Improvement of NADPH-Dependent Bioconversion by Transcriptome-Based Molecular Breeding

Transcriptome data for a xylitol-producing recombinant Escherichiacoli were obtained and used to tune up its productivity. Structuralgenes of NADPH-dependent D-xylose reductase and D-xylose permeasewere inserted into an Escherichia coli chromosome to constructa recombinant strain producing xylitol from D-xylose for useas a model system for NADPH-dependent bioconversion. Transcriptomeanalysis of xylitol-producing and nonproducing conditions forthe recombinant revealed that xylitol production down-regulated56 genes. These genes were then selected as candidate factorsfor suppression of the NADPH supply and were disrupted to validatetheir functions. Of the gene disruptants, that resulting fromthe deletion of yhbC showed the best bioconversion rate. Also,the deletion accelerated cell growth during log phase. The featuresof the mutant could be maintained in jar fermenter-scale productionof xylitol. Thus, our novel molecular host strain breeding methodusing transcriptome analysis was fully effective and could beapplied to improving various industrial strains.

INTRODUCTION

Top

INTRODUCTION

Bioconversion using recombinant cells is a useful method forsynthesis of valuable chemical compounds and drug metabolites.Growing cells are preferable in cases where the supply of cofactorsis critical for biosynthetic reactions, because cofactors areexpensive to add externally (5, 12, 20). Growing cells alsomake it possible to maintain unstable enzymes such as membrane-associatedmultiprotein complexes (21).

Bioconversion using oxidoreductases is generally coupled withenzymatic cofactor regeneration systems (29). In this case,resting cells were used and NAD⁺ or NADP⁺ was externally addedto stimulate reactions. Formate dehydrogenase for NADH regeneration(14) and glucose dehydrogenase for NAD(P)H regeneration (23)have often been used in resting-cell oxidoreductase reactions.Although the resting-cell reaction is certainly a powerful toolfor short-term reactions, several disadvantages emerge whenit is applied to longer- term reactions on an industrial scale.In particular, NAD⁺ or NADP⁺ may cause a considerable increasein production costs. In contrast, growing cells can supply cofactorseffectively via their own cellular metabolism. Thus, this nativecofactor supply system is particularly useful for larger industrialreaction volumes. However, the activity of this endogenous cofactorregeneration system is strictly controlled by the robustnessof the cell physiology. There is little research on increasingNAD(P)H availability for bioconversion in growing cells.

In this research, we obtained specialized host cells for NADPH-dependentbioconversion by using a transcriptome data set. Transcriptomeanalysis is a powerful tool for functional analysis of genesin metabolic networks (18, 19, 28). We used a recombinant Escherichiacoli strain producing xylitol from D-xylose as a model of NADPH-dependentbioconversion. In the recombinant, chromosomal xylA (D-xyloseisomerase gene for D-xylose utilization) was disrupted. Andthe E. coli xylose permease gene (xylE) and Kluyveromyces lactisNADPH-dependent xylose reductase gene (XYL1) were inserted intothe chromosome of the recombinant. E. coli is a good host forfunctional analysis of genes, because a collection of single-gene-knockout(single-gene-KO) mutants is available. Transcription data setsfrom xylitol-producing and nonproducing conditions of the recombinantstrain were compared. Genes down-regulated during xylitol productionwere hypothesized as suppressors of the supply of NADPH, andtheir defective mutants were tested for xylitol production.Among the tested mutants, a yhbC-deficient strain showed improvementof xylitol production and additional acceleration of log-phasegrowth. Thus, the combination of transcriptome analysis andphenotype tests of single-gene-knockout mutants is a good methodfor screening important negative regulators for strictly controlledcofactor supply.

MATERIALS AND METHODS

Top

MATERIALS AND METHODS

Construction of xylitol-productive strain AK1.
The primers and plasmids used for construction of AK1 were shownin Table 1 and 2. XYL1 (GenBank accession no. L36993), xylE(GenBank accession no. U00096) and a tetracycline-resistantgene (Tc^r; GenBank accession no. AY528506) were amplified byPCR with primers in Table 2. A XYL1 fragment was digested withBspHI and BamHI, a xylE fragment containing a Shine-Dalgarnosequence at its 5' terminus was digested with BamHI, and a Tc^rfragment was digested with NheI. The digested fragments of XYL1and xylE were tandem inserted between the NcoI and BamHI sitesof pQE60 expression vector (Qiagen); also, the Tc^r fragmentwas inserted at the NheI site. The resulting plasmid was designatedpXKET6 in that T5 promoter, and the lac operator, XYL1, xylE,and Tc^r were concatenated in that order to compose a xylitol-productionoperon. This synthetic operon was amplified using PCR with 40-bphomologous sequences at both ends. The amplified fragment wastransformed to replace a chromosomal xylA gene (10) to constructa xylitol-producing strain, AK1. The resulting chromosomal regionfor xylitol production was easily transferred into other strainsby P1-phage transduction (17).

View this table:
[in this window]
[in a new window]

TABLE 1. Strains and plasmids used in this study

View this table:
[in this window]
[in a new window]

TABLE 2. Primers used in this study

Bacterial strains and culture.
The strains and plasmids used in this study are shown in Table1. Single-gene-knockout mutants of E. coli BW25113 (1) wereobtained from Hirotada Mori (Nara Institute of Science and Technology).A kanamycin-resistant (Km^r) gene of a yhbC knockout mutant waseliminated according to a method using Saccharomyces cerevisiaeFLP recombinase (10). In all experiments, seed cultures wereincubated overnight at 30°C and 250 rpm using LB-glucosemedium (5 g liter^–1 yeast extract, 10 g liter^–1peptone, 10 g liter^–1 NaCl, 10 g liter^–1 D-glucose,and proper antibiotics if necessary) and then inoculated intoM9 minimal medium (pH 6.8) (6.8 g liter^–1 Na₂HPO₄, 3 gliter^–1 KH₂PO₄, 0.5 g liter^–1 NaCl, 1 g liter^–1NH₄Cl, 0.247 g liter^–1 MgSO₄·7H₂O, 0.028 g liter^–1FeSO₄·7H₂O, 0.015 g liter^–1 CaCl₂·2H₂O)for xylitol production. The cell growth was monitored by measuringoptical density at 660 nm (OD₆₆₀) with a general photometer.

DNA microarray analysis.
A seed culture of xylitol production strain AK1 was inoculatedat 1% volume into M9 medium that included 1% glucose and 0.1mM isopropyl-ß-D-thiogalactopyranoside (IPTG), andthe cells were cultured in 5-ml medium in test tubes at 30°Cand 250 rpm. At the middle of the exponential phase (when theOD₆₆₀ was about 1.0), the culture was divided into two portionsand D-xylose (final concentration, 2%) was added to one aliquot.After 2 h from division of the culture, cells were harvestedand total RNAs were prepared separately from two divided cultureswith an RNeasy Protect bacterial Mini kit (Qiagen). Fluorescentlylabeled cDNA was prepared from the total RNA with an RNA fluorescencelabeling core kit (Takara Bio). Competitive hybridization ofthe cDNAs (13) onto IntelliGene E. coli CHIP version 2.0 (TakaraBio) was done according to the attached instructions. The microarraywas scanned with a microarray scanner (GenePix 4000B; Axon Instruments).All fluorescence intensity data were statistically analyzedwith data analysis software (GeneSpring 6.1; Agilent).

Monosaccharide and sugar alcohol analysis.
All cultures for the analysis were centrifuged at 16,000 x gfor 5 min at 4°C in a 5415R centrifuge (Eppendorf), andtheir supernatants were assayed. Concentrations of glucose,xylose, and xylitol were analyzed using HPAEC-PAD and a DXc-500system (Dionex). Analysis was performed using a Dionex CarboPacPA-1 column (4 by 250 mm) equipped with a guard column. Conditionswere isocratic (50 mM NaOH for 20 min at a 1 ml min^–1flow rate and 30°C).

Large-scale bioconversion of xylitol by use of jar fermentation.
Cells from the seed culture in LB-glucose medium were inoculatedinto the M9 medium with 10 g liter^–1 glucose, cultivatedovernight again, and inoculated at 3% volume into 1 liter ofthe M9 medium containing 30 g liter^–1 glucose and 0.1mM IPTG in a 2-liter bench top fermenter (BMJ-PI; Able). Wefound this concentration of IPTG to be appropriate for the reaction.The initial fermentation conditions were 30°C, 500-rpm agitation,an aeration rate of a 1.0 ratio of air volume to liquid volumeper min, and pH 7.0. The culture pH was controlled by automaticaddition of 14% NH₃, and dissolved oxygen in the culture waskept above 1.0 ppm by agitation change automatically. When theOD₆₆₀ reached about 1.0, 60 g liter^–1 xylose was added.Also, after 28.5 h additional glucose (30 g liter^–1) wasadded.

RESULTS

Top

RESULTS

Bioconversion of D-xylose to xylitol in E. coli AK1.
AK1 is a xylitol production strain based on E. coli BW25113(Fig. 1). AK1 possesses a synthetic operon consisting of theK. lactis NADPH-dependent xylose reductase gene XYL1 (3) andthe extra E. coli D-xylose permease gene xylE (11) instead of,and at the chromosome site of, the E. coli D-xylose isomerasexylA (Fig. 1A). Because XylA catalyzes the initial step of E.coli D-xylose metabolism, isomerizing it to D-xylulose (25),AK1 does not metabolize D-xylose. XYL1 and xylE were placedin tandem under the control of an IPTG-inducible promoter. Whenthe strain was cultivated in a mixed-sugar culture of D-xyloseand glucose, XylE imported D-xylose into the cell and XYL1 reducedit to xylitol by use of intracellular NADPH. It should be emphasizedthat the NADPH was autonomously supplied through cellular glucosemetabolism. An overview of xylitol production from xylose coupledwith cellular NADPH recycling in AK1 is shown in Fig. 1B. Infact, the sum of the D-xylose and xylitol concentrations inAK1 culture was almost constant (about 22.5 g liter^–1)during a 72-hour reaction (Fig. 2), indicating that D-xylosewas stoichiometrically converted to xylitol without any loss.Thus, NADPH availability for the reaction was easily assayedas xylitol productivity even in the case of long-term bioconversionusing AK1.

View larger version (23K):
[in this window]
[in a new window]

FIG. 1. Xylitol-productive strain AK1. (A) Schematic illustration of homologue recombination, placing xylA into a synthetic xylitol-producing operon. Expression of XYL1 and xylE is induced and controlled by addition of IPTG. (B) Schematic illustration of xylitol production in AK1. Xylose in culture was imported into the cell by XylE and reduced to xylitol by XYL1 and NADPH. The NADPH was supplied from glucose metabolism. IM, inner membrane.

View larger version (20K):
[in this window]
[in a new window]

FIG. 2. Stable bioconversion of xylitol in AK1. Stoichiometrical bioconversion of xylose to xylitol was performed using an AK1 reaction culture. AK1 was cultured in 5 ml of M9 minimal medium that included 10 g liter^–1 glucose and 22.5 g liter^–1 xylose in a test tube at 30°C and 250 rpm for 72 h. Concentrations of xylose ( ${blacksquare}$ ) and xylitol ( ${blacklozenge}$ ) and of a combination of these ( ${blacktriangleup}$ ) in the culture are shown.

Transcriptome analysis of AK1.
Transcriptions of AK1 with and without xylitol production werecompared using DNA microarray slides. It was found that expressionlevels of 56 genes were strongly repressed less than 0.5-foldin xylitol-producing AK1 (see Table S1 in the supplemental material).Eight sets of genes (cydAB, glf/rbfX, fruKB, atoEB, yhiQ/prlC,dppCA, glySQ, and ibpBA) were thought to be located in the sameoperons (i.e., cydA and cydB in the same operon, glf and rbfXin the same operon, etc.). According to functional classificationof 56 genes, many belonged to the metabolic pathway; 8 encodedenergy metabolism enzymes or related transporters (cydAB, ybiW,focA, gly, fruKB, and garK), 7 encoded amino acid metabolismenzymes or related transporters (leuL, glnP, trpC, sdaA, aroF,gltB, and glnA), and 5 encoded fatty acid-phospholipid metabolismenzymes or related transporters (prpE, atoEB, fabB, and lgt).There were six genes in the transcription or translation categories(rpmF, argS, galR, glySQ, and zur). Interestingly, there werealso four genes for dipeptide- or oligopeptide-related functions(dcp, prlC, and dppCA) and three genes for adaptation to temperatureshift (cspA and ibpBA). Since there were 17 essential genesamong the 56 down-regulated genes, only 39 single-gene-knockoutmutants were available in a KO collection of E. coli gene disruptants(1). The 39 available deletions were transferred into AK1 byuse of P1-phage transduction. The cell growth and xylitol productivityof 39 constructed variants of AK1 were measured and comparedwith those of AK1 after 36 h of cultivation (see Table S1 inthe supplemental material). Xylitol accumulation in three genedisruptants, ${Delta}$ glnA, ${Delta}$ leuL, and ${Delta}$ trpC, was apparently decreasedbecause of a strong growth defect in the M9 minimal medium.These results were understandable, because glnA and trpC encodeenzymes involved in amino acid biosynthesis (glutamine synthetase[8] and indole-3-glycerol phosphate synthase-phosphoribosylanthranilateisomerase [9], respectively) and leuL encodes a leader peptideof a leucine biosynthesis operon; thus, its disruption may affecttranscription of the leu operon. Of these 39 variants, the yhbC-defectiveAK37 strain was the best for xylitol production (10.3 g liter^–1)and also reached the highest cell density (2.6 at OD₆₆₀).

Genetic features of yhbC and overview of NADPH-dependent bioconversion in its mutant.
The yhbC is the second gene of the metY-rpsO operon and encodesa 15-kDa protein of unknown function (22). Many of the genesbelonging to the operon aside from yhbC are essential to E.coli survival and are well conserved among representative prokaryotes(Fig. 3A). In particular, yhbC and nusA genes are perfectlyconserved in those strains, including the thermophilic bacteriaAquifex aeolicus and Thermotoga maritima. The most investigatedhomologue of E. coli YhbC is that of Streptococcus pneumoniaeSP14.3 (30), having 60% similarity (Fig. 3B). The protein structureof SP14.3 was elucidated, and a three-dimensional structuralhomology search showed the existence of a nucleic acid-bindingdomain at its C terminal and a protein-interacting domain atits N terminal. AK1 and its ${Delta}$ yhbC mutant AK37 showed differentprofiles of growth, glucose consumption, and xylitol productionin a 72-hour cultivation (Table 3). Although all measured parameterswere higher in AK37 than in AK1 during the whole cultivationperiod, AK37 showed prominently higher activity levels at anearly stage of cultivation (2.1-fold higher in its OD₆₆₀ value,3.4-fold higher in glucose consumption, and 2.7-fold higherin xylitol production at 24 h). It was hypothesized that enhancementof whole-cell activity, including better glucose utilizationand active cell division, resulted in higher NADPH-dependentxylitol production in AK37.

View larger version (29K):
[in this window]
[in a new window]

FIG. 3. Alignment of the metY-rpsO operon in prokaryotes and YhbC homologues. (A) Comparison of the conserved gene arrays of the metY-rpsO operon among various prokaryotes. (B) Sequence alignment of E.coli YhbC and S. pneumoniae SP14.3 by CLUSTALW. The underline below the SP14.3 sequence shows the N-terminal domain (solid line) and C-terminal domain (dashed line). The bottom row indicates the positions of residues which are conserved (*), strongly similar (:) and weakly similar (.).

View this table:
[in this window]
[in a new window]

TABLE 3. AK1 and AK37 during xylitol bioconversion^a

Observation of cell growth in yhbC mutants.
The results of assays of growth acceleration after yhbC disruptionare shown in Fig. 4. The growth rates of AK1 and AK37 were measuredaccording to xylitol-producing or nonproducing status. Underxylitol-producing conditions, two strains showed growth slowerthan that seen nonproducing conditions (Fig. 4A). Both withand without xylitol production, AK37 showed obvious growth improvementcompared to AK1, especially in the early log phase. Thus, eliminationof yhbC in AK1 always accelerated growth in minimal media. Nextwe checked the effect of yhbC deletion in another genetic backgroundwithout XYL1 and xylE. BW25113 is the parental strain of theKO collection and the parental strain of AK1. A deletion mutantof yhbC from BW25113 in the KO collection was designated hereBW25113 ${Delta}$ yhbC. Growth profiles of BW25113 and BW25113 ${Delta}$ yhbC werecompared (Fig. 4B). In this case, the yhbC-deletion counterpartalso showed a superior growth curve in the early log phase.It was concluded that yhbC deletion improved growth of E. coliregardless of the presence or absence of xylitol-producing genes.The results of a yhbC complementation experiment using a plasmidare summarized in Fig. 4C. A plasmid with yhbC was able to restorethe phenotype of BW25113 ${Delta}$ yhb to that of the parental strain(compare TY37 with TC1 in Fig. 4C). This result indicates thatphenotypical changes of yhbC mutants are derived from the depletionof YhbC protein and that rearrangement of chromosomes causedby yhbC deletion results in little effect.

View larger version (14K):
[in this window]
[in a new window]

FIG. 4. Acceleration of log-phase growth in yhbC mutants. (A) Measurement of growth curves of AK1 and AK37 in two different cultures. For the first experiment, AK1 ( ${blacksquare}$ ) and AK37 ( ${blacklozenge}$ ) were cultured in M9 minimal medium that included 10 g liter^–1 glucose. For the second experiment, AK1 ( ${square}$ ) and AK37 ( ${diamond}$ ) were cultured in M9 minimal medium that included 10 g liter^–1 glucose and 22.5 g liter^–1 xylose. (B) Measurement of growth curves of E. coli BW25113-based strains. The cell growth characteristics of BW25113 ( ${triangleup}$ ) and BW25113 ${Delta}$ yhbC ( ${circ}$ ) were observed. (C) yhbC complementation experiment. TC1 ( ${triangleup}$ ), TY1 ( ${blacktriangleup}$ ), TC37 ( ${circ}$ ), and TY37 (•) were cultured in M9 minimal medium that included 10 g liter^–1 glucose. All strains were shaken in L-shaped test tubes at 30°C and 70 rpm.

Fermentation of AKP37 in a jar fermenter.
Xylitol production by a yhbC mutant was tested using a 2-literjar fermenter (Fig. 5). AKP37, a variant of AK37 with kanamycinresistance (Km^r) removed, was used for the experiment to avoidinvolvement of a Km^r promoter in expression of downstream genesof yhbC. Even in a jar fermenter, the yhbC mutant showed betterearly growth (Fig. 5A), glucose consumption (Fig. 5B), and xylitolproduction (Fig. 5C) than the wild type, a result similar tothat observed using flask-scale bioconversion (Table 3). Xylitolproduction is proportional to cell density in log phase, andthe slopes of these linearizations were very similar for AK1and AKP37 (0.98 and 0.97, respectively). These results indicatethat the cellular NADPH supply depends on the cell growth ratein log phase. Therefore, the growth acceleration of AKP37 directlyincreases the xylitol production rate. Though the final celldensities of AK1 and AKP37 were about the same (OD₆₆₀s of 12.9and 13.9), total glucose consumption and xylitol accumulationlevels were higher in AKP37 than in AK1. AK1 and AKP37 showed43.4 and 51.7 g liter^–1 of xylitol accumulation at 64h and 0.68 and 0.81 g liter^–1 h^–1 of xylitol productivity,respectively. These results indicate that yhbC disruption wasstill effective in accelerating cellular NADPH supply at thestationary phase as well.

View larger version (16K):
[in this window]
[in a new window]

FIG. 5. Practical scale of xylitol bioconversion in batch-fed cultivation. (A to C) OD₆₆₀ (A), glucose consumption (B), and xylitol production (C) of AK1 ( ${square}$ ) and AKP37 ( ${diamond}$ ) in jar fermenters are shown. In addition, xylose consumption of AK1 ( ${blacksquare}$ ) and AKP37 ( ${blacklozenge}$ ) is shown in panel C.

DISCUSSION

Top

DISCUSSION

We constructed a xylitol-producing strain, AK1, as a model ofreductive bioconversion using an NADPH supply from glucose metabolismin E. coli (Fig. 1B). At an early stage of this study, it wasdifficult to overcome catabolite repression of E. coli (4).Cells expressing XYL1 alone did not import xylose or producexylitol at all in media containing both glucose and xylose (datanot shown). Gene expression of xylose transporters, includingxylE and ABC transporter xylFGH, was repressed with glucose(17, 29). AK1 did import xylose even in the presence of glucosewhen IPTG was added for artificial expression of xylE insertedin its chromosome (Fig. 1B). And our strain does produce xylitolat a level of productivity similar to that seen with anotherxylitol-producing E. coli recombinant recently reported (7).That xylitol producer possesses NADPH-dependent xylose reductasefrom Candida boidinii and exhibits a disruption of xylulokinase(xylB) and a mutation in the cyclic AMP receptor protein (crp)for avoiding catabolite repression (7).

We tried to find regulators for the NADPH supply by use of transcriptomecomparison between two AK1 states with and without xylitol production.We thought that genes in which transcription levels were depressedduring xylitol production were factors repressing NADPH availabilityunder normal conditions. Using this strategy we successfullyfound a yhbC mutant with enhanced NADPH availability. This indicatesthat transcriptome analysis is effective as an initial screeningof genes for targets of molecular breeding. Besides yhbC, genescydAB, encoding components of terminal oxidases of E. coli (27),were notably down-regulated. Down-regulation of cydAB may resultin suppression of electron transfer in the respiratory chainto save cellular reduction power. Functional classificationshows that many of the genes selected (see Table S1 in the supplementalmaterial) belonged to the metabolic pathway. This suggests thatglobal modification of cellular metabolism occurs when NADPHis used for xylitol production.

A gene disruption test showed that deletion of yhbC was mosteffective for xylitol production. This functionally unknowngene is well conserved in many prokaryotes (Fig. 3A). Sinceone of YhbC homologues, S. pneumoniae SP14.3, has a nucleicacid-binding domain and a protein-interacting domain (Fig. 3B),YhbC would act as a transcriptional or translational regulatoryfactor. The metY-rpsO operon, including yhbC, consists of manywell-known transcription- and translation-related genes, andmost of these are essential for E. coli. Expression of the metY-rpsOoperon is strictly regulated by several promoters and terminators(15). It has been demonstrated that the transcriptional regulatorgene nusA in the operon does autoregulate itself in E. coli(6, 16, 26). At low temperature, the expression of the metY-rpsOoperon was increased due to the cold shock response caused bythe presence of CspA-family RNA chaperones (2). This indicatesthat the genes belong to the metY-rpsO operon and act to adaptto low-temperature stress, but the adaptive mechanism is stillunknown. It is possible that YhbC may also play an importantrole in autoregulation of the metY-rpsO operon to adapt to somestresses, because yhbC expression was repressed in oxidativeconditions by NADPH consumption in our experiment.

Our molecular breeding process to increase NADPH supply forbioconversion is a completely new approach. We have successfullyovercome catabolite repression by use of inducible D-xylosepermease to construct a recombinant xylitol producer, AK1. AK1is a good component of a system to analyze cellular NADPH supply.Target genes for modification were initially selected usingtranscriptome data sets of AK1 under two different sets of conditionswith and without xylitol production. Down-regulated genes weregood candidates for disruption to accelerate the NADPH supply.Consequently it was found that deletion of yhbC activates xylitolproduction. Disruption of yhbC can be used to improve productivityof other NADPH-dependent bioconversion activities. Since deletionof yhbC results in accelerated log-phase growth, combining yhbCmutants and cultivation methods for a longer log phase wouldachieve higher productivity.

ACKNOWLEDGMENTS

We thank Hirotada Mori (Nara Advanced Institute for Scienceand Technology) for providing strains and genes.

This study was carried out as a part of the Project for Developmentof a Technological Infrastructure for Industrial Bioprocesseson R&D of New Industrial Science and Technology Frontiersby Ministry of Economy, Trade & Industry and supported bythe New Energy and Industrial Technology Development Organization.

FOOTNOTES

* Corresponding author. Mailing address: Biofrontier Laboratories, Kyowa Hakko Kogyo Co. Ltd., 3-6-6 Asahimachi, Machida, Tokyo 194-8533, Japan. Phone: 81-42-726-2555. Fax: 81-42-726-8330. E-mail: hmori{at}kyowa.co.jp

Published ahead of print on 5 October 2007.

Supplemental material for this article may be found at http://aem.asm.org/.

REFERENCES

Top