|Home | About | Journals | Submit | Contact Us | Français|
AKR/J mice all display a hair interior defect (hid) phenotype for which the molecular basis is unknown. To investigate application of hair shaft proteomics to study of such diseases, pelage from AKR/J and two other mouse strains without this defect were analyzed by shotgun proteomics. The results permitted identification of 111 proteins from tryptic digests of total hair from AKR/J-hid/hid mice, which were predominantly keratins and keratin-associated proteins. From the non-solubilizable (cross-linked) fraction of the hair remaining after extensive detergent extraction, 58 proteins were identified. The majority were keratins and keratin-associated proteins, but junctional and other membrane proteins, cytoplasmic proteins and histones were also identified. The results indicate incorporation of a multitude of proteins into highly cross-linked material. Comparison of unique peptides generated among hair samples from AKR/J-hid/hid, FVB/NJ +/+, and LP/J +/+ mice indicated these inbred strains could be distinguished by proteomic pattern. Transmission electron microscopy after mild treatment in detergent and reducing agent permitted visualization of projections of cortex cells, with characteristic filament patterns, into adjoining medulla cells. Hair shafts from AKR/J mice were deficient in these projections and also exhibited relatively low levels of trichohyalin, a possible contributor to or marker for the hid phenotype.
The mature hair shaft is an intricate structure consisting of an outer layer of cuticle cells surrounding an annular layer of cortex cells with a column of medulla cells in the center. This complex structure consists primarily of keratin and keratin associated proteins stabilized by disulfide bonds. Providing further stabilization are numerous isopeptide bonds, arising from transglutaminase action, that prevent solubilization of ≈ 20% of the total protein in human hair by powerful denaturing conditions. In addition to the keratins and keratin associated proteins, the isopeptide cross-linked fraction of human hair consists of junctional and other membrane proteins and cytoplasmic proteins (Lee et al., 2006). In the extensively extracted hair shaft, cells of the cortex appear essentially empty, while those in the cuticle appear largely resistant to extraction, and those in the medulla exhibit remnant nuclei and much amorphous condensed protein as well as empty spaces (Rice et al., 1994). The degree of resistance to extraction can be greatly affected by aberrant transglutaminase activity in human hair and by other genetic defects in human and mouse hair (Rice et al., 1999; Rice et al., 1996).
Mice with hair defects provide a wealth of models with which to explore the genetic basis of hair and skin diseases (Sundberg and King, 1996). Among these, all AKR/J mice exhibit a defect in organization of medulla cells in the hair shaft due to a mutation in the hid (hair interior defect) locus on Chromosome 1 (Giehl et al., 2009). Compared to those in other strains, the medulla cells appear irregularly spaced instead of forming regular columns and rows and lack projections from cortical cells that appear to preserve regular spacing (Sundberg, 1994; Trigg, 1972). To complement genetic efforts to find the molecular basis for this condition, present work has focused on examining mouse hair proteins by shotgun proteomics. A semi-quantitative comparison of hair proteins from AKR/J-hid/hid mice with those from two other closely related strains that do not show this defect (FVB/NJ +/+, LP/J +/+) was done. In the process, relative amounts of the various proteins identified were estimated and their propensity to undergo transglutaminse cross-linking was revealed.
Pelage hair of AKR/J-hid/hid mutant mice were compared to those of LP/J +/+ and FVB/NJ +/+ strains which do not display the hair interior defect phenotype. The AKR/J-hid/hid mutant hair is readily distinguishable from the hair of the other strains by gross inspection or by light microscopy as previously observed (Giehl et al., 2009). For examination by transmission electron microscopy, the hair shafts were permitted to swell for 2–2.5 hr at room temperature in 2% sodium dodecyl sulfate at neutral pH in the presence of 10 mM dithioerythritol before fixation. Figure 1 illustrates representative electron micrographs of longitudinal sections of the hair of all three strains. These show characteristic cross-linked protein features previously seen after more extensive extraction (Giehl et al., 2009) and, with the minimal extraction presently employed, illustrate much filamentous detail in addition. While individual hair shafts exhibited variability, samples from LP/J and FVB/NJ strains displayed deep projections from cells of the cortex into the medulla cells. An example of such a structure at higher magnification (Figure 2, right panel) shows the pattern of filaments in it oriented largely perpendicular to the filaments in the rest of the cortex, which are parallel to the length of the shaft. A consistent strain difference was the paucity of such indentations in hair from AKR/J-hid/hid mice. The indentations in rare instances looked like those in hair from the other two strains, but in most cases were much smaller, barely noticeable or not visible (Figure 2, left panel). This observation parallels extensive findings by white and polarized light microscopy (Giehl et al., 2009).
Shotgun proteomic analysis of unfractionated hair shaft samples of mouse pelage proved satisfactory to identify >100 proteins comprising major and minor components (Table S1). Semi-quantitative estimates of relative abundance (average of 3 samples from AKR/J-hid/hid mice) using the exponentially modified protein abundance index (Ishihama et al., 2005) indicated that ≈ 95% of the protein consisted of keratins and keratin-associated proteins (Figure 3). A variety of cytoplasmic proteins were also identified (2–3%), as well as several histones (1–2%) and junctional and other membrane proteins (≈ 0.5%). Table 1 shows the most prominent proteins identified based on emPAI values. A complete list is given in Table S1.
Previous analysis of human hair shaft showed that the components in the cross-linked fraction, isolated by repeated extraction of solubilizable proteins with ionic detergent under reducing conditions, differ in prominence from those that are detergent-extractable (Lee et al., 2006). Present analysis gives a similar result with mouse hair and includes a semiquantitative comparison based on emPAI. As shown in Figure 3, keratins and keratin-associated proteins comprised nearly 50% of the non-solubilizable material while histones comprised nearly 40%. Cytoplasmic proteins comprised most of the remainder (≈ 10%), with junctional and membrane proteins comprising ≈ 2%. As shown in Table 1, estimates for individual proteins can vary considerably in the total versus the cross-linked fractions. Overall, more proteins were identified in the unfractionated hair samples (111) than in the cross-linked fraction (58), with 55 being identified in both. Figure S1 shows the distribution of proteins identified in these two fractions.
Protein profiles of pelage hair from the other two mouse strains (LP/J and FVB/NJ) were generated in parallel. These latter two strains are genetically close (FVB/NJ) or very close (LP/J) to AKR/J-hid/hid by descent (Petkov et al., 2004), so few variations not due to the hid mutation were anticipated. Despite the same general lineup of identified proteins, however, close inspection revealed that the relative abundances of certain constituents varied considerably, with hair from the three strains showing characteristic patterns of unique peptides. The differences were most obvious in unique peptides tabulated from the digests of total hair. Of the 124 identified proteins, 65 were significantly different in abundance across strains. Of these, 42 were significantly different between the two normal strains (FVB/NJ +/+ and LP/J +/+), 51 were significantly different between LP/J +/+ and AKR/J-hid/hid, and 15 were significantly different between FVB/NJ +/+ and AKR/J-hid/hid. Complete results of the statistical testing are included as Table S2.
Table 2 lists some of the proteins among those exhibiting significant differences above along with the number of unique peptides identified (from Table S1). As illustrated, hair from the LP/J+/+ strain was relatively low in several keratins (e.g., Krt81 and 33a) and keratin associated proteins (Krtap13-1, 14, 7-1), while hair from the FVB/NJ strain exhibited relatively low levels of transglutaminase 3 (Tgm3). The latter hair also had lower levels of trichohyalin (Tchh) than hair from the LP/J strain, but hair from the AKR/J strain had by far the lowest Tchh levels. On the other hand, 7 proteins were detected at low levels in hair from the AKR/J strain that were not clearly present in the other two strains. Of these, one appears to have primarily a structural function (novel Krtap4 family member), three participate in protein synthesis, two are kinases and one is a heat shock protein.
The proteins identified in unfractionated mouse pelage hair were overwhelmingly keratins and keratin-associated proteins, while junctional and other membrane proteins, cytoplasmic proteins and histones were minor components. The cross-linked fraction, comprising ≈ 20% of the total, had substantially higher levels of the last three categories. Although differing in estimated abundance in the two categories, nearly all the proteins identified in the cross-linked fraction (55 of 58) were also identified in the total fraction. This observation emphasizes the propensity of the corneocytes of hair to incorporate a variety of proteins into cross-linked structures, likely based on their availability as well as suitability as transglutaminase substrates, a variation of the “dustbin” hypothesis for formation of epidermal cross-linked envelopes (Michel et al., 1987) as proposed for those formed in culture (Robinson et al., 1997). Some differences in protein components are evident in comparison with previous results from human hair. Inasmuch as the medulla of human hair is a minor constituent, the much higher relative levels of trichohyalin and histones in mouse hair reflect the much higher proportion of the total protein being derived from medulla cells.
Transmission electron microscopy permitted verification of a major defect in AKR/J-hid/hid hair shaft compared to the other strains. The observed lack of projections from cells of the cortex into those of the medulla could account for the tendency of the latter not to form as regular arrays as ordinarily seen in mouse pelage hair. As previously observed by us (Giehl et al., 2009; Rice et al., 1994) and others, sectioning unmodified, mature hair shafts is quite difficult due to poor embedding. Allowing the hair shaft to swell briefly in detergent with reducing agent before fixation permitted effective embedment and consequent visualization of filament protein networks in the cortex. Electron micrographs revealed a contrast in cortical filament organization in the cortical projections at the medulla boundary, as previously demonstrated in rat hairs (Morioka, 2004). This contrast suggested that altered structure could reflect altered protein content, a possibility that was realized. The possibility that the relatively low level of trichohyalin in the AKR/J-hid/hid mice affects proper orientation of the medulla cells merits consideration either as a causative factor in the phenotype or as a downstream marker for the genetic defect. Further study of mice with potential trichohyalin defects (e.g., matted) would thus be of interest.
The identified keratin components are largely consonant with the known keratin content of human hair (Langbein et al., 2007; Schweizer et al., 2007). Several noteworthy discrepancies are the lack of keratins 36, 37, 38 and 40 and the finding of keratins 16, 17 and 84. Since the present study focused on prominent constituents, the first group of keratins could have escaped detection due to low expression levels in mouse hair and/or low yields due to intrinsic properties of the generated peptides. Degradation or cross-linking during the shaft maturation process could also contribute, possibly modifying relative yields of detected proteins as well. The second group of keratins was detected in present work at low levels. Keratin 75 has been detected in the medulla (Wang et al., 2003), but the lack of keratins 25, 26 and 28, also found in medulla (Langbein et al., 2006), is curious. In a careful study, keratin 84 has been detected in mouse tail scales and dorsal tongue papillae but not in the hair (Langbein et al., 2001). In human hair shaft, we previously detected keratins 25, 38, 40 and 84 at low levels using an additional step of peptide fractionation (ion exchange chromatography) prior to mass spectrometry (Lee et al., 2006), and more recently keratins 36, 37, 38 and 40 have been detected at low levels using the present procedure (Shimomura et al., 2009). We anticipate that quantitative methods will eventually reconcile such discrepancies that probably reflect low expression levels and/or peptide yields,
Quantitation of protein abundance in shotgun proteomics is a major challenge. Present measurements were designed primarily to identify prominent proteins of the mouse hair shaft. For this purpose, ranking proteins in order of emPAI values is a convenience, although the values are only semi-quantitative (Ishihama et al., 2008). Future efforts to develop a targeted approach focusing on proteotypic peptides could provide more accurate quantitation (Lange et al., 2008). Nevertheless, if the demonstrated differences in hair from mouse strains are indicative of hair variations in human races, this approach may be of value for human populations.
Hair samples were obtained by shearing with electric clippers from adult AKR/J (JR#648, 45 days old), FVB/NJ (JR#1800, 41 days old) and LP/J (JR#676, 82 days old) mice from The Jackson Laboratory (Bar Harbor, ME). Since all the hairs in AKR/J mice exhibit the hid defect, they were not separated into the 4 major types.
Samples of mouse hair (several mg) were incubated at room temperature in 2 ml of 0.1 M sodium phosphate buffer containing 2% sodium dodecyl sulfate and 10 mM dithioerythritol. When the hair became swollen, in comparison to a sample incubated in parallel without dithiothreitol (2–2.5 hr), it was immersed in Karnovsky’s fixative, postfixed with osmium tetroxide, and embedded in epoxy resin combination (Araldite and EMBed). Thin sections (80 nm) were stained with uranyl acetate and lead citrate and examined by transmission electron microscopy using a Philips CM120 instrument.
As previously described (Lee et al., 2006) with minor modifications, a sample of hair shaft (6–8 mg) was rinsed briefly in 2% SDS to remove loosely adhering contaminants, incubated overnight at 70°C in 5 ml of 2% SDS – 0.1 M sodium phosphate (pH 7.8) – 20 mM DTE and pulverized by stirring for several hours with a small magnetic stirring bar. Insoluble material was recovered from parallel samples (23–47 mg of hair shaft) by centrifugation, extracted 4 more times, with the protein content of the extract being monitored to ensure complete extraction. The unfractionated hair and the insoluble fractions (18 ± 7% of the total) were alkylated with iodoacetamide, digested for three days at room temperature with reductively methylated trypsin (Rice et al., 1977) in fresh 0.1 M ammonium bicarbonate containing either 10% acetonitrile or 0.05% RapiGest (Waters Corp, Milford, MA). The undigested residue of the unfractionated hair was negligible, while that of the insoluble fraction was 11 ± 2% of the starting material judging by ninhydrin reaction after digestion with 10% sulfuric acid. Total hair samples were analyzed from two males and one female of each strain. The cross-linked samples were derived from two males of each strain (FVB/NJ, LP/J) or two males and one female (AKR/J).
Our previous hair analysis subjected the digest to ion exchange and reverse phase column chromatography prior to mass spectrometric protein identification (Lee et al., 2006). For the present purpose, a streamlined approach using only the reverse phase separation has proven sufficient to identify numerous prominent constituents and to permit semi-quantitative comparison among the samples. The NanoLC-2D system (Eksigent, Dublin, CA) coupled with a Thermo Finnigan (San Jose, CA) LTQ ion trap mass spectrometer was used with a home-made fritless reverse phase micro-capillary column (75μm×180mm; packed with Magic C18AQ, 3μm 100Å; Michrom BioResources) in a vented configuration. 15 μg of each digested sample was loaded on a trap column (Zorbax 300SB-C18, 5 μm, 0.3 mm ×5 mm: Agilent, Santa Clara, CA) and desalted. Peptides were then eluted from the trap, separated with the reverse phase capillary column by gradient elution at a flow rate of 300 nl/min, and directly sprayed into the mass spectrometer. An MS survey scan was followed by ten consecutive MS/MS scans over the liquid chromatography gradient. The LTQ parameters were as follows: electrospray potential, 1.8 kV; source temperature, 180°C; collision energy, 35%; dynamic exclusion duration, 1 min. A 107 min gradient (2–40% B for 95 min, 40–80% B for 12 min) was used with 0.1% formic acid in water (buffer A) and 0.1% formic acid in acetonitrile (buffer B).
Tandem mass spectra were extracted using extractmsn.exe (Xcalibur, v2.0 SR2, Thermo). Database searching was performed on MASCOT (v.2.2.0) against the IPI mouse database (version 7/25/08). Search parameters were as follows: peptide tolerance, 2.5 Da; MS/MS tolerance, 0.6 Da; fixed carbamidomethylation on Cys; variable oxidation on Met; enzymatic digestion, semitrypsin. Scaffold (version Scaffold-2_00_02, Proteome Software Inc., Portland, OR) was used to validate MS/MS based peptide and protein identifications. Peptide identifications, manually curated, were accepted if they could be established at greater than 95.0% probability as specified by the Peptide Prophet algorithm (Keller et al., 2002). Protein identifications were accepted if they could be established at greater than 99.0% probability and contained at least 2 identified peptides. Protein probabilities were assigned by the Protein Prophet algorithm (Nesvizhskii et al., 2003). Proteins that contained similar peptides and could not be distinguished based on MS/MS analysis alone were grouped according to parsimony. Numbers of unique peptides were tabulated as a basis for selecting prominent proteins for further analysis. From Mascot database reports (cut-off score 35), emPAI values were tabulated for the prominent proteins across three samples in a given category. Estimates of relative molar amount were calculated by normalizing to the total emPAI values for a given category (Ishihama et al., 2005).
We used a method similar to that of (Ishihama et al., 2005) to accomodate the semi-quantitative nature of the data. The method is based on a response that is the number of unique peptides for each sample for each identified protein. Although this does not measure concentration, for a given protein it should be correlated with concentration. We assumed that the number of unique proteins in a sample is Poisson distributed, a typical assumption for count data, with an average frequency that is a monotonic function of the concentration of the protein. We used Poisson regression (glm function in R) to test for differences among and between strains. This is a one-way ANOVA, but with Poisson regression instead of ordinary least squares, and can be accomplished by a wide variety of statistical packages. We considered the strains to be statistically distinct if the p-value from the three-strain comparison was less than 0.05. The data and the code in R to perform the analysis in the paper are available from Dr. Rocke.
Numbers of proteins in each segment of the Venn diagram are listed for the AKR/J-hid/hid strain.
Listed are the proteins identified, numbers of proteins matched, IPI accession numbers (Accession), molecular weights in AMU (MW), location categories (Loc) of K (keratin and keratin-associated protein), H (histone), C (intracellular/cytoplasmic) and M (membrane/junctional), number of unique peptides in each sample, emPAI values from Mascot output, average emPAI (Ave), standard deviation of emPAI values (SD), and emPAI value normalized to 100 total (Norm) for the identified proteins. Peptides from 3 samples of total hair are identified as F (FVB/NJ), L (LP/J) or A (AKR/J) with the mouse number and dorsal (D) or ventral (V) source of hair. To the right of the samples of total hair are given unique peptides obtained from the cross-linked fraction of 3 AKR/J, 2 FVB/NJ and 2 LP/J mice (each D and V). The D and V samples did not show consistent differences and were treated as replicates.
Listed are the proteins identified, IPI accession numbers (Accession) and p-values from Poisson regression for differences overall (All) or in pairwise comparisons among FVB/NJ (F), LP/J (L) and AKR/J (A). Values <0.05 are highlighted in pink.
This work was supported by Public Health Service Grants 2 P42 ES04699, NHGRI R01-HG003352 and AR047204. We thank Grete Adamson and Pat Kysar for assistance in electron microscopy.
CONFLICT OF INTEREST
The authors state no conflict of interest.