Biochemical Journal

Error message

  • Notice: PHP Error: Undefined index: custom_texts in highwire_highwire_corrections_content_type_render() (line 33 of /opt/sites/jnl-ppbiochemj/drupal-highwire/releases/20151123210458/modules/highwire/plugins/content_types/
  • Notice: PHP Error: Undefined offset: 1 in nice_menus_block_view() (line 306 of /opt/sites/jnl-ppbiochemj/drupal-webroot/releases/20151123210521/sites/all/modules/contrib/nice_menus/nice_menus.module).
  • Notice: PHP Error: Undefined offset: 1 in nice_menus_block_view() (line 306 of /opt/sites/jnl-ppbiochemj/drupal-webroot/releases/20151123210521/sites/all/modules/contrib/nice_menus/nice_menus.module).

Research article

A conserved hydrogen-bond network in the catalytic centre of animal glutaminyl cyclases is critical for catalysis

Kai-Fa Huang, Yu-Ruei Wang, En-Cheng Chang, Tsung-Lin Chou, Andrew H.-J. Wang


QCs (glutaminyl cyclases; glutaminyl-peptide cyclotransferases, EC catalyse N-terminal pyroglutamate formation in numerous bioactive peptides and proteins. The enzymes were reported to be involved in several pathological conditions such as amyloidotic disease, osteoporosis, rheumatoid arthritis and melanoma. The crystal structure of human QC revealed an unusual H-bond (hydrogen-bond) network in the active site, formed by several highly conserved residues (Ser160, Glu201, Asp248, Asp305 and His319), within which Glu201 and Asp248 were found to bind to substrate. In the present study we combined steady-state enzyme kinetic and X-ray structural analyses of 11 single-mutation human QCs to investigate the roles of the H-bond network in catalysis. Our results showed that disrupting one or both of the central H-bonds, i.e., Glu201···Asp305 and Asp248···Asp305, reduced the steady-state catalysis dramatically. The roles of these two COOH···COOH bonds on catalysis could be partly replaced by COOH···water bonds, but not by COOH···CONH2 bonds, reminiscent of the low-barrier Asp···Asp H-bond in the active site of pepsin-like aspartic peptidases. Mutations on Asp305, a residue located at the centre of the H-bond network, raised the Km value of the enzyme by 4.4–19-fold, but decreased the kcat value by 79–2842-fold, indicating that Asp305 primarily plays a catalytic role. In addition, results from mutational studies on Ser160 and His319 suggest that these two residues might help to stabilize the conformations of Asp248 and Asp305 respectively. These data allow us to propose an essential proton transfer between Glu201, Asp305 and Asp248 during the catalysis by animal QCs.

  • glutaminyl cyclase (glutaminyl-peptide cyclotransferase
  • QC)
  • pyroglutamate (pGlu)
  • hydrogen-bond network
  • proton transfer
  • site-directed mutagenesis
  • synchrotron
  • X-ray crystallography


QCs (glutaminyl cyclases; glutaminyl-peptide cyclotransferases, EC catalyse the conversion of protein N-terminal glutamine (or glutamic acid) residue into pGlu (pyroglutamate), a reaction that is important during the maturation of many bioactive peptides, hormones and proteins in their secretory pathway [1,2]. The enzymes have been isolated from several animal and plant sources [3,4] and the genes coding for QC have been identified in numerous organisms [5,6]. In human, QCs were thought to be related to the progression of some amyloidotic disorders, since QC could catalyse N-terminal pGlu formation in several amyloid-β peptides in vitro [7,8], a process that might enhance the hydrophobicity, proteolytic stability and neurotoxicity of these peptides [9,10]. Moreover, a number of studies have suggested that genetic variations, altered expressions and DNA methylation of QPCT, the human gene coding for QC [Q (glutaminyl)-peptide cyclotranferase], might correlate with some pathological processes, such as osteoporosis, rheumatoid arthritis and melanoma [1113].

We have published the crystal structure of human QC [8]. The active site of the enzyme (Figure 1A) has a catalytically essential zinc ion, lying at the bottom of the active-site pocket and tetrahedrally co-ordinating to Asp159, Glu202, His330 and a water molecule. Contiguous to the zinc centre, there are three acidic residues, i.e. Glu201, Asp248 and Asp305, with orientations that point to each other, likely forming the unusual COOH···COOH H-bonds (hydrogen bonds) among them (Figure 1B). Mutations on these three residues decreased the enzyme activity dramatically [8]. The structure of human QC in complex with the substrate glutamine t-butyl ester showed that Glu201 and Asp248 are H-bonded to the α-amino nitrogen atom and the γ-amide amino group of the substrate glutamine residue respectively [8]. These findings suggest that Glu201, Asp248 and Asp305 are important for substrate binding and catalysis by human QC. Since human QC shares a conserved active-site structure and the zinc-binding residues with those of exopeptidases, a catalysis mechanism that employed Glu201 as the general base and acid was proposed on the basis of the mechanism known for exopeptidases [14]. However, other possible pathways, such as the participation of Asp248 and Asp305, in the catalysis process, could not be completely ruled out thus far.

Figure 1 Crystal structure of human QC and comparisons with other animal QCs and the double-zinc exopeptidases

(A) The overall structure of human QC [8]. (B) A close-up view of the catalytic region. The residues that participate in the unusual H-bond network or co-ordinate to the zinc ion are present and labelled, with oxygen, nitrogen, and carbon atoms being coloured red, blue, and grey respectively. The magenta balls depict the water molecules. The H-bonds and co-ordination bonds are represented with dotted lines coloured green and yellow, respectively. (C) Sequence alignment of human QC with several representative animal QCs. The amino acid sequences surrounding the H-bond network forming residues (shaded) are present, with numbering according to human QC. The red balls below the sequences mark the completely conserved residues in these QCs. GenBank® accession numbers of these sequences are as follows: NP_036545 [ man (Homo sapiens)], XP_001167364 [chimpanzee (Pan troglodytes)], NP_803472 [ox (Bos taurus)], XP_532934 [domestic dog (Canis familiaris)], XP_233812 [brown rat (Rattus norvegicus)], XP_128770 [house mouse (Mus musculus)], XP_419527 [chicken (Gallus gallus)], BAA34290 [Brazilian pit viper (Bothrops jararaca)], AAH80397 [South African clawed toad (Xenopus laevis)], CAK05164 [angelfish (Danio rerio)], XP_395412 [honey-bee (Apis mellifera)], AAF51611.1 [fruitfly (D. melanogaster)], XP_318733 [malarial mosquito (Anopheles gambiae)], T23125 (the nematode worm Caenorhabditis elegans) and P43599 [baker's yeast (Saccharomyces cerevisiae)]. (D) Superimposition of human QC structure with the structures of Aeromonas proteolytica aminopeptidase (magenta) (PDB code 1AMP), Streptomyces griseus aminopeptidase (green) (1XJO), and carboxypeptidase G2 (cyan) (1CG2). The residues corresponding to the H-bond network forming residues of human QC are present and labelled, with dotted lines depicting the possible H-bonds.

In addition to the Glu201···Asp305 and Asp248···Asp305 H-bonds, Asp248 is also H-bonded to Ser160, and Asp305 is H-bonded to His319 and a water molecule, together creating an unusual H-bond network in the active site of human QC (Figure 1B). By sequence alignment (Figure 1C), those five residues (Ser160, Glu201, Asp248, Asp305 and His319) that participate in the H-bond network are fully conserved throughout the animal, and even yeast, QCs. By contrast, Ser160, Asp305 and His319 are not found in the active sites of exopeptidases (Figure 1D), i.e., a similar H-bond network does not exist in exopeptidases. Therefore, the H-bond network found in the active site of human QC represents a unique structural feature of animal QCs.

In the present study the residues that participate in the H-bond network have been mutated systematically to different residues. The steady-state enzyme kinetic parameters of the mutant human QCs were determined by using a fluorescence and pyroglutamyl peptidase-coupled assay method [15], in contrast with the spectrophotometric method we used previously [8]. Moreover, the high-resolution (1.66–2.18 Å) (1 Å=0.1 nm) X-ray structures of 10 out of the 11 human QC mutants were solved and refined carefully. The results provide new and deeper insights into the structure–function relationships of human QC and thus improve quite significantly our understanding of the catalytic mechanism of animal QCs.


Site-directed mutagenesis

A thioredoxin-fusion expression vector of human QC, which was derived from a commercial vector (pET32a; Novagen) and contained the human QC cDNA (encoding residues 33–361) with an engineered Factor Xa cleavage site and an additional hexahistidine tag at the N-terminus, was available from previous studies [16]. Using the expression vector as a template, 11 human QC mutants were constructed on the basis of the protocol of the QuikChange® Site-Directed Mutagenesis Kit (Stratagene).

Expression and purification of human QCs

The wild-type and mutant human QCs were expressed in Escherichia coli cells according to a previous protocol [16], but with several modifications. Briefly, the expression constructs were transformed into E. coli BL21(DE3) cells (Novagen). The bacteria were grown in LB (Luria–Bertani) medium containing 200 μg/ml ampicillin at 37 °C until a cell-density attenuance (D600) of 0.8–1.2 was reached. The cultures were induced with 1 mM IPTG (isopropyl β-D-thiogalactopyranoside) for about 30 h at 20 °C. The cells were harvested by centrifugation (8983 g for 30 min at 4 °C), followed by freezing at −80 °C. Frozen bacterial pellets were resuspended in 50 mM Tris/HCl, pH 8.0, containing 150 mM NaCl and 20 mM imidazole, and then lysed using a French press. The lysates were clarified by centrifugation (104630 g for 60 min at 4 °C) and then loaded on to an Ni-NTA (Ni2+-nitrilotriacetate; Amersham Pharmacia) column. The thioredoxin-fusion QCs were eluted from the column with a linear gradient of 0–250 mM imidazole, pooled, and then digested with Factor Xa (Novagen) in a dialysis bag against the same buffer without imidazole at 25 °C for about 40 h. The cleaved QC products were obtained by using a second Ni-NTA column and elution with a linear gradient of 0–100 mM imidazole. The proteins were shown to be homogeneous by SDS/PAGE and Coomassie Blue staining.

Expression and purification of human PAP I (pyroglutamyl aminopeptidase I)

We have amplified the cDNA coding for human PAP I from a commercial bone-marrow cDNA library (Clontech) by PCR using the forward and backward primers (5′-ATACATATGGAGCAGCCGAGGAAGGCGGTG-3′ and 5′-GTGCTCGAGGTGTTTGTGGCAATAGTTGAT-3′ respectively) as described previously [17]. After confirmation by DNA sequencing, the cDNA fragment was inserted into the pET23a expression vector (Novagen) via NdeI and XhoI cloning sites. Subsequently, the vector containing PAP I was transformed into E. coli Rosetta-gami(DE3) cells (Novagen), and then the bacteria were grown in LB media with 100 μg/ml ampicillin at 37 °C until a D600 of ≈0.6 was reached. The cultures were induced with 1 mM IPTG overnight at 20 °C. The cells were harvested and then frozen at −80 °C. Frozen bacterial pellets were resuspended in 20 mM Tris/HCl, pH 7.9, containing 50 mM NaCl and 5 mM imidazole, and then lysed using a French press. After clarifying by centrifugation (104 630g for 60 min at 4 °C), the lysates were loaded on to an Ni-NTA column. The bound human PAP I was eluted from the column with a linear gradient of 0–75 mM imidazole. The PAP I fractions were pooled, concentrated, and loaded on to a gel-filtration column packed with Sephacryl S100 (Amersham Pharmacia) and equilibrated with 20 mM Tris/HCl, pH 7.9, containing 50 mM NaCl, 2 mM dithiothreitol, 2 mM EDTA and 10% (v/v) glycerol for further purification. The final purified PAP I had a purity of ≈95% as judged by SDS/PAGE and showed strong hydrolysing activity against the substrate pGlu-βNA (pyroglutamyl 2-naphthylamide) at pH 8.0 at 25 °C.

Enzymatic activity

The activities of wild-type and mutant human QCs were evaluated at 25 °C using two fluorescent substrates, namely Gln-βNA (L-glutaminyl 2-naphthylamide) and Gln-AMC (L-glutaminyl 4-methylcoumarinylamide). The 100 μl reaction mixtures contained 0.21–0.25 mM fluorogenic substrate (see Table 1 below), ≈0.2 unit of human PAP I (one unit is defined as the amount hydrolysing 1 μmol of pGlu-βNA/min under the described conditions), and an appropriately diluted aliquot of wild-type or mutant QC in 50 mM Tris/HCl buffer, pH 8.0. The excitation and emission wavelengths were set at 320 and 410 nm, and 380 and 465 nm, for the substrates Gln-βNA and Gln-AMC respectively. The reactions were initiated by the addition of QC. Enzymatic activity was determined by the amount of released βNA (or AMC), calculated using a standard curve for βNA (or AMC) under the same assay conditions. The measurements were made using a FluoroLog®-3 spectrofluorimeter (Horiba Jobin Yvon, Edison, NJ, U.S.A.).

View this table:
Table 1 Activities and relative activities of the wild-type and mutant human QCs

Results are means±S.E.M. from three assays. Abbreviation: ND, not detectable.

Enzyme kinetic assay

The kinetic constants were determined at pH 8.0 at 25 °C using the substrate Gln-βNA. Three sets (0.1, 1.0 and 5.0 μM) of enzyme concentrations were used on the basis of the known relative activities of these mutants. The reaction was initiated by adding QC to the 100 μl reaction mixture. The initial rate was measured with less than 10% substrate depletion for the first 2–12 min. Since substrate inhibition was observed, the kinetic parameters Km, V and Ki were evaluated by fitting eqn (1) to initial velocity data by non-linear regression using KaleidaGraph software: Embedded Image (1) where v0 is the initial velocity, Vmax the limiting rate, [S] the substrate concentration, Km the Michaelis constant and Ki the inhibition constant. Correlation coefficients better than 0.995 were obtained throughout the fittings. Substrate concentrations used were in the range 0.002–2.8 mM. The kcat values were calculated from Vmax/[E] according to the enzyme concentrations determined from UV absorbance at 280 nm (ϵ=55 190 M−1·cm−1) in the presence of 6.0 M guanidinium chloride. Therefore, the kcat values reported below represent minimal estimates on the assumption that the enzyme active-site concentration is equivalent to its concentration as protein.

Crystallization and X-ray data collection

Purified human QC mutants were concentrated to 10–12 mg/ml and crystallized at 25 °C by the hanging-drop vapour-diffusion method. The crystallization buffers (reservoir solutions), depending on the mutants used, consisted of 2–4% (v/v) dioxan and 1.6–1.8 M (NH4)2SO4 in 100 mM Mes, pH 6.5. For example, the concentrations of dioxan and (NH4)2SO4 for S160A, E201D, E201Q and D248Q were 3% and 1.8 M respectively; for E201L, D248A and D305A they were 4% and 1.6 M; for S160G and H319L they were 3% and 1.7 M; and for D305E they were 2% and 1.8 M. For each crystallization, 2 μl of enzyme solution was mixed with 2 μl of reservoir buffer, and the rhombohedral crystals started to appear within 5–10 days. X-ray-diffraction experiments were performed at several synchrotron radiation centres, i.e. beamline 13B1 of the NSRRC (National Synchrotron Radiation Research Center, Hsinchu, Taiwan), beamline 12B2 of the SPring-8 synchrotron radiation facility (Hyōgo, Japan) and beamline 4.2.2 of the ALS (Advanced Light Source, Lawrence Berkeley National Laboratory, Berkeley, CA, U.S.A.). Before being mounted on the X-ray machine, crystals were briefly soaked in reservoir solutions containing 30% (v/v) glycerol as a cryoprotectant. All diffraction data were processed and scaled by employing the HKL package [18]. The final crystallographic properties obtained are listed in Table 3 below.

Structure determination and refinement

The initial difference Fourier maps for the mutant structures were obtained by employing the published structure of wild-type human QC [PDB (Protein Data Bank) code 2AFM]. After replacement of the wild-type model with the mutant ones, the manual refinements were carried out using the programs O and XtalView [19,20], and subsequent computational refinements were performed using the CNS (Crystallography and NMR System) program [21]. The parameters for ideal protein geometry from Engh and Huber [22] were used for refinements, and the stereochemical quality of the refined structures was checked by the program PROCHECK [23]. The Rfree values were calculated by using 5% reflections. In addition, well-ordered water molecules were located and included in the models. The final refined structures include 646 (S160A, S160G, E201D, E201L and E201Q), 644 (D248A, D248Q and D305A) or 643 (D305E and H319L) out of the 658 total residues in an asymmetric unit, with the other residues located at the disordered regions being excluded from the models. The molecular Figures were produced by using the programs MOLSCRIPT [24], ALSCRIPT [25], and RASTER3D [26].


Mutations exert a more significant effect on catalysis (kcat) than does substrate binding (Km)

The mutant human QCs were expressed in E. coli cells and purified to near homogeneity (see Supplementary Figure S1 at Using the fluorogenic substrates Gln-βNA and Gln-AMC at 210–250 μM, we found that the activities of these mutants fall into five levels (Table 1):

  • (i) S160A has a relatively high activity (44–60% of that of the wild-type enzyme)

  • (ii) H319L is middle (8.4–13%)

  • (iii) S160G, E201D, D248A and D305N are low (1.1–4.4%)

  • (iv) D305A is very low (0.23–0.26%)

  • (v) E201L, E201Q, D248Q and D305E are extremely low (<0.046%) or no activity

Furthermore, the steady-state kinetic properties of these mutants, except for E201L and E201Q, were analysed. The substrate Gln-βNA was chosen because of its higher solubility compared with that of Gln-AMC. Under the assay conditions, a slight substrate inhibition was detected with Ki values higher than 3.5 mM, so that the fitted Km and Vmax values are not true Michaelis–Menten parameters. These mutations, except for S160A, significantly affect the steady-state catalytic activity of the enzyme (Table 2). The kcat value of the enzyme was decreased 5.4–2842-fold by these mutations, whereas the Km value was increased only 2.2–19-fold, implying that the H-bond network is more critical for catalysis of the enzyme than substrate binding.

View this table:
Table 2 Kinetic parameters for the wild-type and mutant human QCs with Gln-βNA as substrate

Results are means±S.E.M. from two assays.

X-ray crystallographic analysis of the human QC mutants

We have crystallized all the human QC mutants except for D305N. The crystals of these mutants have a space group of R32 and a typical unit cell of a=b=119 Å and c=333 Å, in which an asymmetric unit comprises two human QC molecules. We have collected the 1.66–2.18 Å-resolution X-ray data of those crystals and refined the mutant structures to high qualities, where the Rfactor and Rfree values are smaller than 0.185 and 0.223 respectively (see Table 3). As expected, the overall structures of these mutants show no significant conformational changes when compared with the wild-type structure, with rmsds (root-mean-square deviations) in the range of 0.089–0.195 Å for the 643–646 Cα-carbons in an asymmetric unit. According to the final B-factor values (see Table 3) and the omitted electron-density maps (Figure 2), the mutational residues Leu201 and Gln248 in the structures of E201L and D248Q respectively are slightly disordered, and Glu305 in the structure of D305E is highly flexible.

View this table:
Table 3 Summary of crystallographic properties

Values in parentheses correspond to the highest-resolution shell. Rfree values were calculated by using 5% reflections. Abbreviation: ID, identity.

Figure 2 Active-site structures of human QC mutants

The 2FoFc omitted electron-density maps (contoured at 1.5σ) around the H-bond network forming residues are shown. The maps for the mutational residues are further coloured purple. The distances (in Å) for possible H-bonds or some adjacent atoms are indicated.

Structural evidences for the unusual Glu201···Asp305···Asp248 H-bonds

In the wild-type structure of human QC, Glu201, Asp248 and Asp305 have the orientations that point to each other [8]. Similarly, the mutational residues Asp201 and Gln201 in the structures of E201D and E201Q respectively appear to still point to Asp305, with the distances that are reasonable for H-bonding (see Figure 2, and Supplementary Table S1 at The well-visible electron-density maps and the relatively low B-factor values of Asp201 and Gln201 indicate that both residues are quite stable. In contrast, Leu201 in the E201L structure moves slightly away from Asp305 by 0.3 Å and 0.44 Å (for the QC molecules A and B respectively in an asymmetric unit) when compared with similarly sized Asp201 in the E201D structure (Figure 2 and Table S1). This mutation causes Leu201 to become more flexible. Moreover, as observed in the structure of D305A, this mutation makes a slight (32° and 26° for A and B) rotation of Glu201 (see Figure 3A). These results clearly indicate that Glu201 is stabilized in the active site by H-bonding to Asp305.

Figure 3 Superimpositions (stereo view) of the wild-type and mutant human QC structures

(A) Wild-type (grey) versus D305A (blue); (B) wild-type versus S160G (cyan); (C) wild-type versus H319L (green); (D) wild-type versus D248A (red); (E) wild-type versus D305E (yellow); and (F) wild-type versus S160A (magenta) and S160G (cyan).

Likewise, as shown in Figures 2 and 3(A), Asp248 in the D305A structure undergoes an obvious (52° and 48° for A and B) rotation compared with the wild-type structure and forms a new H-bond to His319. The S160G mutation causes Asp248 to swing 0.82 Å (molecule A, and 0.83 Å for B) and also leads to a 0.67 Å (A, and 0.62 Å for B) movement of Asp305 toward Asp248 (see Figures 2 and 3B). Moreover, the H319L replacement induces a drastic (59° and 54° for A and B) rotation of Asp305 and, consistently, also causes Asp248 to undergo an obvious (49° and 56°) rotation, resulting in loss of the H-bond Asp248···Ser160 (Figures 2 and 3C). These data demonstrate that there is indeed an H-bond between Asp248 and Asp305.

Central COOH···COOH H-bonds are critical for catalysis

As shown in Figure 2, the central COOH···COOH H-bonds still remain in the structures of S160A, S160G, E201D and H319L, whereas one or both of the COOH···COOH H-bonds are disrupted by the mutations E201L, D248A and D305A. The high flexibility of Glu305 observed in the structure of D305E suggests that the mutant also lost the central COOH···COOH H-bonds. On the other hand, results from enzyme kinetic analysis show that the kcat values for S60A, S160G, E201D and H319L are in the range 0.8–5.3 s−1 (Table 2). Whereas the mutants D248A, D305A and D305E demonstrate kcat values of ≈0.0019–0.085 s−1, E201L shows no detectable activity (Table 1). These data indicate that disruption of the central COOH···COOH H-bonds strongly affects the steady-state catalytic activity of the enzyme.

With regard to the mutants D305N, D248Q and E201Q, it is expected that one of the central COOH···COOH H-bonds was replaced by the COOH···CONH2 H-bond (Figure 2). However, D305N, D248Q and E201Q showed low (≈1.25%), extremely low (≈0.042%) or no activity respectively toward the substrates Gln-βNA and Gln-AMC (Table 1). The kcat values for D305N and D248Q were 0.03 and 0.0042 s−1, respectively (Table 2). These results indicate that the roles of the central COOH···COOH bonds could not be replaced by the COOH···CONH2 bonds in the catalytic process.

Roles of the central COOH···COOH H-bonds could be partly replaced by the COOH···water bonds

Noteworthily, the mutants that lost one or both of the cental COOH···COOH bonds have various QC activities (Table 1). The activities of D248A and D305A are 30–1500- and 5–260-fold respectively higher than those for E201Q, D248Q and D305E. Consistently, the D248A and D305A kcat values are 20–45- and 16–36-fold respectively higher than those for D248Q and D305E (Table 2). By contrast, although there is a ≈6-fold difference of specific activity between D248A and D305A, their kcat values are almost equivalent. By superimposing the D248A structure on that of wild-type enzyme (Figure 3D), it is apparent that there is an additional water molecule in D248A overlapping with one of the side-chain carboxylic oxygen atoms of the wild-type Asp248. This water molecule forms a H-bond to Asp305, spatially displacing the central Asp248···Asp305 bond observed in the wild-type structure. Since Asp248 was found to bind to the QC substrate directly [8], the observation that D248A has a relatively low Km value (125 μM) suggests that the new water molecule might help to maintain the enzyme–substrate interactions by means of the water-mediated H-bondings. Besides the additional water molecule, the other active-site residues of D248A are overlapped very well by the same residues in the wild-type active site (Figure 3D). This also implies that the new water molecule of D248A is responsible for the restoration of ≈1.6% steady-state catalytic activity of the mutant. Therefore the roles of the central COOH···COOH H-bonds could be partially replaced by COOH···H2O bonds in the catalysis process. A similar situation is also found in the structure of D305A (Figure 2), in which both the central COOH···COOH bonds have been replaced by the COOH···H2O bonds. The significant rotations on Glu201 and Asp248 observed in the D305A structure might provide an explanation for the fact that D305A has a relatively high Km value (1.1 mM).

Asp305 primarily plays a catalytic role

Mutations of Asp305 to alanine, asparagine and glutamic acid residues lower the kcat value of the enzyme 79–2842-fold and increase the Km value 4.4–19-fold (Table 2), suggesting that Asp305 primarily plays a catalytic role. The D305A mutation disrupted the two central COOH···COOH bonds but, in their place, created several COOH···H2O H-bonds, including those of Glu201···water and Asp248···water (Figure 2). Such H-bond substitutions decreased the steady-state catalytic activity of the enzyme ≈79-fold. Interestingly, the D305E replacement, with the side-chain COOH group still remaining, showed a ≈2842-fold decrease in the steady-state catalytic activity (Table 2). Compared with Asp305 in the wild-type structure, the longer side chain of Glu305 in D305E makes the environment around Glu201, Asp248 and His319 more crowded (Figure 2). As a consequence, Glu305 does not have a proper conformation for H-bonding to Glu201 and Asp248, and also blocks the entrance of additional water molecules into the active site. However, the crowding environment might limit the rotations of Glu201 and Asp248 (Figure 3E), thus providing an explanation for the observation that D305E has a relatively low Km value compared with that of D305A (Table 2).

In addition, D305N also demonstrates a low Km value, similar to that of D248Q, a phenomenon presumably attributable to their similar COOH···CONH2 bonds, which stabilize Glu201 or Asp248 in the active site. However, unlike D248Q, the kcat value for D305N is relatively large (Table 2). To answer the question of the catalytic mechanism fully, a greater knowledge of the X-ray structure of D305N will be required.

Ser160 and His319 help to stabilize the conformations of Asp248 and Asp305

Although the S160A replacement caused a loss of the Ser160···Asp248 H-bond (Figure 2), its steady-state catalytic activity was almost completely retained (Table 2), indicating that the Ser160···Asp248 bond is not critical for catalysis. This S160A mutation also causes Asp248 to undergo a slight (27° and 28° for A and B) rotation (see Figure 3F), but it is less significant than the rotation (39° and 40°) induced by the S160G substitution. This structural difference might account for why the Km value of S160G is ≈3.5-fold higher than that for S160A (Table 2). As described above, the orientational change of Asp248 might simultaneously induce a rotation on Asp305 in favour of an H-bonding, the S160G mutation causing the catalytically critical Asp248···Asp305 H-bond to move 0.75 Å and 0.73 Å (for A and B), in contrast with the same H-bond in S160A, which moves 0.48 Å and 0.46 Å (Figure 3F). This difference between S160A and S160G might be the reason for the ≈6.3-fold difference between their steady-state catalytic activities (Table 2).

The H319L mutation caused a 5.4-fold decrease in kcat value, with a 2.8-fold increase in Km (Table 2). Since the Asp305···His319 H-bond is spatially far away from the zinc-atom-containing catalytic centre, and the orientations of Asp248 and Asp305 in H319L structure are quite similar to those of S160G (compare Figures 3B and 3C), we suggest that the decrease in catalytic activity caused by the H319L mutation is due to the drastic orientational changes of Asp248 and Asp305.


Despite the highly conserved structure between human QC and the exopeptidases, it is noteworthy that the human QC active site exhibits a more acidic environment [8]. A number of acidic residues (Asp142, Asp159, Glu201, Glu202, Asp248, Asp305 and Asp306) are located in the zinc-atom-containing catalytic centre of human QC and are fully conserved in the active sites of animal QCs. Conceivably the acidic environment may provide ideal conditions for maintaining protons between Glu201, Asp305 and Asp248.

Unexpectedly, an imidazole molecule in the active site of E201L displaced the zinc-co-ordinated water molecule. Presumably, this imidazole moiety might take part in a hydrophobic interaction with Leu201. Although imidazole is a competitive inhibitor of human QC [27], its inhibitory effect seems quite weak (Ki=103 μM). Since the assay conditions involved 210–250 μM substrate with no addition of imidazole, we believe that the imidazole only slightly attenuated the QC activity of E201L.

Previously performed structural studies [8] showed that the conserved Glu201 of human QC superimposes very well with the catalytic glutamic acid residue of the exopeptidases (Figure 1D). This glutamate residue of exopeptidases was proposed to act as the general base and acid during catalysis [14,28,29]. On the basis of structural similarity, we suggested that Glu201 may play a similar catalytic role [8]. However, the proposed mechanism does not completely accord with several results described here.

  • (i) Mutating Asp305 to asparagine and glutamic acid residues greatly reduced (>180 fold) the steady-state catalytic activity of the enzyme

  • (ii) The D248Q mutation lowered the kcat value ≈1286-fold, but raised the Km value only ≈5.7-fold

  • (iii) The D248A and D305A replacements rescued ≈1% of steady-state activity

  • (iv) There is a sharp correlation between the central COOH···COOH H-bonds and the enzymatic activity

These new findings likely support an alternative pathway in which Asp248 and Asp305, as well as the central COOH···COOH bonds, are involved in catalysis.

As shown in Figure 1(D), Asp305 of human QC is replaced by serine and threonine residues in aminopeptidases and carboxypeptidase G2 respectively, and no H-binding to the catalytic glutamic acid residue could be identified in these proteases. This structural difference between QC and exopeptidases, together with the drastic decrease in QC activities of the mutants D305A, D305E and D305N, preclude the possibility that the Glu201···Asp305 bond only stabilizes the catalytic Glu201. In addition, the close correlation between the central COOH···COOH bonds of human QC with steady-state catalysis is reminiscent of the low-barrier aspartic acid···aspartic acid H-bond in the catalytic centre of pepsin-like aspartic peptidases [30]. Das et al. [31] reported that the inner oxygen atoms of the two catalytic aspartate residues in HIV-1 protease (an aspartic peptidase) were 2.30 Å apart, a distance that is shorter than a normal H-bond distance but is close to the distance (2.53 Å and 2.44 Å for A and B) involved in the Asp248···Asp305 bond of human QC. This consistency suggests that the H-bond network of human QC plays a homologous, proton transfer, role.

In the QC-catalysed intramolecular cyclization, the enzyme assists to position the substrate N-terminal α-amino group in close proximity to the carbonyl group of the scissile γ-amide. Subsequently, the α-nitrogen nucleophilically attacks the γ-carbonyl carbon atom, resulting in the formation of a tetrahedral intermediate (Figure 4A). Prior to collapse of the intermediate, it is necessary to transfer a proton from the positively charged α-amino group to the leaving amino group on the scissile γ-amide. This is reminiscent of the similar proton transfer from a zinc-bound water molecule to the leaving NH group on the scissile peptide bond during catalysis of the double-zinc exopeptidases (see Supplementary Figure S2 at Such a water-to-substrate proton transfer in exopeptidases is well known to be mediated by the catalytic glutamic acid residue [14], which structurally corresponds to Glu201 in human QC. However, as described above, the structural and mutational data presented here do not seem to support the notion that the proton transfer from α- to γ-amino groups in human QC is exclusively attributable to Glu201. Instead, it might be accomplished by a co-operation of Glu201, Asp248, Asp305 and the COOH···COOH bonds that they form. Noteworthily, a previous study of the H-bond network in the active site of amiopeptidase P suggested an alternative proton donation to the leaving NH group [32], which is mediated by an aspartic acid residue, an arginine residue and a tyrosine residue and the H-bond network that they form.

Figure 4 Proposed catalytic mechanism of human QC

(A) The residues Glu201, Asp248 and Asp305 of human QC (ball-and-stick model) guide an essential proton transfer during catalysis, resulting in a proton movement from the positively charged α-amine of the tetrahedral intermediate of substrate (black) to the leaving amino group on the scissile γ-amide. (B) One possible catalytic mechanism for human QC.

The large decreases in the catalytic activity for the mutations D248Q and D305N are consistent with the fact that glutamine and asparagine residues are not chemically equipped to act as proton donors during the proton-transfer process. By contrast, water molecules have been frequently identified to participate in catalytically important H-bond networks, such as those observed in aspartic peptidases and methyltransferases [33,34], the findings being in accord with our results for the mutants D248A and D305A. These observations allow us to suggest a proton-transfer process between Glu201, Asp248 and Asp305 during catalysis by human QC, which leads to a proton movement from the α-amine of the substrate tetrahedral intermediate to the γ-amide amino group (Figure 4B). The H-bond network among Glu201, Asp248 and Asp305 has one net negative charge, thus well accommodating the positively charged α-amino group of the N-terminal glutamine residue of substrate on binding to human QC [35]. Additionally, the enzymatic activity of human QC decreased significantly at pH values higher than 9.0 [36], probably reflecting a breakdown of the H-bond network under the more basic conditions.


We thank Dr. Yuch-Cheng Jean and Dr. Yu-San Huang (NSRRC) and Dr Jay Nix (ALS) for assistance in X-ray data collection at their synchrotron facilities. We are also grateful to Dr Shu-Chuan Jao and Dr Tzu-Ping Ko (Institute of Biological Chemistry, Academia Sinica, Taipei, Taiwan) for helpful discussion of the catalytic mechanism. This work was supported by grants from Academia Sinica and National Core Facility of High-Throughput Protein Crystallography Grant [NSC (National Science Council) 95-3112-B-001 to A. H.-J. W.].


  • The atomic co-ordinates and structure factors for S160A (codes: 2ZED), S160G (2ZEE), E201D (2ZEF), E201L (2ZEG), E201Q (2ZEH), D248A (2ZEL), D248Q (2ZEM), D305A (2ZEN), D305E (2ZEO), and H319L (2ZEP) have been deposited in the Protein Data Bank, Research Collaboratory for Structural Bioinformatics, Rutgers University, New Brunswick, NJ, U.S.A. (

Abbreviations: Gln-AMC, L-glutaminyl 4-methylcoumarinylamide; Gln-βNA, L-glutaminyl 2-naphthylamide; H-bond, hydrogen bond; IPTG, isopropyl β-D-thiogalactopyranoside; LB, Luria–Bertani; Ni-NTA, Ni2+-nitrilotriacetate; PAP I, pyroglutamyl aminopeptidase I; PDB, Protein Data Base; pGlu, pyroglutamate; pGlu-βNA, pyroglutamyl 2-naphthylamide; QC, glutaminyl cyclase (glutaminyl-peptide cyclotransferase, EC; QPCT, the human QC gene coding for Q (glutaminyl)-peptide cyclotranferase; rmsd, root-mean-square deviation


View Abstract