Overview of structure and function of IDE.

(A) Overall structure of IDE. IDE is comprised of four structurally homologous domains: D1 (red, residues 43-285), D2 (orange, residues 286-515), D3 (blue, residues 545-768), and D4 (cyan, residues 769-1,016). The domains are arranged in two roughly hemispherical regions that enclose the catalytic chamber referred to as IDE-N (comprised of the D1 and D2 domains) and IDE-C (comprising the D3 and D4 domains), which are joined by a linker region (residues 516-544). (B) Primary sequence and overall structure of insulin. Insulin cleavage by IDE as revealed by mass spectrometry (Manolopoulou et al., 2009). Cleavage sites are marked in arrows, which occur within α-helical regions and requires the peptide to be unfolded prior to cleavage. (C) Overview of the IDE catalytic cycle with key questions addressed in this study.

Cryo-EM structures.

(A) Overview of the cryo-EM structures. See figure S1 for processing details. (B) Comparison of the open (O), partial open (pO), and partial closed (pC) subunit states present in our cryo-EM structures with domain organization. The distance between the D1 and D4 domain centers-of-mass (D1-D4 COM) along with the dihedral angle formed by the D1-D2-D3-D4 domain centers-of-mass (D1-D2-D3-D4 dihedral) described in Zhang et al. (21) and depicted in Figure 3 – figure supplement 1 were used as biologically important criteria to quantify observed conformations. (C) Insulin density and corresponding model in our cryo-EM structures. Both the A chain (magenta) and B chain (yellow) can fit the density in the exosite and catalytic cleft.

Conformational dynamics of IDE implied by structural heterogeneity.

(A) All-atom MD simulations analysis. The primary source of structural variance (RMSD) results from the IDE-N moving against IDE-C as a rigid body. Rigid bodies were defined as colored for multibody refinement in RELION. (B-C) Multibody analysis. The range of conformational variance described by the top principal component vectors displays an unexpectedly high degree of rotational motion, as measured by the change in D1-D2-D3-D4 dihedral angle across each vector’s gradient of structural heterogeneity, in both the absence (B) and presence (C) of insulin compared to the expected open-close transition pathway predicted from a linear interpolation of the experimentally determined structures of IDE (dashed line, Figure 3 – figure supplement 1). Two dominant components of structural variance are revealed from multibody analysis: (D) where IDE-N swings relative to IDE-N about the inter-domain linker, and (E) where IDE-N rotates against IDE-C. Starting (orange) and ending (red) states of IDE-N shown with pathway depicted by arrows. IDE-C shown as gray surface.

All-atom MD reveals a molecular basis for IDE conformational dynamics.

(A) Measurements of the O subunit D1-D4 distance over the course of six separate microsecond long all-atom MD simulations of WT IDE. Of which, the open subunit closed in 5 of the 6 simulations. (B) Plot of the O subunit D1-D4 distance vs the D1-D2-D3-D4 COM dihedral angle over the course of the simulation of WT IDE. The open subunits displayed a variety of closing pathways and did not close to a consensus structure. Starting structure shown as black dot, pO structure shown as white dot. (C) R668 acts as a guidepost residue, rapidly interacting with D309 or E381. Formation of this interaction is associated with rapid closing, as measured by a decrease in D1-D4 distance (D). (E) Hydrogen-deuterium exchange mass spectrometry highlights the importance of R668 in mediating the open-close transition. In the presence of insulin (panel 1, red), Aβ (panel 2, red), and BDM-44768 (panel 3, red), all of which promote IDE closing, the peptide containing R668 shows reduced deuterium exchange relative to apo-IDE (black), yet in the presence of 6bk (panel 4, red), which does not promote closing, there is no difference in the exchange rates for the R668 containing peptide relative to apo-IDE (black). Helix containing R668 colored by red (increase) – white (no change) – blue (decrease) gradient depicting the degree of deuterium exchange relative to apo-IDE. (F) Measurements of the O subunit D1-D4 distance over the course of six separate microsecond long all-atom MD simulations of IDE R668A. (G) Plot of the O subunit D1-D4 distance vs the D1-D2-D3-D4 COM dihedral angle over the course of the simulation of IDE R668A. The six separate microsecond long simulations indicate that an R668A mutation significantly alters the closing dynamics of IDE (F) and increases the rotational motion (G) relative to WT (panels A and B respectively). Starting structure shown as black dot, pO structure shown as white dot.

R668A alters IDE activity in vitro

(A) Elution profile of WT IDE (blue) compared to the R668A mutant (orange) from a S200 SEC column. (B) Degradation of the fluorescent substrate MCA-RPPGFSAFK(Dnp) by WT IDE and the R668A construct in the presence and absence of ATP. Data represents the average initial velocities of three replicates performed at a protein concentration of 3.125 nM. Error bars (gray) represent the standard error. (C) Inhibition of MCA-RPPGFSAFK(Dnp) degradation by WT IDE (circles, solid fit lines) and IDE R668A (squares, dashed fit lines) in the presence of varying amounts of insulin. Data was fit to the Michaelis-Menten (black) and Hill equations (red). Relevant parameters, Michaelis-Menten: WT: χ2=0.001, Vmax=0.951, Ki=8.3 nM; R668A: χ2=0.005, Vmax=0.892, Ki=52 nM; Hill: WT: χ2=0.009, n=0.55, Ki=51 nM; R668A: χ2=0.055, n=0.61, Ki=198 nM. Error bars represent standard error, data points represent the average of three replicates. (D) Inhibition of MCA-RPPGFSAFK(Dnp) degradation by WT IDE (circles, solid fit lines) and IDE R668A (squares, dashed fit lines) in the presence of varying amounts of insulin. WT data was fit to the Michaelis-Menten (black) and Hill equations (red), R668A data could not be fit to either equation, indicating that the mutation confers substrate-specific altered enzyme kinetics. Relevant WT parameters, Michaelis-Menten: χ2=0.066, Vmax=0.823, Ki=353 nM; Hill: χ2=0.08, n=0.72, Ki=688 nM. (E) SEC-SAXS profile of WT (black) and R668A (red) constructs with Rg values calculated by both the Guinier and Porod methods along with Dmax derived from the P(r) function (F).

Structural basis of closed state conformational dynamics.

(A) Measurements of the pO subunit D1-D4 distance over the course of six separate microsecond long all-atom MD simulations of WT IDE. (B) Plot of the pO D1-D4 distance vs the D1-D2-D3-D4 COM dihedral angle over the course of the simulation of WT IDE. (C) IDE-N/C interface previously solved crystal structures (PDB:2G47 shown) shows side chains are ill-positioned for interaction. (D) IDE-N/C interface formed upon open subunit closing in our MD simulations reveals a complex hydrogen bonding network. (E) Heat map showing conformational geometries that were preferentially sampled in our MD simulations by the open subunits upon closing. Insets highlight how the IDE-N/C interface changes to permit interdomain motion. (F) Plot of the O subunit D1-D2-D3-D4 dihedral angle during a subset of a single WT IDE MD simulation after the open-close transition has been completed. Charge-swapping between residues at the IDE-N/C interface is associated with changes in the D1-D2-D3-D4 dihedral. (G) For most of the simulation, D309 interacts with K483 (black), however, this interaction is broken for ∼100 ns, during which D309 instead interacts with R311 (blue) and R668 (orange). (H) For most of the simulation, D426 interacts with K571 (black), yet this interaction is periodically broken, and D426 instead interacts with K425 (green) and K899 (magenta). When these events of charge-swapping coincide with D309 charge-swapping (G), they are associated with a large change in the D1-D2-D3-D4 dihedral angle (F). When they occur alone, the effect on D1-D2-D3-D4 dihedral is smaller.

Time-resolved cryo-EM of IDE+insulin reveals a new O/O state.

(A) 5.1 Å reconstruction of IDE+Fab rapidly mixed with a 5x molar excess of insulin and vitrified with a mix-to-freeze time of 123 milliseconds. (B) Measurements (as in Fig. 2) of the distance between the D1-D4 centers-of-mass and dihedral angle formed by the D1-D2-D3-D4 centers-of-mass as indicators of the “openness” of the structure depicted in (A). (C) Alignment of the O/O state from our rapid-mixing dataset (domains colored as in Fig. 1) compared to the previously solved O/O states (gray, as labeled). (D) Overview of the conformational states adopted by IDE in all 4 available cryo-EM datasets with corresponding percentage of particles mapping to each state (Zhang et al., 2018)(this work).

Upside analysis of IDE-insulin interactions.

(A) Transient cross-β-strand interactions between insulin and IDE. Secondary structure (gray: coil, green: β-strand, yellow: α-helix) for each insulin residue (Y-axis) over the course (X-axis) of a representative Upside simulation. Points indicate predicted hydrogen bond interactions between IDE and insulin, colored by which domain of IDE that insulin residue is interacting with (D1: red, D2: orange, D3: blue, D4: cyan). Stars below the X-axis indicate frames from which structures were extracted for panel B. (B) Structure of IDE/insulin complex in Upside simulations at frames indicated in (A). IDE domains colored as in (A), insulin shown in magenta (chain A) and yellow (chain B). Inset highlights the intermolecular cross-β sheets formed between IDE and insulin within each of the four homologous IDE domains. (C) Individual domains of IDE, colored as in (A), extracted from (B) and oriented to demonstrate the structural conservation of intermolecular cross-β-sheet formation with bound insulin (colored as in B).

Model for the catalytic cycle of IDE.

The details are described in discussion. For simplicity, only the scenario involving the addition of a second insulin molecule is illustrated, although this is not required for the catalytic cycle. The mechanism by which IDE transitions between closed and open states remains unknown and is therefore indicated with question marks.