ANALYTICALBIOCHEMISTRY

Analytical Biochemistry 327 (2004) 91–96

www.elsevier.com/locate/yabio

Tools for metabolic engineering in Escherichia coli: inactivationof panD by a point mutation

Jonathan Kennedy and James T. Kealey¤

Kosan Biosciences, Inc., 3832 Bay Center Place, Hayward, CA 94545, USA

Received 9 September 2003

Abstract

L-Aspartate-�-decarboxylase (PanD) catalyzes the decarboxylation of aspartate to produce �-alanine, a precursor of Coenzyme A(CoA). The pyruvoyl-dependent enzyme from Escherichia coli is activated by self-cleavage at serine 25 to generate a 102-residue �subunit with the pyruvoyl group at its N terminus and a 24-residue � subunit with a hydroxy at its C terminus. A mutant form of thepanD gene from E. coli in which serine 25 was replaced with an alanine (S25A) was constructed. Assays conducted in vitro andin vivo conWrmed that the mutant version was completely inactive and was incapable of undergoing self-cleavage to generate theactive form of the enzyme. The S25A panD mutant was used to replace the chromosomal copy of panD in BAP1, a strain of E. colimodiWed for polyketide production. Comparison of this strain with panD2 mutant strains derived from E. coli SJ16 showed an equiv-alent dependence on exogenous �-alanine for growth in liquid medium. Unlike the undeWned and leaky panD2 mutation, the panDS25A mutation is deWned and tight. The panD S25A E. coli strain enables analysis of intracellular acyl-CoA pools in both deWnedand complex media and is a useful tool in metabolic engineering studies that require the manipulation of acyl-CoA pools for theheterologous production of polyketides. 2004 Elsevier Inc. All rights reserved.

Keywords: Polyketide; panD; Coenzyme A

The use of engineered Escherichia coli strains for theproduction of complex and bulk organic molecules isattractive from both academic and industrial stand-points. Many tools for genetic engineering of E. coli areavailable and high-density fermentation systems aresimilarly well developed. By manipulating the E. coligenome and/or by introducing heterologous, plasmid-borne genes, strains for the production of severalcommodity chemicals [1,2], 6-methylsalicylic acid [3],and 6-deoxyerythronolide B (6-dEB)1 [4] have beendeveloped. The latter two compounds are examples ofthe polyketides, a therapeutically important class ofnatural products that are synthesized from simple acyl-Coenzyme A (CoA) precursors, including acetyl-CoA,

malonyl-CoA, propionyl-CoA, methylmalonyl-CoA,and ethylmalonyl-CoA. Many of the acyl-CoA polyke-tide precursors are not naturally produced in E. coli orare present at very low concentrations. Moreover, thecomposition of the acyl-CoA pool may vary consider-ably throughout the growth phase and the polyketideproduction phase.

To fully develop E. coli as a host for the productionof polyketides [3–5], the acyl-CoA pools must be deter-mined accurately to facilitate their manipulation and/orto ensure that adequate levels of necessary acyl-CoAsare available for polyketide biosynthesis. For example,for production of 6-methylsalicylic acid and 6-dEB inE. coli, it was necessary to manipulate malonyl-CoA ormethylmalonyl-CoA levels either through media optimi-zation or by introduction of heterologous acyl-CoApathways. The need to measure acyl-CoA levels in engi-neerined E. coli led us to develop a convenient, robustand reliable method for the in vivo quantitation of acyl-CoAs.

¤ Corresponding author. Fax: 1-510-732-8401.E-mail address: kealey@kosan.com (J.T. Kealey).1 Abbreviations used: 6-dEB, 6-deoxyerythronolide; carb, carbeni-

cillin; Cm, chloramphenicol; kan, kanamycin; pks, polyketide syn-thase; IPTG, isopropyl �-D-thiogalactoside.

0003-2697/$ - see front matter 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.ab.2003.11.027

92 J. Kennedy, J.T. Kealey / Analytical Biochemistry 327 (2004) 91–96

Pantothenate is an essential precursor for CoA andacyl-carrier protein biosynthesis in E. coli (Fig. 1).Pantothenate is formed from the condensation of D-pan-toate and �-alanine. In bacteria the major route of �-ala-nine production is from the �-decarboxylation ofaspartate, catalyzed by L-aspartate-�-decarboxylase (EC4.1.1.11), the product of the panD gene. Analysis of intra-cellular acyl-CoA levels can be performed using a panDmutant coupled with feeding of radiolabeled �-alanine.The most utilized panD mutant for CoA labeling ispanD2, which has been shown by DNA sequencing notto carry mutations in the structural panD gene [6,7].Moreover, panD2 is known to have a leaky phenotypewhich allows, under certain conditions, growth withoutexogenous �-alanine [6,8]. Thus, the panD2 mutationlikely aVects the regulation of panD, with the level ofauxotrophy varying with the conditions used; e.g., it hasbeen reported that a more complete mutant phenotype isobserved with growth at 42 °C [8]. For this reason wewished to produce a deWned panD mutant with the leastpossible perturbation to the E. coli host system.

L-Aspartate-�-decarboxylase is a pyruvoyl-dependentenzyme, initially made as an inactive proenzyme (� pro-tein), which is subsequently cleaved to � and � subunits[6]. The E. coli � protein is capable of self-cleavage atserine 25 to form the active � subunit (with a pyruvoyl

group at its N terminus) and � subunit (with a hydroxylgroup at its C terminus). We have demonstrated thatmutation of this speciWc serine residue to an alanineresults in an enzyme that is incapable of self-cleavage,rendering it locked in the inactive � protein form. Thereplacement of the wild-type panD gene in E. coli strainBAP1 with this new mutant allele, panD S25A, resultedin a strain with an absolute requirement for exogenous�-alanine. In deWned media the S25A strain (K173-145)allows for incorporation of [3H]�-alanine into acyl-CoAs without isotopic dilution. EYcient labeling ofacyl-CoAs is also achieved in complex media.

Materials and methods

For the following experiments, restriction enzymesites in oligonucleotides are shown in boldface; startand stop codons are underlined. The sequences of allclones generated by PCR were veriWed by DNAsequence analysis. For the selection of plasmids carbeni-cillin (carb) was used at 100 �g/mL, chloramphenicol(Cm) was used at 25 �g/mL, and kanamycin (kan) wasused at 50 �g/mL.

Bacterial strains and growth

Strain K117-60, a BL21(DE3) (Novagen) derivativecarrying the panD2 allele, was used for overexpressionstudies [5]. Strain SJ16 (CGSC 6341) was used as a con-trol for growth experiments. Strain BAP1, a BL21(DE3)derivative modiWed for expression of active polyketidesynthases (PKSs) [4], was used for the gene knockoutexperiments (see Table 1 for genotypes of strains).Strains were grown in LB or M9 medium [9], supple-mented with 5 �M �-alanine and 100 mg/L methionine asneeded. For growth studies 3 mL of M9 medium plussupplements in a 16 £ 100-mm glass tube was inoculatedwith a single colony and grown overnight at 37 °C withshaking. The cells were harvested by centrifugation at3000g for 10 min and washed 3£ with M9 medium lack-ing �-alanine. The equivalent of 0.5 mL of overnightculture was then used to inoculate 50 mL of M9 withand without �-alanine and 50 mL of LB with and with-out �-alanine. Cultures were incubated at 37 °C in 250-mL Erlenmeyer Xasks with shaking at 200 rpm. Aliquotsof the cultures were removed at intervals and OD600was measured using a Spectronic Genesys 5 spectro-photometer.

PCR cloning and expression plasmids

The wild-type panD gene was PCR-ampliWed fromE. coli BL21 genomic DNA using the primers 50 AAGGTA GAA CAT ATG ATT CGC ACG ATG CTG CA30 (JKPAN1) and 50 CCA GCC GCA AGC TTA ACA

Fig. 1. Biosynthetic pathway for Coenzyme A in E. coli.

J. Kennedy, J.T. Kealey / Analytical Biochemistry 327 (2004) 91–96 93

ATC AAG CAA CCT GT 30 (JKPAN2). The PCRproduct was digested with NdeI and HindIII and clonedinto pET22b (Novagen) to give pKOS173-131-21. Forthe C-terminally 6£-histidine-tagged version, panD wasampliWed using primers JKPAN1 and 50 CGC AGGGAT AAA AGC TTA GCA ACC TGT ACC GG 30

(JKPAN3). The PCR product was digested with NdeIand HindIII and cloned into pET22b to give pKOS173-132-31.

The S25A mutant panD gene was PCR-ampliWedusing a two-step recombinant PCR procedure. The 50

end of the gene was PCR-ampliWed using the primersJKPAN1 and 50 AAT CCT GGT CAA TGG CGC ATGCAC CTT CAT AGT GCA GGT CCG CAT 30

(JKPAN5). The 30 end of the gene was PCR-ampliWedusing primers 50 ATG CGG ACC TGC ACT ATGAAG GTG CAT GCG CCA TTG ACC AGG ATT 30

(JKPAN4) and JKPAN2 or JKPAN3 to generate thenatural or histidine-tagged version, respectively. The twoPCR products were annealed, extended, and subjected toa second ampliWcation step using the Xanking primersJKPAN1 and JKPAN2 or 3. The full-length PCR prod-ucts were digested with NdeI and HindIII and subclonedinto pET22b to give pKOS173-133-46 and pKOS173-134-52 for the natural or histidine-tagged version,respectively.

Protein expression and puriWcation

For overexpression studies cells were grown in100 mL LB + carb in a 250-mL Erlenmeyer Xask at 37 °Cto an OD600 of 0.6. Cultures were induced with 1 mMIPTG and grown for a further 3 h at 30 °C. Cell extractswere prepared by treatment with lysozyme, followed bysonication. Histidine-tagged proteins were puriWed fromthe soluble cell extract using nickel spin columns (Qiagen)following the manufacturer’s instructions. Samples wereprepared for SDS–PAGE by incubating at 90 °C for10 min in 2% SDS and protein expression was analyzedusing 16.5% Tris–tricine SDS–PAGE gels (Bio-Rad).

Mutant strain construction

The region of DNA upstream of panD was PCR-ampliWed from BL21 genomic DNA using the primers 50

GTA CCC TTC GTC GAC CAG CGT GAC CAG 30

(JKPAN7) and JKPAN5. The region downstream ofpanD was PCR-ampliWed using the primers JKPAN4and 50 CGA CCT CAA GCG GCC GCT GAG TAA T30 (JKPAN6). The two PCR products were annealed,extended, and subjected to a second ampliWcation stepusing the Xanking primers JKPAN6 and JKPAN7. Thefull-length PCR product, containing the mutant panDgene plus Xanking regions, was subcloned into pSL1190(Pharmacia) as a SalI–NotI fragment to give pKOS173-129-1. The SalI–NotI fragment was excised frompKOS173-129-1 and then subcloned into pKO3 [10] togive pKOS173-136.

The wild-type panD gene in E. coli strain BAP1 wasreplaced with the mutant version using pKOS173-136.The plasmid was introduced into BAP1 and transfor-mants were grown on LB + Cm plates at 44 °C to selectfor integrants (pKO3 contains a temperature sensitiveSC101 replicon that is unable to replicate at 44 °C). Col-onies that grew at 44 °C were picked and grown for sev-eral hours in 1 mL LB broth with Cm. Dilutions of thisculture were plated on LB + 5% sucrose plates whichselects for strains that have lost the sacB gene present onpKO3. Sucrose-resistant colonies were screened for sen-sitivity to Cm (CmS). Selected CmS colonies were furtherscreened by PCR with primers JKPAN8 and JKPAN9,followed by digestion with SphI to distinguish betweenwild-type and mutant panD and to assess �-alanine aux-otrophy. Using this procedure the panD S25A mutantstrain KOS173-145 was identiWed.

Acyl-CoA analysis

A 20-�L portion of a 3-mL overnight culture, grownin LB, was used to inoculate 1 mL of LB in a 16 £ 100-mm culture tube. This culture was incubated at 37 °C

Table 1Plasmids and E. coli strains used in this study

Strain/plasmid Description Source

SJ16 (CGSC 6341) F¡ panD2 zad-220::Tn10 gyrA216 (NalR) relA1 spoT1 metB1 E. coli Genetic StockCenter

BL21 (DE3) F¡ ompT hsdSB ( ) gal dcm (DE3) NovagenK117-60 BL21 (DE3) with panD2 [5]BAP1 F¡ ompT hsdSB ( ) gal dcm (DE3) �prpRBCD::T7prom-sfp, T7prom-prpE [4]K173-145 BAP1 panD::panDS25A This studypKOS173-131-21 ColE1 based plasmid for expression of E. coli panD from the T7 promoter (carbR) This studypKOS173-132-31 ColE1 based plasmid for expression of E. coli panD + 6£HIS from the T7 promoter (carbR) This studypKOS173-133-46 ColE1 based plasmid for expression of E. coli panD S25A from the T7 promoter (carbR) This studypKOS173-134-52 ColE1 based plasmid for expression of E. coli panD S25A + 6£HIS from the T7 promoter (carbR) This studypBP130 ColE1 based plasmid for expression of DEBS2 and DEBS3 from the T7 promoter (carbR) [4]pBP144 ColE1 based plasmid for expression of propionyl-CoA carboxylase and DEBS1

from T7 promoters (kanR)[4]

r ¡B m ¡B

r ¡B m ¡B

94 J. Kennedy, J.T. Kealey / Analytical Biochemistry 327 (2004) 91–96

with shaking to an OD600 of 0.5; 10 �Ci of [3H]�-alanine(speciWc activity, 50 Ci/mmol), sodium propionate to 5mM, and IPTG to 0.5 mM were added, and the cultureswere incubated at 22 °C for 24 h at 200 rpm. Cells werechilled on ice, harvested by centrifugation (10,000g,5 min, 4 °C), and then washed twice with 0.5 mL ice-coldLB. The washed cell pellet was resuspended in 300 �L ofice-cold 10% trichloroacetic acid containing 5 �L CoAstandard mix (0.5 mM CoA and 1.6 mM each of malo-nyl-, methylmalonyl-, succinyl-, acetyl-, and propionyl-CoA), and 150 �L of 6-�m glass beads were added. Thesuspension was vortexed four times for 30 s, with coolingon ice in between. Precipitants and cell debris wereremoved by centrifugation (10,000g, 5 min, 4 °C), and100 �L of the supernatant was analyzed for acyl-CoAsby HPLC as described below.

HPLC was performed using a 150 £ 4.6-mm 5-�mODS-3 Inertsil HPLC column (MetaChem Technolo-gies). HPLC buVer A contained 100 mM NaH2PO4, 75mM NaOAc, pH 4.6, and buVer B contained 40% buVerA, 60% methanol. The HPLC column was equilibratedwith buVer A containing 10% buVer B at a Xow rate of1 mL/min. After sample injection, a linear gradient to30% buVer B was formed over 18 min, followed by a lin-ear gradient to 100% buVer B over 4 min; the eluant wasmonitored at 260 nm and, for radioactive samples, theeluant was also analyzed using a Packard 500TR FlowScintillation Analyzer.

Analysis for 6-dEB production

Plasmids pBP130 carrying DEBS2 and DEBS3 [4]and pBP144 carrying DEBS1 and the propionyl-CoAcarboxylase genes for methylmalonyl-CoA production[4] were introduced into strains BAP1 [4] and K173-145.From a 3-mL overnight culture in LB + carb + kan,0.5 mL was used to inoculate 25 mL of LB + carb andkan in a 250-mL Erlenmeyer Xask and incubated at37 °C with shaking. The cultures were grown to anOD600 of 0.5, the cultures were cooled to 22 °C and pro-pionate and IPTG were added to 5 and 1 mM, respec-tively. The cultures were incubated at 22 °C with shakingfor an additional 48 h. Titers of 6-dEB were quantiWedfrom culture extracts by HPLC/MS with evaporativelight scattering detection as previously described [5].

Results and discussion

Complementation of panD2 mutant strain

Strain K117-60 containing the plasmids pKOS173-131-21, pKOS173-132-31, pKOS173-133-46, and pKOS173-134-52 (see Table 1) was analyzed for growth on M9medium plus 1 mM IPTG with and without �-alanine.The wild-type panD and panD-HIS genes present on

pKOS173-131-21 and pKOS173-132-31 complementedthe panD2 mutation present in strain K117-60, resultingin wild type growth on medium lacking �-alanine. TheS25A mutant panD genes present on plasmids pKOS173-133-46 and pKOS173-134-52 had no discernable eVecton growth characteristics, i.e., they failed to complementthe panD2 mutant. This result demonstrates that thewild-type L-aspartate-�-decarboxylase with or withoutthe histidine tag is active in vivo, while the S25A mutant(with or without the histidine tag) is inactive.

Overexpression of native PanD and PanD-S25A

Expression from plasmids pKOS173-131-21, pKOS173-132-31, pKOS173-133-46, and pKOS173-134-52 wasanalyzed by SDS–PAGE of cell extracts. Soluble L-aspartate-�-decarboxylase was clearly observed in allcases (data not shown), demonstrating that lack of com-plementation with the S25A mutant was not due to lackof expression. The overexpressed proteins from strainscarrying the histidine-tagged versions of the wild-typeand mutant panD were puriWed using nickel spin columns.

Native L-aspartate-�-decarboxylase from E. coli is ca-pable of self processing to � and � subunits. However,this process is rather slow, exhibiting an in vitro half-lifeof several days at 37 °C and 16 h at 50 °C [6]. PuriWedproteins were analyzed by SDS–PAGE and processingwas examined by incubation at 50 °C for 16 h (Fig. 2).The unprocessed � protein in control samples is presentin both native and mutant samples as two closely run-ning bands that migrate near to the expected size of15,352 Da. Incubation at 50 °C leads to the processing ofthe native � protein into � and � subunits of the expectedsizes of 12,536 and 2834 kDa, respectively. In contrastthe S25A mutant remains a single polypeptide afterincubation, demonstrating that the S25A mutationprevents self-processing, which results in a completelyinactive enzyme.

Fig. 2. SDS–PAGE of puriWed PanD, demonstrating self-cleavage.PuriWed wild-type PanD-HIS (lane 2) is self-cleaved following 16 h at50 °C (lane 3). The mutant S25A PanD-HIS (lane 4) is unaVected bysimilar incubation at 50 °C (lane 5). Protein molecular weight stan-dards are show in lanes 1 and 6.

J. Kennedy, J.T. Kealey / Analytical Biochemistry 327 (2004) 91–96 95

Construction of mutant strain

The S25A mutation was introduced into the wild-typepanD gene in strain BAP1 using the pKO3 system [8].Mutants were screened by PCR and restriction analysisfor the presence of the mutant allele. Selected mutantstrains were then tested for �-alanine auxotrophy bycomparison to wild-type strains grown on minimalmedium plates. All strains identiWed by PCR to carry themutation were found to grow signiWcantly slower in the

absence of �-alanine. Growth analyses in liquid mediumand in comparison to wild-type and SJ16 E. coli (Fig. 3)demonstrated that the mutant had an absolute require-ment for �-alanine, with growth eVectively stopping atan OD600 of 0.3 (presumably when all the endogenous �-alanine in the inoculum had been utilized). Growth ofthe S25A mutant strain in LB or in M9 supplementedwith 5 �M �-alanine was indistinguishable from that ofthe wild-type strain.

Acyl-CoA analysis

The growth analyses show a complete requirementfor �-alanine in minimal medium. Feeding of [3H]�-ala-nine to strains grown in this medium allows absolutedetermination of intracellular acyl-CoAs as there can beno isotopic dilution of the [3H]�-alanine. Acyl-CoAanalysis using minimal medium conWrmed that acyl-CoAs could be clearly detected radiometrically (data notshown). However, for the production of 6-dEB in E. coli,much higher metabolite titers are observed in complexmedia. Analysis of strains grown in LB (Fig. 4) demon-strates very clear labeling of acyl-CoAs. In this examplethe level of propionyl-CoA increased markedly (to 30%of the total CoA pool), when the E. coli prpE gene wasoverexpressed in the presence of propionate. Thus rela-tive levels of acyl-CoAs can be determined in conditionsused for polyketide production.

As aspartate decarboxylase is a key enzyme for CoAmetabolism we were concerned that the panD mutant

Fig. 3. Growth curves of SJ16 and K173-145. Strain K173-145 (solidline) was grown in LB (M), M9 (�), and M9 + �-alanine (�). Strain SJ16(broken line) was grown in LB (�), M9 (�), and M9 + �-alanine (�).

Fig. 4. Acyl-CoA HPLC chromatograms. Strain K173-145 was grown in LB with propionate and IPTG (broken line) and without (solid line).Induction of prpE with IPTG and addition of propionate leads to the production of signiWcant amounts of propionyl-CoA (31% of total CoAs).

96 J. Kennedy, J.T. Kealey / Analytical Biochemistry 327 (2004) 91–96

strain may adversely aVect polyketide production inE. coli. The mutant strain was analyzed for 6-dEB pro-duction using the two-plasmid system for expression ofthe DEBS PKS as described by Pfeifer et al. [4]. ThepanD mutant strain K173-145 was compared to BAP1 asa host for polyketide production. Six individual trans-formants were analyzed for each strain. BAP1 produced9.75 § 0.4 mg 6-dEB/L and K173-145 produced 11.3 §0.6 mg 6-dEB/L. Thus the panD mutation was foundto have no negative eVect on polyketide production inE. coli when grown in LB.

The BAP1/panD S25A strain has been demonstratedto have an absolute requirement for �-alanine; it hasalso been shown that the panD S25A mutation has nodiscernable eVect on 6-dEB production. As recentlydemonstrated by Murli et al. [11], K173-145 and otherstrains carrying the same mutation will be extremely use-ful for monitoring acyl-CoA pools during heterologouspolyketide production in E. coli.

Acknowledgments

We thank C. Richard Hutchinson for critical readingof the manuscript, Linda Dayem for acyl-CoA analysis,and John Carney and Nina Viswanathan for LC/MSanalysis.

References

[1] G.N. Vemuri, M.A. Eiteman, E. Altman, EVects of growth modeand pyruvate carboxylase on succinic acid production by metabol-

ically engineered strains of Escherichia coli, Appl. Environ. Micro-biol. 68 (2002) 1715–1727.

[2] S.J. Sim, K.D. Snell, S.A. Hogan, J. Stubbe, C. Rha, A.J. Sinskey,PHA synthase activity controls the molecular weight and polydis-persity of polyhydroxybutyrate in vivo, Nat. Biotechnol. 15 (1997)63–67.

[3] J.T. Kealey, L. Liu, D.V. Santi, M.C. Betlach, P.J. Barr, Productionof a polyketide natural product in nonpolyketide-producing pro-karyotic and eukaryotic hosts, Proc. Natl. Acad. Sci. USA 95(1998) 505–509.

[4] B.A. Pfeifer, S.J. Admiraal, H. Gramajo, D.E. Cane, C. Khosla,Biosynthesis of complex polyketides in a metabolically engineeredstrain of E. coli, Science 291 (2001) 1790–1792.

[5] L.C. Dayem, J.R. Carney, D.V. Santi, B.A. Pfeifer, C. Khosla,J.T. Kealey, Metabolic engineering of a methylmalonyl-CoAmutase-epimerase pathway for complex polyketide biosynthesis inEscherichia coli, Biochemistry 41 (2002) 5193–5201.

[6] M.K. Ramjee, U. Genschel, C. Abell, A.G. Smith, Escherichia coliL-aspartate-�-decarboxylase: preprotein processing and observa-tion of reaction intermediates by electrospray mass spectrometry,Biochem. J. 323 (1997) 661–669.

[7] W.K. Merkel, B.P. Nichols, Characterization and sequence of theEscherichia coli panBCD gene cluster, FEMS Microbiol. Lett. 143(1996) 247–252.

[8] N. Dusch, A. Puhler, J. Kalinowski, Expression of the Corynebac-terium glutamicum panD gene encoding L-aspartate-�-decarboxyl-ase leads to pantothenate overproduction in Escherichia coli,Appl. Environ. Microbiol. 65 (1999) 1530–1539.

[9] J. Sambrook, E.F. Fritsch, T. Maniatis, Molecular Cloning: ALabatory Manual, Cold Spring Harbor Laboratory Press, ColdSpring Harbor, NY, 1989.

[10] A.J. Link, D. Phillips, G.M. Church, Methods for generating pre-cise deletions and insertions in the genome of wild-type Esche-richia coli: application to open reading frame characterization,J. Bacteriol. 179 (1997) 6228–6237.

[11] S. Murli, J. Kennedy, L.C. Dayem, J.R. Carney, J.T. Kealey, Meta-bolic engineering of Escherichia coli for improved 6-deoxyeryth-rolonide B production, J. Ind. Microbiol. Biotechnol. 30 (2003)500–509.