APPLIED AND ENVIRONMENTAL MICROBIOLOGY, July 2011, p. 4352–4360 Vol. 77, No. 130099-2240/11/$12.00 doi:10.1128/AEM.02912-10Copyright © 2011, American Society for Microbiology. All Rights Reserved.

Coevolutionary Analysis Enabled Rational Deregulation of AllostericEnzyme Inhibition in Corynebacterium glutamicum

for Lysine Production�

Zhen Chen, Weiqian Meyer, Sugima Rappert, Jibin Sun,§ and An-Ping Zeng*Institute of Bioprocess and Biosystems Engineering, Hamburg University of Technology,

Denickestrasse 15, D-21073 Hamburg, Germany

Received 14 December 2010/Accepted 23 April 2011

Product feedback inhibition of allosteric enzymes is an essential issue for the development of highly efficientmicrobial strains for bioproduction. Here we used aspartokinase from Corynebacterium glutamicum (CgAK), akey enzyme controlling the biosynthesis of industrially important aspartate family amino acids, as a model todemonstrate a fast and efficient approach to the deregulation of allostery. In the last 50 years many researchersand companies have made considerable efforts to deregulate this enzyme from allosteric inhibition by lysineand threonine. However, only a limited number of positive mutants have been identified so far, almostexclusively by random mutation and selection. In this study, we used statistical coupling analysis of proteinsequences, a method based on coevolutionary analysis, to systematically clarify the interaction network withinthe regulatory domain of CgAK that is essential for allosteric inhibition. A cluster of interconnected residueslinking different inhibitors’ binding sites as well as other regions of the protein have been identified, includingmost of the previously reported positions of successful mutations. Beyond these mutation positions, we havecreated another 14 mutants that can partially or completely desensitize CgAK from allosteric inhibition, asshown by enzyme activity assays. The introduction of only one of the inhibition-insensitive CgAK mutations(here Q298G) into a wild-type C. glutamicum strain by homologous recombination resulted in an accumulationof 58 g/liter L-lysine within 30 h of fed-batch fermentation in a bioreactor.

The construction of a minimally mutated strain with a highproduction rate and a high final titer represents a major effortand future direction of strain engineering (14). For such apurpose, only beneficial mutations ought to be integrated intothe genome of a production strain. Unnecessary or detrimentalmutations, which are often introduced during the processes ofrandom mutation and selection, should be avoided in order tominimize their side effects on cellular constitutions such as cellgrowth and substrate consumption. The fast identification andconstruction of beneficial mutations are thus a key issue for thesuccessful construction of a minimally mutated strain. Formany industrial processes, such as amino acid production, thederegulation of allosteric inhibition is the initial and highlychallenging step. A rational approach that can quickly andefficiently identify and introduce targeted mutations to desen-sitize the allosteric inhibition of enzymes is highly desirable tothis end. In the present study, we describe a strategy employinga coevolutionary analysis of protein sequences to circumventthe problem of the allosteric inhibition of enzymes and dem-onstrate its application to create a minimally mutated lysine-producing strain. The allosteric enzyme aspartokinase was se-lected as a model enzyme for this proof of concept.

Aspartokinase is the first and committed enzyme involved in

the biosynthesis of several industrially important amino acidssuch as lysine, threonine, and methionine (2). Aspartokinasecatalyzes the phosphorylation of the �-carboxyl group of as-partate using ATP and is feedback regulated by end products.It has different isozymes and can be regulated in different ways.For example, Escherichia coli has three aspartokinases: two ofthem are conjugated with homoserine dehydrogenases. Aspar-tokinase I-homoserine dehydrogenase I (AKI-HDI) is alloster-ically inhibited by threonine, and its synthesis is repressed bythreonine plus leucine, while aspartokinase II-homoserine de-hydrogenase II (AKII-HDII) is repressed only by methionine(8). The third aspartokinase isozyme in E. coli, aspartokinaseIII (AKIII), is a single functional enzyme that is inhibited andrepressed by lysine (16, 17, 19). In Corynebacterium glutami-cum, there is only one single aspartokinase (CgAK) that isconcertedly inhibited by lysine and threonine (22, 29).

CgAK has an �2�2-type heterotetrameric structure contain-ing equimolar � and � subunits, which are encoded by in-frameoverlapping genes (15). The � subunit consists of two domains,the N-terminal catalytic domain and the C-terminal regulatorydomain. The � subunit is identical to the regulatory domain ofthe � subunit. The � subunit or the regulatory domain of the �subunit consists of two motifs called ACT domains, whichshare a fold of ������ (5, 10). The ACT domain appears inmany allosteric enzymes, especially enzymes in the amino acidand purine synthesis pathways, serving as a small-moleculebinding site for the allosteric regulation of proteins (5). InCgAK, lysine and threonine are also bound at the interfacebetween different ACT domains from the � subunit and �subunit, as indicated in Fig. 1A (30).

The deregulation of CgAK from allosteric inhibitions was

* Corresponding author. Mailing address: Institute of Bioprocessand Biosystems Engineering, Hamburg University of Technology,Denickestrasse 15, D-21073 Hamburg, Germany. Phone: 49-40-42878-4183. Fax: 49-40-42878-2909. E-mail: aze@tu-harburg.de.

§ Present address: Tianjin Institute of Industrial Biotechnology, Chi-nese Academy of Sciences, Tianjing 300308, People’s Republic ofChina.

� Published ahead of print on 29 April 2011.

4352

on August 15, 2019 by guest

http://aem.asm

.org/D

ownloaded from

considered the most important step toward lysine productionby C. glutamicum (8). Mutants with CgAK insensitive to lysineinhibition have been obtained by random mutation and selec-tion with the phenotype of resistance to S-(2-aminoethyl)-L-cysteine (AEC), an L-lysine structural analog (20, 23). Thistraditional method, however, is time-consuming and labor-intensive and not efficient enough. Only a limited number ofpositive mutations within fewer than 10 amino acid positionshave been identified during the past 40 years, and these pointsare the subjects of more than 20 patents (8). The underlyingmechanism of such mutations is nevertheless unknown. Thesepositions, however, may represent only a small percentage ofthe large sequence space that may influence the allosteric con-trol of the enzyme. A fast and systematic method is desirable toexplore the sequence space and to better understand thismechanism. To this end, we attempted to use the method ofstatistical coupling analysis (SCA) (9, 18, 26), a powerful co-evolutionary analysis approach, to clarify the allosteric inter-action network within the regulatory domain of CgAK. Thisallosteric interaction network was then used as the basis for thedesign of efficient targeted mutations.

SCA is a sequence-based statistical method for analyzingconserved cooperative actions between amino acids (8, 26).Based on a multiple-sequence alignment of protein families,SCA can identify a small subset of residues that are coevolvedduring protein evolution. These coevolved or correlated resi-dues are considered to be especially important for the struc-ture or function of proteins (1, 3). The prediction of thesecorrelated residues can greatly reduce the sequence space ofproteins and facilitate the identification of a potential allostericinteraction network that mediates protein function. SCA hasbeen successfully used to clarify the interconnected allostericnetworks of several protein families, such as PDZ domains(18), G proteins, and serine proteases (12, 18). However, thismethod has not been used for guiding the targeted mutagen-esis of industrially important allosteric proteins such as aspar-tokinases.

In this study, we used the SCA method to define a cluster ofevolutionarily correlated residues within the regulatory do-main of CgAK, which is responsible for allosteric inhibition.This small cluster of residues includes most of the previouslyreported positions for efficient mutations and enabled us to

FIG. 1. Structure and properties of aspartokinase from C. glutamicum (CgAK) and results of coevolutionary analysis of aspartokinase familyproteins. (A) Overall structure of CgAK. The catalytic domain and regulatory domain of the � subunit are indicated green and cyan, respectively.The � subunit is shown in magenta. Lysine and threonine are presented by the Corey-Pauling-Koltun (CPK) coloring model. CgAK has an�2�2-type heterotetrameric structure. Only one �� dimer from the protein reported under Protein Data Bank (PDB) accession number 3AAWis indicated here. (B) SDS-PAGE of purified CgAK. M, marker; P, purified protein. (C) Profiles of inhibition of wild-type CgAK by lysine andthreonine. The specific activity under standard assay conditions (Table 2) was taken as 100%. (D) SCA matrix of the regulatory domain ofaspartokinase family proteins. The position numbering is indicated according to the sequence of CgAK. The color scale linearly maps thecorrelated value from 0 to 2. (E) Mapping of the 25 highly correlated positions into the structure of CgAK.

VOL. 77, 2011 DEREGULATION OF ALLOSTERIC ENZYME 4353

on August 15, 2019 by guest

http://aem.asm

.org/D

ownloaded from

quickly design another 14 deregulated CgAK mutants. Themapping of these residues onto the recently solved crystalstructure of CgAK (30) helped us to understand the potentialeffects of these point mutations on the regulation of this en-zyme. Finally, we integrated one such point mutation (Q298G)into wild-type C. glutamicum ATCC 13032, which resulted in aremarkably high level of lysine production (58 g/liter) in fed-batch fermentation based on only one point mutation. Themutant showed growth behavior similar to that of the wild-typestrain and can be used as a basis for further strain develop-ment.

MATERIALS AND METHODS

Multiple-sequence alignment. Sequences of the aspartokinase family proteinswere collected from the UniRef90 database in the UniProt Knowledgebase(http://www.uniprot.org/). The sequences were aligned with MUSCLE (7) andClustalX (27), followed by structure-guided manual adjustment (6). The se-quences of the regulatory domains of the aspartokinase family were truncatedfrom this data set according to the sequence of the � subunit of CgAK. Anysequence sharing �90% similarity to another sequence was then removed inorder to achieve a diverse distribution of samples. Sequence positions with gapfrequencies higher than 20% were deleted. The final alignment consisted of 500sequences and 160 amino acid positions.

Statistical coupling analysis. A detailed procedure for the SCA calculationwas previously described (11, 26). The METLAB script was described previouslyby Halabi et al. (11). In summary, SCA gives a matrix that calculates the weightedsequence correlation between any two positions, Cij

ab � �i�j� f ijab � f i

af jb),

where f ia denotes the observed frequency of amino acid a at position i and �i is

the positional conservation-based weight. The positional conservation is definedas Di � ln[ fi(1 �qa)/(1 � fi )qa], where qa is the background frequency of thisamino acid (�i � �D/�f ).

Preparation of expression vectors for CgAK and site-directed mutagenesis.For the overexpression and enzyme purification of CgAK, the lysC gene was amplifiedfrom genomic DNA of C. glutamicum ATCC 13032 by PCR using primers lysC_ndeI_F(5-GAGCCTCATATGGCCCTGGTCGTACAGAAATA-3) and lysC_xhoI_R (5-ATTAATCTCGAGGCGTCCGGTGCCTGCATAAA-3). The DNA fragment wasdouble digested with NdeI and XhoI and ligated with NdeI-XhoI-digested plasmidpET-22b() (Novagen). The expression vector was designated pElysC. Fourteenpoint mutations of CgAK derived from the SCA calculation were created by usingQuikChange site-directed mutagenesis kits (Stratagene) according to standard protocolsand were named pElysC(I272E), pElysC(D274A), pElysC(E278V), pElysC(F283R),pElysC(Q298G), pElysC(N299L), pElysC(T336L), pElysC(M365A), pElysC(N372A),pElysC(N374A), pElysC(I375P), pElysC(E382A), pElysC(R384L), and pElysC(S386A).All of the CgAKs, including the wild type and mutants, had His6 tags at their C termini.

Protein expression and purification of CgAK and its mutants. CgAK and itsmutants were overexpressed in E. coli BL21-CodonPlus (DE3)-RIL cells (Strat-agene). The recombinant cells bearing the expression vectors were grown in LBmedium supplemented with 100 �g/ml ampicillin at 37°C. When the absorbanceof the culture at 600 nm reached 0.6, gene expression was induced by the additionof 0.1 mM isopropyl-�-D-thiogalactopyranoside (IPTG), and the culture wascontinued for an additional 12 to 14 h at 20°C. The cells were harvested andwashed with buffer A (20 mM Tris-HCl, pH 7.5) and suspended in buffer B (20mM Tris-HCl [pH 7.5], 150 mM NaCl, 0.5 M ammonium sulfate). Suspendedcells were disrupted by sonication (10 times for 30 s at 200 W with a 60-s coolinginterval) and centrifuged at 20,000 � g for 1 h. The supernatant was purified byuse of a Ni2-nitrilotriacetic acid (NTA) column (GE Healthcare Bio-Sciences,Piscataway, NJ) for obtaining samples for the enzyme activity assay.

Enzyme assay. The enzyme activity of aspartokinase was detected according toa method described previously by Black and Wright (2). The quantity of aspar-tate hydroxamate formed in the presence of hydroxylamine was measured at 540nm. The standard assay reaction mixture in 1 ml contained 200 mM Tris-HCl(pH 7.5), 10 mM MgSO4 � 6H2O, 10 mM aspartate, 10 mM ATP, 160 mMNH2OH � HCl (neutralized with KOH), and appropriate amounts of enzyme.After incubation at 30°C for 30 min, the reaction was stopped by mixing themixture with 1 ml of a 5% (wt/vol) FeCl3 solution, and the absorbance at 540 nmwas monitored. The kinetic analyses were carried out under the same conditionsexcept for the substrate concentrations. For the characterization of the profilesof inhibition of CgAKs, different concentrations of lysine and threonine wereadded to the standard reaction mixtures. The activities were expressed as relativeactivities normalized by the specific activities of CgAKs without inhibitions. One

unit of enzyme activity is defined as the formation of 1 micromole of aspartatehydroxamate per minute under assay conditions. All measurements were re-peated three times.

Generation of lysine-producing strain C. glutamicum LC298 by homologousrecombination. The feedback-insensitive CgAK with a mutation of Q298G wasintroduced into strain C. glutamicum ATCC 13032 via homologous recombina-tion. The fragment bearing a point mutation of Q298G in the lysC gene wasamplified from the above-described plasmid pElysC(Q298G) by using two prim-ers: 5-GAGCCTGAATTCGCCCTGGTCGTACAGAAATA-3 and 5-ATTAATGGATCCGCGTCCGGTGCCTGCATAAA-3. The PCR product wascloned into the EcoRI-BamHI site of suicide plasmid pK18mobsacB (8), and theresulting plasmid was designated pKlysC(Q298G). This plasmid was electropo-rated into C. glutamicum ATCC 13032 cells, and the recombinants with single-crossover homologous recombination were selected for kanamycin resistance.The kanamycin-resistant transformant was grown without kanamycin for 1 day,and the new recombinants with the second homologous recombination wereselected via a sucrose tolerance marker. Sequence analysis was performed for thevalidation of the second recombination. The strain with the correct mutation wasdesignated C. glutamicum LC298.

Fed-batch cultivation and analytical methods. The medium for the precultureand fermentation of C. glutamicum was described elsewhere previously (20). Themain fermentation was carried out with a 2-liter fermentor at 34°C. The feedingsolution contained 50% (wt/vol) glucose, 4.5% (wt/vol) NH4Cl, and 0.5 mg/literD-biotin. The feeding started after the initial sugar was consumed, and thefeeding rate was controlled to keep the glucose concentration lower than 10g/liter. The agitation speed was changed during the fermentation to keep asufficient oxygen supply. The pH was maintained at 7.3 by the addition ofNH4OH. The fed-batch fermentations were repeated three times for both strains(C. glutamicum ATCC 13032 and LC298).

The cell concentration was determined as the optical density at 660 nm and bydry biomass measurements. The glucose concentration was determined by high-performance liquid chromatography (HPLC). L-Lysine was quantified as de-scribed previously by Hsieh et al. (13).

RESULTS

Identification of highly correlated residues within the regu-latory domain of CgAK. The regulatory domain of CgAK wasconsidered to be primarily responsible for the allosteric inhi-bition (15, 29). All of the lysine-insensitive mutations wereidentified within the regulatory domain of CgAK (20, 23, 29).Thus, we performed SCA of the regulatory domain of theaspartokinase family in order to identify a small subset ofresidues that would be potential targets for point mutations.After deleting positions with high gap frequencies, the finalsamples for SCA calculations contained 500 sequences and 160amino acid positions corresponding to residues 250 to 409 ofCgAK. The evolutionary correlation between these 160 posi-tions is shown in the form of an SCA matrix in Fig. 1D.Residue pairs that are evolutionarily correlated would havehigh correlation values (red), while uncorrelated residue pairsshow low correlation values (blue).

The 25 highly correlated positions are summarized in Table1. Strikingly, most of the previously reported mutation pointsof feedback-resistant CgAKs are within these identified posi-tions. Based on this cluster of residues, we have created an-other 14 point mutations (Table 1). All of the constructedCgAK mutants are partially or completely insensitive to theconcerted inhibition of lysine and threonine (Fig. 2 and 3),indicating that these highly correlated residues are essential forthe allosteric regulation of CgAK.

Based on the recently reported crystal structure of CgAK(30), we are able to map these 25 highly correlated residuesinto the three-dimensional structure (Fig. 1E). Interestingly,although they are sparsely distributed within the sequence ofCgAK, this small cluster of residues forms an interconnected

4354 CHEN ET AL. APPL. ENVIRON. MICROBIOL.

on August 15, 2019 by guest

http://aem.asm

.org/D

ownloaded from

network mediated mainly by Van der Waals interactions orhydrogen bonds. This interconnected network not only coversthe most important residues of the inhibitors’ binding sites butalso provides a route linking the two threonine binding sitesand one lysine binding site, indicating their central roles inmediating the concerted inhibition of CgAK. For a better un-derstanding of their roles in allosteric regulation, we describethe mutational effect of these residues according to their loca-tions in more detail in the following sections.

Mutation of highly correlated residues within inhibitors’binding sites. Purified wild-type CgAK contains equimolarconcentrations of the � and � subunits (Fig. 1B) and isstrongly inhibited by the concerted action of threonine andlysine (Fig. 1C). Threonine or lysine alone shows only aslight inhibition of CgAK. The mutation of points withinthreonine (or lysine) binding sites to reduce the bindingaffinity of threonine or lysine would be expected to destroytheir concerted action upon CgAK and thus reduce theconcerted inhibition.

CgAK has two threonine binding sites per �� dimer, whichare located at the interface between the regulatory domain ofthe � subunit and the � subunit (30) (Fig. 1A). The two thre-onine binding sites are equal at the sequence level if we con-sider that the sequence of the � subunit is identical to theregulatory domain of the � subunit. The highly correlatedresidues within threonine binding sites are shown in Fig. 2A.Two mutations, G277A and A279V, were previously reportedfor the deregulation of CgAK by concerted inhibition, and theC. glutamicum mutant with the latter point mutation showedAEC resistance (29). Both G277 and A279 participate in thehydrogen bond network with threonine through water mole-cules. Mutations of other highly correlated residues within

threonine binding sites in our experiment, either by destroyingthe hydrogen bonds or by introducing steric hindrance, alsoreduced the concerted inhibition of CgAK by lysine and thre-onine (Fig. 2C). Mutations of I272E, Q298G, N372A, andI375P not only affected threonine binding but also reducedlysine inhibition. The concerted inhibitions of these mutantswere also significantly removed.

Although CgAK has two potential lysine binding units per�� dimer, only one lysine molecule has been found in thecrystal structure, which is located at the interface between the� subunit and the � subunit (30) (Fig. 1A). The highly corre-lated residues within lysine binding sites are shown as Fig. 2B.Two mutations, T361A (29) and S381F (23), were previouslyreported for the deregulation of CgAK by concerted inhibition,and the C. glutamicum mutant with the latter point mutationshowed AEC resistance. The results are understandable, asboth of these two point mutations would destroy the corre-sponding hydrogen bonds, and the binding of lysine would beinhibited. We constructed another mutation of highly corre-lated residues within this region, E382A. Strangely, this pointmutation does not change the inhibition profile of CgAK bylysine but change its inhibitions by threonine alone and bylysine plus threonine (Fig. 2C). Threonine partially inhibitedthe mutant, while the concerted inhibition was reduced. Thereason is unclear. The effects of mutations on I291, I293, andD294 have not been examined in this study. However, muta-tions of the corresponding points in aspartokinase III from E.coli release lysine inhibition (4a).

Mutation of highly correlated residues in other regions ofthe protein. Some highly correlated residues located in otherregions of the regulatory domain are not involved directly inthe inhibitors’ binding (Fig. 3A). These residues, however,

TABLE 1. Locations of the 25 highly correlated positions and point mutations

Point mutation Location of mutated points Correlation valuea Other property Reference

I272E Threonine binding sites 1.75 This studyD274A Threonine binding sites 1.51 This studyG277A Threonine binding sites 1.45 29E278V Threonine binding sites 1.49 This studyA279V Threonine binding sites 1.58 AEC resistant 29F283R Other positions 1.78 This studyI291b Lysine binding sites 1.47I293b Lysine binding sites 1.39D294b Lysine/threonine binding sites 1.88Q298G Threonine binding sites 1.88 This studyN299L Other positions 1.45 This studyS301F Other positions 1.55 AEC resistant 29T308A Other positions 1.75 AEC resistant 29T311I Other positions 1.86 AEC resistant 20T336L Other positions 1.39 This studyT361A Lysine binding sites 1.59 29F364A Other positions 1.57 29M365A Other positions 1.67 This studyN372A Threonine binding sites 1.75 This studyN374A Threonine binding sites 1.55 This studyI375P Threonine binding sites 1.75 This studyS381F Lysine binding sites 1.76 AEC resistant 23E382A Lysine binding sites 1.76 This studyR384L Other positions 1.74 This studyS386A Other positions 1.88 This study

a The highest values are shown.b The corresponding mutant has not been constructed or reported.

VOL. 77, 2011 DEREGULATION OF ALLOSTERIC ENZYME 4355

on August 15, 2019 by guest

http://aem.asm

.org/D

ownloaded from

FIG. 2. Distribution of highly correlated residues within inhibitors’ binding sites and results of point mutations. (A) Distribution of highlycorrelated residues within threonine binding sites. Residues from the � subunit are shown by asterisks and numbered as their correspondingpositions of the � subunit. (B) Distribution of the highly correlated residues within lysine binding sites. (C) Inhibition profiles of CgAKs by pointmutations of the highly correlated residues within inhibitors’ binding sites. The specific activities under standard assay conditions (Table 2) weretaken as 100%.

4356 CHEN ET AL. APPL. ENVIRON. MICROBIOL.

on August 15, 2019 by guest

http://aem.asm

.org/D

ownloaded from

may play important roles in signal transduction betweeninhibitor binding sites and catalytic domains (24, 30). Forexample, CgAK with a mutation of S301F binding bothlysine and threonine has been crystallized, indicating thatthe inhibitor’s binding was not destroyed upon such a mu-tation (30). Thus, the potential reason for this mutation’sdownregulation of CgAK is that the mutation destroys theallosteric interaction network and thus affects allosteric sig-nal transduction. Several AEC resistance mutants were re-

ported previously to have mutation points belong to thisgroup of residues, including S301F (29), T308A (29), andT311I (20). We also created several other point mutationsbelonging to this group, as shown in Fig. 3B. The mutationof T336L does not change the inhibition profiles of CgAK bylysine and threonine alone but reduces the concerted inhi-bition, revealing that it may destroy the signal transductionpathway of threonine, as threonine works with lysine only ina concerted manner. The mutation of F283R completelyremoved the inhibitions of CgAK by lysine and threonine,indicating that this residue may be crucial for the signaltransduction of both lysine and threonine. Interestingly, mu-tations of N299L, M365A, R384L, and S386A transfer thre-onine into an allosteric activator and reduce the concertedinhibition. This phenomenon was also illustrated previouslyfor the mutation of G277A (29). The reason for this findingis unclear now. All of these aspartokinase mutants havingthis property, however, showed relatively low activities com-pared to that of wild-type aspartokinase. One potential rea-son for this could be that these mutations destroy the inter-action between the � subunit and the � subunit, which isessential for enzyme activity. Threonine was reported pre-viously to be able to stabilize the protein complex and thusmay recover part of the enzyme activity (29). A more clearunderstanding of this phenomenon needs the aid of proteindynamics, which is under investigation by our group.

Introduction of the point mutation Q298G into the lysC genefor lysine overproduction. One of the feedback-insensitive mu-tations, Q298G, was introduced into the lysC gene of C. glu-tamicum ATCC 13032 through two-step homologous recom-bination. The resulting strain was designated C. glutamicumLC298. The production of lysine by strain LC298 was investi-gated with fed-batch fermentation using the wild-type strain asa control. All of the experiments were repeated three times.The results are depicted in Fig. 4.

The growth profile of strain LC298 is similar to that of thewild-type strain, although the final growth level was some-what lower. Strain LC298 started exponential growth at 4 hand reached stationary phase after 16 h. Lysine was accu-mulated mainly during the fed-batch stage (8 to 30 h) andreached an average of 58 g/liter within 30 h. After 30 h, thecell entered the decline phase, and lysine production almoststopped. In contrast, there was no extracellular lysine accu-mulation in C. glutamicum ATCC 13032. Thus, the lysineproduction in strain LC298 should be due to the deregula-tion of the allosteric inhibition of CgAK, which enabled theshift of metabolic flux from the tricarboxylic acid (TCA)cycle and biomass to lysine synthesis. The lysine productionrate of strain LC298 is 1.9 g/liter/h for the first 30 h, whichis comparable to that of the classically derived industrialstrain B-6 (20), although the yield is relatively low (0.17 glysine/g glucose and 1.38 g lysine/g biomass). Several factors,including precursor and NADPH availability and by-productformation, would be further engineered to increase lysineproduction and yield.

DISCUSSION

The deregulation of allosteric inhibitions of enzymes is achallenge for strain development (4). Random mutation and

FIG. 3. Distribution of the highly correlated residues within otherregions of the CgAK protein and results of point mutations. (A) Dis-tribution of the highly correlated residues within other regions of theprotein. (B) Profiles of inhibition of aspartokinase with point muta-tions of highly correlated residues within other regions of the protein.The specific activities under standard assay conditions (Table 2) weretaken as 100% (Table 2).

VOL. 77, 2011 DEREGULATION OF ALLOSTERIC ENZYME 4357

on August 15, 2019 by guest

http://aem.asm

.org/D

ownloaded from

selection have been traditionally used for such a purposeand have contributed greatly to the improvement of fermen-tation processes like amino acid production (21). However,this approach has serious disadvantages. For example, itis time-consuming and can introduce unnecessary mutationsat the same time. More importantly, a well-selectable phe-notype such as resistance to the analog of the inhibitor isrequired for this process. In many cases, however, there areno reliable selectable phenotypes for allosteric enzymessuch as phosphoenolpyruvate carboxylase, an importantanaplerotic enzyme inhibited by aspartate and malate. Thus,a more rational approach that can design targeted mutationsis highly desired. We demonstrated in this study that SCAcan be an effective way for solving the problem of the allo-steric regulation of industrially important enzymes.

SCA has proven to be an efficient approach to the clari-fication of the coevolutionary pattern of residues within aprotein family (26). SCA uses two levels of statistical anal-ysis of protein sequences to identify important targets: po-sition-specific conservation (first level) and position-corre-lated conservation (second level). The first level of statisticsis traditionally used in multiple-sequence alignments toidentify conserved residues at certain positions. The conser-vation of residues, however, does not give direct information

on their function. SCA thus focuses more on the secondlevel of statistics to identify highly correlated residues inwhich the mutation of one residue would often result in thecorresponding change of the other during evolution. Theseresidues coevolve and often form an interacting network inorder to keep a certain protein function or regulation. Asindicated in this study, a small cluster of residues within theregulatory domain shows a high correlation during the evo-lution of aspartokinases. These residues form an intercon-nected network linking different inhibitor binding sites andplay important roles in mediating the concerted inhibitionof CgAK by lysine and threonine. The mutation of anyresidues within this network affects the profiles of inhibitionof CgAK. The effects of mutations in threonine and lysinebinding sites are reasonably comprehensible, as they willaffect the binding of lysine or threonine. These point muta-tions can be evaluated if the crystal structure is available andthe inhibitor binding sites can be analyzed. However, muta-tions of residues not involved directly in the binding ofinhibitors, such as F283R and S301F, have also been foundto be able to reduce the concerted inhibition of CgAK. Thecrystal structure of a CgAK S301F mutant showed that thispoint mutation did not affect the inhibitors’ binding. Thesepoint mutations cannot be predicted, and the phenomenacannot be explained by a simple analysis of the static crystalstructures. Allosteric regulation is a dynamic process, andthese residues may play important roles in signal transduc-tion during this process (28). The prediction of these resi-dues illustrates the usefulness of SCA. It should be men-tioned that the crystal structure of CgAK was not availableat the time when we predicted and experimentally verifiedall designed mutations. The crystal structure was combinedduring the preparation of the manuscript, as it facilitatedthe explanation of the phenomena. However, it is not nec-essary for SCA predictions, as SCA is a sequence-basedanalysis. We also expect that the identified positions wouldbe efficient for other members of the aspartokinase family.

After the identification of targeted positions, the choice ofamino acid substitution is also important for the successfulderegulation of enzyme inhibition. In case there is no crystalstructure available, this choice could be made based on ananalysis of the amino acid distribution in the specific posi-tion during evolution (or in the multiple-sequence align-ment). The basic principle is to destroy the allosteric inter-action network. One simple strategy is to change the nativeamino acid to other amino acids that are rarely or neverused in this position. More specifically, changing the aminoacid group to ones having different chemical propertieswould be expected to have a more obvious effect on thederegulation of allosteric inhibition. On the other hand, theselection of the amino acid substitution could also affectthe specific activity of the enzyme. In our cases, since all ofthe targeted residues are located in the regulatory domainof aspartokinase and not involved directly in substrate bind-ing, the mutations did not show a significant influence on theKm values for ATP and aspartate (Table 2). The reductionof the enzyme specific activities of some mutants could beattributed to the diminished Vmax owing to the effect ofprotein folding, stability, or other factors. The predicteddesign of the optimal substitution to largely reduce alloste-

FIG. 4. Profiles of fed-batch fermentation (L-lysine production, cellgrowth, and glucose consumption) by C. glutamicum wild-type strainATCC 13032 and mutant strain LC298. Data represent mean valuesand standard deviations from three independent cultures. OD660, op-tical density at 660 nm.

4358 CHEN ET AL. APPL. ENVIRON. MICROBIOL.

on August 15, 2019 by guest

http://aem.asm

.org/D

ownloaded from

ric inhibition and minimize the loss of enzyme activity wouldbe the future direction.

The identification of these feedback-insensitive CgAKmutants enabled us to create a new lysine producer based ononly one point mutation in the lysC gene. This point muta-tion (Q298G), however, enabled C. glutamicum LC298 toproduce 58 g/liter lysine within 30 h. The production rate iscomparable to those of some industrial lysine producers,such as C. glutamicum B-6 (20). A similar point mutation onthe lysC gene, T311I, was previously identified by compar-ative genomics (20). The introduction of this mutation intowild-type C. glutamicum ATCC 13032 achieved a lysine pro-duction rate of 55 g/liter in 30 h. Compared with that strain,our strain showed a slightly higher level of lysine productionbut better cell growth, indicating that the point mutation ofQ298G on lysC has a lesser effect on cell growth. Thus, thisstrain can be used as a good starting strain for the introduc-tion of more beneficial mutations.

The construction of minimally mutated strains is a direc-tion for future industrial strain development (4, 14, 25). Anattractive idea based on comparative genomics was pro-posed previously by Ikeda et al. (14). That approach, how-ever, has several limitations. For example, it requires theavailability of different industrial strains, and the beneficialmutations should appear in the selected industrial strain.What is more, it is difficult to differentiate the beneficial orunbeneficial mutations without an understanding of theirpotential roles. The strategy herein provides an alternativeway for achieving such a purpose without the need for in-dustrial strains. Our approach is based on more rationaltools combining sequences and structure analysis to improvethe defined enzymes in metabolic pathways. We have usedthe same approach to modify several other important en-zymes, including homoserine dehydrogenase and phosphoe-nolpyruvate carboxylase, which also greatly increase lysineproduction (Z. Chen et al., unpublished data). It can beconcluded that coevolutionary analysis is an efficient ap-

proach to accelerate the process of industrial strain devel-opment.

REFERENCES

1. Agarwal, P. K., S. R. Billeter, P. T. Rajagopalan, S. J. Benkovic, and S.Hammes-Schiffer. 2002. Network of coupled promoting motions in enzymecatalysis. Proc. Natl. Acad. Sci. U. S. A. 99:2794–2799.

2. Black, S., and N. G. Wright. 1954. �-Aspartate kinase and �-aspartyl phos-phate. J. Biol. Chem. 213:27–38.

3. Capra, J. A., and M. Singh. 2007. Predicting functionally important residuesfrom sequence conservation. Bioinformatics 23:1875–1882.

4. Chen, Z., M. Wilmanns, and A. P. Zeng. 2010. Structural synthetic biotech-nology: from molecular structure to predictable design for industrial straindevelopment. Trends Biotechnol. 28:534–542.

4a.Chen, Z., S. Rappert, J. Sun, and A. P. Zeng. 2011. Integrating moleculardynamics simulation and co-evolutionary analysis for reliable prediction andderegulation of aspartokinase for amino acid production. J. Biotechnol.[Epub ahead of print.] doi:10.1016/j.jbiotec.2011.05.005.

5. Chipman, D. M., and B. Shaanan. 2001. The ACT domain family. Curr.Opin. Struct. Biol. 11:694–700.

6. Doolittle, R. F. 1996. Computer methods for macromolecular sequence anal-ysis, volume 266. Academic Press, San Diego, CA.

7. Edgar, R. C. 2004. MUSCLE: multiple sequence alignment with high accu-racy and high throughput. Nucleic Acids Res. 32:1792–1797.

8. Eggeling, L., and M. Bott. 2005. Handbook of Corynebacterium glutamicum.CRC Press, Boca Raton, FL.

9. Estabrook, R. A., et al. 2005. Statistical coevolution analysis and moleculardynamics: identification of amino acid pairs essential for catalysis. Proc. Natl.Acad. Sci. U. S. A. 102:994–999.

10. Grant, G. A. 2006. The ACT domain: a small molecule binding domainand its role as a common regulatory element. J. Biol. Chem. 281:33825–33829.

11. Halabi, N., O. Rivoire, S. Leibler, and R. Ranganathan. 2009. Proteinsectors: evolutionary units of three-dimensional structure. Cell 138:774–786.

12. Hatley, M. E., S. W. Lockless, S. K. Gibson, A. G. Gilman, and R. Ranga-nathan. 2003. Allosteric determinants in guanine nucleotide-binding pro-teins. Proc. Natl. Acad. Sci. U. S. A. 100:14445–14450.

13. Hsieh, C. L., K. P. Hsiung, and J. C. Su. 1995. Determination of lysine withninhydrin-ferric reagent. Anal. Biochem. 224:187–189.

14. Ikeda, M., J. Ohnishi, M. Hayashi, and S. Mitsuhashi. 2006. A genome-based approach to create a minimally mutated Corynebacterium glutamicumstrain for efficient L-lysine production. J. Ind. Microbiol. Biotechnol. 33:610–615.

15. Kalinowski, J., B. Bachmann, G. Thierbach, and A. Puhler. 1990. Aspar-tokinase genes lysC alpha and lysC beta overlap and are adjacent to theaspartate beta-semialdehyde dehydrogenase gene asd in Corynebacteriumglutamicum. Mol. Gen. Genet. 224:317–324.

16. Kikuchi, Y., H. Kojima, and T. Tanaka. 1999. Mutational analysis of thefeedback sites of lysine-sensitive aspartokinase of Escherichia coli. FEMSMicrobiol. Lett. 173:211–215.

17. Kotaka, M., J. Ren, M. Lockyer, A. R. Hawkins, and D. K. Stammers. 2006.Structures of R- and T-state Escherichia coli aspartokinase III. Mechanismsof the allosteric transition and inhibition by lysine. J. Biol. Chem. 281:31544–31552.

18. Lockless, S. W., and R. Ranganathan. 1999. Evolutionarily conservedpathways of energetic connectivity in protein families. Science 286:295–299.

19. Miyata, Y. O., H. Kojima, and K. Sano. 2001. Mutation analysis of thefeedback inhibition site of aspartokinase III of Escherichia coli K-12 andits use in L-threonine production. Biosci. Biotechnol. Biochem. 65:1149–1154.

20. Ohnishi, J., et al. 2002. A novel methodology employing Corynebacteriumglutamicum genome information to generate a new L-lysine-producing mu-tant. Appl. Microbiol. Biotechnol. 58:217–223.

21. Omori, K., Y. Imai, S. I. Suzuki, and S. Komatsubara. 1993. Nucleo-tide sequence of the Serratia marcescens threonine operon and analysisof the threonine operon mutations which alter feedback inhibition of bothaspartokinase I and homoserine dehydrogenase I. J. Bacteriol. 175:785–794.

22. Sano, K., and I. Shiio. 1970. Microbial production of L-lysine III. Productionby mutants resistant to S-(2-aminoethyl)-L-cysteine. J. Gen. Appl. Microbiol.16:373–391.

23. Sindelar, G., and V. F. Wendisch. 2007. Improving lysine production byCorynebacterium glutamicum through DNA microarray-based identificationof novel target genes. Appl. Microbiol. Biotechnol. 74:677–689.

24. Smock, R. G., and L. M. Gierasch. 2009. Sending signals dynamically. Sci-ence 324:198–203.

25. Stephanopoulos, G. 1999. Metabolic fluxes and metabolic engineering.Metab. Eng. 1:1–11.

TABLE 2. Specific activities and kinetic parameters ofaspartokinase mutants

Mutation Mean sp acta

(mU/mg protein) SD

Mean Km (mM) SD for:

ATPb Aspartateb

Wild type 280 15 0.48 0.08 0.25 0.03I272E 146 11 0.56 0.14 0.26 0.08D274A 238 19 0.51 0.22 0.31 0.04E278V 242 33 0.49 0.10 0.28 0.11F283R 88 11 0.55 0.12 0.38 0.08Q298G 294 15 0.46 0.17 0.25 0.06N299L 118 16 0.51 0.04 0.28 0.08T336L 205 12 0.50 0.09 0.27 0.05M365A 131 13 0.49 0.25 0.31 0.07N372A 174 12 0.48 0.08 0.27 0.12N374A 199 23 0.52 0.12 0.38 0.11I375P 125 15 0.60 0.18 0.32 0.13E382A 257 18 0.50 0.13 0.24 0.05R384L 101 14 0.53 0.11 0.29 0.11S386A 97 8 0.63 0.08 0.31 0.15

a The standard assay mixtures contained 200 mM Tris-HCl (pH 7.5), 10 mMMgSO4 � 6H2O, 10 mM aspartate, 10 mM ATP, and 160 mM NH2OH � HCl.

b Kinetic analyses were carried out under the same conditions except for thesubstrate concentrations.

VOL. 77, 2011 DEREGULATION OF ALLOSTERIC ENZYME 4359

on August 15, 2019 by guest

http://aem.asm

.org/D

ownloaded from

26. Suel, G. M., S. W. Lockless, M. A. Wall, and R. Ranganathan. 2003. Evo-lutionarily conserved networks of residues mediate allosteric communicationin proteins. Nat. Struct. Biol. 10:59–69.

27. Thompson, J. D., T. J. Gibson, F. Plewniak, F. Jeanmougin, and D. G.Higgins. 1997. The CLUSTAL_X Windows interface: flexible strategies formultiple sequence alignment aided by quality analysis tools. Nucleic AcidsRes. 25:4876–4882.

28. Tsai, C. J., A. Sol, and R. Nussinov. 2009. Protein allostery, signal transmis-

sion and dynamics: a classification scheme of allosteric mechanisms. Mol.Biosyst. 5:207–216.

29. Yoshida, A., et al. 2007. Structural insight into concerted inhibition of �2�2-type aspartate kinase from Corynebacterium glutamicum. J. Mol. Biol. 368:521–536.

30. Yoshida, A., T. Tomita, T. Kuzuyama, and M. Nishiyama. 2010. Mechanismof concerted inhibition of �2�2-type heterooligomeric aspartate kinase fromCorynebacterium glutamicum. J. Biol. Chem. 285:27477–27486.

4360 CHEN ET AL. APPL. ENVIRON. MICROBIOL.

on August 15, 2019 by guest

http://aem.asm

.org/D

ownloaded from