Photolysis of recombinant human insulin in the solid state: formation of a dithiohemiacetal product at the C-terminal disulfide bond.

Olivier Mozziconacci,† Jessica Haywood,†Eric M. Gorman,†Eric Munson,†,$Christian Schöneich*,†

†Contribution from the Department of Pharmaceutical Chemistry, 2095 Constant Avenue, University of Kansas, Lawrence, KS 66047, $present address:Department of Pharmaceutical Sciences, University of Kentucky, Lexington, KY 40506

Figure S1.Comparison of the polarized light microscopyof the crystalline bovine insulin (left) and the native human insulin powder from Roche (right).

Figure S2.HPLC method to measure the crystalline content of zinc-insulin according to the 2005 U.S. Pharmacopeia National Formulary. Black curve: human insulin from Roche. Red curve: remaining crystal insulin from the human insulin after extraction of the amorphous phase (or soluble microparticles). Blue curve: extraction buffer used to extract the amorphous phase of insulin.

Figure S3.Scanning electronic microscopy (SEM) of human insulin powder supplied by Roche (A) and Millipore (B). SEM of human insulin from Roche (C) and Millipore (D) after application of the procedure for microcrystal preparation.

A
/ B

C
/ D

Experiment setup process for SSNMR analysis:

The relaxation properties of a sample are typically measured to ensure that the best spectra are collected. For 13C CP/MAS, these often include 1H T1 or spin-lattice relaxation time and the cross-polarization dynamics, consisting of both the TCH and 1H T1p or spin-lattice relaxation time in the rotating frame. Due to the presence of spin diffusion, all of the peaks of an intimate mixture should exhibit a common 1H T1. Since the peaks in the insulin spectra are actually made up of many overlapping peaks it is impossible to accurately analyze individual peaks; therefore, regions associated with different carbon types have been analyzed (i.e., carbonyl, aromatic, aliphatic carbons). The results of the 1H T1 analysis for both human insulin samples are shown in Fig. S4. The integration region for each analysis is show, the first six categories are the “individual peaks”, the next is the average of the results from the “individual peaks”, and the last category is from the integration of the entire spectrum. The error bars of the “individual peaks” and the entire spectrum are the uncertainty of the non-linear regressions, while the error bars of the peak average are the standard deviation in the results from the “individual peaks”. No significant difference in 1H T1 is observed between the two human insulin samples, the small difference between the two samples when integrated from 135-126 ppm is likely due to an artifact in the data. When analyzing the cross-polarization dynamics, it is customary to analyze individual peaks because different carbons in the same molecule will exhibit different dynamics. However, the insulin samples exhibit insufficient resolution of individual peaks to allow accurate measurement; thus regions representing different carbon types have been integrated and their cross polarization profiles are shown in Fig. S5. There appear to be no significant differences in the cross polarization dynamics between the two human insulin samples. In all of the plots shown in Fig. S5, with the exception of the integration of 135-126 ppm, the profiles of both samples overlay very well with one another. In the case of the 135-126 ppm integration, the lyophilized human insulin preparation has a background signal (as with the 1H T1 data set, which was caused by the Torlon® end caps that were used) that served to increase the absolute peak area in this region of the spectrum. However, the general shape of the profile for both samples are very similar and do overlay very well when the peak areas are normalized; thus there do not seem to be any differences in the samples.

Figure S4.Comparison of1H T1 relaxation rates between thehumaninsulin from Roche (red) and the lyophilized human insulin preparation (blue). Integrated areas are shown for “individual peaks” or the entire spectrum (186-7 ppm) with error bars showing the error of the non-linear fit from the data analysis, and the peak average is the average of the “individual peaks” with the error bars showing the standard deviation of the “individual peaks”.

Figure S5.Comparison of the cross polarization profiles between the human insulin from Roche (red) and the lyophilized human insulin preparation (blue). 179-167 ppm (top left), 135-126 ppm (top right), 65-50 ppm (bottom left), and 40-20 ppm (top right).

Figure S6.Product I. CID mass spectrum obtained by means of a FT-MS mass spectrometer of the product with m/z 1607.7 generated by UV-irradiation of solid human insulin under Ar-atmosphere.After derivatization with NEM and digestion of insulin with endoproteinase Glu-C, the peptide corresponds to the sequence FVNQHLC(+125)GSHLVE. The residue C+125 corresponds to the cysteine residue derivatized with one molecule of N-ethylmaleimide.

Figure S7.Product II. CID mass spectrum obtained by means of a FT-MS mass spectrometer of the product with m/z 990.4 generated by UV-irradiation of solid human insulin under Ar-atmosphere. After derivatization with NEM and digestion of insulin with endoproteinase Glu-C, the peptide corresponds to the sequence ALYLVC(+125)GE. The residue C+125 corresponds to the cysteine residue derivatized with one molecule of N-ethylmaleimide.

Figure S8.Circular dichroism spectra of insulin samples measured in a Jasco J-815 spectropolarimeter equipped with a 2 mm path cell. Native insulin (…, dotted line), insulin irradiated at 254 nm in aqueous solution (- - -, dashed line), solid state insulin irradiated at 254 nm (-, plain line). Insulin from Millipore was UV-irradiated at a concentration of 1 mg/mL in H2O. The circular dichroism spectra were recorded with insulin samples at a concentration of 0.1 mg/mL. The UV-irradiated samples were diluted in milliQ water.

Native insulin (…, dotted line) and solid insulin photo-irradiated (-, plain line) - The presence of the positive absorption band at 190 nm with the negative absorption bands at 208 nm and 222 nm characterize the α-helix structures. UV-exposure of solid insulin does not affect the secondary structure of insulin.

Photo-irradiated insulin in solution (- - -, dashed line) –The absence of the absorption bands at 208 nm and 222 nm reveals the loss of helicity of the protein.