Exchange Broadening Underlies the Enhancement of IDE-Dependent Degradation of Insulin by Anionic Membranes.

Insulin-degrading enzyme (IDE) is an evolutionarily conserved ubiquitous zinc metalloprotease implicated in the efficient degradation of insulin monomer. However, IDE also degrades monomers of amyloidogenic peptides associated with disease, complicating the development of IDE inhibitors. In this work, we investigated the effects of the lipid composition of membranes on the IDE-dependent degradation of insulin. Kinetic analysis based on chromatography and insulin's helical circular dichroic signal showed that the presence of anionic lipids in membranes enhances IDE's activity toward insulin. Using NMR spectroscopy, we discovered that exchange broadening underlies the enhancement of IDE's activity. These findings, together with the adverse effects of anionic membranes in the self-assembly of IDE's amyloidogenic substrates, suggest that the lipid composition of membranes is a key determinant of IDE's ability to balance the levels of its physiologically and pathologically relevant substrates and achieve proteostasis.

Introduction

Insulin-degrading enzyme (IDE) is a ubiquitous zinc metalloprotease that has been implicated in the regulation of the steady-state levels of physiologically and pathologically important proteins, making it an appealing target for the development of therapeutic strategies for two common late-onset diseases: type 2 diabetes (T2D) and Alzheimer’s disease (AD).[1−3] IDE’s substrates include key metabolic hormones (e.g., insulin and glucagon) and amyloidogenic peptides (e.g., islet amyloid polypeptide (IAPP) linked to T2D[4] and Aβ42 associated with AD[5]). Of this diverse array of substrates, kinetic studies have shown that IDE is highly effective at degrading insulin.[6,7]

Since its discovery in 1949,[8] several key structural features of IDE have been identified. IDE is composed of an N-terminal domain (IDE-N) and a C-terminal domain (IDE-C) (Figure ) that come together to form a catalytic chamber, the volume of which limits the substrate to a monomer less than 80 residues in size.[9,10] The enzyme exists in two major conformational states during its catalytic cycle:[1,11] open-state IDE facilitates substrate capture and product release, and closed-state IDE allows proteolysis to take place. In addition to size, specificity is also provided by specific interactions between the substrate and IDE’s exosite located ∼30 Å away from the catalytic Zn2+ ion.[7,10,12] Allosteric regulation of IDE’s activity has been shown to be driven by endogenous molecules, including ATP,[13,14] carnosine,[15] dynorphin,[16] somatostatin,[17] and bradykinin.[16] Kurochkin et al. showed that allosteric mutagenesis can be an attractive strategy for increasing the activity of IDE toward Aβ.[18] More recently, we showed that resveratrol sustains IDE’s activity toward Aβ42[19] but has no effect on the enzyme’s ability to degrade insulin,[20] suggesting that resveratrol is a substrate-selective activator of IDE. In spite of the structural and mechanistic advances discussed above, however, outstanding challenges remain in current efforts to develop IDE-centric therapeutics for T2D and AD. Importantly, regulatory processes that govern IDE’s ability to degrade its monomeric substrates in physiological and pathological settings are not understood.

Figure 1

Closed conformational state of the insulin-degrading enzyme (PDB ID: 2WBY). IDE is composed of an N-terminal half (IDE-N) and a C-terminal half (IDE-C) joined together by an unstructured linker (magenta). When in its closed conformation, IDE forms a catalytic chamber that can only accommodate small monomeric substrates such as insulin and Aβ42. IDE-N contains a conserved exosite (yellow), which has been hypothesized to anchor the substrate prior to degradation. IDE has the HXXEH motif, which contains the two histidine residues (H108 and H112) that coordinate Zn2+ and the catalytically important glutamate residue (E111). The arrow in the lower half of IDE-N indicates the relative location of the catalytic Zn2+ ion.

The subcellular localization of IDE in vivo is not well known.[21] Nonetheless, in vitro studies have shown that the enzyme is localized primarily in the cytosol, as discussed in recent reviews.[1,3] IDE has also been found to be associated with assemblies that contain lipid bilayers, including the cell membrane[22−26] and membrane-enclosed organelles, including peroxisomes[27−29] and endosomes.[30] IDE may also be present in the extracellular space in association with exosomes[31−34] that also contain lipid bilayers.[35] The presence of IDE in the extracellular space seems to be well supported by several in vitro proteolysis studies. Selkoe and co-workers[36−38] and Hersh and colleagues[39] identified IDE as the primary protease responsible for the degradation of Aβ secreted from neuronal and non-neuronal cells.

Insulin is a hormone that is essential for the metabolism of glucose.[40] It is a predominantly α-helical protein that is stored as a hexamer in β-cell secretory granules, presumably to protect it from unregulated or unwanted degradation.[41] The key steps in the physiological journey of insulin in the body are:[42,43] (a) secretion from β cells; (b) clearance in the liver where 75–80% of secreted insulin is cleared; (c) distribution to target tissues where it promotes glucose uptake; and (d) degradation in the kidney. The functional form of the protein is a monomer that initiates its function by binding to its membrane receptor.[44−46] Interestingly, the binding of insulin to its receptor is also the initial step in its clearance by the liver.[47−49] Two clearance pathways have been proposed. In the extracellular pathway, insulin is degraded by IDE without internalization of the substrate.[26,50,51] In the intracellular pathway, some of the receptor-bound insulin is shunted to the endolysosomal system for degradation,[52,53] where IDE degrades insulin in the neutral environment of early endosomes.[30,54]

Regulatory mechanisms that underlie the dissociation of insulin monomer from hexameric insulin—important for its function and degradation—are not well understood. Here, we show that membranes that contain anionic lipids significantly enhance the IDE-dependent degradation of insulin through a mechanism that involves increased chemical exchange between insulin oligomers and degradable insulin monomer.

Results and Discussion

Recombinant human IDE was expressed and purified in-house, as previously described.[7,19,55] Small unilamellar vesicles (SUVs) of varying lipid compositions, including 100% zwitterionic DOPC (1,2-dioleoyl-sn-glycero-3-phosphocholine), 100% anionic DOPS (1,2-dioleoyl-sn-glycero-3-phospho-l-serine), and DOPC/DOPS (7:3 mol/mol), which mimics the ratio of zwitterionic to anionic lipids found in cells relevant to metabolism including insulin-producing β cells,[56] were also prepared. We use SUVs as model membranes in our in vitro studies of the biophysical and biochemical properties of amyloidogenic proteins such as insulin and Aβ because of their high membrane curvatures that facilitate biologically relevant membrane-induced protein conformational transitions.[57] Insulin solutions were prepared at pH 7.4. At this pH and in the presence of Zn2+, it is well known that insulin self-assembles to hexamers.[58,59] In our hands, we found that the dominant oligomeric state of insulin depends on its concentration. Analytical gel filtration indicates that as the concentration of insulin was increased from 15 to 170 μM, the dominant oligomeric state of insulin changes from dimers to hexamers (Figure S1).

The IDE-dependent degradation of insulin was carried out at 37 °C. Figure S2 presents time-dependent circular dichroic spectra showing that IDE remains active at 37 °C for 48 h. To quantitatively assess the effect of SUVs on the IDE-dependent degradation of insulin, we used a combination of chromatographic and spectroscopic techniques. Figure S3 presents representative time-dependent chromatograms of quenched insulin digestion reactions in the absence or presence of SUVs. Each data set shows the gradual loss of the insulin peak accompanied by the appearance of peaks at short retention times, consistent with proteolysis of the substrate. The amount of undigested insulin was determined using eq and plotted against digestion time, as shown in Figure S4. We noted that the loss of insulin with digestion time in samples that contain SUVs composed of 100% DOPC is similar to that in control samples which do not contain SUVs. In contrast, insulin in the presence of SUVs containing anionic DOPS was degraded more rapidly in a manner that correlates with the anionic lipid content of the vesicles. Figure S4 shows that after 1 h of digestion, <10 and ∼45% of insulin remains in digests that contain SUVs composed of 100% DOPS and DOPC/DOPS (7:3 mol/mol), respectively. In contrast, almost 60% of insulin remain in samples that contain SUVs composed of 100% DOPC and in samples that do not contain SUVs.

Next, we determined the Michaelis–Menten kinetic constants for the IDE-dependent degradation of insulin using circular dichroism (CD) spectroscopy. Recently, we demonstrated that the dichroic spectrum of insulin is sensitive to the kinetics of its proteolysis by IDE.[7,60] As degradation proceeded, the intensity of insulin’s ellipticity at 222 nm, widely used as a measure of the helical content of proteins,[61,62] decreased with an increase in digestion time.[7] We calculated the amount of digested insulin using eq where [DI] is the amount of digested insulin at time t, [I]0 is the initial amount of undigested insulin, and [θobs(222 nm)] and [θobs(222 nm)]0 are the observed ellipticities at time t and time = 0, respectively. Linear regression analysis of the plot of [DI] against digestion time yielded V0, the initial rate of insulin proteolysis.[7]Figure shows that V0 plotted against insulin concentration resulted in a hyperbolic plot fitted by the Michaelis–Menten equation. Table presents the steady-state kinetic constants KM (Michaelis constant), kcat (turnover number), and kcat/KM obtained from the Michaelis–Menten plots. IDE’s catalytic efficiency, indicated by kcat/KM, in the presence of SUVs composed of 100% DOPC was similar to the control. However, IDE’s efficiency increased by 67 and 94% in the presence of SUVs composed of DOPC/DOPS (7:3 mol/mol) and 100% DOPS, respectively. Together, our kinetic analysis based on chromatography (Figure S3) and insulin’s helical circular dichroic signal (Figure ) show that membranes containing anionic lipids enhance the IDE-dependent degradation of insulin.

Figure 2

IDE-dependent degradation of insulin in the absence or presence of SUVs follows Michaelis–Menten kinetics. Plots of V0 against insulin concentration are hyperbolic. Each data point represents the mean from three kinetic trials, and the error bars are standard deviations. The lines are fits to the Michaelis–Menten equation.

Table 1

Steady-State Kinetic Parameters for the IDE-Dependent Degradation of Insulin in the Absence and Presence of SUVs

SUVs	KM (M)	kcat (s–1)	kcat/KM (M–1 s–1)
none	3.8 ± 0.3 × 10–5	0.027 ± 0.0006	7.2 ± 0.5 × 102
100% DOPC	3.2 ± 0.5 × 10–5	0.027 ± 0.003	8.5 ± 0.5 × 102
100% DOPS	2.1 ± 0.4 × 10–5	0.028 ± 0.001	1.4 ± 0.03 × 103
DOPC/DOPS (7:3 mol/mol)	2.1 ± 0.03 × 10–5	0.026 ± 0.0009	1.2 ± 0.05 × 103

To explain the increased rate of insulin degradation in the presence of anionic lipids, we used solution-state 1H NMR spectroscopy. Insulin solutions at a concentration of 100 μM were prepared in deuterated buffer to prevent H2O-peak-induced distortion of the peaks from insulin. We noted that the aromatic region (i.e., 6.6–7.4 ppm) of the 1H NMR spectra of the insulin digests is most sensitive to degradation. First, we determined the sensitivity of 1H NMR to the kinetics of degradation by IDE. Figure presents portions of the 1H NMR spectra of insulin digests in the absence and presence of SUVs. In the absence of IDE, the peaks in the aromatic regions of the spectra are broad. In the presence of IDE, the broad peaks are increasingly replaced by sharp peaks with digestion time, consistent with the production of insulin fragments that tumble rapidly in solution. The spectra of the digests in the absence of SUVs (Figure A) are similar to the spectra recorded for the samples that contain 100% DOPC SUVs (Figure B). Furthermore, the intensities of the sharp peaks in the 3-h spectra increase in the order of increasing amounts of DOPS, i.e., DOPC < DOPC/DOPS < DOPS, indicating that the degradation of insulin is enhanced by the presence of anionic lipids. Together, our NMR results are consistent with the kinetic results obtained by chromatographic (Figure S3) and circular dichroic spectroscopic methods (Figure and Table ).

Figure 3

IDE-dependent degradation of insulin monitored by NMR. One-dimensional 1H NMR spectra of insulin digestion reactions in (A) the absence of SUVs, and the presence of SUVs composed of (B) 100% DOPC, (C) DOPC/DOPS (7:3 mol/mol), and (D) 100% DOPS. In the absence of IDE, the broadening of the peaks in the aromatic region (6.6–7.4 ppm) in samples that contain membranes with anionic DOPS increased. In the presence of IDE, the broad peaks in the aromatic region are replaced by sharp peaks due to insulin fragments that tumble rapidly in solution. All spectra were recorded at 37 °C. The NMR samples were incubated at 37 °C in between the acquisition of spectra. The concentration of insulin in all samples was set at 100 μM. At this concentration, insulin exists mainly as hexamers. The insulin-to-lipid and the substrate-to-enzyme molar ratios were set at 1:50 and 100:1, respectively.

Next, we analyzed the spectra of insulin in the absence of IDE to decipher the mechanism for the enhancement of degradation by anionic lipids. At a concentration of 100 μM, insulin in the presence of Zn2+ and at neutral pH exists mainly as a mixture of monomers and oligomers that are predominantly hexamers (Figure S1), consistent with results obtained by others using dynamic light scattering,[63] size-exclusion chromatography,[64] and analytical ultracentrifugation.[65] Because insulin at a concentration of 100 μM is degradable by IDE (Figure ) and the enzyme only degrades monomeric substrates,[9,10] insulin monomer must be in dynamic exchange with insulin oligomers (Figure ). Insulin monomer thus samples two environments: one in which it is in association with other molecules of itself and the other in which it is by itself. In NMR spectroscopy, this is known as chemical exchange,[66] also known as magnetic site exchange.[67] Chemical exchange is characterized by an exchange rate, denoted by kex (Figure ). The relative values of kex and Δω, where Δω is the difference in frequency between the two sites, define the slow and fast exchange limits of chemical exchange:

Figure 4

Chemical exchange in insulin. Oligomeric insulin and insulin monomer are in dynamic equilibrium with one another. This equilibrium is characterized by the exchange rate kex. The distribution of oligomers is indicated by n, which ranges from 2 to 6. At the concentration of insulin used in the NMR studies, hexamers are the dominant oligomers.

Because the peaks in the aromatic region of the 1D 1H NMR spectra of insulin in the absence of IDE are broad (Figure ), and sharp peaks expected for a protein of the size of the insulin monomer, i.e., ∼5.8 kDa were not detected, the exchange between insulin oligomers and insulin monomers must be in the intermediate timescale. This conclusion is also supported by 2D NOESY NMR of insulin in the absence of IDE (Figure S5), which shows the absence of cross peaks due to insulin, in sharp contrast to NOESY spectra we reported for insulin at pH 2 in the absence of SUVs.[68] Additionally, our NOESY data (Figure S5) indicate that multidimensional relaxation-based NMR methods that were recently used to investigate the interaction of Aβ42 with IDE do not apply.[69] In the presence of SUVs containing anionic lipids and in the absence of IDE, the broadening of the peaks in the aromatic regions of the spectra shown in Figure C,D increased. This is exchange broadening[70] that results from an increase in the exchange rate between insulin oligomers and insulin monomer. In turn, the increase in exchange rate leads to increased IDE-dependent degradation of insulin monomer.

Our work shows that the lipid composition of membranes is an efficient regulator of the dissociation of insulin monomer from oligomeric insulin and, thus, of the ensuing IDE-dependent degradation of insulin monomer. When anionic lipids are present, the exchange rate between insulin oligomer and insulin monomer increases, leading to increased availability of insulin monomer for degradation by IDE. Interestingly, levels of anionic phospholipids in islet cells increase significantly after glucose stimulation.[71] We speculate that glucose stimulation in vivo also leads to an increase in anionic-phospholipid levels in β-cell membranes, resulting in increased release of insulin monomer.

Finally, in light of IDE’s ability to degrade insulin, Aβ42, and IAPP, it is no surprise that there has been a strong interest in the development of small molecules that modulate IDE’s activity.[1−3,72] Several inhibitors have been reported to elevate insulin levels in cells or mice, including BMD44768,[73] B35,[74] 6bk,[75] Ii1,[76] and P12-3A.[77] In spite of these significant advances, outstanding challenges remain.[1−3,72] Most importantly, the inhibition of IDE may lead to increased levels of Aβ42 and IAPP. Attractive strategies in response to this challenge include development of substrate-selective IDE inhibitors,[78] identification of allosteric ligands,[79] and allosteric mutagenesis of IDE.[18] These strategies require consideration of the mechanisms that regulate the levels of the monomeric states of IDE’s substrates. In the case of insulin, we have shown here that anionic lipids in membranes increase the levels of IDE-degradable insulin monomer. However, this finding does not apply to all of IDE’s substrates. The self-assembly of IDE’s amyloidogenic substrates has been shown to be enhanced by anionic lipids in membranes.[4,57,80,81] Zhang et al. showed that even low levels of anionic lipids promote the aggregation of cationic IAPP and facilitate IAPP-induced leakage of membranes.[82] Clusters of anionic GM1 gangliosides in membranes accelerate Aβ aggregation to form assemblies with increased cytotoxicity.[83] Aggregation is preceded by binding of Aβ to anionic lipids, which is mediated by cationic lysine residues (K16 and K28).[84,85] The differential effects of anionic lipids in membranes in modulating the levels of the monomeric states of IDE’s substrates indicate that the lipid composition of membranes is a key determinant of IDE’s ability to balance the levels of its physiologically and pathologically relevant substrates. This finding should be considered in the development of IDE-centric therapeutic strategies for T2D and AD.

Experimental Methods

Materials

Phospholipids, including 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC) (>99% pure) and 1,2-dioleoyl-sn-glycero-3-phospho-l-serine (DOPS) (>99% pure) were purchased from Avanti Polar Lipids, Inc. (Alabaster, Alabama). Human insulin (99% pure, 0.4% zinc) was purchased from Sigma-Aldrich (St. Louis). Stock solutions of insulin were prepared in 10 mM phosphate buffer (pH 7.4). To overcome the low solubility of insulin at pH 7.4, the solutions were incubated overnight at 37 °C, followed by centrifugation for 1 min at 16 000g at room temperature to separate undissolved insulin. The concentration of insulin was determined by UV absorbance at 275 nm using a molar extinction coefficient of 6190 M–1 cm–1.[68]

Analytical Gel Filtration Chromatography

Five insulin solutions with concentrations of 500, 300, 200, 100, 50, and 25 μM in 50 mM Tris at pH 7.4 were prepared. To perform the chromatography, a precalibrated Superdex 75 HR 10/300 GL column connected to an ÄKTA pure FPLC system was used. The concentration of the fractions containing the sample that yielded maximum absorbance at 275 nm was determined by UV spectroscopy, as described above.

Preparation of Small Unilamellar Vesicles

SUVs containing 100% DOPC, 100% DOPS, or DOPC/DOPS (7:3 mol/mol) were prepared as previously described.[57] Briefly, lipids were dissolved in methanol/chloroform (1:1, v/v). After drying, the lipid films were hydrated with 1 mL of 10 mM phosphate buffer (pH 7.4). The resulting suspension of multilamellar vesicles was then subjected to 10 freeze–thaw cycles to increase their size homogeneity. This was followed by sonication using a VCX750 Vibra-Cell ultrasonic liquid processor equipped with a tapered microtip (Sonics and Materials, Inc., Newtown, CT) to produce SUVs. The SUVs were separated from titanium particles by centrifugation for 15 min at 16 000g at room temperature.

IDE Overexpression and Purification

Recombinant human IDE was expressed and purified as previously described.[7,19,55] Briefly, glutathione-S-transferase tagged IDE (GST-IDE) was overexpressed in E. coli BL21-CodonPlus RIL cells. After cell lysis, GST-IDE was separated using a GSTrap Fast Flow column connected to an ÄKTA pure FPLC system and eluted with phosphate-buffered saline containing 10 mM glutathione. GST PreScission protease was then used to cleave the GST tag. IDE was further purified using standard gel filtration.[7,19,55] The concentration of IDE was determined by UV absorbance at 280 nm using its molar extinction coefficient of 113 570 M–1 cm–1.[86]

IDE-Dependent Degradation of Insulin Monitored by Reversed-Phase HPLC

Four proteolysis samples, each with a total volume of 300 μL, were prepared in triplicates. These samples included (1) insulin in the absence of SUVs; (2) insulin in the presence of 100% DOPC SUVs; (3) insulin in the presence of 100% DOPS SUVs; and (4) insulin in the presence of DOPC/DOPS (7:3 mol/mol) SUVs. The concentration of insulin was set at 40 μM. The ratio of insulin to lipid was set at 1:50 (mol/mol). The substrate-to-enzyme ratio was set at 100:1 (mol/mol). IDE was added last to initiate the reaction, followed by incubation of the reaction solutions at 37 °C. At the 1-min, 1-h, 3-h, 6-h, 24-h, and 48-h time points of proteolysis, 50 μL of the reaction solution was removed and transferred into an Eppendorf tube containing 23 μL of 1% (v/v) trifluoroacetic acid in water to quench the reaction.

All quenched samples were analyzed using a Varian ProStar 210 HPLC system equipped with a ProStar 325 Variable Wavelength UV–Visible Detector. Fractionation of insulin digests was carried out at room temperature using an Agilent AdvanceBio mAb C4 column. Solvent A was 0.1% (v/v) formic acid in H2O, whereas solvent B was 0.1% (v/v) formic acid in acetonitrile. Water used for the mobile phase was obtained by a Millipore Milli-Q 185 Plus system (Millipore, Bedford, MA). Acetonitrile of chromatographic grade was supplied by Fisher Scientific. Aliquots (20 μL) of solutions of quenched insulin digests were injected into the HPLC manually and eluted with a 15-min linear gradient of 0–100% B at a flow rate of 1 mL/min. Elution of analytes was monitored by UV absorbance at wavelengths 214 and 254 nm. By utilizing the integrals of the insulin peak, the amount of undigested insulin remaining at a specific time point of the proteolysis reaction was calculated using eq and plotted against digestion time.

Kinetics of IDE-Dependent Degradation of Insulin by Circular Dichroism Spectroscopy

To obtain the steady-state kinetic parameters for the IDE-dependent degradation of insulin in the presence of SUVs, we used a kinetic assay that takes advantage of the loss of insulin’s helical circular dichroic signal at 222 nm with digestion time.[7] Briefly, seven insulin solutions in 10 mM phosphate buffer (pH 7.4) with concentrations ranging from 15 to 110 μM (i.e., 15, 20, 25, 30, 50, 80, and 110 μM) were prepared. The insulin solutions were prepared in the absence and presence of SUVs with the insulin/lipid ratio set at 1:50 (mol/mol). The reaction was initiated with the addition of IDE at a concentration of 1 μM. After mixing, the solution was transferred into a quartz cuvette with a path length of 1 mm and loaded into the sample holder of our JASCO J-815 spectropolarimeter set at 37 °C. The ellipticity at 222 nm ([θobs(222 nm)]) was then recorded for 5 min. The real-time [θobs(222 nm)] data were then used to calculate the amount of digested insulin ([DI]) using eq . Linear regression analysis of plots of [DI] against digestion time (up to 5 min) yielded V0, the initial rate or velocity of the IDE-catalyzed degradation of insulin. Michaelis–Menten plots were then generated, from which the kinetic constants KM, Vmax, kcat, and kcat/KM were determined by curve-fitting to the Michaelis–Menten equation (eq )

IDE-Dependent Degradation of Insulin Monitored by 1H NMR Spectroscopy

Stock solutions of insulin and SUVs in 10 mM phosphate buffer were prepared using 99.9% D2O (Sigma-Aldrich). The concentration of insulin in all NMR samples was set at 100 μM. In samples that contain SUVs, the insulin/lipid ratio was set at 1:50 (mol/mol). The substrate/enzyme ratio was set at 100:1 (mol/mol). All 1D and 2D 1H NMR spectra were recorded at 37 °C using a Varian INOVA spectrometer operating at 400 MHz. For chemical shift referencing, the methyl peak of the internal standard 2,2-dimethyl-2-silapentane-5-sulfonate was set to 0 ppm. NOESY spectra were recorded in phase-sensitive mode using mixing times ranging from 100 to 400 ms. All samples were incubated at 37 °C in between spectral acquisitions.