Báo cáo khoa học: Mammalian glutaminyl cyclases and their isoenzymes have identical enzymatic characteristics docx

Because the differences in subcellular localization and tissue distribution might reflect other functions and catalytic specificity, it was the aim of this study to express heterologously

Mammalian glutaminyl cyclases and their isoenzymes

have identical enzymatic characteristics

Anett Stephan1, Michael Wermann1, Alex von Bohlen2, Birgit Koch1, Holger Cynis1,

Hans-Ulrich Demuth1 and Stephan Schilling1

1 Probiodrug AG, Halle ⁄ Saale, Germany

2 Institute for Analytical Sciences, Dortmund, Germany

Introduction

In addition to proteolytic cleavage, glycosylation and

amidation, N-terminal formation of 5-oxoproline

(pyro-glutamate, pGlu) is a common post-translational event

during the biosynthesis of secretory peptides and

proteins, such as thyrotropin-releasing hormone

(TRH), gastrin, fibronectin and neurotensin [1–3]

Glu-taminyl cyclases (QCs) have been identified in mammals,

invertebrates and plants, catalyzing pGlu formation from glutaminyl precursors [4–8] Moreover, compel-ling evidence suggests an involvement of QCs in diseases such as osteoporosis and Alzheimer’s disease (AD) [9,10] QCs have been shown to catalyze pGlu formation at the N-terminus of amyloid peptides from glutamyl precursors, rendering them hydrophobic and

Keywords

Alzheimer’s disease; glutaminyl cyclase

isoenzyme; glycosylation; Golgi apparatus;

Pichia pastoris

Correspondence

S Schilling, Probiodrug AG, Weinbergweg

22, 06120 Halle ⁄ Saale, Germany

Fax: +49 345 5559901

Tel: +49 345 5559911

E-mail: stephan.schilling@probiodrug.de

(Received 30 June 2009, revised 17 August

2009, accepted 28 August 2009)

doi:10.1111/j.1742-4658.2009.07337.x

Glutaminyl cyclases (QCs) catalyze the formation of pyroglutamate resi-dues at the N-terminus of several peptides and proteins from plants and animals Recently, isoenzymes of mammalian QCs have been identified In order to gain further insight into the biochemical characteristics of iso-QCs, the human and murine enzymes were expressed in the secretory pathway of Pichia pastoris Replacement of the N-terminal signal anchor

by an a-factor prepropeptide from Saccharomyces cerevisiae resulted in poor secretion of the protein Insertion of an N-terminal glycosylation site and shortening of the N-terminus improved isoQC secretion 100-fold A comparison of different recombinant isoQC proteins did not reveal an influence of mutagenic changes on catalytic activity An initial character-ization showed identical modes of substrate conversion of human isoQC and murine isoQC Both proteins displayed a broad substrate specificity and preference for hydrophobic substrates, similar to the related QC Likewise, a determination of the zinc content and reactivation of the apo-isoQC revealed equimolar zinc present in QC and apo-isoQC Far-UV CD spectroscopic analysis of murine QC and isoQC indicated virtually identi-cal structural components The present investigation provides the first enzymatic characterization of mammalian isoQCs QC and isoQC repre-sent very similar proteins, which are both prerepre-sent in the secretory path-way of cells The functions of QCs and isoQC probably complement each other, suggesting a pivotal role of pyroglutamate modification for protein and peptide maturation

Abbreviations

AD, Alzheimer’s disease; BMGY, buffered glycerol complex medium; BMMY, buffered methanol complex medium; GST, glutathione transferase; IMAC, immobilized metal ion affinity chromatography; pGlu, pyroglutamate; QC, glutaminyl cyclase; TRH, thyrotropin-releasing hormone; TXRF, total X-ray fluorescence.

more prone to aggregation, probably contributing to

AD pathology [11–14] A chronic inhibition of QCs

has been shown to attenuate AD-like pathology in

mouse models, introducing QC activity as a target for

drug development [15]

Recently, we have isolated an isoenzyme of human

QC (human isoQC) [16] Database mining led to the

identification of a protein of 382 amino acids, which

shares a sequence identity with human QC of 45%

In contrast with QC, a signal anchor was identified in

isoQC, which putatively mediates the retention of the

class II transmembrane protein in the Golgi apparatus

Thus, the protein shares interesting similarity to

glyco-syltransferases with regard to subcellular localization,

and may act in concert with those in the secretory

pro-tein maturation process This hypothesis is supported

by the ubiquitous expression pattern of isoQC, which

contrasts with the differential QC expression in glands

and neuronal tissue [16] Because the differences in

subcellular localization and tissue distribution might

reflect other functions and catalytic specificity, it was

the aim of this study to express heterologously and

characterize the isoQCs of human and murine (murine

isoQC) origin in the methylotrophic yeast Pichia

pastoris The method required extensive optimization

of expression, which might have implications for other

proteins

Results

Expression of human isoQC

The isoQC proteins are localized within the Golgi

complex in their native forms The retention in the

compartment is mediated by the N-terminus, which

includes a membrane anchor directing the nascent

pep-tide chain into the secretory pathway (Fig 1) In order

to provoke efficient secretory expression in P pastoris,

the N-terminal region was substituted by the a-factor

prepropeptide from Saccharomyces cerevisiae, which

should result in the secretion of the protein into the

expression medium [17] The leader sequence thus

functions in a similar manner to the signal peptides

in native QC proteins (Fig 1) A coding sequence of

six His residues was additionally attached to the 3¢ end

of the human isoQC open reading frame, facilitating

purification The expression of this construct should

result in a secreted isoQC starting with phenylalanine

48 (isoQC(F48)C-His) (Figs 1 and 2) An isoQC

expres-sion construct was generated by applying the vector

backbone of pPICZaA, linearized and used for the

electroporation of P pastoris Unexpectedly, a

charac-terization of 50 stably transformed clones for the

presence of isoQC activity in the medium revealed only very low concentrations of the recombinant protein (Fig 3) According to this observation, several sequence modifications were considered in order to improve the solubility and, potentially, the secretion process In contrast with isoQC, QCs contain one (mouse, rat) or two (human, bovine) sites of N-glyco-sylation One N-terminal site is conserved in all mam-malian QCs and is also present in a secreted QC from Drosophila melanogaster, all of which were successfully expressed in yeast [4,5,18,19] Hence, in order to improve the secretion of human isoQC, a potential gly-cosylation site was introduced at position 73 by the exchange of isoleucine with asparagine (isoQC(F48;I73N), Fig 2), i.e resembling the glycosylation site of QC as suggested by a multiple sequence alignment (not shown) A software-based algorithm predicted a high probability of derivatization (http://www.cbs.dtu.dk/ services/NetNGlyc/) of the introduced asparaginyl residue in the modified isoQC

The I73N variant of human isoQC was expressed carrying an N-terminal or C-terminal poly-His fusion tag (isoQC(F48;I73N) N-His, isoQC(F48;I73N) C-His) The comparison of the activity in the expression medium showed an up to 10-fold higher isoQC secretion of the now glycosylated variant in comparison with the unmodified human isoQC(F48)C-His (Fig 3A) No sig-nificant difference in isoQC activity was observed between expressed proteins with the different poly-His fusions, indicating that the tag did not influence the activity or production process

In addition to the introduction of the N-terminal glycosylation site, a cysteine residue present in human isoQC was exchanged for an alanine The mutation appears to be conceivable, because the residue is nei-ther conserved in murine isoQC nor in onei-ther animal QCs As the two conserved cysteines have been shown

to form a disulfide bond in QC, the third cysteine in human isoQC is potentially free and might therefore interfere with the protein production in yeast as a result of oxidation However, further improvement

of enzyme secretion into the medium was observed fol-lowing expression of the protein (isoQC(F48;I73N;C369A) C-His, isoQC(F48;I73N;C369A)N-His) (Fig 2)

In a final approach to further improve the yield of the secreted protein, the N-terminus of the recombi-nant protein was shortened by 59 amino acids in total The shortening results in complete deletion of the transmembrane region, which spans amino acid posi-tions 35–52, 34–55 or 40–60 according to predicposi-tions

of HMMTOP, SUSOI or TMpred [20–22] (all avail-able at http://www.expasy.ch) The plasmid constructs were expressed encoding a human protein containing

the glycosylation site and the Cys⁄ Ala mutation

(isoQC(E60;I73N;C369A) N-His), or with only one of

the modifications, either the glycosylation site

(isoQC(E60;I73N) N-His) or the mutated cysteine

(iso-QC(E60;C369A) N-His) (Fig 2) The human isoQC

con-structs encoded for an additional poly-His fusion

at the N-terminus The N-terminal shortening of the

construct resulted in a 10-fold higher protein yield in

comparison with the initial construct (Fig 3A) As

observed previously, the introduction of the

glycosyla-tion site further improved the amount of isoQC in the

medium The highest protein yield was finally obtained

after expression of isoQC(E60;I73N;C369A) N-His, which

revealed a 100-fold higher isoQC concentration in the

medium compared with the initial expression

con-struct The broad distribution in activity levels in the

expression medium of some constructs (e.g

iso-QC(E60;I73N;C369A)) are caused by the transformation of yeast As a result of the integration of the recombinant DNA into the genome of the host, large clonal varia-tions occur with respect to expression, caused, for instance, by multicopy insertion into the genome (Fig 3A)

The results of the determination of the isoQC activ-ity in the medium were corroborated by western blot analysis of the expression medium, applying an isoQC-specific antiserum (Fig 3A, inset) The most intense immunostaining was observed following the expression

of isoQC(E60;I73N;C369A)N-His

From all expressed constructs, isoQC(F48;I73N;C369A) C-His, isoQC(E60;I73N;C369A) N-His and isoQC(E60;I73N) N-His were used for a scale-up of expression in shake

Fig 1 Amino acid sequence alignment of human and murine isoQCs, human QC (hQC), mouse QC (mQC) and Drosoph-ila melanogaster QC (Drome QC) The signal sequences of QC and the signal anchor of isoQCs are highlighted in bold italics The residues for complexation of zinc ions in the active site (Asp-Glu-His) (bold) and the core structure surrounding the active site, con-taining a conserved disulfide bond (bold and underlined), are conserved in all enzymes In addition, the secreted QC proteins contain sites of N-glycosylation (bold, italics, under-lined) The position of the introduced glyco-sylation site by mutation of an isoleucine residue in isoQC is shown in bold italics The alignments were prepared using CLUSTALW at EMBnet-CH (http://www expasy.ch).

flasks The protein was purified to virtual homogeneity

by initial Ni2+-immobilized metal ion affinity chroma-tography (IMAC), followed by hydrophobic

interac-Cytosolic sequence Membrane anchor Luminal catalytic domain Yeast secretion signal

isoQC (E60;I73N)

COOH

NH2

Native isoQC

isoQC (F48)

isoQC (E60; I73N;C369A)

↓

isoQC (F48;I73N;C369A)

Cys →Ala

↓

isoQC (F48;C369A)

COOH

isoQC E61 I74N

NH

2

NH2

COOH

isoQC (F49)

Human isoQC

Murine isoQC

Fig 2 Schematic representation of the

different constructs for the secretory

expression of human and murine isoQCs in

Pichia pastoris The native isoQC with the

cytosolic tail and the membrane anchor is

shown for comparison This N-terminal

region is exchanged for the a-leader

prepro-sequence of Saccharomyces cerevisiae as a

secretion signal In some constructs, an

isoleucine is mutated into an asparagine,

generating an N-glycosylation site

Further-more, a third cysteine at the C-terminus of

human isoQC is mutated into an alanine.

Finally, the constructs differed in terms of

their N-terminus, i.e the N-terminal amino

acid corresponding to the open reading

frame of isoQC was either a phenylalanine

(Phe48 and Phe49 of human and murine

QC, respectively) or a glutamic acid residue

(position 60 or 61).

C -H is

(F 48

)

is oQ

C

C -H is

(F 48

36 9A )

is oQ

C

N -H is

(E 60

36 9A )

is oQ

C

N -H is

(F 48

;I7 )

is oQ C

C -H is

(F 48

;I7 )

is oQ C

N -H is

(E

;I7 )

is oQ C

N -H is

(F 48

;I7

;C 36 9A )

is oQ C

C -H is

(F 48

;I7

;C 9A )

is oQ C

N -H is

(E

;I7

;C 9A )

is oQ C

0

1

2

3

4

5

6

0.001 0.01 0.1 1 10

1 2 3 4 5 6 7 8 9 10

36 kDa

A

250

150

100

75

50

37

25

20

15

5

Fig 3 Characterization of human isoQC expression in Pichia pasto-ris (A) Determination of the QC activity in the medium of P pastoris expressing the different constructs At least 50 clones of each con-struct were checked with regard to QC activity in the culture medium after small-scale expression The inset shows western blot analysis

of the expression medium and a logarithmic scatter plot of the acti-vity determined for each yeast clone investigated The logarithmic plot of the QC activity data points to a similar variation of expression after transformation with the different plasmid constructs, which is caused by differences in transcriptional efficacy and insertion events

of the expression constructs into the genome of yeast In western blot analysis, the proteins were visualized using an isoQC antibody.

A total amount of 4 lg of protein was applied to each lane Lane 1, isoQC (F48) C-His; lane 2, isoQC (F48;C369A) C-His; lane 3,

iso-QC(E60;C369A)N-His; lane 4, deglycosylated isoQC(F48;I73N)N-His; lane

5, deglycosylated isoQC(F48;I73N)C-His; lane 6, deglycosylated

iso-QC (E60;I73N) N-His; lane 7, deglycosylated isoQC (F48;I73N;C369A) N-His; lane 8, deglycosylated isoQC(F48;I73N;C369A) C-His; lane 9, deglyco-sylated isoQC(E60;I73N;C369A)N-His; lane 10, 400 ng of purified and deglycosylated isoQC (F48;I73N;C369A) as a positive control The degly-cosylated protein migrates at 37 kDa (B) SDS-PAGE analysis illus-trating the purification of human isoQC(F48;I73N;C369A) C-His after expression in shake flasks Proteins were visualized by Coomassie staining: lane 1, molecular mass standards (kDa) (Dual Color, Bio-Rad); lane 2, supernatant after expression; lane 3, isoQC-containing fractions after initial affinity chromatography on immobilized Ni 2+

ions; lane 4, human isoQC after hydrophobic interaction chromato-graphy; lane 5, purified protein after desalting The isoQC protein corresponds to a band migrating between 50 and 70 kDa The degly-cosylation causes a shift to 37 kDa (lane 6) The upper band (75 kDa)

in lane 6 corresponds to the endoglycosidase Hf.

tion chromatography A purification process illustrated

by SDS-PAGE analysis is shown in Fig 3B As a

result of the competitive inhibition of isoQC and QC

by imidazole, the elution was performed with His,

which is much less inhibitory to QC and isoQC

Inacti-vation of isoQC, e.g caused by zinc complexation

dur-ing IMAC, was not observed

Typical expression in a final culture volume of 4 L

resulted in the isolation of 7 mg of isoQC(E60;I73N)

N-His and 14 mg of isoQC(E60;I73N;C369A) N-His

variant Expression of isoQC(F48;I73N;C369A) C-His was

carried out in a total volume of 8 L, resulting in the

isolation of 7.3 mg of human isoQC protein The

over-all yield of purification was 60%

Expression and purification of murine isoQC

On the basis of human isoQC expression in P pastoris,

three different murine isoQC constructs were

gener-ated The introduction of a glycosylation site and the

shortening of the N-terminus (isoQC(E61;I74N) N-His)

resulted in an increase in secretory expressed murine

isoQC, similar to that observed with human isoQC In

addition to the poly-His-tagged proteins, an untagged

protein was generated A comparison of the

N-termi-nally His-tagged and untagged protein did not reveal a

significant influence on isoQC expression (Fig 4A)

The isoQC(E61;I74N) variant was successfully expressed

by fermentation using a 5 L bioreactor, which typically

results in the harvesting of 2 L of isoQC-containing

medium Homogeneous protein was obtained after

purification, applying two different hydrophobic

interaction chromatographic matrices, followed by

anion-exchange chromatography and size exclusion

chromatography The purity of murine isoQC was

analyzed by SDS-PAGE (Fig 4B) A typical

purifica-tion process resulted in the isolapurifica-tion of 8 mg of murine

isoQC protein

Characterization of the substrate and inhibitor

specificity

In order to rule out an influence of the different

modi-fications of human isoQC on substrate conversion, the

kinetic parameters for the turnover of H-Gln-bNA

(Q-bNA), H-Gln-Glu-OH (QE) and H-Gln-Gln-OH

(QQ) and the competitive isoQC inhibitors

benzimid-azole and benzylimidbenzimid-azole were analyzed The

evalua-tion of the influence of sequence shortening to

glutamic acid 60, glycosylation of the protein and

mutation of cysteine 369 into alanine did not show a

considerable effect on the kinetic parameters or on the

inhibitory specificity (cf Tables 1 and 2) Accordingly,

the substrate and inhibitor specificities of the human isoQC(F48;I73N;C369A) C-His variant were analyzed and compared with the data obtained with a glutathione transferase (GST)-fusion protein, which was expressed

in Escherichia coli previously [16] Interestingly, the substrate specificity of the isoQCs expressed in P pas-toris and E coli was virtually identical for the tested substrates, i.e the highest specificity was obtained for substrates containing hydrophobic residues, e.g bNA, Tyr-Ala-OH (QYA) or H-Gln-Glu-Tyr-Phe-NH2(QEYF) (Table 2) A comparison of

N- His ( F49 ) C

N - His ( E 61;

I7 4N )

i soQ isoQ C

(E 61

4 N)

i s o Q

C

0 0

0 5

1 0

1 5

2

A

B

0

kD a

150

100

75

50

37

25

20

0.001 0.01 0.1

1

10

Fig 4 Expression and characterization of murine isoQC in Pichia pastoris (A) Determination of QC activity in the medium of P pas-toris expressing the indicated constructs The analysis was per-formed as described for human isoQC in Fig 3 (B) SDS-PAGE illustrating the purification of murine isoQC(E61;I74N)after fermenta-tion Proteins were visualized by Coomassie staining Lane 1, molecular mass standards (kDa) (Dual Color, Bio-Rad); lane 2, supernatant after expression; lane 3, isoQC-containing fractions after initial hydrophobic interaction chromatography applying expanded bed absorption; lane 4, isoQC after hydrophobic inter-action chromatography; lane 5, isoQC after anion exchange chroma-tography; lane 6, isoQC after desalting and deglycosylation of the protein isoQC corresponds to a protein between 50 and 70 kDa The deglycosylated protein migrates at 37 kDa and the endoglycosi-dase Hfis at 75 kDa.

the kcat⁄ Km values of both human isoQC variants, however, revealed a prominent difference The specific-ity constant obtained with isoQC expressed in P pas-toris was two- to three-fold higher for almost every substrate tested, suggesting a potential influence of the expression system or of the isoQC fusion construct As

a result of the very similar Michaelis constants for the isoQCs expressed in E coli and P pastoris, the influ-ence on substrate specificity was mainly caused by kcat Similar to Km, no significant difference was observed

in the Kivalues between isoQCs expressed in yeast and

E coli (Table 2) Presumably, the expression and puri-fication of isoQC in P pastoris resulted in efficient recovery of the active protein

The relative substrate specificity of murine isoQC was similar to that of the human enzyme (Table 3) Most substrates, however, were more efficiently verted into products by the murine protein, which con-trasts with the related QCs [5] The higher proficiency was caused by the lower Michaelis constants and higher turnover numbers

In order to further characterize the activity of mur-ine isoQC in comparison with QC, the pH dependence

of catalysis was assessed under first-order conditions, i.e at [S] << Km Both enzymes displayed a pH opti-mum of kcat⁄ Km of between 7.5 and 8 (Fig 5) The kinetic data of the pH dependence were evaluated by applying a model, which considers three dissociating groups, one of the substrate and two of the enzyme The pKa value of the applied substrate H-Gln-AMC (6.83 ± 0.01) has been determined previously and matches well with the acidic inflection points of the

pH dependences of isoQC and QC The nonsymmetric character of the bell-shaped curve was calculated assuming two dissociating groups of the enzyme Eval-uation of these data resulted in pKa values of 9.5 ± 0.3 and 8.2 ± 0.4 for murine isoQC and 9.0 ± 0.2 and 8.3 ± 0.3 for murine QC Apparently, all dissociating groups, which influence the catalytic process in QC and isoQC, are conserved, supporting

an identical catalytic mechanism

Characterization of metal dependence and structural elements

The animal QCs and isoQCs share a structural relationship to bacterial double-zinc aminopeptidases [23–25] Although the coordinating residues of the aminopeptidases are also conserved in QCs, it has been shown that only one zinc ion is bound to QC [5,26] Therefore, murine isoQC without an affinity tag was analyzed for the presence of transition metal ions using total X-ray fluorescence (TXRF) spectroscopy

Km

kcat

1 )

Ki

Km

kcat

1 )

Ki

Km

k cat

1 )

KI

Substrates

Inhibitors Benzimidazole

A typical spectrum is displayed in Fig 6A Three

inde-pendently prepared enzyme samples showed a

signifi-cantly increased zinc concentration The averaged zinc

content in murine isoQC was 0.99 ± 0.38 moles of

zinc per mole of enzyme, clearly supporting

stoichiom-etric zinc binding of murine isoQCs Accordingly, full

reactivation of the murine isoQC apo-enzyme was

obtained in the presence of equimolar zinc

concentra-tions (Fig 6B) In addition to zinc, reactivation was also obtained with cobalt ions, also suggesting an equi-molar stoichiometry of metal to protein necessary for activity With regard to zinc content and reactivation, therefore, the isoQCs are very similar to QCs A cata-lytic role of the transition metal ion is also suggested

by a comparison of the far-UV CD spectrum of murine isoQC and its apo-enzyme (Fig 6C) The CD

Table 3 Kinetic evaluation of peptide substrates by mouse isoQC Protein was expressed in Pichia pastoris a ND, not determined.

Compound

Km(m M ) kcat(s)1) Ki(m M ) kcat⁄ K m (m M )1Æs)1) k

cat ⁄ K m (m M )1Æs)1) K

i (m M ) Substrates

Inhibitors

a Reactions were carried out in 0.05 M Tris ⁄ HCl, pH 8.0, at 30 C b Substrate inhibition c Data from [5].

Table 2 Kinetic parameters of substrate conversion by recombinant human isoQC obtained from different host systems a

Compound

Km(m M ) kcat(s)1) Ki(m M )

kcat⁄ K m

(m M )1Æs)1)

kcat⁄ K m

(m M )1Æs)1) K

i (m M ) Substrates

Inhibitors

a Reactions were carried out in 0.05 M Tris ⁄ HCl, pH 8.0, at 30 C b Substrate inhibition c Data from [16].

spectra were virtually identical between isoQC,

apo-isoQC and QC Thus, the loss of the metal ion does

not result in large structural rearrangements The

virtually identical spectra further support the strong

similarity between QC and isoQC globular domains

Finally, we characterized whether the conserved

cysteines in human isoQC form a disulfide bond, as

described for human QC [4] The characterization of

the disulfide status of the protein by SDS-PAGE

clearly suggests disulfide bond formation (Fig 7)

Therefore, the structurally conserved features of isoQC

and QC proteins not only contain metal complexation

and general fold, but also the formation of a disulfide

bond close to the active center of the protein

In addition to the two conserved cysteines, human

isoQC contains a third cysteine residue in the

C-ter-minal part of the protein (Fig 1) The presence of a

third cysteine might imply the formation of dimers in

the secretory pathway because, in an oxidative

envi-ronment, cysteine does not usually appear unbound

In the case of isoQC, it should be noted that the

cysteine residue is not conserved in murine (Fig 1),

bovine (UniProt: Q0V8G3) and macaque (UniProt:

Q4R942) isoQCs Initial expression of murine and

human isoQCs in human HEK293 cells and an

accompanying western blot analysis involving a

reducing or nonreducing sample preparation did not

reveal the formation of homo- or heterodimers (not

shown) The data thus imply that dimerization

involv-ing a covalent interaction does not occur in the Golgi

complex

Discussion

In a first approach to characterize the recently discov-ered isoQCs from mouse and humans, we aimed to

C

Wavelength in nm

–15 000 –10 000 –5000 0 5000

10 000

15 000

20 000

25 000

m-isoQC apoenzyme m-isoQC reactivated mQC

A

Counts 2000 4000

Photon energy, keV

0

m-isoQC Cl

Zn

Zn Br

B

Ratio metal to enzyme

0 2000 4000

Cobalt Manganese Nickel Calcium

Fig 6 Spectroscopic analysis of murine isoQC (A) TXRF spectrum

of a murine isoQC preparation IsoQC was dissolved in 10 m M Tris ⁄ HCl, pH 7.6 The evaluation of the measurements revealed equimolar amounts of zinc bound to the enzyme (B) Reactivation

of murine apo-isoQC with different divalent metal ions The enzyme was inactivated by 1,10-phenanthroline and subjected to dialysis against Chelex-treated buffer Reactivation was carried out by the addition of different concentrations of transition metal ions to the inactivated protein The ratios of the concentrations of transition metal and apo-enzyme are indicated on the x-axis (C) CD spectro-scopic analysis of murine isoQC, murine QC and murine apo-isoQC The protein was dissolved in 10 m M potassium phosphate buffer,

pH 6.8 There was virtually no difference in the spectra between the apo- and holo-enzymes and murine QC, supporting a low influence of the active site zinc on the structure of isoQC.

pH

kcat

/Km

0

20

40

60

80

100

120

140

160

180

200

220

240

260

m-isoQC

mQC

Fig 5 pH dependence of murine isoQC catalysis Determination

of the specificity constant k cat ⁄ K m for the conversion of H-Gln-AMC

by purified murine isoQC (h) and QC (O), determined under

first-order rate law conditions ([S] << Km) The substrate concentration

was 0.005 m M and the reactions were carried out at 30 C in a

buf-fer consisting of 0.2 M Tris ⁄ HCl, 0.1 M Mes and 0.1 M acetic acid.

express and characterize the proteins in P pastoris in

order to compare the catalytic properties with those of

the sister enzyme QC In contrast with E coli, P

pas-toris has been shown to exert many post-translational

modifications, such as N-glycosylation and disulfide

formation [27–31], some of which have been shown to

be present in animal and plant QCs [4,5,19,32] and

may also be important for the expression of isoQCs

On the basis of these results, we aimed to obtain the

secretory expression of murine and human isoQCs in

P pastoris In the native state, isoQC is expressed as a

class II transmembrane protein, which is anchored by

an N-terminal signal peptide in the membrane of the

endoplasmic reticulum and retained in the Golgi

appa-ratus In order to obtain secreted protein, the short

cytosolic tail, including the major part of the

mem-brane anchor, was deleted and substituted by an a-leader

prepro sequence of yeast, which should direct the

protein efficiently into the secretory pathway

Although the remaining protein shares 51% sequence

identity with the mature QCs, isoQC expression

resulted in media virtually devoid of QC activity In

contrast with the QC proteins, human and murine

iso-QCs lack a conserved N-terminal site for

N-glycosyla-tion (Fig 1) It is known that glycosylaN-glycosyla-tion may lead

to an increase in protein solubility [33] and decreased

aggregation propensity [34] Furthermore,

deglycosyla-tion of human QC resulted in protein precipitadeglycosyla-tion

(results not shown) Therefore, we introduced an

N-glycosylation site into isoQC for heterologous

expression Glycosylation resulted in a significant

increase in isoQC activity in the medium Although

the reason for the improvement was not investigated

in detail, an increase in protein solubility and, in turn,

a decrease in hydrophobic interactions between the

protein might be a primary cause Consistent with this

hypothesis, an enzymatic deglycosylation of the isoQC

protein resulted in a dramatic decrease in solubility to

about 1 mgÆmL)1, whereas the glycosylated protein did not precipitate up to 30 mgÆmL)1 (not shown) In addition to glycosylation, the N-terminal truncation to glutamic acid 60/61 of the proteins further improved the expression of human and murine isoQCs Based on sequence comparisons, the finally deleted region corre-sponds mainly to unstructured parts of the protein, which might also exert an influence on efficient protein production The sequential optimization of the protein construct used for expression ended in a 100-fold improvement of the protein yield, which was secreted into the medium of the cells and could be purified by two- to three-step protocols The secretory expression

in yeast facilitates an efficient purification process, because yeast – in spite of a fully developed secretory machinery – secretes only a few proteins into the extra-cellular space [35] Because the heterologous protein reaches a high specific activity in the expression med-ium, secretion can be regarded as a separation step, allowing efficient recovery of the protein of interest, even without an affinity tag, as shown here for murine isoQC

Moreover, the expression of proteins in the secretory pathway facilitates the appropriate formation of post-translational structural elements, such as disulfide bonds As shown in this study, the two cysteine resi-dues, which are conserved in murine and human isoQCs, form a disulfide bond, which is reminiscent of the disulfide bond formation in human QC Appar-ently, the disulfide bridge is an evolutionary conserved structural element of QC and isoQC (Fig 1) Accord-ing to the crystal structure of human QC [25], the cysteine residues are close to the active site, probably exerting a stabilizing effect on the flexibility in that region

The high degree of sequence similarity between the isoQCs and their sister enzyme QC was finally mir-rored by the characterization of the secondary struc-ture, metal dependence and catalytic activity A virtually identical catalytic activity was demonstrated

by the comparison of murine isoQC and murine QC, which were both expressed in P pastoris The data clearly suggest that the active site of both enzymes has

a very similar structure, forming identical secondary interactions with the substrate to facilitate binding and turnover Apparently, the core features of both pro-teins, i.e the active site and the general fold, are iden-tical, a hypothesis which is also supported by an alignment of the proteins, which shows a high degree

of conservation of the inner core structures between

QC and isoQC and a weaker similarity in the connect-ing loops (Fig 1) The far-UV CD spectra of QC and isoQC are shown to be identical, confirming the

kDa 1 2 3 4 5 6 7 8 9 10

50

37

Fig 7 Analysis of disulfide formation in human isoQC expressed

in Pichia pastoris Lanes 1 and 10, molecular mass standards (kDa);

lanes 2, 3, 8 and 9, sample prepared under reducing conditions

(5% b-mercaptoethanol); lanes 4–7, samples prepared under

non-reducing conditions Electrophoresis (15% gels) was performed at

a constant voltage of 200 V for 1.5 h, and protein was visualized by

Coomassie staining.

above-mentioned assumptions According to the

pro-posed high degree of similarity, isoQCs and QCs

con-tain one zinc ion in the active site, which is responsible

for the catalytic activity of the enzyme As shown here

by titration experiments, zinc can be replaced by

cobalt, which results in a less active enzyme Similar

results have been reported previously for human QC

[24] The results of the titration experiments clearly

suggest that the binding of one transition metal ion is

necessary and sufficient for the exertion of full

enzy-matic activity, and that the metal ion does not exert a

direct effect on the structure of the protein

Thus, with regard to the recombinant expression

strategy introduced in the present study, it appears

that the various mutagenic changes to the isoQCs did

not exert a relevant influence on the catalytic activity

of the soluble, heterologous proteins However, a

potential influence of the deletion of the N-terminal

signal anchor in isoQC cannot be fully excluded In its

native state, the protein is a membrane-bound enzyme

of the Golgi complex, and membrane anchoring might

potentially affect substrate turnover, perhaps by

prox-imity to the membrane or other interacting proteins

The present results thus mirror well the catalytic

potential of the globular domain, especially in relation

to the sister enzyme QC, but cannot be translated into

the in vivo situation without caution

The characterization of QC and isoQC revealed that

evolution apparently resulted in globular proteins with

a very similar catalytic power, virtually identical

sub-strate specificity and similar subcellular localization

within the secretory pathway Most likely, the proteins

had a common ancestor It still remains unclear,

however, whether the proteins are responsible for the

conversion of the same substrates, i.e pGlu-modified

proteins and⁄ or peptide hormones and pGlu-modified

amyloid peptides in neurodegenerative disorders

The localization of QC and isoQC in the secretory

pathway enables the conversion of secretory proteins

by both enzymes – QC is probably transported within

the regulated pathway [8] and isoQC exerts its function

as a resident enzyme of the Golgi apparatus [16]

Although the localization of both proteins appears to

be virtually identical at first glance, the presence of QC

in vesicles of the regulated secretory pathway might

indicate the responsibility of QC for the conversion

of substrates requiring extensive post-translational

processing, e.g the neuropeptide TRH [36–38] The

liberation of the substrate in secretory vesicles requires

the presence of QC in the same compartment, because

N-terminal pGlu formation represents a finishing

reaction in the post-translational maturation of these

hormones In contrast, many proteins do not require

such processing of the precursor; the N-terminal glutaminyl residue is directly generated by signal pepti-dase cleavage in the endoplasmic reticulum, e.g in ribonuclease or a-amylase These proteins are also likely to be secreted via the constitutive pathway from the Golgi complex Therefore, the primary converting enzyme of these proteins might be isoQC Taken together, it appears that the liberation of N-terminal glutamine in the secretory pathway results inevitably in N-terminal pGlu formation The broad and similar substrate specificity of QC and isoQC might therefore

be important for the conversion of these different substrates Finally, the similar cellular distribution of two proteins with virtually identical specificity, as shown here, might also point to an overall important role of pGlu protein formation for physiology The elucidation of the physiological function might have implications for drug development, as a partial complementation of QC and isoQC might compensate for the side-effects of potential, isoform-specific drug candidates Indeed, it is likely that the protein func-tions of QC and isoQC complement each other, as QC knockout mice do not exhibit an apparent phenotype (S Schilling et al., unpublished results)

In summary, the first detailed heterologous expres-sion and characterization study of mammalian isoQCs was accomplished Expression was mainly optimized

by the insertion of an artificial glycosylation site into the isoQC protein, resulting in efficient protein secre-tion by the yeast P pastoris, resembling the subcellular localization in the native tissue of origin These results might have implications for the expression of other mammalian proteins, which display a high tendency to aggregation and are, therefore, difficult to express The isolation of the isoQC protein represents a basis for structural investigations and drug candidate profiling

Materials and methods Materials

The E coli strain DH5a was applied for all plasmid con-struction and propagation; P pastoris strain X-33 (AOX1, AOX2) was used for the expression of the different isoQC variants Yeast was grown, transformed and analyzed according to the manufacturer’s instructions (Invitrogen,

obtained from Bachem (Bubendorf, Switzerland) or synthe-sized as described elsewhere [39] Recombinant pyroglutamyl

purchased from Qiagen (Hilden, Germany) and glutamic dehydrogenase from Fluka (Seelze, Germany) The