G. A. Digenis, R.G. Buice, B. Gold, and R.A. Lodder,
College of Pharmacy
University of Kentucky Medical Center
Lexington, KY 40536-0082

The fifteen-minute presentation began with a general discussion of molecular structure and the near-IR absorbance signals that various protein molecular structures generate. Once the relevant signals were described, near-IR imaging spectrometry of hard gelatin capsules (HGCs) was demonstrated.

The aim of this research was to test the hypothesis that chemical crosslinking reactions contribute to the observed decrease in the ability of the formaldehyde-stressed HGCs to absorb water. Spatially resolved water absorption can be determined by near-IR imaging spectrometry of HGCs (figure shown).

A fluorimetric assay was used to determine formaldehyde concentration inside the desiccator chamber in which HGCs were incubated. Figure 1 represents the standard curve (r2=0.992), which included formaldehyde concentrations from 25 to 200 ng/ml (ppb). This represented a range of 50 - 400 ng/ml of formaldehyde standard solution as prepared, since the latter solutions were diluted with an equal volume of formaldehyde reagent. The fluorescence intensities of the solutions containing the 1 h desiccator-sampled atmosphere from each of the five experiment groups (2.25, 4.60, 9.24, 16.0, and 24.0 h exposure of 12 HGCs to formaldehyde vapor) were determined. From the standard curve (Figure 1), concentration values of the five experiment groups were extrapolated, the mean of which was 145.4 ng/ml of formaldehyde present in the desiccator atmosphere (145.4 ppb). This value agreed closely with the theoretical calculation of 151.7 ppb, in which the mass of formaldehyde (400.7 ng) was divided by the volume of the desiccator (2640 ml) into which the formaldehyde was introduced. The determination of formaldehyde in some methods of analysis may be complicated by its existence in both monomeric and polymeric forms. However, the assay utilized in the present work which involves the reaction of one mole ammonia, two moles 2,4-pentanedione, and one mole formaldehyde, requires the latter to be in its monomeric form. Since formaldehyde hydrate (monomer) is consumed in the aforementioned reaction, the equilibrium of formaldehyde in water shifts toward the monomer, as predicted by the law of mass action. Therefore, whether the formaldehyde sampled from the desiccator chamber atmosphere was in polymeric or monomeric form is of little consequence, because the reaction equilibrium assured that all formaldehyde which was present in the aqueous solution of ammonia and 2,4-pentanedione would be available for assay. In Figure 2, the obtained dissolution profiles (pH 1.2 medium) for amoxicillin from HGCs are shown. With increasing exposure time (up to 16 h) of the HGCs to the 150 ppb formaldehyde atmosphere, the percentage of amoxicillin dissolved at any time point decreased (Figure 2 and Figure 3). However, the dissolution curves of HGCs, exposed to either 16 or 24 h of 150 ppb formaldehyde, were virtually superimposable (Figure 2). The similarity of the two dissolution curves (16 and 24 h, Figure 2) may be the result of the fact that, given ample time (16 or more hours), the formaldehyde (150 ppb) inside the desiccator was the limiting reagent with respect to the reactive sites, E-amino and guanidino functionalities of lysine and arginine residues, respectively, within the gelatin polypeptide. Further, the presence of significant amoxicillin in the 90 minute dissolution samples from the HGCs exposed for 16 and 24 h to 150 ppb formaldehyde (~40% amoxicillin dissolved at 90 minutes, respectively, Figure 2) supports the hypothesis that not all of the reactive sites on the gelatin molecule had been modified by formaldehyde. In general, reaction of gelatin with excessive quantities of formaldehyde produces crosslinked, hydrophobic peptide chains of increased average molecular weight which are no longer soluble in aqueous media and through which drug is not able to migrate.

Figure 4 is the NIR reflectance spectra of fresh and formaldehyde stressed HGCs. Because the NIR spectra contained baseline shifts which were both additive and multiplicative, simple calibration methods were not suitable for data analysis. Instead, principal component regression (PCR) was used to analyze the NIR spectra. PCR reduces the dimensionality of a data set, in which there are a large number of correlated variables, by linear transformation into a new set of uncorrelated variables called principal components (PCs). The PCs are structured so that the first few retain most of the variation contained in all the original variables. Thus, the first PC contains information from the constituent which contributes most to the total NIR spectral variation of the data. The second PC is orthogonal to the first and weights most heavily the wavelengths which contribute the most variation to the spectra after removal of the first PC. Progressively smaller contributions to the spectral variation are described by additional, orthogonal PCs. Most of the variation in the NIR spectra (96%) was described by six principal components. Examination of the transformation matrix connecting wavelength and PC hyperspace showed that the signal on the first PC was due to water. The latter has NIR absorbances at 1450 and 1940 nm, and these peaks' intensities were found to decrease with the capsules' increasing time spent in the formaldehyde atmosphere (Figure 4). Since water is a product of many crosslinking phenomena, including that which occurs between formaldehyde and gelatin, the exclusion of water from the HGC shell with increasing time of exposure to formaldehyde (NIR spectra, Figure 4) substantiates the idea that covalent crosslinking is indeed occurring. The first six PCs of the capsule spectra were regressed against percentage of amoxicillin dissolved at 45 minutes from the HGCs exposed to 150 ppb formaldehyde (Figure 5).

The NIR spectra of the capsules show a strong linear correlation to actual dissolution of amoxicillin from the formaldehyde-stressed HGCs (Figure 5, SEE (standard error of estimate) = 6.23%, SEP (standard error of prediction) = 7.67%). The dissolution experiments to which the NIR spectra were correlated were conducted in acidic medium because of amoxicillin's superior solubility at low pH. Thus, correlation was established not between the inherent solubility of drug and the NIR spectral data, but rather, between the dissolution from the formaldehyde-stressed capsule shell and its respective NIR spectrum. The effect of acid on the methylene crosslink, which covalently links the E-amino nitrogen of lysine with the guanidino nitrogen of arginine to form an aminal, has not been established definitively. Recently, it has been demonstrated, using 13C-NMR and 13C enriched formaldehyde, that crosslinking of aqueous gelatin solutions does not occur as readily at pH 2 as in neutral (pH 7) or even alkaline (pH 13) media. This suggests that the methylene crosslink (aminal) in gelatin behaves similar to an acetal and thus, may be susceptible to acid hydrolysis. Correlation of dissolution data with NIR spectra may therefore have been further improved had dissolution occurred in neutral medium, where the formaldehyde-produced methylene crosslink would be less likely to undergo hydrolysis.

After removal of the first principal component, which was determined by matrix transformation to be due to water, the capsule spectra were reconstructed from PCs where crosslinking appears, with wavelength plotted against standard deviation units (Figure 6). The NIR spectra of fresh control HGCs was selected as the zero standard deviation point on the graph in Figure 6. At 1734 and 1782 nm, most of the spectra of the formaldehyde-stressed capsules deviate from spectra of control (unstressed) HGCs in a positive direction (Figure 6). Conversely, the capsule spectra show both positive and negative deviations from spectra of unstressed HGCs at 1760 nm (Figure 6). Infrared spectra (2500-15000 nm) are the result of rhythmical changes in the dipole moment of molecules which absorb vibrational energy. These fundamental molecular vibrations, representing either bond stretching or bending, give rise to harmonic and overtone frequencies which are observed in the near infrared (1000-2500 nm) region. Thus, simple IR spectra (a few absorption frequencies) often give rise to more complicated (many absorption frequencies) NIR spectra. More information can therefore be obtained about chemical environment from the NIR spectral region than from the IR region. The chemical bonds broken and formed during formaldehyde-induced crosslinking reactions in gelatin absorb vibrational energy and give rise to fundamental IR frequencies as well as harmonics in the NIR region.

Images showing water absorption by HGCs were obtained by near-IR spectrometry every 15 min for 1 hour. Exposure to water occurred by sealing capsules into scintillation vials in which a piece of gauze, saturated with water, was fastened into the cap. The vials were maintained on their sides throughout the experiment to prevent droplets of water from falling onto the capsules. In this configuration, one end of the capsule is closer to the water-saturated gauze than the other end, making it a simple matter to observe a gradient in water absorption in the images, and demonstrating the ability of near-IR imaging to achieve spatial resolution of the analyte. In this case, near-IR imaging is able to resolve water uptake even at wavelengths well away from the water absorbance peak at 1940 nm. This ability arises from the intense background absorbance of water, which in turn arises from the strong dipole of the water molecule.

Graphing HGC water exposure for whole capsules (every 15 min for one hour) using absorbance images at obtained at 1671 nm yields a curve that strongly resembles that of diffusion-limited absorption.

CONCLUSIONS

In vitro dissolution of amoxicillin from the capsules at pH 1.2 decreased with increasing time of exposure to the formaldehyde atmosphere. Water content of the capsules was the largest determinant in the variation between HGC spectra at each exposure time, with NIR absorbances of the capsules at 1450 and 1944 nm decreasing with increasing exposure time to formaldehyde.

The hypothesis that chemical reactions contribute to the decreased ability of the formaldehyde-stressed HGCs to imbibe water was supported by NIR spectral reconstruction from principal components where crosslinking occurred. At 1734 and 1782 nm, most of the spectra of the formaldehyde stressed capsules deviate from spectra of control (unstressed) HGCs in a positive direction.

The NIR spectra of the capsules show a strong correlation to actual dissolution of amoxicillin from the formaldehyde-stressed HGCs (SEE (standard error of estimate) = 6.23%, SEP (standard error of prediction) = 7.67%). The dissolution experiments to which the NIR spectra were correlated were conducted in acidic medium because of amoxicillin's superior solubility at low pH. Correlation was established not between the inherent solubility of drug and the NIR spectral data, but rather, between the dissolution from the formaldehyde-stressed capsule shell and its respective NIR spectrum.

Return to ASRG Home Page