COMPONENT DISTRIBUTIONS IN KERATINS FROM AMINO ACIDS 25 portion of low-sulfur protein (62% as compared to 44% for hair) and a much lower proportion of high-sulfur proteins, especially when the cuticle fraction is added in. The other mammalian keratins also show reasonable calculated compositions, though agreement is worse (higher R values) as lower mammalian orders are reached. Notable in the proportions of components is an increase in both cuticle and high glycine-tyrosine proteins. A component richer in proline and glycine, and deficient in serine and thre- onine, seems to be missing in the lower orders. A taxonomic correlation of this kind, with more extensive data, may be of some interest. Two examples are given of keratins not closely related to mammalian fibrous keratins: a feather keratin shown in Table IV, and a lizard claw in Table V. Although the regression fit is poor for both, the component proportions are of interest. The feather has a large amount of cuticle-like protein, along with high glycine, as can be seen from the analysis in Table IV. Aside from structural arguments, such a composition for feather may have useful hydrophobic behavior. The lizard claw is unusual in its apparent absence of low-sulfur protein and the high proportion of cuticle-like component. Such a composition would suggest a tough non-brittle polymer. Although detailed calculated analyses are not shown, the component composition of wool orthocortex and paracortex, samples 24 and 25 in Table V, show expected dif- ferences, notably a higher sulfur level for the paracortex region. Human hair, exclusive of cuticle, is thought to be paracortex, though not necessarily of this composition. The human hair samples in Appendix II, with component proportions summarized in Table V, are all from different laboratories. In spite of this, regression fits are,,fairly good, ranging from R's of. 13 to .47. An R of 0.5 has been arbitrarily selected as about the limit of reasonably good fit to the data beyond this there is the suggestion that an additional component may be needgd or that the analyses are not accurate. From the average of 6 human hairs, the component proportions are 44% low sulfur, 35% high and ultra-high sulfur, and 19% cuticle, with 2% high glycine-tyrosine proteins. Since there is considerable correlation between the composition of the. high and ultra-high sulfur components and the cuticle (see Table II), it is not certain how much component spillover may occur. Normal hair has been found by fractionation to contain about 40% high-sulfur proteins (27), in good agreement with the calculated amounts. A similar situation occurs with the wool samples of Table V, namely fairly large component proportion differences among samples from various laboratories. The orig- inal analyses in Appendix II bear this out. Whether one can take seriously individual differences among analyses, in view of their different origins, is uncertain. To do this properly would require a number of analyses from the same laboratory using samples prepared identically. Under such conditions, it should be possible to attempt significant correlation between the composition and the mechanical or chemical behavior of normal or damaged hair. For example, since many hair treatments require substantivity to the hair surface, an estimate of the amount of cuticle fraction may be of use. Similarly, as argued before, hair with an overabundance of microfibrils may be more brittle, or hair with an excess of matrix may be tough but flabby. Coupling of component proportions with appropriate mechanical models should provide a base of knowledge to facilitate safer and more effective treatments. Before leaving the problem of component definition, one other set of experiments should

26 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS be mentioned, because of their somewhat puzzling place in the general picture so far given. By progressive hydrolysis (20) of wool, a soluble fraction can be removed, leaving an insoluble gel fraction whose composition changes linearly with the amount of sol- ubilization (19,49). After extensive hydrolysis, the gel composition strongly resembles that of the high and ultra-high fractions already described. Theory based on a simplified model (50) predicts that the sulfur or half-cystine content of the gel should increase to twice its original value, and this is observed. The interesting part of these results is that this composition split is dependent exclusively on the cleavage of peptide linkages rather than on intermolecular crosslinks. Nevertheless, the high-sulfur component com- positions seem to be quite similar. Furthermore, the fiber behaves according to the simplified theory as though the crosslinks were fairly uniformly distributed throughout the fiber. However, other alternative explanations are possible for this observation (14). Although further experimental work is needed, a more careful modeling of the hydro- lysis process (14,51) should shed further light on the crosslink distribution. SUMMARY The evidence is fairly strong that aqueous swelling in keratin fibers occurs at least as much in the microfibrils as in the matrix, though there is great need for further study and understanding of the location and composition of various structural and morpho- logical components. Estimation of component distributions directly from amino acid analyses appears to be a fruitful area for study. The present work is far from definitive, since it relies on fractionation information and analyses that are only approximate. The complexity of the high-sulfur proteins is the most difficult problem. Although several more inde- pendent components might be added, and regression methods still applied, it would be desirable to keep the number as low as possible. This would facilitate the use of proportion estimation as a kind of fingerprint of the keratin structure, possibly easing the problem of collating information on mechanical properties and chemical reactivity for a wide range of fibrous keratins, or perhaps other families of proteins. ACKNOWLEDGEMENT The author thanks Zotos International, Inc., for its continued support and encourage- ment of this work. APPENDIX I: REGRESSION METHOD FOR COMPONENT PROPORTIONS We assume that a given keratin composition can be reproduced as the linear combi- nation of N components, where, in our case, N is between 2 and 5. Three components are used to illustrate the method. For each amino acid present, we can write a material balance x = ax• + bx2 + (1 - a - b) x 3 In this expression, x• is the fraction of a particular amino acid in component 1, x2 is the fraction of the same amino acid in component 2, and x 3 the fraction in component