The Dynamics and Infrared Spectrocopy of Monomeric and Dimeric Wild Type and Mutant Insulin
TThe Dynamics and Infrared Spectrocopy ofMonomeric and Dimeric Wild Type and MutantInsulin
Seyedeh Maryam Salehi, Debasish Koner and Markus Meuwly ∗ Department of Chemistry, University of Basel, Klingelbergstrasse 80 , CH-4056 Basel,Switzerland.
E-mail: [email protected] a r X i v : . [ phy s i c s . b i o - ph ] S e p eptember 7, 2020 Abstract
The infrared spectroscopy and dynamics of -CO labels in wild type and mutantinsulin monomer and dimer are characterized from molecular dynamics simulations us-ing validated force fields. It is found that the spectroscopy of monomeric and dimericforms in the region of the amide-I vibration differs for residues B24-B26 and D24-D26,which are involved in dimerization of the hormone. Also, the spectroscopic signa-tures change for mutations at position B24 from phenylalanine - which is conserved inmany organisms and known to play a central role in insulin aggregation - to alanineor glycine. Using three different methods to determine the frequency trajectories -solving the nuclear Schr¨odinger equation on an effective 1-dimensional potential en-ergy curve, instantaneous normal modes, and using parametrized frequency maps -lead to the same overall conclusions. The spectroscopic response of monomeric WTand mutant insulin differs from that of their respective dimers and the spectroscopy ofthe two monomers in the dimer is also not identical. For the WT and F24A and F24Gmonomers spectroscopic shifts are found to be ∼
20 cm − for residues (B24 to B26)located at the dimerization interface. Although the crystal structure of the dimer isthat of a symmetric homodimer, dynamically the two monomers are not equivalent onthe nanosecond time scale. Together with earlier work on the thermodynamic stabilityof the WT and the same mutants it is concluded that combining computational andexperimental infrared spectroscopy provides a potentially powerful way to characterizethe aggregation state and dimerization energy of modified insulins. Insulin is a small, aggregating protein with an essential role in regulating glucose uptake incells. Physiologically, it binds to the insulin receptor (IR) in its monomeric form but thermo-dynamically the dimer is more stable for the wild type (WT) protein.
The storage form2s that of a zinc-bound hexamer with either two or four Zn atoms. Hence, to arrive at thefunctionally relevant monomeric stage, insulin has to cycle through at least two dissociationsteps: from the hexamer to three dimers and from the dimer to the monomer.For pharmacological applications the dimer ↔ monomer equilibrium is particularly relevantbecause for safe insulin administration this equilibrium needs to be tightly controlled. How-ever, reliable experimental physico-chemical information about the relative stabilization ofinsulin monomer and dimer, which is − . is only availablefor the WT and the barrier between the two states is unknown. For mutant insulins, there isno such quantitative information from experiments. On the other hand, insulin has becomea paradigm for studying coupled folding and binding, whether or not association proceedsalong one or multiple pathways, and for the role of water in protein association. Mostof these studies were based on atomistic molecular dynamics (MD) simulations and providedremarkable insight into functionally relevant processes for this important system.Infrared spectroscopy has been proposed and recently demonstrated to provide a way toquantify protein-ligand binding strengths through observation of spectroscopic shifts. Thephysical foundation for this is the Stark effect which is based on the electrostatic interactionbetween a local reporter and the electric field generated by its environment. Using accuratemultipolar force fields it was possible to assign the structural substates in photodissoci-ated CO from Myoglobin whereas more standard, point charge-based force fields are notsuitable for such investigations. The frequency trajectory of a local reporter can be followed in different ways. One of themuses so-called parametrized “frequency maps” which are precomputed for a given reporterfrom a large number of ab initio calculations.
Alternatively, the sampling of the con-figurations and computing frequencies for given snapshots can also be done using the same3nergy function (“scan”). In this approach, the MD simulations are carried out with thesame energy function that is also used for the analysis, which is typically a multipolar repre-sentation for the electrostatics around the spectroscopic probe and an anharmonic (Morse)for the bonded terms.
On each snapshot, the local frequency is determined from eitheran instantaneous normal mode (INM) calculation or by solving the 1D or three-dimensionalnuclear Schr¨odinger equation. Here, the WT proteins and two mutants at position B24 (Phe) are considered. Phenylala-nine B24 is located at the dimerization interface and invariant among insulin sequences. Compared with the WT, the SerB24,
LeuB24, and HisB24 analogues show reducedbinding potency towards the receptor. On the other hand, substitutions such as GlyB24,D-AlaB24, or D-HisB24 are well tolerated as judged from their binding affinity. Nevertheless,substitutions such as GlyB24 (F24G) or AlaB24 (F24A) were found to have reduced stabilityof the modified insulin dimer, both from simulations and experiment, and these are thevariants considered in the present work.In the present work the infrared spectrum in the amide-I stretch region is studied for wildtype (WT) and two mutant insulins in their monomeric and dimeric states using accuratemultipolar force fields. The IR lineshapes are calculated from frequency trajectories calcu-lated by using a normal mode analysis, solving the Schr¨odinger equation from a 1-d scanalong the amide-I normal mode and using previously parametrized maps. First, the meth-ods are presented. Then, results for IR lineshapes and frequency correlation functions fromscanning along the amide-I normal mode are presented and discussed and compared withthe two other approaches. Finally, conclusions are drawn.4 Methods
All molecular dynamics (MD) simulations were carried out using the CHARMM packagetogether with CHARMM36 force field including the CMAP correction and multipolesup to quadrupole on the [CONH]-part of the backbone. The X-ray crystal structure ofthe insulin dimer was solvated in a cubic box (75 ˚A ) of TIP3P water molecules, whichleads to a total system size of 40054 atoms. For the monomer simulations, chains A and Bwere retained and also solvated in a water box (75 ˚A ), the same box size as the dimer. Inthese simulations the multipolar force field is used for the entire amide groups andall CO bonds are treated with a Morse potential V ( r ) = D e (1 − exp( − β ( r − r e ))) . Theparameters are D e = 141 .
666 kcal/mol, β = 2 .
112 ˚A − and r = 1 .
231 ˚A.Hydrogen atoms were included and the structures of all systems were minimized using 2000steps of steepest descent (SD) and 200 steps of Newton Raphson (ABNR) followed by 20ps of equilibration MD at 300 K. A Velocity Verlet integrator and Nos´e-Hoover thermo-stat were employed in the N V T simulations. Then production runs (1 ns or 5 ns) werecarried out in the
N pT ensemble, with coordinates saved every 10 fs for subsequent analysis.For the
N pT simulations an Andersen and Nos´e-Hoover constant pressure and tempera-ture algorithm was used together with a leapfrog integrator. a coupling strength forthe thermostat of 5 ps and a damping coefficient of 5 ps − . All bonds involving hydrogenatoms were constrained using SHAKE. Nonbonded interactions were treated with a switch-ing function between 10 and 14 ˚A and for the electrostatic interactions, the Particle MeshEwald (PME) method was used with grid size spacing of 1 ˚A, characteristic reciprocal length κ = 0 .
32 ˚A − , and interpolation order 4. Figure 1A shows the insulin dimer highlightingsome of the CO labels studied in the current work with particular attention to the -CO labelsat the protein-protein interface (B24-B26) and (D24-D26).5igure 1: Panel A: Structure of wild type insulin dimer with the -CO labels that are specif-ically probed in the present work. The dimerization interface involves residues B24-B26and D24-D26. Panels B and C show the displacement vectors for the two scan approachesconsidered to construct 1D potentials along the CO and CONH directions, respectively.
Anharmonic transition frequencies can be determined from calculating the 1-d potential en-ergy along the CO or amide-I normal mode (from a normal mode analysis on N-methylacetamide (NMA) in the gas phase) and solving the nuclear Schr¨odinger equation (SE) foreach snapshot using a discrete variable representation (DVR) approach. It was shown pre-viously for NMA that frequency trajectories obtained from solving the SE on the 1-d PESscanned along either the CONH (amide-I) or the CO mode (see Figure 1B and C) result insimilar decay times with frequencies shifted by some ∼
15 cm − . Here, scans were performedfor each snapshot for 61 points along the CO normal mode vector around the minimum en-ergy structure using the same energy function as that used for the MD simulations, i.e. amultipolar representation of the electrostatics and an anharmonic Morse potential for theCO-bond. An RKHS representation of the 1-d PES is then constructed from these energiesand the SE is solved on a grid ( − .
53 ˚A < r < .
53 ˚A ) using a reduced mass of 1 amu. Fordirect comparison, scans along the amide-I mode were also carried out for selected residues.6 .3 Instantaneous Normal Mode
The instantaneous (harmonic) frequencies for each snapshot of the trajectory from the
N P T simulation were calculated for the same snapshots for which the scan along the CO normalmode was carried out, see above. Such instantaneous normal modes (INM) are determinedby minimizing CO or [CONH] while keeping the environment (protein plus solvent) fixed.Next, normal modes were calculated from the “vibran” facility in CHARMM.
The frequency map used in the present work is that parametrized by Tokmakoff and cowork-ers. It requires MD simulations to be run with fixed CO bond length and is based on theexpression ω i = ω + aE C i + E N i (1)where ω i is the instantaneous frequency for the i th vibrational label, E C i is the electric fieldon the C atom in the i th label along the C=O bond direction, and E N i is that on the N atom.Parameters ω , a , and b were fitted such that they optimally reproduce the experimental IRabsorption spectra of NMAD. The optimized backbone map is ω i = 1677 . . E C i − . E N i (2)In this equation, ω i is in cm − and E C i and E N i are in atomic units.7 .5 Frequency Fluctuation Correlation Function and Lineshape From the harmonic or anharmonic frequency trajectory ω i ( t ) or ν i ( t ) for label i its frequencyfluctuation correlation function, h δω (0) δω ( t ) i is computed. Here, δω ( t ) = ω ( t ) − h ω ( t ) i and h ω ( t ) i is the ensemble average of the transition frequency. From the FFCF the line shapefunction g ( t ) = Z t Z τ h δω ( τ ) δω (0) i dτ dτ . (3)is determined within the cumulant approximation. To compute g ( t ), the FFCF is numericallyintegrated using the trapezoidal rule and the 1D-IR spectrum is calculated according to I ( ω ) = 2 < Z ∞ e i ( ω −h ω i ) t e − g ( t ) e − tα T dt (4)where h ω i is the average transition frequency obtained from the distribution, T = 0 . α = 0 . For extracting time information from the FFCF, h δω ( t ) δω (0) i is fitted to an empirical ex-pression h δω ( t ) δω (0) i = a cos( γt ) e − t/τ + n X i =2 a i e − t/τ i + ∆ (5)where a i are amplitudes, τ i are decay times and ∆ is an offset for long correlation times.The cos − term allows to capture a short-time recurrence (anticorrelation) that may or maynot be present in the correlation function. This minimum at very short time ( t ∼ . and can be related to the strength of the interactionbetween solute and solvent or between the spectroscopic probe and its environment(as in the present case). The decay times τ i of the frequency fluctuation correlation functionreflect the characteristic time-scale of the solvent fluctuations to which the solute degreesof freedom are coupled. In most cases the FFCFs were fitted to an expression containing8wo decay times using an automated curve fitting tool from the SciPy library. Only if thequality of the resulting fit was evidently insufficient, a third decay time was included.
The results section is structured as follows. First, a brief account is given of representa-tive structures along the trajectories for the different simulation conditions used. Next, theamide-I spectroscopy for the WT monomer and dimer using the “scan” approach is given.This is followed by the spectroscopy for the mutant monomer and dimer compared with theWT systems. Then, a comparative discussion of the results for WT and mutant monomerand dimer is given for the three methods to determine the frequency trajectories (“scan”,“INM” and “map”) and finally, the FFCFs from the “scan” and “INM” frequency trajecto-ries are discussed.
The root mean squared deviation between the reference X-ray structure and those of themonomer and dimer structure of the WT protein in solution is reported in Figure 2. Typi-cally, the RMSD is around 1.5 ˚A which is indicative of a stable simulation on the nanosecondtime scale. Such RMSD values have also been reported from simulations in smaller waterboxes. R M S D ( Å ) A WT−MOWT−DI
Time (ps) B Figure 2: The structural RMSD between the reference X-ray structure and the Wild typemonomer and dimer insulin for A) flexible and B) constrained CO.With constrained CO (as is required for using the frequency maps) the structure of themonomer is equally well maintained whereas for the dimer it starts to deviate from the ref-erence structure by ∼ To set the stage, the Amide-I spectroscopy for the WT monomer and dimer is discussedfrom frequency trajectories obtained by scanning along the CO normal mode for each snap-shot. Figure 3 reports the lineshapes for all CO-labels for the WT monomer. Lineshapes10or chain A are solid lines and those for chain B are dashed. The overall lineshape for themonomer (black solid line) is centered at 1630.5 cm − and has a full width at half maximumof ∼
30 cm − , compared with a center frequency of ∼ − and a FWHM of ∼
30 cm − from experiments. When comparing the position of the frequency maximum it should benoted that the present parametrization is for NMA and slight readjustments of the Morseparameters could be made to yield quantitative agreement. However, for the present purposesuch a step was deemed unnecessary.On the other hand, scanning the 1-dimensional potential along the amide-I normal modeshifts the frequencies by about 30 cm − to the blue (see Figure S1A). The correlation be-tween scanning along the CO and amide-I normal modes is high, as Figure S1C shows. Inaddition, the full 1D infrared spectrum was also calculated from scanning along the amide-Inormal mode (Figure S2) and confirms the overall shift to the blue by 25 cm − while main-taining the shape and width of the total lineshape from scanning along the CO normal mode.11 I n t e n s i t y ( a . u . ) A A1A2A3A4A5A6A7A8A9A10A11A12A13A14A15A16A17A18A19A20B1B2B3B4B5 B6B7B8B9B10B11B12B13B14B15B16B17B18B19B20B21B22B23B24B25B26B28B29SUM
Frequency (cm -1 ) I n t e n s i t y ( a . u . ) B WT-MOWT-DIWT-DI-M1WT-DI-M2
Figure 3: Panel A: 1D IR spectra for all residues in WT monomer based on “scan” for thefrequency calculation. The labels for the individual line shapes are given in the panel andthe overall sum is the solid black line. Panel B: The total lineshape for all CO probes of themonomer (black) compared with that of M1 (green) and M2 (blue) within the dimer andwith the dimer itself (red). All lineshapes are scaled to the same maximum intensity. Theline shapes are determined from 1 ns simulations and the snapshots analyzed are separatedby 10 fs.Most notably, the center frequencies for each of the labels cover a range from 1612.5 cm − (residues B24, B29) to 1647.5 cm − (residue B11) although the bonded potential (Morse) forthe CO stretch is the same for all 51 labels. Hence, the multipolar charge distribution usedfor the electrostatics and its interaction with the environment leads to the displacements ofthe center frequencies. The linewidths also vary for the -CO probes at the different locationsalong the polypeptide chain and cover a range from 10 cm − (Residues A10, A16, A18, B18,B21) to 28 cm − (Residue A5). 12elected lineshapes for the monomer and each of the two monomers within insulin dimer fromscanning along the CO normal mode are reported in Figures 4 and S3 and the individual andtotal lineshapes for the two monomers (M1 and M2) within the dimer are shown in FiguresS4 and S5. For the dimer it is noted that some probes at symmetry related positions withinthe dimer structure typically have their maxima at different frequencies. In other words,structurally related -CO probes sample different environments in the hydrated system atroom temperature. The overall lineshapes of M1 and M2 are directly compared with that ofthe isolated monomer and the dimer in Figure 3B. The lineshape of M1 and M2 differ whichconfirms the asymmetry noted earlier from X-ray experiments. Also, the spectroscopy ofthe isolated monomer differs from that of M1 and M2 within the dimer. Notably, the -COgroups involved in the hydrogen bonding motif of the insulin dimer (B24 to B26 and D26to D24) display frequency maxima that differ by ∼
10 cm − . Other -CO reporters, such asB20 and D20, have their maxima only ∼ − apart.13 I n t e n s i t y ( a . u . ) A B24B25B26
Frequency (cm -1 ) I n t e n s i t y ( a . u . ) B B24B25B26D24D25D26 -1 ) [WT-MO] 161016151620162516301635164016451650 F r e qu e n cy ( c m - ) [ W T - D I ] WT-MO/DI (A,B)WT-MO/DI (C,D)D23D24 D25D26 D22 D21D20 C6D6 C19,B26B24 B21B20B22A19B23,B25 B6 A6 C Figure 4: 1D IR spectra for WT monomer (panel A) and dimer (panel B) for residues at thedimerization interface (B24-B26) and (B24-B26, D24-D26), respectively, based on “scan” forfrequency calculation. Panel C compares the maximum frequency of the 1D IR spectra forthe selected residues (A6, A19, B6, B20-B26, C6, C19, D6, D20-D26) between WT monomerand dimer.It is also observed that the absolute frequency maximum of the same reporter in the monomerand in the dimer can differ. For example, while the maximum frequency of -CO at positionB24 in the monomer is at 1612.5 cm − the maxima for B24 and D24 in the dimer are at1625.5 cm − and 1620.5 cm − . Hence, in addition to a splitting in the dimer spectrum alsoan overall shift of the frequencies compared with the monomer is found. Again, these effectsare largest for the dimerization motif and for residues A/C6, see Figure 4C.The close agreement of the computed overall spectrum with the experimentally measuredone (see above) and the fact that the same computational model was successful in describingthe spectroscopy and dynamics of hydrated NMA provides a meaningful validation ofthe present approach. Amide-I Spectroscopy of Wild Type and Mutant Monomers:
Mutation at position B24 con-14iderably influences the dimerization behaviour of the hormone. Hence, the dynamics ofthe hydrated F24A and F24G monomers was first considered. The infrared lineshapes forresidues along the dimerization interface and the same selected -CO probes for the WTmonomer are reported in Figure S6. For the two mutant monomers (Figure S6A for F24Aand Figure S6B for F24G) the frequency maximum for -CO at position B24 is shifted from1612.5 cm − (WT) to 1614.5 cm − (F24A) and 1628.5 cm − (F24G), respectively. Theamide-I band maxima at positions B25 and B26 show differences for the the F24A mutantbut not for F24G and for position A19 the frequency maxima shift to the blue (7 cm − ) forF24A and to the red (6 cm − ) for F24G compared to WT. For all other -CO labels in themonomer the differences between F24A and F24G are less than 14 cm − . I n t e n s i t y ( a . u . ) A B24B25B26
Frequency (cm -1 ) I n t e n s i t y ( a . u . ) B B24B25B26 -1 ) [WT-MO] 161016151620162516301635164016451650 F r e qu e n cy ( c m - ) [ B A / B G - M O ] WT/B24A-MOWT/B24G-MOB24 B25B26 B22B21B20 B23B24 B25B26 B22B21B20B23 A6B6A19A19 A6B6 , ,, , C Figure 5: 1D IR spectra for monomeric mutants at position B24. Panels A and B reportspectra for F24A (panel A) and F24G (panel B) for residues (B24-B26) at the dimerizationinterface, based on “scan” for frequency calculations. Panel C compares the maximum fre-quency of the 1D IR spectra for selected residues (A6, A19, B6, B20-B26) between monomericWT and mutants F24A and F24G.A direct comparison of the maxima between the WT and the two mutant monomers is givenin Figure 5C for selected -CO probes, as for WT monomer and dimer (see Figure 4). The15ost pronounced differences in the maximum absorbances occur around the mutation sitewhereas away from it they are minor, except for -CO at position A19. Interestingly, residueTyrA19 is structurally close to PheB24 (see Figure 1A) which explains the dynamical cou-pling between the two sites that leads to a shift of ∼ ± − and is also consistent withrecent work on the stability of B24-mutated insulin. Amide-I Spectroscopy of Wild Type and Mutant Dimers:
The peak frequencies for residuesat the dimerization interface for the WT and the F24A mutant are reported in Figures 6Aand B and directly compared for a larger number of residues, see Figure 6C and S7. Asfor the monomer, there are specific differences such as for TyrA19, PheB25, and PheD25which shift by up to 15 cm − between the two systems. For other residues the differences areconsiderably smaller. For the F24G mutant differences persist, but are in general smaller,see Figure S8. What is found from simulations for both mutants is that residues are notnecessarily symetrically affected, in particular for those along the dimerization interface.Also, depending on the modification at position B24 the effects differ and may allow todistinguish between the different insulin variants.16 I n t e n s i t y ( a . u . ) A B24B25B26D24D25D26
Frequency (cm -1 ) I n t e n s i t y ( a . u . ) B B24B25B26D24D25D26 -1 ) [WT-DI] 161016151620162516301635164016451650 F r e qu e n cy ( c m - ) [ B A - D I ] WT/B24A-DI (A,B)WT/B24A-DI (C,D)B23 C6B25 B26 B22B21B20 A6B6A19 C19D21D23D6D22D24 D20D26D25 , C ,B24 Figure 6: 1D IR spectra for WT (panel A) and the F24A (panel B) dimer for residues atthe dimerization interface (B24-B26, D24-D26), based on “scan” for frequency calculation.Panel C compares the maximum frequency of the 1D IR spectra for selected residues (A6,A19, B6, B20-B26, C6, C19, D6, D20-D26) between the WT and F24A mutant dimer.
The three approaches to determine frequency trajectories considered here (“scan”, “INM”,and “map”) differ considerably in terms of computational expense and the formal approxima-tions in applying them. Scanning along the CO or amide-I normal mode for every snapshotis computationally expensive as it requires for every snapshot to carry out a 1-dimensionalscan of the PES, representing it as a RKHS, and solving the nuclear Schr¨odinger equation.As this needs to be done for ∼ snapshots per nanosecond, such an approach does notscale arbitrarily to larger systems and long time scales ( µ s or longer). Compared to “scan”,determining instantaneous normal modes is computationally less demanding and the “map”approach is also computationally efficient. In the following, the lineshapes from the fre-quency trajectory for the WT monomer using the three methods are compared.17igure 7 reports the 1d lineshapes for all residues of the WT monomer from INM. As for“scan” the maxima of the individual line shapes cover a range between 1625.5 cm − and1657.5 cm − and the average spectra over all individual lineshapes is centered at 1640.5 cm − with a FWHM of 26 cm − , compared with 1630.5 cm − and a FWHM of ∼
30 cm − from“scan”, see Figure 3. A direct comparison of the frequency maxima for the WT monomerfrom “scan” and INM is reported in Figure S9A. Frequency (cm -1 ) I n t e n s i t y ( a . u . ) WT-MO (INM)
A1A2A3A4A5A6A7A8A9A10A11A12A13A14A15A16A17A18A19A20B1B2B3B4B5 B6B7B8B9B10B11B12B13B14B15B16B17B18B19B20B21B22B23B24B25B26B28B29SUM
Figure 7: 1D IR spectra for all residues for the WT monomer from INM for the frequencycalculations. The black line shows the superposition of all CO spectra compared with othersingle CO spectrum.The individual and total lineshapes from using the “map” frequencies are reported in Figure8. Again, the individual frequency maxima span a range of ∼
50 cm − and the FWHM differfor the residues. Contrary to the overall line shape for the monomer from “scan” and “INM”,using the frequency map leads to an infrared spectrum with two peaks. This shape is notconsistent with the experimentally observed IR spectrum. Also, the frequency maximaare somewhat displaced to higher frequencies and do not correlate particularly well with thefrequency maxima from “scan” (see Figure S9B). One possibility for these differences maybe the fact that for using “map” simulations with constrained -CO are required. Also, themap used in the present work was parametrized with respect to experiments and using a18oint charge-based force field whereas the simulations in the present work used multipoles.
Frequency (cm -1 ) I n t e n s i t y ( a . u . ) WT-MO (MAP)
A1A2A3A4A5A6A7A8A9A10A11A12A13A14A15A16A17A18A19A20B1B2B3B4B5 B6B7B8B9B10B11B12B13B14B15B16B17B18B19B20B21B22B23B24B25B26B28B29SUM
Figure 8: 1D IR spectra for all residues in WT monomer based on “map” for the frequencycalculation. The labels for the individual line shapes are given in the panel and the overallsum is the solid black line. The line shapes are determined from 1 ns simulations and thesnapshots analyzed are separated by 10 fs.Next, the lineshapes for the residues involved in the dimerization interface and the selectionof other residues already considered until now are analyzed for WT monomer and dimerfor INM and “map”, see Figures 9, 10, S10, and S11. When using INM it is again foundthat for the residues at the dimerization interface the location of the frequency maxima inthe two monomers differ and also change compared with the isolated monomer (see Figure9C). These effects are not only observed for residues at the interface but also away fromit. Splitting for B/D24, B/D25, and B/D26 are comparable or larger than with “scan” andblue/red shifts are consistent for the two methods.19 I n t e n s i t y ( a . u . ) A B24B25B26
Frequency (cm -1 ) I n t e n s i t y ( a . u . ) B B24B25B26D24D25D26 -1 ) [WT-MO] 162016251630163516401645165016551660 F r e qu e n cy ( c m - ) [ W T - D I ] WT-MO/DI (A,B)WT-MO/DI (C,D)D23D24 D25D26 D22D21 D20 C6D6C19,B26B24 B21 B20B22 A19 B23,B25 B6 A6 C Figure 9: 1D IR spectra from INM for residues (B24-B26) and (B24-B26, D24-D26) atthe dimerization interface for WT monomer (panel A) and WT dimer (panel B). PanelC compares the maximum frequency of the 1D IR spectra for the residues (A6, A19, B6,B20-B26, C6, C19, D6, D20-D26) between WT monomer and dimer.For the analysis using “map” in Figure 10 it is important to note that they do not use thesame structures for analysis as for “scan” and INM because the -CO bond lengths wereconstrained. As for the other two methods the frequency maxima for B24 to B26 do notcoincide for the monomer (Figure 10A) and the -CO labels in the two monomers have theirmaxima at different frequencies in the dimer (Figure 10B). However, the actual frequencymaxima between the three methods differ. The effect of constrained and flexible -CO in theMD simulations is reported in Figure S12. For a comparison of the maximum frequencies forthe three methods for B24 to B26 and D24 to D26 for direct numerical comparison, see Table1. Figure S13 reports a comparison of the map used here and an alternative parametriza-tion. Consistent with earlier work that compared the performance of different maps, itis found that the two correlate quite well (within a few cm − ) except for residue B20 forwhich they differ by ∼
25 cm − . It is noteworthy that for both, scanning along the [CONH]normal mode (Figure S1) and for using “map” (Figure S9) compared with scanning alongthe CO mode, the frequency maxima are shifted towards the blue, in accord with experiment20frequency maximum ∼ − ). Table 1: Position of the frequency maxima of the 1D IR spectra for WT monomerusing the three different approaches (“scan”, “INM”, and “map”). For “scan”and INM the CO probes are flexible while for ”map” the structures were thosefrom a simulation with constrained CO bond length.
Residue Scan INM MapB24 1612.5 1625.5 1682.5B25 1619.5 1634.5 1680.5B26 1616.5 1631.5 1670.5 I n t e n s i t y ( a . u . ) A B24B25B26
Frequency (cm -1 ) I n t e n s i t y ( a . u . ) B B24B25B26D24D25D26 -1 ) [WT-MO] 165516601665167016751680168516901695 F r e qu e n cy ( c m - ) [ W T - D I ] WT-MO/DI (A,B)WT-MO/DI (C,D)D23 D24D25D26D22D21 D20C6 D6C19B22 B26 B24B20B22 A19 B23 B25B6A6 C Figure 10: 1D IR spectra from ”map” for residues (B24-B26) and (B24-B26, D24-D26) atthe dimerization interface for WT monomer (panel A) and WT dimer (panel B). Panel Ccompares the maximum frequency of the 1D IR spectra for the residues (A6, A19, B6, B20-B26, C6, C19, D6, D20-D26) between WT monomer and dimer. The CO bond length isconstrained in the MD simulations.Using “map” the labels at B/D25 and B/D26 show splittings comparable to those from“scan” and INM whereas for B/D24 the splitting is only 1 to 3 cm − which is considerablysmaller than for the two other methods. Nevertheless, the results from “map” also indicatethat the spectroscopic signatures of the residues at the dimerization interface are not iden-tical and differ from the monomer whereas for the other residues considered the differences21etween monomer and dimer and the two monomers within the dimer are smaller.In summary, all three methods agree in that a) the individual labels have their frequencymaxima at different frequencies and b) in going from the WT monomer to the dimer the IRspectra of the labels involved in dimerization split and shift. The magnitude of the splittingand shifting differs between the methods which is not surprising given their very differentmethodologies. For the two mutants F24A and F24G the IR lineshapes using “scan” weredetermined for the residues involved in the dimerization interface and a selection of otherresidues, see Figure 1A. Compared with the WT monomer and dimer, characteristic shiftswere found. The frequency fluctuation correlation functions that can be computed from the frequencytime series contain valuable information about the dynamics around a particular site con-sidered, here the -CO groups of every residue. Specifically, FFCFs were analyzed for labelsalong the dimerization interface, for WT and the two mutant monomers and dimers, fromusing frequencies determined from “scan” and INM. Before discussing the FFCFs their con-vergence with simulation time is considered as it has been observed that an extensive amountof data is required. For this, the first 1 ns and the entire 5 ns run for WT insulin monomer was analyzed using“scan”. For the 1 ns simulation snapshots every 10 fs and every 2 fs were analyzed (seeFigure S14 top and middle row) and every 10 fs for the 5 ns simulations (Figure S14 bottomrow). The computational resources required for such an analysis are considerable. Using 8processors, the analysis of the 1 ns simulation for 10 snapshots (saved every 10 fs) takes 400hours for a single spectroscopic probe. Figure S14 shows that except for one feature at ∼ τ ) can differ by up to 30 % (B26) and the offset ∆ can differ by a factor of two or more. To balance computational expense and quality of data,the remaining analysis was carried out with data from the 1 ns simulation with snapshotsrecorded every 10 fs. 23 .11100.1110 FF C F ( p s - ) Figure 11: Comparison of the FFCFs for WT monomer and dimer for residues B24 to B26at the dimerization interface. The frequencies are based on “scan” and snapshots from the1 ns simulation, saved every 10 fs were analyzed.The FFCFs for B24 to B26 of the insulin monomer and the two monomers within the dimerare reported in Figure 11 together with the fits to Eq. 5. For the three labels from themonomer simulations the FFCFs differ in the longest decay time and the offset ∆ . As forthe infrared spectra, the three -CO labels exhibit different environmental dynamics. Whencompared with the two monomers in the insulin dimer these differences are even more pro-nounced. In general, all decay times increase to between 1 ps and ∼
13 ps and the offset canbe up to 5 times larger than for the monomer. This is owed to the considerably restraineddynamics of the residues at the dimerization interface compared with the free monomer.Comparing the two monomer mutants with the WT it is found that the picosecond com-24 able 2: Parameters from fitting the FFCF to Eq. 5 for frequencies from “scan”for the selected residues (B24-B26 and D24-D26). The amplitudes a to a inps − , the decay times τ to τ in ps, the parameter γ in ps − , and the offset ∆ in ps − . For residues D24 in monomer M2 from the WT dimer and B26 in theF24G monomer the third time scale is required for a good fit. a γ τ a τ ∆ a τ WT monomer
B24 4.64 25.44 0.025 0.75 0.74 0.21B25 4.94 22.19 0.028 0.66 0.98 0.37B26 4.97 21.82 0.019 0.62 0.62 0.07
WT dimer M1
B24 4.80 14.50 0.080 0.23 4.72 0.13B25 3.92 27.74 0.023 0.49 1.15 0.12B26 4.12 16.36 0.038 0.44 2.51 0.55
WT dimer M2
D24 0.30 17.59 0.56 3.68 0.039 0.18 0.19 4.08D25 3.17 0.033 0.41 2.32 1.50D26 5.32 13.49 0.040 0.42 2.10 0.27
F24A monomer
B24 4.94 29.51 0.020 0.51 0.61 0.07B25 4.37 13.45 0.020 0.64 0.79 0.21B26 4.11 25.05 0.027 0.63 1.38 0.45
F24A dimer M1
B24 4.90 14.61 0.046 0.33 1.68 0.41B25 3.19 0.028 0.62 1.81 1.08B26 2.15 0.040 0.34 1.89 0.48
F24A dimer M2
D24 1.27 0.043 0.31 1.17 0.36D25 1.39 0.031 0.32 1.10 0.51D26 4.91 13.72 0.039 0.60 1.40 0.63
F24G monomer
B24 4.72 29.74 0.032 0.43 1.24 0.26B25 4.57 16.46 0.019 0.58 0.81 0.24B26 4.60 25.73 0.022 0.59 0.54 0.04 0.51 7.89
F24G dimer M1
B24 1.42 0.028 0.21 1.02 0.37B25 3.70 0.018 0.48 1.18 0.24B26 3.87 0.029 0.68 1.90 0.88
F24G dimer M2
D24 2.50 38.76 0.016 0.30 1.29 0.17D25 1.53 0.030 0.25 1.70 0.65D26 3.94 5.32 0.042 0.27 2.14 0.2325onent is comparable whereas ∆ is similar (for F24A) or somewhat larger (for F24G), seeTable 2. When moving to the mutant dimers, the differences with their monomeric coun-terparts are considerably smaller than for the WT system. This is likely to be related to aweakening of the F24A and F24G dimers which also allows water to penetrate more or lessdeeply into the dimer interface. Overall, the dynamics still is slowed down in the mutantdimers by up to a factor of two compared with the mutant monomer but the effects areconsiderably less pronounced than for the WT systems.FFCFs were also determined from frequency trajectories determined from the INMs for thethree residues at the dimerization interface, see Table S3. The findings are similar to thosefrom analyzing frequencies from “scan” whereas the actual numerical values for amplitudes,decay times and offset differ somewhat.
The present work demonstrates that WT insulin monomer and dimer and mutant monomersand mutant dimers lead to different spectroscopic and dynamical signatures for residues alongthe dimerization interface. This is found - to different extent - for all three approaches usedfor computing the frequency trajectory (“scan”, INM, “map”) and suggests that the overallfindings do not depend strongly on the way how these frequencies are determined. The centerfrequency and FWHM for insulin monomer are in qualitative (scan along CO INM) or evenquantitative (scan along [CONH] INM) agreement with experiment which, together withearlier investigations of the spectroscopy and dynamics of and around NMA, providea validation of the computational model. It is noteworthy that using one single parametriza-tion for the -CO stretch and the multipoles on the [CONH] moiety of the peptide bond theexperimentally observed FWHM for the protein is correctly described.26he fact that the stability differences between WT and mutant (here at position B24) insulin dimer are also reflected in the spectroscopy and dynamics of WT and mutant in-sulin monomers and dimers suggests that spectroscopic investigations can be used to pro-vide information about the association thermodynamics. This follows earlier suggestions forcharacterizing protein-ligand binding which are supported by atomistic simulations. Forinsulin this is particularly relevant because except for the WT dimer direct thermodynamicinformation about its stability appears to be missing. Replacing a thermodynamic approachby a spectroscopic characterization is an attractive alternative. The present work suggeststhat by combining quantitative simulations with modern experiments is a potentially usefulway to obtain pharmacologically relevant information such as the strength of the modifiedinsulin dimers.
Supporting Information
The supporting information reports further comparison of the infrared spectra for WT andmutant insulin monomer and dimer. Additional validation of the FFCF and comparisons oftwo different spectroscopic maps are provided as well.
Acknowledgments
This work was supported by the Swiss National Science Foundation grants 200021-117810,200020-188724, the NCCR MUST, and the University of Basel which is gratefully acknowl-edged. The authors thank Profs. T. la Cour Jansen and A. Tokmakoff for valuable corre-spondence. 27 eferences (1) Strazza, S.; Hunter, R.; Walker, E.; Darnall, D. W. The thermodynamics of bovineand porcine insulin and proinsulin association determined by concentration differencespectroscopy.
Archives of Biochemistry and Biophysics , , 30–42.(2) Tidor, B.; Karplus, M. The contribution of vibrational entropy to molecular association:the dimerization of insulin. J. Mol. Biol. , , 405–414.(3) Zoete, V.; Meuwly, M.; Karplus, M. Study of the insulin dimerization: Binding freeenergy calculations and per-residue free energy decomposition. Proteins: Structure,Function, and Bioinformatics , , 79–93.(4) Baker, E.; Blundell, T.; Cutfield, J.; Cutfield, S.; Dodson, E.; Dodson, G.; Hodgkin, D.;Hubbard, R.; Isaacs, N.; Reynolds, C. The structure of 2Zn pig insulin crystals at 1.5˚A resolution. Phil. Trans. R. Soc. Lond. B Biol. , , 369–456.(5) Zhang, X.-X.; Tokmakoff, A. Revealing the Dynamical Role of Co-solvents in the Cou-pled Folding and Dimerization of Insulin. J. Phys. Chem. Lett. , , 4353–4358.(6) Banerjee, P.; Mondal, S.; Bagchi, B. Insulin dimer dissociation in aqueous solution: Acomputational study of free energy landscape and evolving microscopic structure alongthe reaction pathway. J. Chem. Phys. , .(7) Antoszewski, A.; Feng, C.-J.; Vani, B. P.; Thiede, E. H.; Hong, L.; Weare, J.; Tok-makoff, A.; Dinner, A. R. Insulin Dissociates by Diverse Mechanisms of Coupled Un-folding and Unbinding. J. Phys. Chem. B , , 5571–5587.(8) Raghunathan, S.; El Hage, K.; Desmond, J. L.; Zhang, L.; Meuwly, M. The Role ofWater in the Stability of Wild-type and Mutant Insulin Dimers. J. Phys. Chem. B , , 7038–7048. 289) Wang, P.; Wang, X.; Liu, L.; Zhao, H.; Qi, W.; He, M. The Hydration Shell ofMonomeric and Dimeric Insulin Studied by Terahertz Time-Domain Spectroscopy. Bio-phys. J. , , 533–541.(10) Banerjee, P.; Bagchi, B. Dynamical control by water at a molecular level in proteindimer association and dissociation. Proc. Natl. Acad. Sci. , , 2302–2308.(11) Suydam, I. T.; Snow, C. D.; Pande, V. S.; Boxer, S. G. Electric Fields at the ActiveSite of an Enzyme : Direct Comparison of Experiment with Theory. Science , ,200–204.(12) Mondal, P.; Meuwly, M. Vibrational Stark Spectroscopy for Assessing Ligand-BindingStrengths in a Protein. Phys. Chem. Chem. Phys. , , 16131–16143.(13) Bereau, T.; Kramer, C.; Meuwly, M. Leveraging Symmetries of Static Atomic MultipoleElectrostatics in Molecular Dynamics Simulations. J. Chem. Theo. Comp. , ,5450–5459.(14) Plattner, N.; Meuwly, M. The Role of Higher CO-Multipole Moments in Understandingthe Dynamics of Photodissociated Carbonmonoxide in Myoglobin. Biophys. J. , , 2505–2515.(15) Nutt, D.; Meuwly, M. Theoretical investigation of infrared spectra and pocket dynamicsof photodissociated carbonmonoxy myoglobin. Biophys. J. , , 3612–3623.(16) Wang, L.; Middleton, C. T.; Zanni, M. T.; Skinner, J. L. Development and Validationof Transferable Amide I Vibrational Frequency Maps for Peptides. J. Phys. Chem. B , , 3713–3724.(17) Jansen, T. l. C.; Dijkstra, A. G.; Watson, T. M.; Hirst, J. D.; Knoester, J. Modelingthe amide I bands of small peptides. J. Chem. Phys. , .2918) Jansen, T. l. C.; Dijkstra, A. G.; Watson, T. M.; Hirst, J. D.; Knoester, J. Modelingthe amide I bands of small peptides (vol 125, 044312, 2006). J. Chem. Phys. , .(19) Reppert, M.; Tokmakoff, A. Electrostatic frequency shifts in amide I vibrational spectra:Direct parameterization against experiment. J. Chem. Phys. , .(20) Lee, M. W.; Carr, J. K.; G¨ollner, M.; Hamm, P.; Meuwly, M. 2D IR Spectra of Cyanidein Water Investigated by Molecular Dynamics Simulations. J. Chem. Phys. , ,054506.(21) Cazade, P.-A.; Bereau, T.; Meuwly, M. Computational Two-Dimensional Infrared Spec-troscopy without Maps: N-Methylacetamide in Water. J. Phys. Chem. B , ,8135–8147.(22) Salehi, S. M.; Koner, D.; Meuwly, M. Vibrational Spectroscopy of N − in the Gas andCondensed Phase. J. Phys. Chem. B , , 3282–3290.(23) Hua, Q. X.; Shoelson, S. E.; Kochoyan, M.; Weiss, M. A. Receptor binding redefinedby a structural switch in a mutant human insulin. Nature , , 238–241.(24) Shoelson, S.; Fickova, M.; Haneda, M.; Nahum, A.; Musso, G.; Kaiser, E.; Ruben-stein, A.; Tager, H. Identification of a mutant human insulin predicted to contain aserine-for-phenylalanine substitution. Proc. Natl. Acad. Sci. , , 7390–7394.(25) Haneda, M.; Kobayashi, M.; Maegawa, H.; Watanabe, N.; Takata, Y.; Ishibashi, O.;Shigeta, Y.; Inouye, K. Decreased biologic activity and degradation of human [SerB24]-insulin, a second mutant insulin. Diabetes , , 568–573.(26) Tager, H.; Thomas, N.; Assoian, R.; Rubenstein, A.; Saekow, M.; Olefsky, J.; Kaiser, E.Semisynthesis and biological activity of porcine [LeuB24] insulin and [LeuB25] insulin. Proc. Natl. Acad. Sci. , , 3181–3185.3027) ˇZ´akov´a, L.; Kletv´ıkov´a, E.; Veverka, V.; Lepˇs´ık, M.; Watson, C. J.; Turkenburg, J. P.;Jir´aˇcek, J.; Brzozowski, A. M. Structural integrity of the B24 site in human insulin isimportant for hormone functionality. J. Biol. Chem. , , 10230–10240.(28) Chen, H.; Shi, M.; Guo, Z.-Y.; Tang, Y.-H.; Qiao, Z.-S.; Liang, Z.-H.; Feng, Y.-M. Fournew monomeric insulins obtained by alanine scanning the dimer-forming surface of theinsulin molecule. Protein Engineering , , 779–782.(29) DeFelippis, M. R.; Chance, R. E.; Frank, B. H. Insulin self-association and the rela-tionship to pharmacokinetics and pharmacodynamics. Critical Reviews in TherapeuticDrug Carrier Systems , .(30) Brooks, B. R.; Brooks III, C. L.; MacKerell Jr., A. D.; Nilsson, L.; Petrella, R. J.;Roux, B.; Won, Y.; Archontis, G.; Bartels, C.; Boresch, S. et al. CHARMM: TheBiomolecular Simulation Program. J. Comp. Chem. , , 1545–1614.(31) J. A. MacKerell, et. al. . All-Atom Empirical Potential for Molecular Modeling andDynamics Studies of Proteins. J. Phys. Chem. B , , 3586–3616.(32) Mackerell, A.; Feig, M.; Brooks, C. Extending the treatment of backbone energetics inprotein force fields: Limitations of gas-phase quantum mechanics in reproducing proteinconformational distributions in molecular dynamics simulations. J. Comp. Chem. , , 1400–1415.(33) MacKerell, A.; Feig, M.; Brooks, C. Improved treatment of the protein backbone inempirical force fields. J. Am. Chem. Soc. , , 698–699.(34) Cazade, P.-A.; Hedin, F.; Xu, Z.-H.; Meuwly, M. Vibrational Relaxation and EnergyMigration of N-Methylacetamide in Water: The Role of Non bonded Interactions. J.Phys. Chem. B , , 3112–3122. 3135) Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.; Impey, R. W.; Klein, M. L.Comparison of Simple Potential Functions for Simulating Liquid Water. J. Chem. Phys. , , 926–935.(36) Kramer, C.; Gedeck, P.; Meuwly, M. Atomic Multipoles: Electrostatic Potential Fit,Local Reference Axis Systems and Conformational Dependence. J. Comp. Chem. , , 1673–1688.(37) Swope, W. C.; Andersen, H. C.; Berens, P. H.; Wilson, K. R. A Computer SimulationMethod for the Calculation of Equilibrium Constants for the Formation of PhysicalClusters of Molecules: Application to Small Water Clusters. J. Chem. Phys. , ,637–649.(38) Nos´e, S. A Unified Formulation o the Constant Temperature Molecular-Dynamics Meth-ods. J. Chem. Phys. , , 511–519.(39) Hoover, W. G. Canonical Dynamics: Equilibrium Phase-Space Distributions. Phys.Rev. A , , 1695–1697.(40) Andersen, H. C. Molecular Dynamics Simulations at Constant Pressure and/or Tem-perature. J. Chem. Phys. , , 2384–2393.(41) Nos´e, S.; Klein, M. L. Constant Pressure Molecular Dynamics for Molecular Systems. Mol. Phys. , , 1055–1076.(42) Hairer, E.; Lubich, C.; Wanner, G. Geometric Numerical Integration Illustrated by theSt¨ormer/Verlet Method. Acta Numerica , , 399–450.(43) Gunsteren, W. V.; Berendsen, H. Algorithms for Macromolecular Dynamics and Con-straint Dynamics. Mol. Phys. , , 1311–1327.(44) Steinbach, P. J.; Brooks, B. R. New Spherical-Cutoff Methods for Long-Range Forcesin Macromolecular Simulation. J. Comp. Chem. , , 667–683.3245) Darden, T.; York, D.; Pedersen, L. Particle Mesh Ewald: An Nlog(N) Method for EwaldSums in Large Systems. J. Chem. Phys. , , 10089–10092.(46) Colbert, D. T.; Miller, W. H. A novel discrete variable representation for quantummechanic al reactive scattering via the S-matrix method. J. Chem. Phys. , ,1982–1991.(47) Koner, D.; Salehi, S. M.; Mondal, P.; Meuwly, M. Non-conventional force fields forapplications in spectroscopy and chemical reaction dynamics. J. Chem. Phys. , , 010901.(48) Hamm, P.; Zanni, M. Concepts and Methods of 2D Infrared Spectroscopy ; CambridgeUniversity Press: New York, 2011.(49) Moller, K.; Rey, R.; Hynes, J. Hydrogen Bond Dynamics in Water and Ultrafast InfraredSpectroscopy: A Theoretical Study.
J. Phys. Chem. A , , 1275–1289.(50) Li, S.; Schmidt, J. R.; Piryatinski, A.; Lawrence, C. P.; Skinner, J. L. VibrationalSpectral Diffusion of Azide in Water. J. Phys. Chem. B , , 18933–18938.(51) Virtanen, P.; Gommers, R.; Oliphant, T. E.; Haberland, M.; Reddy, T.; Courna-peau, D.; Burovski, E.; Peterson, P.; Weckesser, W.; Bright, J. et al. SciPy 1.0: Fun-damental Algorithms for Scientific Computing in Python. Nature Methods , ,261–272.(52) Zoete, V.; Meuwly, M.; Karplus, M. A comparison of the dynamic behavior ofmonomeric and dimeric insulin shows structural rearrangements in the active monomer. J. Mol. Biol. , , 913–929.(53) Dhayalan, B.; Fitzpatrick, A.; Mandal, K.; Whittaker, J.; Weiss, M. A.; Tokmakoff, A.;Kent, S. B. H. Efficient Total Chemical Synthesis of C-13=O-18 Isotopomers of HumanInsulin for Isotope-Edited FTIR. Chem. Bio. Chem. , , 415–420.3354) Zhang, X.-X.; Jones, K. C.; Fitzpatrick, A.; Peng, C. S.; Feng, C.-J.; Baiz, C. R.;Tokmakoff, A. Studying Protein-Protein Binding through T-Jump Induced Dissocia-tion: Transient 2D IR Spectroscopy of Insulin Dimer. J. Phys. Chem. B , ,5134–5145.(55) Falconi, M.; Cambria, M.; Cambria, A.; Desideri, A. Structure and stability of theinsulin dimer investigated by molecular dynamics simulation. J. Biomol. Struc. Dyn. , , 761–772.(56) Desmond, J. L.; Koner, D.; Meuwly, M. Probing the Differential Dynamics of theMonomeric and Dimeric Insulin from Amide-I IR Spectroscopy. J. Phys. Chem. B , , 6588–6598.(57) Reppert, M.; Roy, A. R.; Tokmakoff, A. Isotope-enriched protein standards for compu-tational amide I spectroscopy. J. Chem. Phys. , .TOC Graphic34 upporting Information: The Dynamics andInfrared Spectrocopy of Monomeric and DimericWild Type and Mutant Insulin Seyedeh Maryam Salehi, Debasish Koner and Markus Meuwly ∗ Department of Chemistry, University of Basel, Klingelbergstrasse 80 , CH-4056 Basel,Switzerland.
E-mail: [email protected] S1 a r X i v : . [ phy s i c s . b i o - ph ] S e p Additional Lineshapes I n t e n s i t y ( a . u . ) A B24B25B26B24B25B26
Frequency (cm -1 ) I n t e n s i t y ( a . u . ) B COCONH -1 ) [SCAN: CO] 1610162016301640165016601670 F r e qu e n cy ( c m - ) [ S CAN : C O NH ] WT-MOB24B25B26 B22B21B20B23 ,A19 A6B6 C Figure S1: Comparison for scanning along the CO (solid line) and CONH (amide-I, dashedline) normal modes for ”scan” for the insulin monomer. Panel A: 1D IR spectra for residues(B24-B26), panel B: the sum frequency of all the residues and panel C: Comparison of themaximum frequency of the 1D IR spectra for the selected residues (A6, A19, B6, B20-B26).The black dashed line shows the linear regression with regression coefficient (slope) of 0.81and correlation coefficient of 0.95. The analysis is done for 1 ns simulation and the snapshotsanalyzed are separated by 10 fs. The frequency maxima from scanning along the [CONH]INM are shifted to the blue, in accord with the experimental observations. S2 Frequency (cm -1 ) I n t e n s i t y ( a . u . ) WT-MO (SCAN)
A1A2A3A4A5A6A7A8A9A10A11A12A13A14A15A16A17A18A19A20B1B2B3B4B5 B6B7B8B9B10B11B12B13B14B15B16B17B18B19B20B21B22B23B24B25B26B28B29CONHCO
Figure S2: 1D IR spectra for all residues in WT monomer based on “scan” for frequencycalculations along the amide-I normal mode. The labels for the individual line shapes aregiven in the panel and the overall sum is the solid black line. The line shapes are determinedfrom 1 ns simulations and the snapshots analyzed are separated by 10 fs. I n t e n s i t y ( a . u . ) A A6A19B6B20B21B22B23B24B25B26
Frequency (cm -1 ) I n t e n s i t y ( a . u . ) B A6A19B6B20B21B22B23B24B25B26C6C19D6D20D21D22D23D24D25D26 -1 ) [WT-MO] 161016151620162516301635164016451650 F r e qu e n cy ( c m - ) [ W T - D I ] WT-MO/DI (A,B)WT-MO/DI (C,D)D23D24 D25D26 D22 D21D20 C6D6 C19,B26B24 B21B20B22A19B23,B25 B6 A6 C Figure S3: 1D IR spectra for WT monomer (panel A) and dimer (panel B) for selectedresidues (A6, A19, B6, B20-B26) and (A6, A19, B6, B20-B26, C6, C19, D6, D20-D26), using“scan” for the frequency calculation. Panel C: Comparison for the maximum frequency ofthe 1D IR spectra for the selected residues between WT monomer and dimer.S3
Frequency (cm -1 ) I n t e n s i t y ( a . u . ) WT-DI-M1 (SCAN)
A1A2A3A4A5A6A7A8A9A10A11A12A13A14A15A16A17A18A19A20B1B2B3B4B5 B6B7B8B9B10B11B12B13B14B15B16B17B18B19B20B21B22B23B24B25B26B28B29SUM
Figure S4: 1D IR spectra for all residues of monomer 1 (M1) in the WT dimer based on“scan” for the frequency calculation along the CO normal mode. The labels for the individualline shapes are given in the panel and the overall sum is the solid black line. The line shapesare determined from 1 ns simulations and the snapshots analyzed are separated by 10 fs.
Frequency (cm -1 ) I n t e n s i t y ( a . u . ) WT-DI-M2 (SCAN)
C1C2C3C4C5C6C7C8C9C10C11C12C13C14C15C16C17C18C19C20D1D2D3D4D5 D6D7D8D9D10D11D12D13D14D15D16D17D18D19D20D21D22D23D24D25D26D27D28D29SUM
Figure S5: 1D IR spectra for all residues of monomer 2 (M2) in the WT dimer based on“scan” for the frequency calculation along the CO normal mode. The labels for the individualline shapes are given in the panel and the overall sum is the solid black line. The line shapesare determined from 1 ns simulations and the snapshots analyzed are separated by 10 fs.S4 I n t e n s i t y ( a . u . ) A A6A19B6B20B21B22B23B24B25B26
Frequency (cm -1 ) I n t e n s i t y ( a . u . ) B A6A19B6B20B21B22B23B24B25B26 -1 ) [WT-MO] 161016151620162516301635164016451650 F r e qu e n cy ( c m - ) [ B A / B G - M O ] WT/B24A-MOWT/B24G-MOB24 B25B26 B22B21B20 B23B24 B25B26 B22B21B20B23 A6B6A19A19 A6B6 , ,, , C Figure S6: 1D IR spectra for the F24A (panel A) and F24G (panel B) mutant monomers forselected residues (A6, A19, B6, B20-B26) using “scan”. Panel C: Comparison for the maxi-mum frequency of the 1D IR spectra for the selected residues between WT and F24A/F24Gmonomers. I n t e n s i t y ( a . u . ) A A6A19B6B20B21B22B23B24B25B26C6C19D6D20D21D22D23D24D25D26
Frequency (cm -1 ) I n t e n s i t y ( a . u . ) B A6A19B6B20B21B22B23B24B25B26C6C19D6D20D21D22D23D24D25D26 -1 ) [WT-DI] 161016151620162516301635164016451650 F r e qu e n cy ( c m - ) [ B A - D I ] WT/B24A-DI (A,B)WT/B24A-DI (C,D)B23 C6B25 B26 B22B21B20 A6B6A19 C19D21D23D6D22D24 D20D26D25 , C ,B24 Figure S7: 1D IR spectra for the WT (panel A) and F24A (panel B) mutant dimers forselected residues (A6, A19, B6, B20-B26, C6, C19, D6, D20-D26) based on “scan” for thefrequency calculations. Panel C: Comparison between maximum frequency of 1D IR spectrafor the selected residues between WT and F24A mutant dimers.S5 I n t e n s i t y ( a . u . ) A A6A19B6B20B21B22B23B24B25B26C6C19D6D20D21D22D23D24D25D26
Frequency (cm -1 ) I n t e n s i t y ( a . u . ) B A6A19B6B20B21B22B23B24B25B26C6C19D6D20D21D22D23D24D25D26 -1 ) [WT-DI] 161016151620162516301635164016451650 F r e qu e n cy ( c m - ) [ B G - D I ] WT/B24G-DI (A,B)WT/B24G-DI (C,D)B23C6B25 B26 B22B21B20 A6B6A19 C19D21D23D6D22D24 D20 D26D25 ,B24 , C Figure S8: 1D IR spectra for the WT (panel A) and F24G (panel B) mutant dimers forselected residues (A6, A19, B6, B20-B26, C6, C19, D6, D20-D26) based on “scan” for thefrequency calculations. Panel C: Comparison between maximum frequency of 1D IR spectrafor the selected residues between WT and F24G mutant dimers. -1 ) [SCAN]161016201630164016501660167016801690 F r e qu e n cy ( c m - ) [ I N M ] WT-MOB24B25B26 B22B21B20 B23, A19 A6B6 A -1 ) [SCAN] 161016201630164016501660167016801690 F r e qu e n cy ( c m - ) [ M A P ] WT-MOB24 B25B26 B22B21 B20B23A19 A6B6 B Figure S9: Comparison of the maximum frequency of the 1D IR spectra between “scan”and “INM” (panel A) and ”scan” and “map” (panel B) for the selected residues (A6, A19,B6, B20-B26, C6, C19, D6, D20-D26) for WT monomer. The CO probes are flexible in thesimulations analyzed with “scan” and “INM” and constrained for the one using “map”.S6 I n t e n s i t y ( a . u . ) A A6A19B6B20B21B22B23B24B25B26
Frequency (cm -1 ) I n t e n s i t y ( a . u . ) B A6A19B6B20B21B22B23B24B25B26C6C19D6D20D21D22D23D24D25D26 -1 ) [WT-MO] 162016251630163516401645165016551660 F r e qu e n cy ( c m - ) [ W T - D I ] WT-MO/DI (A,B)WT-MO/DI (C,D)D23D24 D25D26 D22D21 D20 C6D6C19,B26B24 B21 B20B22 A19 B23,B25 B6 A6 C Figure S10: 1D IR spectra from INM. Panel A: WT monomer and panel B: WT dimer forselected residues (A6, A19, B6, B20-B26) and (A6, A19, B6, B20-B26, C6, C19, D6, D20-D26), respectively. Panel C: Comparison between maximum frequency of 1D IR spectra forthe selected residues between monomer and dimer. I n t e n s i t y ( a . u . ) A A6A19B6B20B21B22B23B24B25B26
Frequency (cm -1 ) I n t e n s i t y ( a . u . ) B A6A19B6B20B21B22B23B24B25B26C6C19D6D20D21D22D23D24D25D26 -1 ) [WT-MO] 165516601665167016751680168516901695 F r e qu e n cy ( c m - ) [ W T - D I ] WT-MO/DI (A,B)WT-MO/DI (C,D)D23 D24D25D26D22D21 D20C6 D6C19B22 B26 B24B20B22 A19 B23 B25B6A6 C Figure S11: 1D IR spectra from ”map” from simulations with constrained CO bond length.Panel A: WT monomer and panel B: WT dimer for selected residues (A6, A19, B6, B20-B26)and (A6, A19, B6, B20-B26, C6, C19, D6, D20-D26), respectively. Panel C: Comparisonbetween maximum frequency of 1D IR spectra for the selected residues between monomerand dimer. S7 I n t e n s i t y ( a . u . ) A B24B25B26
Frequency (cm -1 ) I n t e n s i t y ( a . u . ) B B24B25B26D24D25D26 -1 ) [WT-MO] 1650166016701680169017001710 F r e qu e n cy ( c m - ) [ W T - D I ] WT-MO/DI (A,B)WT-MO/DI (C,D)D23 D24D25D26D22 D21 D20C6 D6C19, B26 B24B21 B20B22 A19 B23 B25B6A6 C Figure S12: 1D IR spectra from ”map” from simulations with flexible CO probe. Panel A:WT monomer and panel B: WT dimer for residues at the dimerization interface (B24-B26)and (B24-B26, D24-D26), respectively. Panel C: Comparison between maximum frequencyof 1D IR spectra for residues (A6, A19, B6, B20-B26) and (A6, A19, B6, B20-B26, C6, C19,D6, D20-D26), respectively between WT monomer and dimer.
As a separate test, a different map is used in which the frequency shift due to the dihedralangles ( φ, ψ ) between neighboring peptide units are included. Here the map parametrizationis ω i = 1684 + 7729 E C i + 3576 E N i (1)and the local frequency is ω bi = ω i + ∆ ω N ( φ i − , ψ i − ) + ∆ ω C ( φ i +1 , ψ i +1 ) (2)Based on the ( φ, ψ ) angles for i th chromophore, ∆ ω N and ∆ ω C are the contributions from( i − i + 1)th residues. S8 able S1: Position of the frequency maxima from applying two maps (Skinner and Tokmakoff ) to the same trajectory for WT insulin monomer for selectedresidues. Map FrequenciesResidue Skinner Map Tokmakoff MapA6 1654.50 1660.50A19 1668.50 1665.50B6 1671.50 1671.50B20 1715.50 1683.50B21 1660.50 1663.50B22 1666.50 1661.50B23 1677.50 1672.50B24 1676.50 1682.50B25 1692.50 1680.50B26 1656.50 1670.50 -1 ) [Skinner]16501660167016801690170017101720 F r e qu e n cy ( c m - ) [ To k m ak o ff] WT-MO B22B26 B24 B20B21 A19 B23 B25B6A6
Figure S13: Comparison between maximum frequency of 1D IR spectra for residues (A6,A19, B6, B20-B26, C6, C19, D6, D20-D26) based on two different maps for WT monomer.Snapshots from the same trajectory, run with constrained CO, were analyzed.S9
Validations for FFCFs FF C F ( p s - ) Figure S14: FFCF for residues at the dimerization interface (B24-B26) from frequency trajec-tories based on ”scan” for WT monomer. The FFCF is shown based on different simulationlengths (1 ns and 5 ns) and computing frequencies from snapshots saved every 2 or 10 fs(savc2 and savc10). The overall shape of the FFCFs changes little whereas the noise leveldecreases especially for longer simulation times. Also, the magnitude of the static componentincreases for longer simulation times. S10 FF C F ( p s - ) Figure S15: Fitting the FFCF for residue B24 for the three analyses from Figure (S14). Thefitting parameters for the different FFCFs are summarized in Table S2.
Table S2: Parameters obtained from fitting the FFCF to Eq. 5 from “scan” fre-quencies for residues (B24-B26) for WT monomer based on different simulationlength and different time separations between coordinates analyzed (every 2 or10 fs - nsavc2 and nsavc10). The amplitudes a to a are in ps − , the decay times τ to τ in ps, the parameter γ in ps − , and the offset ∆ in ps − . a γ τ a τ ∆ B24 4.64 25.44 0.025 0.75 0.74 0.21B25 4.94 22.19 0.028 0.66 0.98 0.37B26 4.97 21.82 0.019 0.62 0.62 0.07
B24 4.64 25.77 0.023 0.65 0.68 0.13B25 4.40 20.96 0.025 0.62 1.05 0.54B26 4.89 25.63 0.020 0.77 0.70 0.17
B24 4.74 17.46 0.023 0.69 0.83 0.35B25 4.24 0.00 0.022 0.60 1.07 0.94B26 4.96 16.09 0.021 0.59 0.92 0.47S11 able S3: Parameters obtained from fitting the FFCF to Eq. 5 for frequenciesfrom INM for residues at the dimerization interface (B24-B26 and D24-D26).The amplitudes a to a in ps − , the decay times τ to τ in ps, the parameter γ in ps − , and the offset ∆ in ps − . a γ τ a τ ∆ a τ WT monomer
B24 4.80 27.19 0.026 0.69 0.75 0.21B25 4.66 22.60 0.030 0.62 1.02 0.33B26 4.58 22.32 0.020 0.53 0.68 0.06
WT dimer M1
B24 4.65 15.90 0.070 0.24 3.80 0.15B25 3.62 29.25 0.026 0.45 1.19 0.15B26 4.03 17.16 0.039 0.40 2.47 0.50
WT dimer M2
D24 0.34 17.51 0.54 3.62 0.039 0.14 0.19 3.96D25 2.98 0.037 0.41 2.56 1.67D26 5.13 14.17 0.043 0.42 2.09 0.27
B24A monomer
B24 4.81 28.91 0.021 0.50 0.60 0.06B25 4.10 15.70 0.021 0.64 0.71 0.11B26 3.87 25.00 0.030 0.60 1.43 0.52
B24A dimer M1
B24 4.81 14.85 0.050 0.32 1.46 0.43B25 2.98 0.029 0.63 1.82 0.70B26 2.03 0.041 0.27 1.78 0.44
B24A dimer M2
D24 1.19 0.043 0.29 1.24 0.43D25 1.28 0.035 0.29 1.13 0.33D26 4.67 15.50 0.043 0.56 1.46 0.46
B24G monomer
B24 4.57 29.61 0.033 0.41 1.31 0.27B25 4.34 17.06 0.019 0.56 0.80 0.27B26 4.41 25.09 0.022 0.53 0.51 0.17 0.37 4.30
B24G dimer M1
B24 1.28 0.032 0.19 1.21 0.32B25 3.55 0.019 0.48 1.19 0.23B26 3.71 0.030 0.69 2.05 0.93
B24G dimer M2
D24 2.46 38.53 0.015 0.30 1.22 0.14D25 1.31 0.034 0.23 1.82 0.62D26 3.94 5.32 0.042 0.27 2.14 0.23S12 eferences (1) Dhayalan, B.; Fitzpatrick, A.; Mandal, K.; Whittaker, J.; Weiss, M. A.; Tokmakoff, A.;Kent, S. B. H. Efficient Total Chemical Synthesis of C-13=O-18 Isotopomers of HumanInsulin for Isotope-Edited FTIR.
Chem. Bio. Chem. , , 415–420.(2) Zhang, X.-X.; Jones, K. C.; Fitzpatrick, A.; Peng, C. S.; Feng, C.-J.; Baiz, C. R.;Tokmakoff, A. Studying Protein-Protein Binding through T-Jump Induced Dissociation:Transient 2D IR Spectroscopy of Insulin Dimer. J. Phys. Chem. B , , 5134–5145.(3) Wang, L.; Middleton, C. T.; Zanni, M. T.; Skinner, J. L. Development and Validationof Transferable Amide I Vibrational Frequency Maps for Peptides. J. Phys. Chem. B , , 3713–3724.(4) Reppert, M.; Tokmakoff, A. Electrostatic frequency shifts in amide I vibrational spectra:Direct parameterization against experiment. J. Chem. Phys. ,138