al. 2003). In this study, some amino acids surrounding the accepter binding pocket and active site were mutated. Mutation of some sites was confirmed to increase enzyme activity and substrate tolerance of KDERA. S238D substitution increased the d-glyceraldehyde-binding capacity of KDERA. F200I and M185V substitutions strengthened the hydrophobic environment around the catalytic center and increased the enzyme activity towards d-glyceraldehyde. Deletion of Y259 increased the substrate tolerance. In addition, we further tested the effects of all of these mutant sites combined on enzyme activity and substrate tolerance of KDERA. Mutant KDERAK12 (S238D/F200I/ΔY259) exhibited a 3.15-fold improvement in enzyme activity and a 1.54-fold increase in substrate tolerance compared with wild type (Table 2).

Although mutant KDERAK12 showed stronger resistance towards acetaldehyde and d-glyceraldehyde compared with wild type, its substrate tolerance did not meet the requirement for large-scale production of 2-deoxy-d-ribose. Whole-cell transformation strategy has been widely used to synthesize target products for higher substrate consumption and product formation rates (Bäumchen and Bringer-Meyer 2007). In addition, the presence of a relatively rigid cell wall enables high tolerance and robustness to chemicals and organic solvents (Olmo et al. 2011). Therefore, a whole-cell transformation strategy using the BL21(pKDERA12) strain was applied to further increase substrate tolerance of KDERA. We confirmed that this strategy helps to increase 2-deoxy-d-ribose yield (from 0.41 mol/mol d-glyceraldehyde to 0.81 mol/mol d-glyceraldehyde) compared with that of in vitro assays, in addition to increasing the substrate resistance concentration to 3 M. Furthermore, higher cell density in the medium contributed to higher 2-deoxy-d-ribose production rate and substrate resistance to acetaldehyde and d-glyceraldehyde. Eventually, 2.14 M (287.06 g/L) 2-deoxy-d-ribose was produced with a yield of 0.71 mol/mol d-glyceraldehyde; the average productivity was 0.13 mol/L·h (17.94 g/L·h). These results demonstrated the potential for large-scale production of 2-deoxy-d-ribose using the BL21(pKDERA12) strain. Therefore, we believe that the whole-cell transformation strategy could also be effective at increasing the tolerance of KDERA to other non-phosphorylated aldehydes such as chloroacetaldehyde.

The most attractive feature of DERA is its ability to catalyze sequential self or cross-aldol additions. Jennewein et al. (2006) found that tandem addition of two equivalents of acetaldehyde to one equivalent of glycolaldehyde or chloroacetaldehyde resulted in the statin drug precursor (3R, 5S)-2,4-dideoxyhexose and (3R, 5S)-6-chloro-2,4,6-trideoxyhexose, respectively. Other 2-deoxysugars, such as 2-deoxy-l-ribose and 2-deoxyglucose, also have wide applications in the pharmaceutical industry (Helanto et al. 2009; Nakajima et al. 2004; Ben Sahra et al. 2010). Therefore, the ability of the BL21(pKDERA12) strain to produce two polyhydroxy aldehydes, namely l-glyceraldehyde and d-erythrose, was further tested. In this study, we discovered that only 2-deoxy-d-altrose was produced from acetaldehyde and d-erythrose. However, when using l-glyceraldehyde, KDERAK12 lost its stereoselectivity to aldol products and produced not only the natural product 2-deoxy-l-xylose (3R, 4S) but also the non-natural product 2-deoxy-l-ribose (3S, 4S), an important l-nucleoside precursor. The property of loose stereochemical control has also been described in dihydroxyacetone phosphate (DHAP)-dependent aldolases (Fessner et al. 1991; Li et al. 2011a, b). Thus, this property of aldolases helped to synthesize multiple polyhydroxylated compounds with non-natural stereoconfiguration.

In summary, a few KDERA mutants were generated by multi-site-directed mutagenesis, and these mutants were found to exhibit higher enzyme activity and tolerance to polyhydroxy aldehydes compared with wild type. Furthermore, a whole-cell transformation strategy using resting cells of the BL21(pKDERA12) strain was found to further increase product yield and substrate tolerance. With further optimization of transformation conditions, a large amount of 2-deoxy-d-ribose could be produced. In addition, the BL21(pKDERA12) strain showed the ability to produce other 2-deoxysugars from l-glyceraldehyde and d-erythrose.