A Monomeric Variant of Insulin Degrading Enzyme (IDE) Loses Its Regulatory Properties

Background Insulin degrading enzyme (IDE) is a key enzyme in the metabolism of both insulin and amyloid beta peptides. IDE is unique in that it is subject to allosteric activation which is hypothesized to occur through an oligomeric structuture. Methodology/Principal Findings IDE is known to exist as an equilibrium mixture of monomers, dimers, and higher oligomers, with the dimer being the predominant form. Based on the crystal structure of IDE we deleted the putative dimer interface in the C-terminal region, which resulted in a monomeric variant. Monomeric IDE retained enzymatic activity, however instead of the allosteric behavior seen with wild type enzyme it displayed Michaelis-Menten kinetic behavior. With the substrate Abz-GGFLRKHGQ-EDDnp, monomeric IDE retained ∼25% of the wild type activity. In contrast with the larger peptide substrates β-endorphin and amyloid β peptide 1–40, monomeric IDE retained only 1 to 0.25% of wild type activity. Unlike wild type IDE neither bradykinin nor dynorphin B-9 activated the monomeric variant of the enzyme. Similarly, monomeric IDE was not activated by polyphosphates under conditions in which the activity of wild type enzyme was increased more than 50 fold. Conclusions/Significance These findings serve to establish the dimer interface in IDE and demonstrate the requirement for an oligomeric form of the enzyme for its regulatory properties. The data support a mechanism where the binding of activators to oligomeric IDE induces a conformational change that cannot occur in the monomeric variant. Since a conformational change from a closed to a more open structure is likely the rate-determining step in the IDE reaction, the subunit induced conformational change likely shifts the structure of the oligomeric enzyme to a more open conformation.


Introduction
Insulin-degrading enzyme (IDE) also known as insulysin, is a zinc metalloprotease first described based on its ability to cleave insulin [1][2][3]. Mutations in IDE are linked to a type II diabetic phenotype in the GK rat [4]. In addition IDE has been shown to play a key role in Alzheimer's disease in that it is one of the major enzymes responsible for amyloid b peptide (Ab) clearance in the brain. Studies from this [5] and other laboratories [6] have shown that mice lacking IDE activity through gene disruption accumulate Ab in the brain in a gene dose dependent manner. A number of studies have linked the IDE gene to both type 2 diabetes [7,8] and AD [9,10] although in the latter case a genetic association has yet to be firmly established.
We first reported that IDE is unique among the zinc metalloproteases in that it exhibits allosteric kinetic behavior [11]. The reaction of IDE with the internally quenched fluorogenic substrate 2-aminobenzyl-glycyl-glycyl-phenylalanyl-leucyl-arginyllysyl-histidyl-glycyl-asparaginyl-ethylenediamine-2,4-dinitrophenol (Abz-GGFLRKHGQ-EDDnp) was shown to exhibit both substrate induced (homotrophic) activation as well as activation produced by small peptides (heterotrophic activation). In addition we reported that polyphosphates such as ATP and triphosphate act as non-substrate heterotrophic activators, binding at a site distinct from the active site and distinct from the peptide activation site [12].
Recently the structures of IDE both liganded and unliganded were determined by Tang and co-workers [13][14][15]. These structures show that larger peptides make extended binding interactions at both the active site and at a distal site. We have recently determined the structure of rat IDE and found that this distal binding site serves as the site in which small peptide activators bind (Nicholas Noinaj, Sonia K. Bhasin, Eun Suk Song, Kirsten Scoggin, Louis B. Hersh, and David W. Rodgers, manuscript in preparation).
IDE exists as a mixture of monomers, dimers, and tetramers with the dimer the predominant species [11]. From the IDE structure of Shen et al [13] and confirmed in our IDE structure an interface between the two monomeric units involving elements from domains 3 and 4 of the four-domain protein was identified. We have taken advantage of this structural data to generate a monomeric variant of IDE. Studies on this monomeric IDE support our previous hypothesis [11] that an oligomeric form of IDE is required for allosteric activation.

Results
IDE is known to exist in a monomer-dimer-tetramer equilibrium with the dimer being the predominant form [11]. We examined our crystal structure of rIDE for a possible dimerization domain and found that molecular packing within the rIDE crystal lattice consistent with the presence of a dimer interface at the Cterminal region of the enzyme. This possibility is supported by the conservation of this interface in the crystals of hIDE [13] even though the human enzyme crystallizes in a different form. The proposed dimer interface buries about 1,400 Å 2 of accessible surface area and involves elements from domains 3 and 4 of the four structurally related domains that make up the enzyme, Figure 1. A key set of contacts occurs between regions near the Cterminus of each monomer, residues 1002-1006, which make nonpolar and hydrogen bond interactions across the interface. rIDE has only 13 residues C terminal to this interaction region, and only eight (to residue 1014) are seen in the crystal structure, which suggests that the C-terminal region of IDE could be removed without greatly compromising the stability of the enzyme. Such a deletion should destabilize the dimer interface sufficiently to make monomeric IDE the predominant form in solution. To test this hypothesis and to study the properties of monomeric IDE we generated a mutant in which residues 1002 to 1019 were deleted yielding rIDE DC . This mutant, as well as wild type rIDE, were expressed in insect cells as hexahistidine fusion proteins and purified by nickel affinity chromatography.
We demonstrated that rIDE DC is monomeric by molecular sieve chromatography. As shown in Figure 2, chromatography of wild type rIDE on a Superdex S200 column revealed the presence of tetrameric, dimeric, and monomeric species, with the dimer being the predominant form as expected. In contrast chromatography of rIDE DC on the same column at the same protein concentration showed a single monomeric species.
The kinetics of the rIDE DC reaction was compared to wild type rIDE using the fluorogenic substrate Abz-GGFLRKHGQ-EDDnp. rIDE DC retained enzymatic activity, however it no longer exhibited the sigmoidal kinetics seen with wild type rIDE, Figure 3. The kinetic constants, summarized in Table 1, show that rIDE DC exhibits about 25% of the k cat of wild type rIDE, but has a lower substrate K M . In contrast to wild type rIDE, where the reaction is cooperative with a Hill coefficient of 2, rIDE DC exhibits classical Michaelis-Menten kinetics with a Hill coefficient of ,1.
Similarly we identified two additional cleavage sites in bendorphin with rIDE DC , Figure 4B. IDE normally cleaves bendorphin at the Leu 17 -Phe 18 and Phe 18 -Lys 19 bonds [16]. The rIDE DC mutant cleaved at these sites, but in addition cleaved at Thr 6 -Ser 7 plus Phe 18 -Lys 19 and at Thr 6 -Ser 7 plus Asn 20 -Ala 21 .
Time course experiments showed that these products could not be detected with wild type IDE throughout the entire time course of its reaction with b-endorphin. Additional cleavage sites in bendorphin (Met 5 -Thr 6 and Lys 19 -Asn 20) were previously observed in the reaction of rIDE mutants containing substitutions of the active site glutamate [16].
We previously reported that substrates for IDE such as bradykinin and dynorphin B-9 act as activators of the hydrolysis of Abz-GGFLRKHGQ-EDDnp [11]. Based on our hypothesis that activation of IDE by activators occurs through conformational changes transmitted from one subunit to the other, we would predict that the monomeric rIDE DC variant would not exhibit this activation. That this is the case is shown for bradykinin and dynorphin B9 as the effectors, Figure 5. With wild type rIDE, bradykinin and dynorphin B-9 activate the reaction while with rIDE DC these peptides inhibited the reaction slightly.
Polyphosphates, including ATP and triphosphate (PPPi) also activate the IDE dependent hydrolysis of small substrates including Abz-GGFLRKHGQ-EDDnp [12]. The ATP binding site was shown to be distinct from both the substrate binding site as well as the distal site where peptide activators such as bradykinin and dynorphin B-9 bind. We thus determined whether ATP or PPPi [12] were capable of increasing the rate of the reaction of rIDE DC with Abz-GGFLRKHGQ-EDDnp. As shown in Figure 6 neither ATP nor PPPi could activate the reaction of rIDE DC under conditions in which the reaction of the wild type enzyme was increased more than 100 fold. Interaction of the ATP analog TNP-ATP with IDE can be followed by measuring the change in fluorescence that occurs upon binding [17]. We therefore compared the fluorescence spectra of TNP-ATP in the presence and absence of the monomeric IDE variant. As shown in Figure 7, the monomeric IDE variant produces an increase in TNP-ATP fluorescence and a slight shift in the fluorescence maximum to a lower wavelength, characteristic of TNP binding [17]. However the shift in fluorescence maximum is notably less than observed with wild-type IDE.

Discussion
Contacts in crystals of IDE suggest that the dimer interface is formed by elements of the two C-terminal domains, including a sheet-like contact formed by a terminal b strand. Thus by deleting 18 residues from the C terminus, we anticipated that the intermolecular contacts would be sufficiently weakened that only monomers would be stable. This was indeed the case as evidenced by the finding of only monomeric rIDE in the deletion mutant preparation.
The monomeric rIDE variant exhibits 25% of the catalytic activity of wild type rIDE and no longer exhibits substrate induced homotrophic activation with Abz-GGFLRKHGQ-EDDnp. Substrate induced cooperativity or homotrophic activation, likely results from binding of the substrate at a site distinct from the catalytic site that is part of an extended substrate-binding site (Nicholas Noinaj, Sonia K. Bhasin, Eun Suk Song, Kirsten Scoggin, Louis B. Hersh, and David W. Rodgers, manuscript in preparation). Binding at this region of the extended substrate binding site is suggested to produce a conformational change in the adjacent subunit. Thus the absence of substrate induced cooperativity as shown by change from a sigmoidal substrate versus velocity response seen with wild type IDE to a hyperbolic response seen in monomeric IDE can be explained by the loss of subunit-subunit interactions.
In contrast to the relatively small decrease in k cat observed with Abz-GGFLRKHGQ-EDDnp as substrate, the monomeric rIDE variant exhibits a 100 to 200-fold decrease in the rate of cleavage of the larger physiological peptides b-endorphin and amyloid b peptide. Larger peptides like b-endorphin and amyloid b peptide traverse the protein and bind both at the active site as well as at a distal site [13]. Since a portion of amyloid b peptide (and likely bendorphin) occupies the distal site we suggest these substrates ''selfactivate'' producing the same or similar conformational change in the adjacent subunit and are thus insensitive to heterotrophic activation.
The crystallographic studies of Tang and coworkers [13] show that IDE totally engulfs the bound peptide in the Michaelis complex, with no room for products to dissociate. Thus the rate-determining step must involve a conformational change that opens the enzyme to permit product dissociation. The increase in k cat produced by activators with small synthetic substrates would therefore involve an increase in the rate of enzyme opening produced by the binding of substrate to one subunit and transmitting a conformational change to the adjacent subunit. Clearly this cannot occur in monomeric IDE, leading to a decrease in kcat.
With the larger substrates b-endorphin and amyloid b peptide the extended binding interactions would be expected to make the conformational change needed for product release energetically more difficult than with small substrates. We suggest that one possibility is that hydrolysis of the substrate in the active site of one oligomeric subunit produces the driving force that leads to a conformational change transmitted to the adjacent subunit. This conformational change increases the rate of enzyme opening and product release. In the monomeric enzyme the absence of an induced conformational change would be a major contributing factor to the slow rate of cleavage of physiological peptide substrates. An alternative hypothesis is that substrate binding in itself is sufficient to translate a conformation change from one monomeric unit to the other which shifts the equilibrium of one subunit to the open conformation and promotes product release.  The observation that both Ab 1-40 and b-endorphin are cleaved at unique sites by monomeric rIDE further suggests that substrate binding is different in the monomeric enzyme. Thus interactions between subunits in the oligomeric form of IDE likely contribute to the conformation of the enzyme-substrate complex.
It should be noted that we can not rule out a mechanism in which dimer contacts are required for activation to occur within a subunit, although previous studies [11] with mixed dimers containing one catalytically active and one catalytically inactive subunit weigh against this alternative. It is also possible that loss of the C-terminal 18 residues directly affects the properties of the enzyme independent of the effect on oligomerization. Given the tenuous attachment of these residues to the rest of the enzyme, this possibility seems unlikely.
We find that in contrast to wild type IDE, the activity of monomeric IDE is not increased by small peptides or by polyphosphates. However binding of the nucleotide triphosphate TNP-ATP does occur as shown by an increase in its fluorescence in the presence of monomeric IDE. Thus the absence of activation of monomeric IDE in the presence of binding can be accounted for by the necessity of a conformational change induced by activator binding to one subunit and transmitted to the adjacent subunit. Alternatively or additionally the oligomeric form of IDE might be required for proper activator binding. That this may be a contributing factor is indicated by the observation of a smaller shift of the fluorescence maxima of bound TNP-ATP with monomeric IDE compared to oligomeric IDE.

Preparation of an IDE C-terminal Deletion Mutant
For construction of an IDE mutant with the C-terminal 18 amino acids deleted PCR was used to generate a 360 bp fragment containing a stop codon, which replaced a 415 bp Pst I -Xho I fragment from the rIDE cDNA in pFast Bac HTb.

IDE Expression and Purification
Native rIDE and its C-terminal deletion mutant, rIDE DC were expressed in Sf9 insect cells and purified as hexahistidine fusion proteins on HIS-select Ni-NTA agarose (Sigma) as previously described [16,18]. This procedure generally yielded homogeneous native rIDE, however in those instances when contaminants were detected, anion exchange chromatography on a 1 ml MonoQ