Epoxyeicosatrienoic Acids Activate Transglutaminases in Situ and Induce Cornification of Epidermal Keratinocytes*

The cytochrome P450 CYP2B19 is a keratinocyte-spe-cific arachidonic acid epoxygenase expressed in the granular cell layer of mouse epidermis. In cultured keratinocytes, CYP2B19 mRNAs are up-regulated coordi-nately with those of profilaggrin, another granular cell-specific marker. We investigated effects of the CYP2B19 metabolites 11,12- and 14,15-epoxyeicosatrienoic acids (EETs) on keratinocyte transglutaminase activities and cornified cell envelope formation. Keratinocytes were differentiated in vitro in the presence of biotinylated cadaverine. Transglutaminases cross-linked this substrate into endogenous proteins in situ ; an enzyme-linked immunosorbent assay was used to quantify the biotinylated proteins. Exogenously added or endogenously formed 14,15-EET increased transglutaminase cross-linking activities in cultured human and mouse epidermal keratinocytes in a modified in situ assay. Transglutaminase activities increased (cid:1) 8-fold ( p < 0.02 versus mock control) in human keratinocytes transduced with adenovirus particles expressing a 14 S ,15 R EET epoxygenase (P450 II; Bio-Rad). Nitrocellulose membranes were blocked in 3% milk buffer (3% nonfat dried milk in DPBS) and incubated with antibody diluted in 0.5% milk buffer for (cid:3) 2 h or overnight at 4 ° C. Primary antibodies were anti-human (Biogenesis) or anti-mouse (Co- vance) involucrin at 1:3000 and anti-guinea pig transglutaminase type II (Ab-3, clones CUB 7402 and TG100 at 1:200; Neomarkers). Data were visualized by chemiluminescence using HRP-conjugated secondary antibodies (Renaissance kit; PerkinElmer Life Sciences) and Eastman Kodak Co. X-Omat AR film. Endogenous Epoxyeicosatrienoic Acids— Human keratinocytes were treated as described for adenoviral gene transfer experiments, scraped in the culture medium, and immediately chilled on ice. Triphenylphos- phine was added as a peroxide-reducing agent, and the cells were disrupted by sonification and homogenization (Brinkman Polytron ho- mogenizer; 5 (cid:5) 5-s with 7-mm generator). Organic extracts of the disrupted cells were purified and quantified by gas chromatography mass spectrometry, as described (30, 31). The contents of 14,15-EET and the summed contents of 11,12- and 8,9-EET were measured; 5,6- EET could not be measured reliably by this assay. In the absence of chaotropic agents, there was significant insoluble material in the dis- rupted cells prior to organic extraction. This made protein determina-tions unreliable, so the data were expressed as ng of EET recovered (4-P100s/treatment).

In the outer epidermis, the stratum granulosum gives rise to the stratum corneum. These two cell layers form a functional unit or boundary layer (1) that confers protection against chemical and physical stress, including toxicities, trauma, harmful irradiation, and infections (2,3). It also functions as a semiper-meable membrane and regulates transepidermal water loss (2). Granular keratinocytes comprise the most terminally differentiated, viable cell layer (stratum granulosum) of the epidermis. Their major functions are to synthesize barrier lipid precursors and to undergo two complex processes, cornification and differentiation-induced programmed cell death (4 -7). In vivo, cornification involves extensive, transglutaminase-mediated crosslinking of scaffold proteins and keratin intermediate filaments, to form a ϳ10 -15-nm-thick cornified cell envelope (CE) 1 just beneath the plasma membrane (8). This insoluble structure acquires functional significance as granular cells undergo programmed cell death. The regulation of these processes is poorly understood, and genetic defects resulting in abnormal CEs are associated with skin diseases and defective epithelial barrier functions (9). Granular keratinocytes in mouse epidermis express the cytochrome P450 CYP2B19, which has epoxygenase activity toward arachidonic acid (10,11). Mouse CYP2B19 and its homologues in the rat (CYP2B15, CYP2B12) are all highly expressed in differentiating keratinocytes in the epidermis and skin appendages (11,12). Their metabolites, mainly epoxyeicosatrienoic acids (EETs), are present in epidermis and sebaceous glands (11,12). Evidence that human epidermal keratinocytes generate EETs has been reported (13), but the CYP gene products involved are not yet known, and the physiological significance of epidermal EET synthesis has remained obscure.
P450 epoxygenases are widely expressed in mammalian tissues and are the catalytic source of EETs (14). The substrate arachidonic acid is normally present in cellular glycerophospholipids, where it is released by arachidonic acid-selective lipases and serves as precursor in metabolic and signal transduction pathways. Mechanisms explaining EET biological activities are not yet clearly defined, and specific cellular receptors have not been characterized. Studies of vascular and renal cells showed that EETs function as intracellular signaling molecules regulating cell growth and differentiated cell functions (14). Their pleiotrophic effects include regulation of mitogenesis and cell survival (15)(16)(17)(18)(19), cell adhesion (20), vascular tone, ion channel function, and concentrations of intracellular calcium ions (21)(22)(23).
The restricted expression of the arachidonic acid epoxygenase CYP2B19 and its homologues to the late stages of keratinocyte differentiation (11,12) led us to investigate its potential physiological roles in the epidermis. We report evidence indicating that this 14,15-EET epoxygenase functions to regulate epidermal cornification. Cutaneous P450 epoxygenases provide new investigative approaches to elucidate the complex mechanisms regulating keratinocyte cornification.

EXPERIMENTAL PROCEDURES
Cell Culture-Epidermal keratinocytes were isolated by trypsinization (24) from outbred mice (Ͻ3 days old, CD-1 strain; Charles River) or discarded, human newborn foreskins. Keratinocytes were maintained in 6-well plates as primary cultures (mouse) or passaged three or four times (human). Proliferating cultures were fed every other day with 2 ml of complete Epi-life medium (0.06 mM Ca 2ϩ ; Cascade). Differentiation was induced at ϳ70% confluence (day 0) by adjusting Ca 2ϩ in complete keratinocyte growth medium (Clonetics) to 0.125 mM (mouse) or 1.4 mM (human). Differentiating cultures were fed daily, and drugs were added as indicated. The transglutaminase inhibitor cystamine and amiloride were from Sigma. Arachidonic acid (NuCheck) in ethanol was mixed with 2% fatty acid-free bovine serum albumin (10 l/ml medium; Sigma) when used as medium supplement (25). Epoxy fatty acids were purified by high pressure liquid chromatography. Proliferating cultures were infected with recombinant adenovirus particles at 60 -70% confluence, for 20 Ϯ 2 h. Virus particles were removed by washing twice with Dulbecco's phosphate-buffered saline (DPBS), and the cells were refed with differentiation medium (day 0). Mock-transduced cultures were treated identically but received no virus particles.
Recombinant Adenoviruses-Replication-defective, recombinant adenoviruses (serotype 5; Ad5) were generated by homologous recombination in 293A cells (Qbiogene) as described (26). Open reading frames encoding P450 BM3v (27) and ␤-galactosidase were subcloned into the plasmid pACCMV.pLpA under control of a cytomegalovirus promoter (28). P450 BM3v is a mutated (F87V) form of P450 BM3 originally constructed in the Peterson laboratory (27). P450 BM3v encodes a cytochrome P450 domain fused to a P450 reductase domain, and it functions as a self-sufficient 14S,15R-EET epoxygenase. Recombinant virus particles were subjected to three rounds of plaque purification. The absence of wild type virus particles was evaluated by PCR (29), using Ad2 genomic DNA (Invitrogen) as the positive control template. Virus particles were purified from 293A cell lysates using a single cesium chloride gradient and desalted on P10 Sephadex columns (Amersham Biosciences) (26). Viral expression of recombinant proteins was confirmed by Southern blot, PCR, and Western immunoblotting using rabbit anti-␤-galactosidase (1:2000; Eppendorf 5 Prime Inc.) and rabbit anti-P450 BM3v. Secondary antibodies were horseradish peroxidase (HRP)-conjugated.
Endogenous Epoxyeicosatrienoic Acids-Human keratinocytes were treated as described for adenoviral gene transfer experiments, scraped in the culture medium, and immediately chilled on ice. Triphenylphosphine was added as a peroxide-reducing agent, and the cells were disrupted by sonification and homogenization (Brinkman Polytron homogenizer; 5 ϫ 5-s with 7-mm generator). Organic extracts of the disrupted cells were purified and quantified by gas chromatography mass spectrometry, as described (30,31). The contents of 14,15-EET and the summed contents of 11,12-and 8,9-EET were measured; 5,6-EET could not be measured reliably by this assay. In the absence of chaotropic agents, there was significant insoluble material in the disrupted cells prior to organic extraction. This made protein determinations unreliable, so the data were expressed as ng of EET recovered (4-P100s/treatment).
In Situ Transglutaminase Enzyme Activities-In situ transglutaminase activities were measured in differentiating keratinocyte cultures by modifying the conventional in situ assay (32). Cells were differentiated in the continuous presence of 2 mM biotinylated cadaverine (5-(biotinamido) pentylamine, 50 mg/1.5 ml DPBS; Pierce). Since this cell-permeant transglutaminase substrate competes efficiently with endogenous transglutaminase substrates, transglutaminase-mediated protein biotinylation is controlled by transcriptional and posttranslational regulatory mechanisms and the metabolic conditions existing in the differentiating keratinocytes.
Cell lysates were harvested and Western blotted as described. Streptavidin-HRP (1:8000 -10,000; Pierce) was used to visualize biotinylated proteins electrotransferred to membranes. For the in situ ELISA (33), triplicate wells in 96-well assay plates were coated overnight at 4°C with 0.2 l of cell lysate diluted into 100 l of Tris buffer (0.1 M Tris-HCl, pH 8.3). The wells were washed with buffer three times and blocked for 1 h with 150 l of 3% milk buffer (3% nonfat dried milk in Tris buffer). Wells were washed again and incubated with 100 l of streptavidin-HRP for 4 h (1:1000 in 0.5% milk buffer). Wells were washed again and incubated with 150 l of TMB peroxidase substrate for 5-15 min (Kirkegaard & Perry). Reactions were stopped by adding 150 l of 1.5 N HCl. Absorbance values measured at 450 nm (A 450 ) were linear with respect to g of protein, and the data were expressed as A 450 /g. The limit of detection (A 450 ) was calculated as the mean ϩ 2 S.D. values for lysates containing no exogenous biotin. Norbiotinamine and ⑀-biotinoyl-L-lysine were from Molecular Probes, Inc. (Eugene, OR). The data represent cumulative cross-linking activities, measured as the cellular content of biotinylated protein that accumulated from day 0 until cell lysis.
Protein Matrix-assisted Laser Desorption⁄Ionization (MALDI) Mass Spectrometry-Proteins were separated by one-dimensional SDS-PAGE and stained with Coomassie Brilliant Blue G-250 (Bio-Rad). Excised protein bands were equilibrated in 100 mM NH 4 HCO 3 , reduced with dithiothreitol (3 mM in 100 mM NH 4 HCO 3 , 60°C for 30 min), and alkylated with iodoacetamide (6 mM in 100 mM NH 4 HCO 3 for 30 min in the dark). Gel slices were then equilibrated and dehydrated with 50% acetonitrile, 100 mM NH 4 HCO 3 , and the liquid was discarded after 15 min. They were further dehydrated with acetonitrile and rehydrated with 20 l of 25 mM NH 4 HCO 3 containing 0.02 g/l modified trypsin (Promega), and trypsin digestion was carried out overnight at 37°C. Peptides were extracted with 60% acetonitrile, 0.1% trifluoroacetic acid, dried by vacuum centrifugation, and reconstituted in 10 l of 0.1% trifluoroacetic acid. Peptides were then desalted and concentrated into 2-3 l of 60% acetonitrile, 0.1% trifluoroacetic acid using ZipTipC18 pipette tips (Millipore Corp.). Samples (1 l) were mixed with an equal volume of ␣-cyano-4-hydroxycinnamic acid matrix and applied to a target plate. MALDI-time-of-flight mass spectrometry was carried out using a Voyager 4700 mass spectrometer (Applied Biosystems, Foster City, CA) operated in reflectron mode. The mass spectrum was calibrated to within 50 ppm using trypsin autolytic peptides in the samples (m⁄z ϭ 842.50 and 2211.09 daltons). Ions ([M ϩ H]) corresponding to peptide masses were entered into the MASCOT data base search algorithm (available on the World Wide Web at www.matrixscience.com), which compared the data against proteins present in the NCBI nr data base (allowing for complete carbamidomethylation of cysteine).
Cornified Cell Envelopes-CEs were counted in keratinocyte cultures that cornified spontaneously during in vitro differentiation (n ϭ 3 P60s/treatment) (34). Cells in each dish were trypsinized and resuspended in 1 ml of DPBS, and duplicate aliquots (175 l) were adjusted to 2 mM EDTA to minimize clumping. These cells were centrifuged (4000 ϫ g, 5 min), resuspended in 1.5 ml of dissociation buffer, and boiled for 5 min. The dissociation buffer was 2% SDS, 20 mM dithiothreitol, 5 mM EDTA, 0.1 M Tris-HCl, pH 8.5. Detergent-insoluble CEs were cooled, centrifuged, and resuspended in 50 l of normal saline. The CEs were visualized by phase-contrast microscopy and counted in a hemacytometer. Total cells and total CEs per dish were calculated, and the data were expressed as total CEs/total cells ϫ 100. To count ionophore-inducible CEs, duplicate aliquots of resuspended cells were adjusted to 1.8 mM Ca 2ϩ . Cells were incubated for 2 h at 37°C with gentle rocking after adding X537A to 50 g/ml (generously provided by Robert Rice) (34); then they were treated exactly as described for spontaneous CEs.
Analyses of Messenger RNAs-Poly(A) ϩ RNAs were isolated using a Fast Track kit (Invitrogen). Reverse transcription and PCR were performed as directed (RNA PCR kit; PerkinElmer Life Sciences). For ribonuclease protection assays, total RNAs and [ 32 P]UTP-cRNAs were hybridized as directed (RPA III kit; Ambion). Protected fragments were visualized using a PhosphorImager (Amersham Biosciences). Isotopic counts associated with protected fragments were quantified using ImageQuant software (Amersham Biosciences) and were linear with respect to g of RNA. Gene-specific oligonucleotides (Table I) were used to determine which transglutaminases were expressed in differentiating keratinocytes. The reverse transcription-PCR products were concentrated under vacuum, size-fractionated on agarose gels, and visualized with ethidium bromide. They were evaluated based on predicted sizes and actual DNA sequences.

The CYP2B19 Metabolite 14,15-EET Activates Keratinocyte
Transglutaminases-The CYP2B19 metabolites 11,12-and 14,15-EET are present in measurable quantities in fresh mouse epidermis (11). The Cyp2b19 gene is expressed in the outer epidermis, and its cell type-specific and (late) differentiationspecific expression patterns are known (10,11). Cyp2b19 is first activated transcriptionally in fetal mouse epidermis (approximately embryonic day 15) in the same cell layers as profilaggrin, and this expression persists throughout postnatal development. These in situ hybridization results showed that, like profilaggrin, mouse CYP2B19 is a specific marker for granular keratinocytes. This was corroborated by Northern blot analyses of mouse keratinocyte cultures showing that differentiation induced the coordinate expression of CYP2B19 and profilaggrin mRNAs (11). Between 48 and 96 h, ϳ15-20-fold increases in mRNA levels were achieved, measured by ribonuclease protection assays (Fig. 1). From these expression studies, it seemed unlikely that CYP2B19 would function to induce keratinocyte differentiation, because its up-regulation does not precede other granular cell-specific gene products temporally or spatially. We hypothesized that CYP2B19 regulates integral cell functions in keratinocytes, having already acquired a terminally differentiated phenotype.
Granular keratinocytes in the outer epidermis express CE component proteins and catalytically active transglutaminases (9,35). Transglutaminases cross-link the scaffold proteins involucrin, loricin, and many others, forming insoluble CEs (8,9). We measured effects of CYP2B19 metabolites (11,12-and 14,15-EET) on in situ cross-linking activities as keratinocytes actively differentiated in vitro. This distinction is important because in vitro cross-linking activities (i.e. occurring in disrupted or fractionated cells) are not governed by normal transcriptional and post-translational control mechanisms; nor are they subject to the hormonal and metabolic milieus that regulate transglutaminase isoforms in cells (36,37). Biotinylated cadaverine was the transglutaminase substrate, and in situ cross-linking activities were measured as the content of biotinylated proteins that accumulated intracellularly. Human keratinocytes differentiated in the presence of 14,15-EET (2 M) contained more biotinylated proteins at day 4 and day 6 of treatment compared with the vehicle (Fig. 2). The 14,15-EET regioisomer appeared more biologically active than 11,12-EET, suggesting that keratinocytes discriminate between structurally similar CYP2B19 metabolites.
In human keratinocytes, Ad-BM3v increased in situ transglutaminase activities, compared with controls (Fig. 3A). Like 14,15-EET (Fig. 2), transglutaminase activation was apparent within 4 days. At day 4, the content of biotinylated proteins in P450 BM3v-expressing keratinocytes averaged ϳ8-fold greater (p ϭ 0.01; n ϭ 5), compared with mock cultures (Fig. 4A). At day 4 and day 6, endogenous 14,15-EET in the P450 BM3vexpressing keratinocytes averaged ϳ25-37-fold greater (p ϭ 0.002), compared with controls (Fig. 4B). No differences were observed among treatments in the levels of endogenous 8,9-   I  GCA GCT GGA GAT GGC ACC ATC  CCA CTG GGA TGA TCA CGT G  II  GTT GTT GTC CCA GCG (T/G)CC CAG  AGC TTT GTG CTG GGC CAC TTC  III  TGG AAA AAA TCT GGC TTC AG  GTA GTA CAC ATC CAC ACT GAG  V  AGC TAC CAT GGC CCA AGG GCT AG  CGT CTG GCG CGT TGT TCC AG FIG. 1. Calcium-induced differentiation of mouse epidermal keratinocytes up-regulates CYP2B19 and profilaggrin mRNA levels within 48 h. This is evidence that CYP2B19 is expressed in granular keratinocytes; these cells comprise the most differentiated viable cell layer in the epidermis. Ribonuclease protection assays were used to quantify relative mRNA levels in primary cell cultures. Data are expressed as percentage of values for proliferating keratinocytes (day 0), and they corroborate previous results from Northern blot and in situ hybridization studies (11). G, CYP2B19 mRNAs; f, profilaggrin mRNAs. plus 11,12-EET. These data show that P450 BM3v was catalytically active and generated 14,15-EET from endogenous arachidonic acid in the differentiating keratinocytes. The increased 14,15-EET formation could explain the increased crosslinking activities in P450 BM3v-expressing cells.
In mouse keratinocyte cultures, Ad-BM3v (Fig. 3B) and exogenous EET (not shown) also increased in situ transglutaminase activities. Maximal responses occurred earlier than in human cultures, within 2 or 3 days, and were greater in arachidonic acid-supplemented media (Fig. 3B). These differences were not surprising, since mouse cultures were viable and differentiated over 1-2 weeks, compared with 2-3 weeks for human cultures. The in situ ELISA results corresponding to Fig. 3B showed that the P450 BM3-expressing mouse cultures had nearly ϳ2-fold greater levels of biotinylated proteins compared with mock cultures. In mouse cells, the magnitude of the Ad-BM3v response varied among different primary keratinocyte cultures, but it was repeatable whether medium calcium was optimal (0.125 mM; Fig. 3B) or suboptimal (0.09 mM; not shown) for the expression of late differentiation genes (e.g. CYP2B19, filaggrin, and loricrin) (11,38).
Whether 14,15-EET was provided exogenously (Fig. 2) or generated endogenously (Fig. 3A) in human keratinocytes, a protein of ϳ140 kDa apparent size was preferentially biotinylated in situ, consistent with previous reports using [ 3 H]putrescine or a cadaverine analog (39). The apparent sizes of proteins biotinylated in situ were different for mouse keratinocyte cultures (Fig. 3B), and no single protein appeared preferentially biotinylated, as reported previously (39).
Transglutaminase Substrate Specificity in Situ-Human keratinocytes were treated with amiloride or cystamine to obtain evidence that transglutaminases catalyzed the protein biotinylation measured in these studies. Transglutaminase activities require Ca 2ϩ as cofactor (36,37,40). Amiloride blocks cell Ca 2ϩ influx, and by this mechanism it has been reported to inhibit Ca 2ϩ -induced keratinocyte differentiation and CE formation (41). Cystamine is a water-soluble primary amine that inhibits transglutaminase activities by competing with ⑀-lysines in endogenous protein substrates (42,43). We postulated that cystamine would also compete with biotinylated cadaverine in the in situ assay; it may also inactivate transglutaminase proteins directly (42). Amiloride and cystamine were added simultaneously with biotinylated cadaverine to the differentiation media. At day 6, amiloride at 1 and 10 M reduced protein biotinylation to 87 and 68% of control (vehicle) values, respectively. At day 6, cystamine at 0.1 and 1 mM reduced protein biotinylation to 54 and 70% of control values, respectively. Cystamine was more effective at the dosages tested, and the kinetics of maximal inhibition depended on the dosage (not shown). Cystamine cytotoxicity was observed at 10 mM.
We evaluated whether Ad-BM3v would activate keratinocyte transglutaminases when other biotinylated substrates were present. Kinetically, cadaverine analogs are excellent transglutaminase substrates, and free lysine is a relatively poor substrate (44,45). When biotinylated L-lysine was substituted for cadaverine, levels of biotinylated proteins were consistently low, and the A 450 values were near the ELISA limit of detection (Fig. 4D, day 4). When visualized by Western blotting, the sizes of the weakly detected protein bands appeared qualitatively similar to those obtained using biotinylated cadaverine (not shown). Nonspecific or spurious biotinylated species were not apparent. Cytotoxicity was not apparent for any of the substrates (2 mM) used in these studies, based on cell morphology, cell numbers, and protein contents.
When norbiotinamine was the substrate, absolute A 450 values were lower than those obtained using biotinylated cadaverine, but P450 BM3v-expressing cultures still contained more biotinylated proteins compared with controls (Fig. 4, C and D). Moreover, the sizes of biotinylated proteins were qualitatively similar regardless of how efficiently these substrates appeared to be utilized (cadaverine Ͼ norbiotinamine Ͼ lysine). Relatively high catalytic efficiency was expected for the cadaverine analog based on previous biochemical and kinetic studies of transglutaminase substrate specificities (44 -46). This could explain the greater content of biotinylated proteins, higher A450 values, and increased assay sensitivity obtained using biotinylated cadaverine. Structurally, the primary amino group in cadaverine is separated from the nitrogen of the amide bond by the same number of carbons as in the ⑀-amino group in lysine (47). Also, there is greater distance between the primary amino group and the biotin moiety, compared with norbiotinamine. These results are consistent with transglutaminasemediated protein biotinylation, in response to Ad-BM3v.

The 14,15-EET Epoxygenase P450 BM3v Induces Cross-linking of Physiological Transglutaminase Substrates and Cornified Cell Envelope
Formation-Streptavidin detected a dominant protein of ϳ140 kDa apparent size that was preferentially biotinylated in human keratinocytes (arrow, Fig. 3A), whereas increasing exposure times revealed lower levels of other biotinylated proteins in the whole cell lysates. By Western immunoblotting (Fig. 5A), the ϳ140-kDa biotinylated protein was indistinguishable from immunoreactive involucrin, as reported previously (39). Human involucrin has a predicted size of 68 kDa, but it migrates anomalously in denaturing gels (39,48). Involucrin is a physiological transglutaminase substrate and one of the first scaffold proteins cross-linked in cornifying epidermal cells (49,50). Its preferential biotinylation in situ would provide evidence that P450 BM3v-induced cross-linking is mediated by transglutaminases, mimicking cornification in vivo. To test this hypothesis, we heat-purified involucrin from human keratinocyte cytosols as described (51) and analyzed the heat-treated supernatants by Western blotting with streptavidin or anti-involucrin. The purified involucrin fractions contained the dominant ϳ140-kDa protein, visualized by Coomassie Brilliant Blue (not shown). This protein band was excised and digested in-gel with trypsin. Peptides were subjected to MALDI mass spectrometry, and 21 ions were identified as being derived from human involucrin (Table II). When keratinocytes were differentiated in the absence of exogenous biotin, the ϳ140-kDa protein in the involucrin fractions interacted with anti-involucrin but not streptavidin (Fig. 5B). When biotinylated cadaverine was present, the ϳ140-kDa protein in the involucrin fractions interacted with anti-involucrin and streptavidin (Fig. 5B). These results provide evidence that P450 BM3v induced the cross-linking of biotinylated cadaverine and involucrin, a physiological transglutaminase substrate.
Finally, reverse transcription-PCR was used to identify transglutaminase isoforms expressed in the human keratinocyte cultures. Fig. 7, A and B, shows cDNA products formed from 10 ng of poly(A) ϩ RNA templates. Products having expected sizes were detected for types I, III, and V but not type II. The faster migrating band in Fig. 7B may represent an alternatively spliced type V transcript reported previously (52). These results show that keratinocytes differentiated for 6 days expressed mRNAs encoding the transglutaminases (types I, III, and V) known to have important functions in epidermal differentiation and CE formation (8). Whereas no type II message was amplified (Fig. 7A), low levels of type II enzyme were detected by Western immunoblotting, compared with the positive control corneal cells (Fig. 7C). DISCUSSION The cytochrome P450 CYP2B19 is a P450 epoxygenase originally discovered in fetal mouse skin (11). It generates biologically active EETs from arachidonic acid. While P450 epoxygenases are present in many mammalian tissues, CYP2B19 is distinguished by its transcriptional regulation, which restricts it to keratinocytes and the later stages of differentiation. In vivo, CYP2B19 mRNAs localize to the outer (granular) cell layers of mouse epidermis (11). We present evidence that the CYP2B19 metabolite 14,15-EET activates transglutaminases in situ and increases CE formation in mouse and human epidermal keratinocytes. This is the first evidence for physiological roles for P450 epoxygenases in keratinocytes. These results have important implications for the barrier functions of keratinizing epithelia in vivo.
In differentiating keratinocytes, transglutaminases crosslink scaffold proteins and keratin intermediate filaments, forming a ϳ10 -15-nm-thick CE just inside the plasma membrane (8). Involucrin is a preferred substrate for the predominant cross-linking activities in cultured epidermal keratinocytes, which localize to particulate or membrane fractions of these cells (49). Involucrin is synthesized as a soluble protein, and this CE scaffold protein is cross-linked into the insoluble CE quite early during CE formation (50,51). The mechanisms regulating transglutaminase cross-linking activities in the epidermis are complex and incompletely understood. They involve transcriptional regulation of several distinct genes and  6. The 14,15-EET epoxygenase P450 BM3v increased CE formation in differentiating human keratinocytes. Human keratinocytes were mock-treated (M) or infected with adenovirus particles that express P450 BM3v (B, Ad-BM3v) or ␤-galactosidase (G, Ad-␤gal). Cornified cell envelopes were counted on day 6 and expressed as percentage of total cells (mean Ϯ S.D.; n ϭ 6). A, percentage of keratinocytes that formed CEs spontaneously as a result of in vitro differentiation. Phenotypes were scored as "fragile" (crumbled, irregular, and rough) and "rigid" (rounded or polygonal, regular, and smooth) CEs, respectively. B, percentage of ionophore-treated keratinocytes that formed CEs. For comparison, the data in A were replotted. posttranslational regulation including phosphorylation, proteolysis, subcellular compartmentalization, myristoylation, or palmitoylation (8,36,37). Additional regulation is imposed by hormonal and metabolic milieus, controlling intracellular free calcium, nucleotides, and other factors. Consequently, epidermal transglutaminases can be present in latent inactive forms or catalytically active (53), and the biochemical mechanisms controlling this switch remain to be elucidated.
We aim to study the effects of CYP2B19 metabolites (EETs) on transglutaminase-mediated cross-linking in situ to approximate, as closely as possible, the regulation operative in keratinocytes during differentiation in vitro. This goal cannot be achieved using in vitro cross-linking assays (34), in which the patterns of endogenous proteins cross-linked are less restricted (54). The glutamine residues utilized were also less restricted when binding interactions among transglutaminase and involucrin occurred in solution, rather than in synthetic membranes (55). These results emphasize the importance of endogenous factors such as spatial and subcellular localization in specifying transglutaminase substrates and products.
Various artificial substrates, which contain detection tags, have been used to assay in situ transglutaminase activities. Most common are analogs of cadaverine, which is biosynthesized by decarboxylation of lysine in cultured cells; cadaverine does not accumulate intracellularly or participate in polyamine biosynthesis like putrescine (56). Putrescine has been used widely for in vitro cross-linking assays (34), but it is technically difficult to label cells with [ 3 H]putrescine in situ because polyamine pools can be quite large and because entry is carriermediated and thus limited (39,44,57). In contrast, cadaverine analogs are excellent transglutaminases substrates, having been used extensively for in situ and in vitro assays (33, 35, 39, 43-45, 54, 58). In keratinocyte and neuronal cell cultures, in situ cross-linking activities visualized with cadaverine analogs co-localized with sites of transglutaminase immunoreactivity (52,59). This was validated and developed into a diagnostic test used to classify autosomal recessive congenital ichthyoses in humans (60). As further biological validation, in situ crosslinking activities visualized with cadaverine analogs were re-stricted to the interface of the stratum granulosum and stratum corneum in human epidermis (35). This corresponds to the same cell layers where CYP2B19 is expressed in mouse epidermis (11), suggesting that transglutaminases are active in the same cells that express P450 epoxygenases and biosynthesize EETs.
Mouse and human keratinocytes generate EETs endogenously and express P450 epoxygenases (11,12). Although we have not yet identified epoxygenases in human skin, the expression of one or more CYP genes is inferred, because cytochrome P450s are the only known enzymatic source of EETs. The greater responsiveness of human keratinocytes to 14,15-EET may be attributed to the ability of these cells to differentiate longer and more completely in vitro, compared with mouse keratinocytes. Despite these differences, the effects of 14,15-EET and P450 BM3v appeared qualitatively similar in humans and mice. This suggests that the basic biochemical mechanisms regulating keratinocyte cornification are conserved in these species and that genetically altered mice (61) may be useful models to study mechanisms regulating aspects of epithelial differentiation in humans.
Whereas cross-linking itself does not prove transglutaminase catalysis (43), several lines of evidence indicate that the in situ protein biotinylation in these studies was transglutaminase-dependent. First, protein biotinylation was partially inhibited by the transglutaminase inhibitor cystamine and the regulator of Ca 2ϩ influx, amiloride. Second, the levels of protein biotinylation were substrate-specific and consistent with previous reports of transglutaminase substrate structures and catalytic efficiencies (45). Furthermore, spurious protein biotinylation was not observed in the absence of exogenous biotin; nor was it apparent with the poor substrate biotinylated lysine. Third, different substrates used at the same concentrations produced qualitatively similar patterns of protein biotinylation although the absolute levels of biotinylated proteins varied among substrates, suggesting a common catalytic mechanism. Fourth, involucrin was preferentially biotinylated regardless of the exogenous substrate. Involucrin is a preferred, endogenous substrate for the major transglutaminase activities in cultured keratinocytes (49). Since transglutaminases are highly selective for their endogenous acyl donors (peptide-bound glutamines) (36), it seems unlikely that nontransglutaminase cross-linking would produce the same biotinylated proteins. Further, even transglutaminase-independent cross-linking was substrate-specific when lymphocytes were labeled in situ with [ 3 H]putrescine. Hypusine derived from [ 3 H]putrescine was incorporated pref- FIG. 7. Differentiating human keratinocytes expressed mRNAs encoding transglutaminase isoforms (types I, III, and V) involved in CE formation. A and B, the template for cDNA synthesis was poly(A) ϩ mRNA isolated from human epidermal keratinocytes differentiated in vitro for 6 days. Templates (10 ng) were reacted with reverse transcriptase (RTϩ) or water (RTϪ) and amplified using genespecific primers (Table I). Molecular ladders were 100 bp (A) or 1 kb (B). Products of the expected size were observed for types I (261 bp), III (161 bp), and V (2 kb). Type II was not detected (347 bp). Faint bands in the RTϪ lanes (types I and III) were larger than expected and may arise from genomic contamination. Faster migrating products in the type V, RTϩ lane, may represent a splice variant (52). C, Western immunoblot showing interaction of anti-transglutaminase II with human keratinocyte (KC) proteins; cultures were differentiated for 0, 4, and 6 days. The positive controls were transformed human corneal cells.
Finally, we describe novel modifications of the in situ assay that increased assay sensitivity, such that cross-linking could be visualized even with poorly utilized substrates like biotinylated lysine. This was achieved by studying keratinocytes differentiated in the continuous presence of substrate and measuring cumulative in situ activities. This is important because the effects of 14,15-EET were chronic and integral to the differentiation program of epidermal keratinocytes, and they were not detectable using conventional, short term (Ͻ4 hours) exposures to substrates (32,33,63). The chronic addition of substrate increased the sensitivity and dynamic range of the in situ assay with no apparent loss in specificity or cytotoxicity.
Keratinocytes expressing P450 BM3v formed more CEs as result of their greater cross-linking activities in situ. The percentages of cells forming CEs spontaneously, as a result of in vitro differentiation, were in the range reported previously for submerged cultures lacking serum and fatty acids (34,64), and mRNAs encoding the transglutaminases (types I, III, and V) implicated in CE formation were detected in the cultures (8).
Since higher percentages of cells formed CEs after ionophore treatment (versus spontaneous CEs), it seems likely that P450 BM3v increased CE formation by perturbing mechanisms regulating cross-linking activities and that the cellular content of transglutaminases and their endogenous substrates were not rate-limiting factors.
In summary, we report novel evidence suggesting physiological functions for 14,15-EET epoxygenases in regulating epidermal cornification by mechanisms involving activation of transglutaminase enzyme activities. Transglutaminases have essential roles in epithelial differentiation and function (8,40). Inactivating or null mutations in the human TGM1 gene result in severe skin disorders and cause a form of lamellar ichthyosis (65,66). In mice, targeted disruption of Tgm1 is neonatal lethal (67). These results and our studies suggest that the underlying biochemical mechanisms regulating cornification are conserved in humans and mice. Future studies are needed to determine whether the effects of 14,15-EET on transglutaminase activities and CE formation involve transcriptional or posttranscriptional mechanisms or other regulatory factors. It is not possible to speculate on the mechanism of 14,15-EET actions in keratinocytes, because there is no unifying hypothesis explaining their diverse biological activities in different cell systems (14), and they have not been studied previously in the epidermis. Results of these studies are relevant to keratinizing and nonkeratinizing epithelia in various organ systems because isoforms of transglutaminases and P450 epoxygenases are widely expressed in human epithelial tissues (8,14,40).