Structural diversity of angiotensin-converting enzyme

Insights from structure–activity comparisons of two Drosophila enzymes


R. E. Isaac, Faculty of Biological Sciences, Miall Building, University of Leeds, Leeds LS2 9JT, UK
Fax: +44 113 34 32835
Tel: +44 113 34 32903


The crystal structure of a Drosophila angiotensin-converting enzyme (ANCE) has recently been solved, revealing features important for the binding of ACE inhibitors and allowing molecular comparisons with the structure of human testicular angiotensin-converting enzyme (tACE). ACER is a second Drosophila ACE that displays both common and distinctive properties. Here we report further functional differences between ANCE and ACER and have constructed a homology model of ACER to help explain these. The model predicts a lack of the Cl-binding sites, and therefore the strong activation of ACER activity towards enkephalinamide peptides by NaCl suggests alternative sites for Cl binding. There is a marked difference in the electrostatic charge of the substrate channel between ANCE and ACER, which may explain why the electropositive peptide, MKRSRGPSPRR, is cleaved efficiently by ANCE with a low Km, but does not bind to ACER. Bradykinin (BK) peptides are excellent ANCE substrates. Models of BK docked in the substrate channel suggest that the peptide adopts an N-terminal β-turn, permitting a tight fit of the peptide in the substrate channel. This, together with ionic interactions between the guanidino group of Arg9 of BK and the side chains of Asp360 and Glu150 in the S2′ pocket, are possible reasons for the high-affinity binding of BK. The replacement of Asp360 with a histidine in ACER would explain the higher Km recorded for the hydrolysis of BK peptides by this enzyme. Other differences in the S2′ site of ANCE and ACER also explain the selectivity of RXPA380, a selective inhibitor of human C-domain ACE, which also preferentially inhibits ACER. These structural and enzymatic studies provide insight into the molecular basis for the distinctive enzymatic features of ANCE and ACER.


angiotensin-converting enzyme


Drosophila melanogaster angiotensin-converting enzyme


Drosophila melanogaster angiotensin-converting enzyme-related




angiotensin I


angiotensin II


o-aminobenzoic acid



Angiotensin-converting enzyme (ACE, EC is a zinc peptidyl-dipeptidase, which is best known for catalysing the last step in the synthesis of the vasoconstrictor angiotensin II (AII) from angiotensin I (AI) and for the metabolic inactivation of the vasodilator bradykinin (BK) [1]. The somatic form of the enzyme is a glycosylated type I membrane protein comprising two homologous domains, generally known as the N-domain and C-domain, arranged in tandem and joined by a short connecting peptide sequence [2]. Each domain is catalytically active, and both are capable of cleaving AI and BK. The ACE gene also gives rise to a second mammalian ACE, known as either testis (tACE) or germinal ACE, through the use of an intragenic promoter that drives expression in developing spermatocytes. It is a single-domain enzyme that is identical with the C-domain of somatic ACE, apart from a peptide insert encoded by the testis-specific exon 13 of the ACE gene [2]. ACE knockout mice display renal abnormalities, low blood pressure, anaemia and male infertility, confirming the important role of this enzyme in development, blood homoeostasis and reproduction [2].

Although N-domain and C-domain are highly similar in protein sequence and share many enzymatic properties, they can be differentiated by substrate and inhibitor preferences and by the extent to which they are activated by Cl[3–5]. The haemoregulatory peptide, N-acetyl-Ser-Asp-Lys-Pro (AcSDKP), another in vivo substrate for mammalian ACE, is hydrolysed more efficiently by the N-domain, as is the internally quenched fluorogenic substrate Abz-SDK(Dnp)P [6,7]. Cl can stimulate the activity of both ACE domains, but the C-domain active site is more sensitive to changes in Cl concentration [3]. The level of activation, as well as the concentration of Cl required for maximal stimulation, is dependent on pH and the peptide substrate. The two domains of mammalian ACE can also be distinguished by the N-domain-selective inhibitor RXP407 [8], the C-domain-selective inhibitor RXPA380 [9], and several BK-potentiating peptides [10].

A homologue of ACE, known as ACE2, has been characterized as a single-domain type I glycoprotein [11,12]. It is important for normal contractility of heart muscle [13]. The important enzymatic feature of ACE2 is that, unlike ACE, it is a carboxypeptidase, removing a single residue from the C-terminus of peptides that have either a Pro or Leu in the P1 position, e.g. angiotensin II, apelin 13 and dynorphin A 1–13 [14]. The activity of ACE2 is greatly enhanced in the presence of NaCl [15,16]. Therefore Cl activation is a common feature of the mammalian members of the ACE family of peptidases.

In vertebrates, the number of ACE genes appears to be limited to ACE and ACE2, but in some insects there has been a much greater expansion of this gene family. For example, in the mosquito, Anopheles gambiae, and in Drosophila melanogaster there are nine and six ACE genes, respectively [17,18]. Of the six Drosophila genes, only ANCE and ACER have been confirmed to produce functional metallopeptidases [19,20]. They are both single-domain proteins with ≈ 40% amino-acid sequence identity and 60% similarity to each of the two domains of mammalian ACE. ANCE and ACER have distinct tissue expression patterns, indicating different physiological roles [21,22]. ANCE appears to have a role in embryogenesis, metamorphosis and reproduction [20,23,24]. A function for ACER has not been established, but the protein is associated with the developing heart in embryos and in the brain and reproductive tissues of adults (A. Carhan, R.E. Isaac and A.D. Shirras, unpublished results). The two enzymes share some enzymatic properties, such as peptidyl-dipeptidase activity towards hippuryl-l-histidyl-l-leucine (Hip-His-Leu), and BK, and inhibition by inhibitors of mammalian ACE [19,20,25]. However, compared with ANCE, ACER displays more restricted substrate specificity. Although both ANCE and ACER hydrolyse Hip-His-Leu, only the ANCE activity is enhanced in the presence of NaCl [20,25]. Another interesting difference between ACER and ANCE is that the ACER active site, but not that of ANCE, can accommodate an N-domain-specific inhibitor (RXP407), indicating common active-site features for ACER and the N-domain of human ACE [17].

Recent descriptions of the high-resolution molecular structure of ACE–inhibitor complexes for both human tACE [26,27] and Drosophila ANCE [28] have revealed the molecular details of the active site and how ACE inhibitors bind with high affinity. These studies confirm many of the predictions regarding the identity of the active-site residues and, in the case of tACE, identify other side chains involved in the binding of Cl at two sites (Cl1 and Cl2) positioned outside of the active site. The crystal structure of human ACE2, with and without bound inhibitor, has also recently been reported [29] and has provided a structural explanation for why ACE2 is a carboxypeptidase and not a peptidyl-dipeptidase. The structure of the native ACE2 identified a single Cl-binding site that corresponded to the Cl1 site of tACE. No bound Cl was recognized in the crystal structure of Drosophila ANCE, and it has been proposed that the equivalent Cl-binding sites in ANCE are substantially different and, in the case of Cl2, may be absent [26], which may explain the weaker effect of Cl on enzyme activity reported for this enzyme. In ACE2, the Cl2 site also does not exist, which leaves only Cl1 as a recognized Cl-binding site [29]. Interestingly, an alternative, but undefined, binding site for Cl has been suggested, which may be influential in the conformational movement that occurs on formation of the ACE2 ES complex [26,29].

Comparative molecular and biochemical studies of members of the ACE family are likely to provide new insights into the evolution of the ACE active site, the structural basis for differences in substrate specificity and the mechanisms by which Cl can have profound effects on enzyme activity. In this respect, Drosophila ANCE and ACER appear to be good examples of two family members that have diverged in structure and substrate specificity and are therefore likely to provide valuable information. We now report on additional biochemical differences between ANCE and ACER regarding substrate specificity, the effect of Cl on enzyme activity, and inhibition by new domain-selective inhibitors of human ACE. A model of the structure of ACER has been generated, which provides explanations for some of these biochemical differences.


Hydrolysis of AI

The effect of NaCl on the conversion of AI into AII by ANCE was determined at two pH values. At pH 7, increasing the concentration of NaCl resulted in a faster rate of conversion, which reached a plateau at 150–200 mm NaCl (Fig. 1A). At pH 8, maximal activity was achieved in the absence of NaCl, which had a weak inhibitory effect on the hydrolysis of AI as the salt concentration increased from 0 to 200 mm NaCl (Fig. 1A). To further examine the effect of NaCl and pH on the mechanism of ANCE activation, the kinetic constants of AI hydrolysis were determined in the presence and absence of 100 mm NaCl at pH 7 and 8 (Table 1). The activation by NaCl at pH 7 was the result of a 330% increase in kcat/Km, which was solely attributable to a lowering of the Km. A similar rise in the kcat/Km was observed when the pH was increased from 7 to 8 in the absence of NaCl, but in this case the greater catalytic efficiency was achieved by a combined increased kcat and a lower Km. Although AI is an extremely poor substrate for ACER, it was possible to determine kinetic constants for this reaction (Km 1.58 ± 0.28 mm; kcat 0.01 ± 0.001 s−1), which showed that this marked difference between ACER and ANCE was due to the very low kcat for AI hydrolysis by ACER. This weak peptidyl-dipeptidase activity, unlike that of ANCE and mammalian ACE, was not stimulated by NaCl (Table 2).

Figure 1.

(A) Effect of NaCl on the conversion of AI (200 µm) into AII by ANCE. Enzyme activity was measured using HPLC to quantify the formation of AII in Hepes buffer (□, pH 7; ▪, pH 8) in the presence of NaCl (0–200 mm) as described in Experimental procedures. The enzyme activity is expressed as percentage of the maximum activity recorded at pH 8 in the absence of NaCl. Values are the mean of three assays and the percentage standard error of the mean was 1–4%. (B) Inhibition of ANCE and ACER by MKRSRGPSPRR. Enzyme activity was determined using Abz-YRK(Dnp)P as described in Experimental procedures and is expressed as a percentage of the uninhibited activity.

Table 1.  Effect of NaCl on the kinetic constants for the conversion of AI into AII by ANCE. Kinetic constants for the conversion of AI into AII were determined as described in Experimental procedures and are expressed as the mean ± SEM (n = 3).
0 mm[NaCl]100 mm[NaCl]
 Km (mm)kcat (s−1)kcat/Km (s−1·µm−1)Km (mm)kcat (s−1)kcat/Km (s−1·m−1)
pH 7.02.70 ± 0.676.84 ± 1.042.53 × 10−30.82 ± 0.146.83 ± 0.468.33 × 10−3
pH 8.01.23 ± 0.1711.06 ± 0.868.99 × 10−31.04 ± 0.2110.78 ± 1.1210.37 × 10−3
Table 2.  Effect of NaCl on the hydrolysis of peptides by ACER. The rate of hydrolysis of peptides (200 µm) was determined in 0.1 m Hepes/10 µm ZnSO4, pH 7 as described in Experimental procedures. Values are mean ± SEM (n = 3).
SubstrateReaction rate (units/h)
0 mm NaCl500 mm NaCl
  • a 

    Units of activity, nmol AII formed per µg ACER.

  • b 

    Units of activity, nmol dipeptide released per µg ACER.

AIa0.033 ± 0.0020.025 ± 0.002
[Leu5]Enkephalinb36.9 ± 1.019.6 ± 1.6
[Met5]Enkephalinb23.8 ± 0.039.9 ± 1.0
[Leu5]Enkephalinamideb10.1 ± 0.4123.2 ± 1.8
[Met5]Enkephalinamideb3.3 ± 0.149.3 ± 4.0

Hydrolysis of enkephalin peptides

[Leu5]Enkephalin, [Met5]enkephalin and their respective C-terminal amidated forms are hydrolysed at the Gly-Phe bond by both ANCE and ACER at neutral pH [20]. The endopeptidase activity of ACER, but not ANCE, towards [Leu5]enkephalinamide and [Met5]enkephalinamide was stimulated in the presence of Cl ions (Table 2). The enhancement of the hydrolysis of the amidated peptides by 500 mm NaCl was 12-fold and 15-fold, respectively, whereas the cleavage of both [Leu5]enkephalin and [Met5]enkephalin was inhibited by ≈ 50% (Table 2). The NaCl-induced activity of ACER was measured at different [Leu5]enkephalinamide and [Met5]enkephalinamide concentrations, which generated anomalous kinetics, including substrate inhibition at peptide concentrations above 150 µm (data not presented).

Hydrolysis of BK and related peptides

Initial velocities for the hydrolysis of the BK peptides were obtained by determining the rate of release of the C-terminal dipeptide (Phe-Arg for BK, [Thr6]BK and Ile-Ser-BK; Tyr-Arg for [Tyr8]BK). ANCE consistently cleaved these peptides with much greater efficiency (kcat/Km) than ACER, mainly because of the lower affinity of ACER for these substrates (Table 3). In the case of ANCE, extending BK at the N-terminus with Ile-Ser had no significant effect on the Km and kcat, and replacing the Phe8 of BK with tyrosine resulted in a modest increase in both the Km and kcat. In contrast, replacing Ser6 of BK with threonine resulted in greatly increased affinity between the substrate and ANCE, but not ACER. Indeed the Km value for the hydrolysis of [Thr6]BK was so low that it was difficult to obtain accurate Km values using HPLC to quantify reaction rates at very low substrate concentrations. We therefore used [Thr6]BK as an inhibitor of the hydrolysis of Abz-YRK(Dnp)P and obtained a Ki value of 23 ± 4 nm, confirming the very high affinity displayed by ANCE for this peptide.

Table 3.  Kinetic constants for the hydrolysis of bradykinin-related peptides by ANCE and ACER. –, No detectable hydrolysis of the peptide and no inhibition of the cleavage of Abz-YRK(Dnp)P by ACER.
Kmm)kcat (s−1)kcat/Km s−1m)−1Kmm)kcat (s−1) kcat/Km (s−1·µm−1)
  1. aKm determined from the IC50 value obtained by measuring initial rates of hydrolysis of the fluorogenic substrate Abz-YRK(Dnp)P (5 µm) in the presence of different concentrations of MKRSRGPSPRR. b Estimated from the initial velocity recorded at a substrate concentration 100 times greater than the Km.

BK (RPPGFSPFR)0.27 ± 0.051.09 ± ± 0.970.54 ± 0.030.09
Ile-Ser-BK (ISRPPGFSPFR)0.30 ± 0.0051.10 ± 0.0053.665.54 ± 0.980.17 ± 0.010.03
[Tyr8]BK (RPPGFSPYR)0.58 ± 0.082.41 ± 0.124.1511.3 ± 3.461.17 ± 0.110.10
[Thr6]BK (RPPGFTPFR)0.073 ± 0.040.24 ± 0.0083.315.67 ± 1.580.60 ± 0.040.11
MKRSRGPSPRR0.37a ± 0.118.8b ± 0.550.81

MKRSRGPSPRR is an invertebrate BK-like peptide predicted to be a cleavage product of a neuropeptide precursor gene in Aplysia californica[30]. HPLC analysis showed that MKRSRGPSPRR was an excellent substrate for ANCE, but was resistant to hydrolysis by ACER. MS confirmed that reaction products were MKRSRGPSP ([M + H]+, m/z 1014.3) and MKRSRGP ([M + H]+, m/z 830.4), generated by the sequential cleavage of Arg-Arg and Ser-Pro. MKRSRGPSPRR was a strong inhibitor of the hydrolysis of Abz-YRK(Dnp)P with a Ki of 185 nm for the inhibition of ANCE (Fig. 1B). In contrast, MKRSRGPSPRR, even at a concentration of 100 µm, did not significantly inhibit ACER activity, measured with the same fluorogenic substrate.

Homology model of the structure of ACER

We generated a model of ACER based on the crystal structure of ANCE. The homology model allowed us to compare the structure of the substrate/inhibitor binding sites between these related enzymes, which are very similar in primary protein structure, but display quite different enzymatic properties. One of the striking differences between ANCE and ACER predicted by our model is a significant change in the electrostatic charge that lines the substrate-binding channel, a change from predominantly negative charges in ANCE to positive charges in ACER (Fig. 2). To gain insight into why BK peptides bind with higher affinity to ANCE than to ACER, we docked BK and [Thr6]BK into the ANCE substrate channel. The modelling predicts that the negatively charged side chain of Asp360, as well as Glu150, forms favourable ionic interactions with the positively charged C-terminal arginine of both substrates (Fig. 3). Interestingly, in ACER, this interaction is lost because Asp360 is replaced with His368 (Table 4). The models of BK and [Thr6]BK bound to ANCE suggest that the extra methyl group of [Thr6]BK occupies a small hydrophobic pocket, which is conserved in both ANCE and ACER. The models also suggest that the two peptides bind in a similar orientation, with a β-turn centred on the residues Pro2-Pro3.

Figure 2.

Surface representations of the electrostatic potential of ANCE and a homology model of ACER. The proteins have been sliced in half to show the internal substrate-binding channel. The N-chamber and C-chamber (N and C) are postulated to bind up to ≈ 7 N-terminal residues and the C-terminal dipeptide of substrate, respectively. Molecular surfaces and electrostatic potential were calculated with the program spock ( ANCE co-ordinates were obtained from the recently determined crystal structure (PDB accession code 1J36). The homology model of ACER was generated in swiss-model using the ANCE structure as a template. Positive and negative charges are represented by shades of blue and red, respectively, with neutral areas coloured white.

Figure 3.

A stick diagram showing predicted electrostatic interactions between the C-terminal Arg9 of BK and ANCE. The interactions between Asp360 of ANCE and the guanidino group of Arg9 of BK will be lost in ACER as Asp360 is replaced with His368.

Table 4.  Comparison of the residues that contribute to the S2′ subsite of human C-domain ACE (the residue numbers for human tACE are in parentheses) with the N-domain of human ACE, ANCE and ACER.
Gln259Gln857 (281)Gln265Gln274
Ser260Thr858 (282)Gln266Ser275
Asp354Glu952 (376)Asp360His368
Ser357Val955 (379)Phe363Tyr371
Thr358Val956 (380)Thr364Val372
Asp393Asp991 (415)Asp399Asp407
Glu431Asp1029 (453)Asp437Ser445
Phe435Phe1033 (457)Phe441Phe449
Phe438Phe1036 (460)Phe444Phe452
Lys489Lys1087 (511)Lys495Lys503
Tyr498Tyr1096 (520)Tyr504Tyr512
Tyr501Tyr1099 (523)Tyr507Tyr515
Phe505Phe1103 (527)Phe511His519

Selective inhibitors of ANCE and ACER

Inhibition constants were determined for RXPA380, RXPA381 and RXPA384 for both ANCE and ACER (Table 5). These values showed that RXPA384 was only slightly more potent as an inhibitor of ACER, whereas RXPA381 was able to distinguish between the two enzymes with a selectivity factor of more than 100 in favour of ACER. RXPA380 inhibited ACER with a Ki of 4.8 µm, but did not inhibit ANCE, even at a concentration of 100 µm.

Table 5.  Potency of RXPA series of compounds as inhibitors of ANCE and ACER. ANCE and ACER activities were measured using the fluorogenic substrate Abz-YRK(Dnp)P (5 µm) as described in Experimental procedures. –, No inhibition with 100 µm RXPA380.
InhibitorKi (nm)
RXPA380 (Cbz-Pheψ[PO2-CH]Pro-Trp-OH)4800
RXPA381 (Cbz-Pheψ[PO2-CH]Ala-Ala-OH)365   3
RXPA384 (Cbz-Pheψ[PO2-CH]Ala-Trp-OH)152  95

To understand the molecular basis behind the selective inhibition of ACER by RXPA380 and RXPA381, these molecules were modelled into the binding sites of ANCE and ACER. The model of RXPA380/ACER shows that RXPA380 is bound in a very similar orientation to the model generated for RXPA380/C-domain ACE [31]. Phe1033 and Phe1103 of C-domain ACE are important in forming a hydrophobic side of the S2′ pocket for binding the tryptophan of RXPA380. Both of these residues are conserved in ANCE, but in ACER, Phe1103 is replaced with His519 (Table 4). The other side of the S2′ pocket is formed by two adjacent valine residues in C-domain ACE (Val955 and Val956). Val955 is replaced by larger phenylalanine and tyrosine residues in ANCE and ACER, respectively, which in our models are pointing away from the inhibitor so that the change in the size of the side chain may have minimal effect on binding. Val956 of C-domain ACE is conserved in ACER as Val372, but in ANCE this is replaced by Thr364, which reduces the hydrophobicity of the ANCE S1′ pocket (Fig. 4A). In ANCE, Gln266 with its large polar side chain replaces Ser275 and Thr858 of ACER and C-domain ACE, respectively (Table 4). In our model, the larger side chain of Gln266 restricts the space available and results in steric hindrance of the large indole ring of RXPA380 (Fig. 4A).

Figure 4.

Representations of enzyme–inhibitor interactions. (A) RXPA380 bound to C-domain ACE (grey), superimposed on the crystal structure of ANCE (yellow), highlighting differences between the proteins at the S2′ pocket. The absence of inhibition of ANCE by RXPA380 can be explained by the replacement of Thr858 with the larger Gln266, and Val956 with the polar Thr364. The combined effects of these changes will be to reduce the hydrophobic nature of the S2′ site and restrict the space available for the large indole ring of RXPA380. In ACER, the equivalent of Thr858 of C-domain ACE is the smaller Ser275, whereas Val956 is conserved as Val372. This is consistent with the inhibition of ACER by RXPA380. (B) Space-filling representation of RXPA381 bound to ANCE (left) and ACER (right) in the S2′ pocket, comparing the differences in packing of the P2′ methyl group of RXPA381 (arrowhead) against Val372 of ACER or the equivalent Thr364 in ANCE. The tyrosine and lysine residues interacting with the C-terminus of the inhibitor are labelled. The figure was generated in pymol.

In RXPA381, the P1′ and P2′ proline and tryptophan residues of RXPA380 are replaced by smaller alanine residues. The models of RXPA381 bound to ANCE and ACER show that the inhibitor is bound in a similar orientation, but with variation in the orientation of the C-terminal residue (Fig. 4B). All S1′ and S2′ residues interacting directly with RXPA381 are conserved between ANCE and ACER except for the aforementioned Val372 (ACER) and Thr364 (ANCE) (Table 5). The molecular dynamic simulations suggest that the methyl groups of the two alanines of RXPA381 pack closely with Val372 of ACER, whereas in ANCE, the methyl group of the terminal alanine residue is orientated away from Thr364, reinforcing the importance of the hydrophobicity of the valine side chain.


We have characterized the effect of Cl on ANCE activity by determining the kinetic constants for the hydrolysis of AI in the absence and presence of NaCl (100 mm). The increased kcat/Km observed at pH 7, was entirely the result of a 3.5-fold lowering of the Km for AI. A similar level of enhancement was also achieved in the absence of NaCl by changing the pH conditions from 7 to 8, although in this case changes in both the Km and kcat contributed to the increased catalytic efficiency. Although these effects are significant, they are modest compared with the activation by NaCl of the AI-converting activities of the C-domain of human ACE [3,32]. ACER hydrolyses AI extremely slowly, an activity that is not stimulated by Cl. Nevertheless, a strong effect of NaCl on the peptidase activity of ACER was observed when either [Leu5]enkephalinamide or [Met5]enkephalinamide was the substrate.

Our observation that NaCl alters the affinity of ANCE for AI suggests that the binding of Cl induces a conformational change in ANCE that influences the hydrolysis of AI. The molecular structures of two Cl-binding sites (Cl1 and Cl2) are known from the structure of human tACE [27], but no Cl anions were identified in the crystal structure of ANCE [28]. The Cl2 Cl-binding site of tACE, 10 Å from the catalytic zinc, is closer to the active site than Cl1 and comprises the side chains of Arg522, Trp220 and Tyr224. Comparing the structures of ANCE, ACER and tACE at the Cl2 binding site suggests that ANCE and ACER would not bind Cl at the Cl2 site. The substitution of Pro519 in tACE by a glutamate in both ANCE and ACER results in the carboxylic acid of this residue residing in the space occupied by Cl in the tACE crystal structure [26].

The Cl1 binding site of tACE lies 20 Å from the catalytic zinc and involves three contacts, Arg186, Trp485 and Arg489. Whereas Arg489 is conserved, Arg186 and Trp485 of tACE are replaced by Tyr170 and Phe469 in ANCE. It has been proposed that the Arg→Tyr substitution may result in a Cl-binding site more similar to the ACE Cl2 binding site [26]. Although the Trp→Phe substitution is expected to reduce the affinity for Cl, it is possible that the Cl1 site in ANCE may still bind the anion and that this interaction is responsible for our observed increase in affinity of ANCE for AI. In ANCE, the potential Cl1 binding site is adjacent to the peptide backbone of Lys495, which our modelling, together with recent site-directed mutagenesis studies on human ACE [33], suggest direct interactions between Lys495 and the C-terminus of the peptide substrate (Fig. 3). The presence of a Cl ion at this site may have a stabilizing effect on binding certain substrates.

In the N-domain of human ACE, and in ACER, the Cl1 site is altered by the replacement of Arg186 of tACE with His164 and His177, respectively, making it unlikely that Cl will bind at this position in both these enzymes [26]. However, there is a possibility that an alternative Cl-binding site exists in the N-domain of human sACE, as the R500Q mutant of the human ACE N-domain, which removes the Cl2 site, responds to 20 mm NaCl by a twofold increase in affinity for AI [32]. The strong NaCl-induced activation of ACER activity towards the amidated enkephalin substrates and the unlikely involvement of the Cl1 and Cl2 sites in this effect suggest that a different anion site may also be present in ACER. A similar proposal for a Cl-binding site, distinct from the two identified in tACE, has been put forward to explain the Cl-enhanced carboxypeptidase activity of human ACE2 [29]. The lack of understanding of the molecular mechanism by which Cl influences the catalytic activity of ACEs is illustrated by the recent characterization of ACE from the leech Theromyzon tessulatum[34]. The residues forming both Cl1 and Cl2 in tACE are absolutely conserved in the leech enzyme, suggesting that this ACE would, like human C-domain, be strongly activated by NaCl. However, the enzyme when expressed in mammalian cells responds with only modest activation (twofold) of the hydrolysis of Hip-His-Leu by NaCl with an optimal Cl concentration of 50 mm, and, thus, resembles the N-domain rather than the C-domain of human ACE.

All the BK peptides used in this study were cleaved by both ANCE and ACER, although ANCE was invariably the more efficient enzyme, displaying kcat/Km values 30–100-fold greater than those obtained with ACER. Our model of ANCE with either BK or [Thr6]BK docked in the substrate channel suggests that the acidic side chains of Asp360, as well as Glu150, form favourable ionic interactions with the positively charged C-terminal Arg of the peptides. These residues are conserved in the human N-domain and C-domain active sites (Table 5), both of which efficiently cleave BK. However, Asp360 of ANCE is replaced with His368 in ACER, and this change in the electrostatic charge in the S2′ pocket is predicted to reduce ionic interactions between ACER and the guanidino group of the C-terminal arginine of the BK peptides. This may explain why the Km values for the hydrolysis of BK, Ile-Ser-BK, [Thr6]BK and [Tyr8]BK by ACER are 20–75-fold higher than the corresponding values for ANCE. The model also suggests that an N-terminal β-turn centred on the residues Pro2-Pro3 of BK and [Thr6]BK allows the peptides to fit tightly into the larger (N chambers) of the two active-site cavities, which may explain why BK peptides bind with much higher affinity to ANCE and ACER than AI. BK adopts a similar conformation in models of BK bound to human C-domain ACE (R. J. Bingham, unpublished work), which would provide an explanation for why BK is the physiological substrate that displays the highest-affinity of any substrate of the human enzyme [2].

The affinity of BK for ANCE is increased almost fourfold by introducing an extra methyl group in [Thr6]BK. It has been shown previously that [Thr6]BK has a markedly different solution structure to BK [35] and has a greater tendency to adopt an N-terminal β-turn, which was also a consistent feature of our molecular modelling. The dynamic structure difference between BK and [Thr6]BK provides a possible explanation for the difference in binding affinity of these two BK peptides to ANCE.

MKRSRGPSPRR is structurally related to mammalian BKs and was shown to be an excellent ANCE substrate. In contrast, this peptide was resistant to hydrolysis by ACER and did not compete with substrate for the enzyme active site. The surface of the ACER active site is predicted to be positively charged, which would present an unfavourable electrostatic environment for Arg/Lys-rich peptides attempting to access the substrate-binding channel. In contrast, the negative charges lining the ANCE substrate channel would be expected to favour interactions with positively charged peptide substrates, especially MKRSRGPSPRR, which has positive charges along the length of the peptide.

RXPA380 (Cbz-Pheψ[PO2-CH]Pro-Trp-OH) is a highly selective inhibitor of the C-domain of somatic ACE, with the pseudo-proline and the tryptophan residues in the P1′ and P2′ positions of the inhibitor being important for determining this selectivity [31]. For both ANCE and ACER, it is clear that proline in the P1′ position does not allow strong inhibitor–enzyme interaction, as the substitution of the P1′ proline of RXPA380 with alanine in RXPA384 (Cbz-Pheψ[PO2-CH]Ala-Trp-OH) makes a much more potent inhibitor of both ANCE and ACER. The proline in RXPA380 probably restricts the orientation of the P2′ side chain to an orientation that is less favourable for interactions in the S2′ pocket of ANCE. Of the 12 residues of the S2′ subsite of C-domain ACE that are predicted to interact with the RXPA380 in a model of the inhibitor–enzyme complex [31], only eight are strictly conserved in the N-domain, nine in ANCE and eight in ACER (Table 4). The adjacent valines (Val955 and Val956) that help form the S2′ pocket of C-domain ACE appear to be involved in binding the tryptophan side chain of RXPA380. It has been proposed that replacement of these two residues in N-domain ACE with polar serine and threonine will limit favourable hydrophobic interactions between inhibitor and enzyme [31]. RXPA380 inhibits ACER, albeit weakly, but not ANCE. Our model of the ACER–RXPA380 complex shows the inhibitor bound in a very similar orientation to that described for C-domain ACE, with the side chain of Val372 (equivalent to Val956 of C-domain ACE) involved in ligand interaction at the S2′ pocket. The replacement of Val372 of ACER with the polar Thr364 in ANCE probably contributes towards the lack of inhibitory activity of RXPA380. This supports the hypothesis that the hydrophobicity of Val956 in C-domain ACE and Val372 in ACER is important for RXPA380 selectivity. In our model, the larger side chain of Gln266 restricts the space available for the large indole ring of RXPA380 and would therefore contribute together with Thr364 towards hindrance of RXPA380 binding to ANCE. In contrast, Thr858 of C-domain ACE is replaced by the smaller Ser275, and ACE Val956 is conserved as Val372 in ACER, which is consistent with the inhibition of ACER by RXPA380. RXPA381, which has alanine in both the P1′ and P2′ positions, inhibits both ANCE and ACER, but displays 100-fold selectivity in favour of ACER. This selectivity is consistent with the observation that RXP407 (Ac-Asp-Pheψ[PO2-CH]Ala-Ala-NH2) and Ac-Asp-Pheψ[PO2-CH]Ala-Ala-OH with a P1′ and a P2′ alanine are also selective inhibitors of ACER [17]. The side chain of Gln266 of ANCE, which forms the back of the S2′ site, is too distal (8 Å) to interact with the P2′ side chains of RXPA381 and RXP407, and therefore will not influence the binding of these less bulky inhibitors.

The unexpected result that ACER is inhibited by both an N-domain-selective and a C-domain-selective inhibitor demonstrates the dangers of classifying ACEs as either N-domain-like or C-domain-like. Molecular models of inhibitors complexed with ANCE and ACER have suggested structural explanations for these observations and provided new insights into how structural diversity in the ACE substrate channel can lead to important differences in enzymatic properties. In addition, our models of BK docked at the ACE active site have provided an explanation for the evolutionarily conserved tight binding of this substrate to ACE.

Experimental procedures

Enzyme substrates and inhibitors

Peptides were purchased from Sigma-Aldrich (Poole, Dorset, UK). RXPA380 (Cbz-Pheψ[PO2-CH]Pro-Trp-OH), RXPA381 (Cbz-Pheψ[PO2-CH]Ala-Ala-OH), RXPA384 (Cbz-Pheψ[PO2-CH]Ala-Ala-OH) were synthesized as described previously [8,31]. Abz-YRK(Dnp)P was a gift from Professor Adriana K. Carmona, Department of Biophysics, Division of Nephrology, Escola Paulista de Medicina, Universidade Federal de Sao Paulo, Sao Paulo, Brazil.

Expression and purification of recombinant ANCE and ACER

Recombinant ANCE and ACER were produced by expression in Pichia pastoris, as described previously [20,25]. Secreted ANCE and ACER were purified to homogeneity from the culture medium by using a combination of hydrophobic interaction and ion-exchange chromatography. (NH4)2SO4 was added to the culture media to a final concentration of 1.5 m, and, after centrifugation and filtration (0.2 µm pore size; Minisart, Sartorius Ltd, Epsom, Surrey, UK), the culture media were applied to a column (12 cm × 2.6 cm) packed with Phenyl-Sepharose Fast Flow 6 (Amersham Biosciences, Chalfont St Giles, Buckinghamshire, UK) pre-equilibrated with 1.5 m (NH4)2SO4/20 mm Tris/HCl, pH 8.0. Protein was eluted with a decreasing gradient of (NH4)2SO4 (1.5–0 m; over 500 mL; flow rate of 5 mL·min−1) and monitored using a UV detector set at 280 nm. Protein-containing fractions were pooled and dialysed against 20 mm Tris/HCl, pH 8.0, before being applied to an ion-exchange column (HiTrap Q HP, 5 mL bed volume; Amersham Biosciences). Protein was eluted using a 200 mL gradient of increasing concentration of NaCl (0–1 m), at a flow rate of 5 mL·min−1. Fractions containing enzyme activity, determined using Hip-His-Leu as the substrate [36], were pooled and dialysed against 100 mm Tris/HCl (pH 7.0)/50 mm NaCl/10 µm ZnCl2, before being concentrated to 1 mg protein per ml of buffer using a centrifugal concentrator (Microsep 10k; Pall Life Sciences, Portsmouth, Hampshire, UK). The final protein concentration was determined by absorbance at 280 nm. Cl-free protein was produced by dialysing 1 mL protein solution (1 mg·mL−1) against 5 L MilliQ water for 24 h followed by dialysis against 100 mm Hepes (pH 8.0)/10 µm ZnSO4 for 24 h.

Enzyme assays

Dipeptidyl carboxypeptidase activity towards peptide substrates was determined by HPLC quantification (214 nm) of the reaction products (AII for the hydrolysis of AI; Phe-Arg for the hydrolysis of BK, Ile-Ser-BK and [Thr6]BK; Tyr-Arg for the hydrolysis of [Tyr8]BK; MKRSRGPSP for the hydrolysis of MKRSRGPSPRR; Tyr-Gly-Gly, Phe-Leu-amide and Met-Leu-amide for [Leu5]enkephalinamide and [Met5]enkephalinamide; Tyr-Gly-Gly, Phe-Leu and Met-Leu for [Leu5]enkephalin and [Met5]enkephalin). Unless otherwise stated, the reactions were carried out at 35 °C in 100 mm Hepes (pH 8.0)/50 mm NaCl/10 µm ZnSO4 in a final volume of 20 µL for AI and larger volumes (200 µL to 1 mL) for BK and BK-related peptides. Reactions were stopped by either addition of trifluoroacetic acid to a final concentration of 2.5% or, for larger volumes, immersion in boiling water for 5 min. HPLC analysis required different reverse-phase columns and elution conditions to achieve peptide separation. The products of AI, MKRSRGPSPRR, and BK hydrolysis were resolved using a Phenomenex Jupiter C18 (5 µm particles, 250 × 4.6 mm; Phenomenex, Macclesfield, Cheshire, UK) column, whereas the separation of BK 1–5 and BK 1–7 required a SuperPac Pep-S column (5 µm particles, 250 mm × 4 mm; Amersham Biosciences). The following elution gradients of acetonitrile in 0.1% trifluoroacetic acid at a flow rate of 1 mL·min−1 were used: 15–36% acetonitrile over 14 min for AII; 6–24% acetonitrile over 22 min for Phe-Arg and MKRSRGPSP; 6–18% acetonitrile for BK 1–5 and BK 1–7 over 20 min; 0–24% acetonitrile over 20 min for the separation of Tyr-Gly-Gly, Phe-Leu, Met-Leu, Phe-Leu-amide and Met-Leu-amide. Identification of peptides by MS was performed using a Q-Tof MS/MS instrument. Hip-His-Leu hydrolysis was assayed as described previously [36].

The kinetics of inhibition of ANCE and ACER by BK, BK-related peptides and phosphinic acid inhibitors were determined by measuring the effects on initial rates of hydrolysis of Abz-YRK(Dnp)P (5 µm) in 100 mm Hepes, pH 8.0, 50 mm NaCl and 10 µm ZnSO4 (final reaction volume, 100 µL). ANCE and ACER hydrolysed Abz-YRK(Dnp)P, a fluorogenic substrate based on the structure of N-acetylSDKP [7], with Km values of 6.64 ± 1.1 µm and 4.60 ± 1.4 µm, respectively. The reactions were performed at 20 °C in 96-well black plastic plates (Corning Life Sciences, High Wycombe, Buckinghamshire, UK) using a Victor2 fluorimeter (PerkinElmer™, Turku, Finland) to quantify the rate of increase in fluorescence (λem 430 nm and λex 340 nm). The reaction was started by adding the substrate to the enzyme in 100 mm Tris/HCl (pH 7.0)/100 mm NaCl/10 µm ZnCl2.

Kinetic parameters and IC50 values were calculated using nonlinear regression curve-fitting programs (figp; Biosoft, Cambridge, UK). Error values are standard deviations of the parameters calculated from the fitted curve by figp. The Ki of inhibition of ANCE by [Thr6]BK was determined by measuring the kinetics of Abz-YRK(Dnp)P hydrolysis in the presence of 0, 10, 20, 50 and 80 nm[Thr6]BK.

Molecular modelling

The model of D. melanogaster ACER was generated in swiss-model[37] using the first approach mode and the crystal structure of ANCE as a template (Protein DataBank accession code 1J36). The zinc atom was manually positioned, co-ordinated by His375, Glu376 and His379, which were deduced to be the co-ordinating residues by sequence alignment. The co-ordinates of BK and [Thr6]BK were generated in pymol (, and manually positioned into the binding channel of ANCE and ACER using the molecular visualization program O. The peptide was aligned such that the carboxy group of the scissile peptide bond was orientated towards the zinc according to the proposed catalytic mechanism [27]. The large N-chamber and C-chamber readily allowed positioning of the peptide with minimal steric clashes. The model was then solvated with explicit water molecules in a 20 Å sphere centred on the peptide. This model was improved by energy minimization and molecular-dynamics simulations using the ds Modelling software (Accelrys, San Diego, CA, USA). All energy calculations were performed using the CHARM22 force field, and were restricted to the 20 Å sphere centred on the peptide. The nonbonded cut-off was set to 12 Å. Initial optimization was performed by two stages of energy minimization, firstly 500 steps of a conjugate gradient minimization, followed by 1000 steps using the adopted basis Newton–Raphson algorithm. This was followed by heating to and equilibrium at 300 K before a 1000-step molecular-dynamics simulation with time steps of 0.001 ps. Co-ordinates of RXPA380 were kindly provided by Philippe Cuniasse, Departement d'Etudes et d'Ingenierie des Proteines, Commissariat a l'Energie Atomique, CE-Saclay, Gif-Sur-Yvette, France. Co-ordinates of RXPA381 were generated in pymol. These co-ordinates were then superimposed on to ANCE and ACER assuming a similar binding orientation to ACE C-domain. This model was then solvated with explicit water molecules in a 20 Å sphere centred on the peptide and then subjected to the molecular modelling scheme described above.


We thank Adriana K. Carmona (Universidade Federal de Sao Paulo) for ACE substrates and Pam Gaunt (University of Leeds) for technical expertise, Alison Ashcroft (University of Leeds) for mass spectrometry, Philippe Cuniasse (Commissariat a l'Energie Atomique, CE-Saclay) for the pdb file of RXPA380, and Pierre Corvol, Tracy Williams and Xavier Houard (College de France, Paris) for Pichia expressing ANCE and ACER. We acknowledge the support of the Biotechnology and Biological Sciences Research Council through a studentship to R.J.B. and a grant to A.D.S. and R.E.I. (No. 89/S19378).