Overall, we identifi ed 362 different proteins across all samples but restricted the statisti- cal comparisons to 217 protein species for which we had confi dent protein identifi cations and that were quantifi ed in at least 75% of the samples. A twofold approach was taken for the analysis of the proteomics data. First, a supervised multivariate clustering method was used to compare overall protein profi les between the different samples. Second, uni- variate analyses were performed to fi nd proteins that had signifi cantly different abundance levels between the two experimental groups. In the supervised multivariate analysis, a clear discrimination of the two experimen- tal groups was observed based on two components (Figure 1). Key observations were as follows: fi rst, many of the duplicates in the pooled samples were in close proximity to each other. This means that the overall protein profi les in these duplicate samples were very similar, which indicates good proteomic profi ling performance. Second, a clear difference between the very straight and the very curly samples was seen when taking into account the overall profi les of the 217 confi dently identifi ed proteins per sample, and this discrimination was mainly based on the fi rst component (x-axis) of the PLS-DA plot, which explains 28% of the variance. Differences in structural proteins are likely to inform us about potential differences in hair performance and those associated with insults and interventions. When looking at the top 20 proteins contributing to the discrimination of the curly versus straight hair shape in the fi rst component (Figure 1, Table I), most can be linked to structural protein components of hair, that is, keratin or KAP families, with keratin K85, the strongest contributor to the distinction of the very straight hair shaft pools, whereas KAP 13-2 has the most discrimination power in the very curly hair pooled samples. Interestingly, K85, one of the fi rst keratins to be expressed during the hair growth process, has been shown to be expressed across the entire developing shaft in straight human hair (20) and in curly Wiltshire wool, but in the latter, its distribution is asymmetric (21). However, no evidence for the link between curly fi bers and KAP 13-2 has previously been reported in the literature. Next, univariate analyses were performed for pairwise comparisons between the two experi- mental groups, which highlighted proteins with different levels of expression between the very straight hair group and the very curly hair group. Interestingly, 14 proteins were identifi ed as signifi cantly differentially expressed (Table II). From these 14 proteins, the majority can be linked to the keratin or KAP families, indicating microstructural com- position differences are important for fi ber shape. The remaining nonstructural proteins are interesting because they may echo differences in processes of fi ber growth and matura- tion (e.g., cornifi cation) between fi bers of different shapes. As hair shafts are known to display marked intra- and between-person variation, we opted to visualize the protein abundance levels of each of the 20 samples measured per experi- mental group to see whether large differences in these abundances could be found across the samples within an experimental group. Interestingly, for many proteins, the detected abundance levels were quite similar within each experimental group, with only minimal differences between the fi rst quantiles, median, and third quantiles in the boxplot, indicat- ing that the sample pooling approach was an effi cient way to minimize variability across samples (Figure 3). The proteins found to have statistically signifi cant fold changes of greater than two are a good starting point for defi ning protein differences between these hair shape sample populations. However, little is known about the functions of most of these proteins in the hair fi ber. JOURNAL OF COSMETIC SCIENCE 258

PROTEINS WITH A PREVIOUS LINK TO CURL IN HUMAN OR SHEEP HAIR Interpretation of the functions of structural proteins (especially keratins and KAPs) is hindered by the less differentiated cortical organization of human hair than that of sheep wool. All mammalian hair is made of modifi ed cell remnants, with the “cells” composed of structural components called macro-fi brils (keratin fi lament bundles with a KAP matrix), digested waste left over from the cytoplasm and nucleus, and sometimes melanin granules. The cells are glued to one another by a cell membrane complex that is composed of highly transformed plasma membranes, cell junction complexes, and extracellular matrix (22). In wool fi bers, cortical cell remnants form clusters of similar structural organization and protein chemistry. The orthocortex contains helically twisted macro-fi brils and is dominated by high–glycine–tyrosine KAP species, and the paracortex contains macro-fi brils with less twist and a matrix dominated by high-sulfur KAP species (23,24). The cortex characteristics of human hair cell types are less clear cut (25), macrofi bril helical twist varies across the cortex (26), and most cells have high sulfur content. Notable among the structural proteins found to be more abundant in the curly human hair samples is K38 (Figure 2, Table II). K38 has a history of association with hair curl in hu- mans and sheep. In follicles from straight human hair, its distribution across the hair shaft cortex is sporadic but uniform (27), but in Western blot studies of curly human hair, K38 was found on the concave side of the curvature and evenly distributed in straight hair (28). In sheep follicles, K38 expression is restricted to orthocortical cell remnants (21) and has been noted as potentially playing a key role in keratin polymerization and structure self- assembly (29). In sheep wool, the orthocortex is dominated by twisted fi bril architectures within which keratin fi laments are embedded in a mixture of high–glycine–tyrosine KAPs (21,24) and, when associated with one side of the cortex, is associated with high curvature fi bers (12,30,31). Because cortical cell types are less clear-cut in human hair than in wool (25), the association between K38 and different KAP families is unlikely to be the same. However, our fi nding does raise the intriguing prospect that K38 may be associated with a particular level of macrofi bril internal twist or with a particular set of KAP species. KAP4-2, KAP4-4, and KAP9-8 are all ultrahigh-sulfur proteins that, like other KAPs, are major constituents of the matrix between keratin fi laments in macro-fi brils. In wool, members of the KAP4 family have been found to be expressed primarily in the paracortex (21,30,32). Thus, the prominence of KAP4-2 and KAP9-8 in the straight hair fi bers would be consistent with them having some aspect in common with wool paracortex, but whether this is structural (e.g., less intense macrofi bril internal twist) or in terms of protein chemistry has yet to be established. KAP13-2, a high-sulfur family KAP, was found in higher amounts in curly hair (Figure 2, Table II). Although no direct link has been reported in the literature, another protein from the KAP13 family, KAP13-1, is present in signifi cantly higher amounts in curly sheep wool fi bers than in the same-diameter straight fi bers (Plowman et al. 2020 unpublished results). PROTEINS NOT PREVIOUSLY ASSOCIATED WITH HAIR CURL K34 and K81 are cortical proteins that have previously been observed to be uniformly distributed across the follicle in both human straight hair and sheep wool. K40 is notably one of the last keratins to be expressed during the hair formation process and is found HAIR SHAPE PROTEOMICS 259