The naturally processed A-chain epitope (KR A1-15) stimulates T cell clones in a reduced environment (Supporting information and Figure S1
). To evaluate the intrinsic conformational flux exhibited by Cys residues (Cys8, Cys9, and Cys13) in KR A1-15, and to understand the importance of a reduced environment [presence of dithiothreitol (DTT) in functional studies] we collected 2D homonuclear and 2D heteronuclear 13
C NMR data in the absence and presence of DTT.
The NMR spectra identified spectrally resolved resonances for each amino acid, which we used to complete the assignment of the peptide (see supporting information, Figure S2 and Table S1
Upon analyzing the 1
H 2D TOCSY and NOESY fingerprint spectra collected in the absence of DTT we determined that the three Cys spin systems in KR A1-15 sampled more than a single conformational state. Three unique conformations were unambiguously observed for Cys8 and Cys13 residue: one major and two minor conformations in a ratio of: 77%:13%:10% (±3%) on the basis of crosspeak intensities observed in homonuclear spectra. For Cys9 residue, one major and one minor conformational states were observed in a ratio 93%:7% (±3%), respectively. These Cys residues were observed in a redox-dependent conformational equilibrium ().
Figure 2 Overlay of Strip-plots from 2D 1H-1H NOESY (τm=300 ms) (colored black), and 2D 1H-1H TOCSY (spin lock =70 ms) (colored red) of KR-A1-15, showing redox spin systems for all three Cys residues. (A) reduced state chemical shifts, (B) oxidized state (more ...)
The unique 1
resonance chemical shifts of the reduced and oxidized states were identified for each Cys residue.16
The major conformation has 1
chemical shifts that are representative of a reduced peptide conformation, i.e.
free thiol (-S-H state) for all three Cys residues ( and Table S1
), while minor conformations for Cys8, Cys9 and Cys13 all possessing 1
chemical shifts representative of Cys residues in an oxidized (-SS-) state ( and Table S1
). The population ratio between two minor oxidized conformations of Cys8 and Cys13 is 0.55:0.45 (± 0.05) ().
These homonuclear assignments were further supported by natural abundance 1
C 2D heteronuclear single quantum correlation (HSQC) data that were acquired in absence of DTT (Figure S3
and ). These spectra illustrate the presence of Cys residue 13
resonances in the 29.0 to 30.0 ppm spectral window, which are highly supportive of the presence of free thiol moieties (). The directly bonded 1
resonances for all three Cys residues support a reduced environment (free thiol state) for these 13
resonances as observed in 2D 1
H TOCSY and NOESY spectra ( and ). Interestingly, no 13
resonances were observed for the oxidized conformation, which suggests that the single reduced conformational (free thiol) state may be more stable, while the two oxidized conformations may be interconverting on the NMR time scale, which we believe may result in exchange broadening.15
In and , some of the 1HN-1Hβ crosspeaks in 2D TOCSY spectrum are not seen for these minor (disulfide) conformations. This is like the result of insufficiencies in the through bond magnetization transfer process, which may have occurred in this relatively fast interconverting oxidized state.
MALDI-TOF mass spectrometry performed on KR A1-15 NMR samples in the absence of DTT identified the presence of monomer/dimer peaks in a ratio of ~1:2 using peak intensity of the representative conformations. These data further support that the reduced and oxidized conformational states are in equilibrium (Figure S4A
). In another sample containing 25 mM DTT we only see monomeric (free thiol) KR A1-15 with no dimer peak (intermolecular disulfide bond formation) observed (Figure S4B
). These data are in good agreement with the functional observations that clearly illustrate approximately a 2-fold increase in T-cell recognition in presence of 25 mM DTT (Figure S1
The two oxidized isomeric states of Cys8 and Cys13 were designated as cis
conformations. This discrimination was based on the observation of a unique 1
intermolecular-NOE between the two conformers of Cys8 that we have identified in a cis:trans
ratio of 55%:45% (). A similar cis:trans
Cys residue equilibrium [ratio of 0.7:0.3 (±0.05)] has been reported previously in a NMR study of a heptapeptide comprised of vicinal cysteine residues.17
The single oxidized state of Cys9 is characterized by a 1
crosspeak that is shifted downfield (with respect to the 1
resonance chemical shifts for of Cys8 and Cys13) to 3.76 ppm (). The chemical shift of this downfield resonance is the result of the unique local environment that is sampled by this Cys residue and is consistent with previous studies of insulin peptides in an oxidizing environment [Table S1(C)
Interestingly, our 2D 1
H NOESY data also allowed us to identify the presence of a sulfhydryl proton (1
) for the oxidized cis
conformation of Cys8 at 7.52 ppm (). These data may suggest that Cys8 possibly participate in a mixed disulfide conformation (-S-SH). A similar observation of the 1
chemical shift (7.6 ppm) has been reported for Cys14 in E.coli
glutaredoxin-3, a residue that was functionally important.18
Furthermore, 1H-1H homonuclear data (in absence of DTT) also identified the presence of alternative amide spin systems for Ser14 and Tyr16, which are likely due to slow exchange properties of these amide protons and/or the redox nature of Cys residues (Cys 8, 9, & 13). The population ratio of major vs minor conformations for Ser14, and Tyr16 is analogous to that observed for the Cys residues (). These data suggest that Ser14 and Tyr16 are in close proximity and thus sterically constrained by the redox state of these Cys residues.
Figure 3 Overlay of contour-plots from 2D 1H-1H NOESY (τm=300 ms) (colored blue), and 2D 1H-1H TOCSY (spin lock =70 ms) (colored red) shows three observed conformations of Tyr16 (the spin-system for major conformation is represented by I, and for two minor (more ...)
The complete assignment of the 2D 1
H NOESY data (in absence of DTT) in the oxidized state aided in our characterization of the multiple redox states of the KR A1-15 peptide. These data identified three weak NOEs between: Cys8Hβ
, and Cys8Hβ
that support the presence of a disulfide bond between Cys8-Cys13. No such NOEs were observed for Cys9. These data propose that Cys9 may also be involved in forming an inter-molecular disufide bond between two KR A1-15 molecules, which may account for the faster thiol-disulfide interconversion rate that is observed in our data. We believe this may also explain the observed line broadening for the single oxidized spin system of Cys9 (). The disulfide arrangement found in KR A1-15 is similar to that in native human insulin molecule ().19,20
To date no prior theoretical investigations have been performed to evaluate the role and importance of Cys proton NMR chemical shifts as redox fingerprints in biomolecules. To further probe the redox state-dependent conformational flux observed for KR A1-15 we performed a number of quantum chemical calculations using state-of-art methods14
(see supporting information
for details). The theoretical investigations were used to evaluate if the observed NMR chemical shifts could be used to predict the unique redox state of the Cys residues in this peptide and possibly translated to other peptide systems. The computational results (vide infra
) confirmed the presence of both a reduced and oxidized state for the KR A1-15 peptide and predicted the chemical shift for the highly unique thiol observed in our experimental NMR data (experimental: 7.52 ppm; calc: 7.50 ppm). The experimental chemical shift range for each of the Cys protons (1
, and 1
) in each of the unique redox states differ modestly (ca. 0.1-0.3 ppm, see ). These data suggest that the observed NMR data result from local effects on the Cys residues alone and provide a basis for using computationally derived molecular models comprised of: Cys-SH, Cys-SS-Cys, and Cys-SHS-Cys to represent the reduced, oxidized, and mixed redox states of the insulin A-chain epitope, respectively.
Comparison of Experimental (NMR) and Computational (Quantum chemical) Results of 1H NMR Shiftsa (ppm)
The first set of non hydrogen bond (non-HB) calculations were performed on models outlined above in the absence of backbone HB partners to evaluate if the oxidation states of the Cys residues alone can reproduce the experimental trend observed: an incremental increase in chemical shift values for the proton(s) in the Cys residues in the order of Cys-SH < Cys-SHS-Cys < Cys-SS-Cys, (). Interestingly, for both the Cα
), the predicted NMR chemical shift data are in excellent agreement with the theoretically derived data (Figure S5
and ). The linear agreement between the calculated quantum chemical values and the average experimentally observed values for each conformation is represented by a correlation coefficient of R2
= 0.998 (standard deviation of 0.05 ppm), Figure S5
and . These results support that the observed chemical shift trend for 1
can be ascribed to the different Cys oxidation states present in the three conformations. However, the predicted amide proton (1
) NMR chemical shift data support a different trend, which we believe can be attributed to differences in the distance between the 1
protons and the backbone 1
versus the sulfur atom involved in the respective redox state. Given that oxidized sulfur atoms are electron deficient and thus attract electrons from surrounding atoms, this reduces the electron density around these atoms and results in decreased chemical shielding and increased NMR chemical shifts as we have observed experimentally.
Given the discrepancy of ~2 ppm between our experimentally observed and theoretically predicted 1
chemical shifts using the Non-HB calculations, we carried out a second set of calculations that included a HB partner for each of the 1
protons (see supporting information
for details). The predicted 1
NMR chemical shifts for the reduced, mixed, and oxidized states are 8.27, 8.47, 8.67 ppm respectively, which correlate well with the average experimental values of: 8.23, 8.57 and 8.58 ppm, respectively ( and ). It is also interesting to note that, the precision of the predicted 1
H NMR chemical shifts of 1
protons was also improved. These proton chemical shifts appear to be independent of the HB distances implemented into our theoretical calculations, which is consistent with their distal proximity to the HB region compared the 1
protons. As illustrated in , there is a significant linear correlation (R2
=0.998) between our experimentally observed and theoretically derived values with a standard deviation of 0.11 ppm or 1.9% of the experimental 1
H NMR shift range observed for KR A1-15.
Correlation plot between the experimental and computational results for 1Hα, 1Hβ, and 1HN of reduced, oxidized, and mixed forms of Cys residues in KR A1-15.
Taken together, our computational data support the presence of three redox states for the Cys residues in the human insulin A-chain epitope, which we have identified in our experimental NMR studies. Each of these redox states can be characterized by unique chemical shift data. In addition, these calculations support that the observed trend for the 1Hα and 1Hβ chemical shifts: Cys-SH < Cys-SHS-Cys < Cys-SS-Cys, which we believe can be attributed to the electronic effect of the Cys residue each redox state (). However, we believe that the observed trend for the 1HN chemical shifts is dependent both on the Cys redox state as well as the relevant HB network within the peptide structure.
In Figure S4(A)
, the relative signal/noise (S/N) ratio of the peak heights in the MALDI spectrum were used to estimate the monomer:dimer ratio. The peptide, in its dimer conformation is comprised of two molecules, which may be inter-disulfide linked via Cys9 in each peptide, while the monomer peak height is representative of a single molecule. The increased peak intensity of the dimer conformation may additionally be due to the presence of some free thiol population of Cys 8 and Cys13 in it, if the formation of dimer has taken place via
inter-disulfide bond formation (Cys9-Cys9) between two KR A1-15 peptide monomer. Importantly, one has to be cautious when comparing MS data from a plate spotted dried peptide sample possessing a fixed molecular composition at the time of the laser shot and NMR data that represents the presence of multiple, dynamic conformations that are sampled simultaneously on the NMR timescale.
The observation of multiple redox dependent states for the Cys residues of KR A1-15 is similar to those previously described for human serum albumin (HSA), in which Cys34 of HSA was observed to be predominantly in a free sulfhydryl conformation (70-80%) with the remaining Cys34 conformations representative of an oxidized and mixed disulfide isomeric conformation.21
The occurrence of multiple states in human insulin has been reported by others previously.22-24
In a NMR study of the human insulin B-chain, Cys7 was present in multiple unique conformations including: a major conformation (>90%), and minor conformations (accounting for <10%).24
Similarly, two different conformations were observed for A-chain and B-chain residues in NMR study of mutant human insulin.23
However, our work provides the first atomic level evidence of the co-existence of the Cys reduced, oxidized, and a mixed disulfide (-S-SH) conformations, which we believe are important redox-dependent conformations responsible for regulating the activity of this peptide and/or its targets.
The observation of a mixed disulfide conformer may represent a redox state that is short lived on the NMR time scale when one considers the signal intensity of the 1
crosspeak versus the crosspeak intensities for the reduced and oxidized conformational states. However, the emergence of the 1
chemical shift may also represent the presence of a thiolate-thiol hydrogen bond pair formed between the Cys8 thiolate and the Cys13 thiol which are the residues responsible for the formation of the disulfide bond; a hypothesis that is supported by our NOESY data. Interestingly, a 1
proton was observed in HSA that does indeed exist in plasma in the mixed disulfide conformation with other redox states.21
There are 89 reports of this unique thiol (1
) chemical shift in the BMRB chemical shift data base: (http://www.bmrb.wisc.edu
). This proton has been assigned chemical shift values between 0.25 ppm and 10.7 ppm depending upon the local protein environment. Indeed, our quantum chemical calculations support that different local environments affect the strength and/or distance between the two sulfurs of the thiolate-thiol hydrogen bonding interaction (SH...S) in the mixed disulfide state. Our calculations identify that the strongest SH...S interaction has a predicted thiol proton shift of 11.13 ppm, which is very close to the experimental upper bound of 10.7 ppm reported in BMRB. This occurs in the fully optimized Cys-SHS-Cys model with no restraints, in which the distance separating the two sulfur atoms is 3.54 Å. However, when the two-sulfur distance is increased due to a specific protein environment, the reduced SH...S interaction shifts this resonance upfield. Calculations were performed at various S(H)...S distances and we determined that when this sulfur-sulfur distance is elongated by 0.35 Å, the predicted thiol proton shift in the HB calculation of the mixed conformed is 7.50 ppm, which is in excellent agreement with our experimental observation of 7.52 ppm. Therefore, this proton shift is also a sensitive probe of the local redox environment of the protein. In contrast to the established distance dependence for the thiol proton shifts, the Cα
, and amide proton shifts are only modestly affected (≤ 0.1 ppm), which is within the computational error margin.
Mannering et al.
, have previously reported that T cell clone recognition of the KR A1-13 peptide required the presence of a vicinal disulfide bond between Cys6-Cys7.4b
The disulfide bond arrangement they proposed was identified in a variant of the peptide in which the third Cys was mutated to a serine ().4b
In our present study of the KR A1-15 peptide that is comprised of all three Cys residues, our NOESY data support the presence of Cys8-Cys13 disulfide bond arrangement within oxidized state conformation, which is similar to that found in native insulin (Cys6-Cys11 disulfide pattern in native insulin) and do not observe the vicinal disulfide (). In addition, T cell clones were generated from the periphery of a DR4+ new onset type 1 diabetes subject using single cell sorting to identify those T cells that proliferated in response to recombinant human pro-insulin using a sensitive CSFE assay. These studies confirmed that these clones responded to the insulin KR A1-13 peptide and that this was a naturally processed epitope4b
. During antigen processing, proteins are cleaved into linear peptides 12-16 amino acids in length by an array of proteases and modifying enzymes in the acidic lysosome that individual expresses and then ultimately loaded into the groove of Class II proteins.4c,26-28
The conformational flux of the insulin A-chain peptide(s) and how they are presented by the Class II cell surface protein can influence the T cell response to the autoantigenic protein.
It is worth noting that in previous NMR studies of a human insulin A-chain analogue (C7S, C20S), carried out in 100% DMSO, a single conformation was observed25
, while studies carried out with a mutant human insulin in aqueous buffer reported the presence of at least two conformations.23,24
These studies provide further support that the redox-dependent conformational states observed for the KR A1-15 peptide are intrinsic properties of the insulin A-chain.