Involucrin Cross-linking by Transglutaminase 1

The transglutaminase 1 (TGase 1) enzyme is essential for the assembly of the cell envelope barrier in stratified squamous epithelia. It is usually bound to membranes, but to date most studies with it have involved solution assays. Here we describe anin vitro model system for characterizing the function of TGase 1 on the surface of synthetic lipid vesicles (SLV) of composition similar to eukaryote plasma membranes. Recombinant baculovirus-expressed human TGase 1 readily binds to SLV and becomes active in cross-linking above 10 μm Ca2+, in comparison to above 100 μm in solution assays, suggesting that the membrane surface is important for enzyme function. Involucrin also binds to SLV containing 12–18% phosphatidylserine and at Ca2+ concentrations above 1 μm. In reactions of involucrin with TGase 1 enzyme in solution, 80 of its 150 glutamines serve as donor residues. However, on SLV carrying both involucrin and TGase 1, only five glutamines serve as donors, of which glutamine 496 was the most favored. As controls, there was no change in specificity toward the glutamines of other substrates used by free or SLV-bound TGase 1 enzyme. We propose a model in which involucrin and TGase 1 bind to membranes shortly after expression in differentiating keratinocytes, but cross-linking begins only later as intracellular Ca2+levels increase. Furthermore, the data suggest that the membrane surface regulates the steric interaction of TGase 1 with substrates such as involucrin to permit specific cross-linking for initiation of cell envelope barrier formation.

The cell envelope (CE) 1 is a highly insoluble structure assembled just inside the plasma membrane of stratified squamous epithelia and is essential for effective barrier function. To form the CE, specialized keratinocyte proteins are expressed and subsequently made insoluble by cross-linking by both disulfide bonds and isopeptide bonds formed by transglutaminases (TGases) (1)(2)(3)(4)(5).
Emerging data suggest that the protein composition of CEs varies widely between epithelia and even different body sites of epithelia such as the epidermis (6,7). However, involucrin seems to be a ubiquitous component of most if not all CEs. Indeed, several types of data imply that it is one of the first proteins to be cross-linked to initiate CE assembly. First, expression studies have revealed that involucrin deposition at the cell periphery precedes all other suspected or confirmed CE protein constituents (8 -15). Second, shadowing and scanning transmission electron microscopy suggest that a monomolecular layer of involucrin is overlayered on the cytoplasmic side by other CE structural proteins (16). Third, extant models of CE structure based on biochemical characterization and protein sequencing indicate that involucrin becomes cross-linked to several cell peripheral proteins including desmoplakin, envoplakin, keratin intermediate filaments, as well as other CE proteins including members of the small proline-rich family, cystatin ␣ and loricrin (6,17,18). Fourth, recent data have shown that involucrin is a major target for the covalent attachment of ceramide lipids from the exterior surface of the CE, which could only occur if involucrin was deposited in the intimate vicinity of the keratinocyte membrane at an early time (19). Human involucrin contains 150 glutamine and 45 lysine residues (20), and it appears that mammalian involucrins have undergone extensive expansion of various glutamine-rich repeating motifs during evolution perhaps to increase the sites suitable or available for TGase-mediated cross-linking (21). However, sequencing studies of this laboratory have shown that only a limited number of these residues are used for cross-linking in vivo (17). Moreover, whereas involucrin appears to be a good substrate for TGases in in vitro reactions (11,22), extant data have provided no information on which of these enzyme(s) are responsible for cross-linking in vivo.
TGases are Ca 2ϩ -dependent enzymes that catalyze an acyl transfer reaction between the ␥-carboxamide group of proteinbound glutamine and the ⑀-amino group of lysine residues. Of the seven known human TGases, four (TGases 1, 2, 3, and X) are expressed in terminally differentiating epithelia such as the epidermis (23,24), but to date only limited data are available on their substrate specificities and relative contributions in CE assembly. Of these, the TGase 2 enzyme is thought to play only a minor role, and the properties of the newly discovered TGase X enzyme await characterization. The TGase 1 and 3 enzymes are essential for the cooperative cross-linking of such substrates as loricrin (25), trichohyalin (26), and small proline-rich proteins 1 (27) and 2 (28). The TGase 3 enzyme is soluble and requires proteolytic activation before it can function (29). The TGase 1 enzyme was first discovered in keratinocytes and is usually anchored to membranes by way of acyl N-myristoyl and S-myristoyl or S-palmitoyl adducts near the amino terminus of the protein (30 -32). However, virtually all studies with the TGase 1 enzyme to date have involved assays with poorly defined keratinocyte particulate fractions (8 -11) or solution assays conducted in the absence of membranes (24 -26, 30, 32).
In the present study, we have used synthetic lipid vesicles (SLV) of composition similar to those of eukaryote plasma membranes in order to explore how membranes affect TGase reaction and residue specificity. First, we show that only the TGase 1 enzyme can associate with SLV. Second, we show that of several proven CE structural proteins available to us, only involucrin associates with SLV and in a Ca 2ϩ -dependent manner. When both TGase 1 and involucrin are attached to membranes, there is remarkable specificity of glutamine residue usage for cross-linking. These data have important implications for CE assembly and ichthyosiform diseases caused by enzyme or substrate abnormalities.

Production of Recombinant TGase 1 and 3 Enzymes-Recombinant
full-length human TGase 1 and TGase 3 enzymes were expressed in Sf9 cells by the BaculoGold system using the pVL1392 plasmid vector (PharMingen, San Diego, CA) as described previously (33). TGase 1 was recovered in the particulate fraction after sonication in lysis buffer (33).
In some experiments this crude particulate fraction was used as the source of enzyme activity in amounts standardized to incorporate 0.7 pmol/min of [ 14 C]putrescine into succinylated casein. In most experiments, it was solubilized from membranes by sonication in lysis buffer with 4% Triton X-100 or in some cases with 1 M NH 2 OH-HCl (32). TGase 3 was recovered in the cytosolic fraction after lysis by sonication. These solutions were clarified by centrifugation and the enzymes subsequently purified by fast protein liquid chromatography on MonoQ Sepharose as before (33). Active fractions were brought to 1 M Na 2 SO 4 and rechromatographed on a 1-ml Resource Phe hydrophobic interaction column (Amersham Pharmacia Biotech) using a gradient from 1 M Na 2 SO 4 , 20 mM Tris-Cl (pH 8.0) to 20 mM Tris-Cl (pH 8.0) in 30 min at a 1 ml/min flow rate. The TGase 1 and 3 enzymes were Ͼ98% pure by SDS-polyacrylamide gel electrophoresis, and could be stably stored as a suspension in 1.5 M Na 2 SO 4 for some weeks at 4°C. Amounts were determined by amino acid analysis following acid hydrolysis. TGase 3 was activated with 0.1 units/100 g dispase for 15 min at 23°C and purified from the protease on a MonoQ column as above.
Expression and Purification of Recombinant Human Involucrin-A full-length cDNA clone of human involucrin was obtained by polymerase chain reaction from human chromosomal DNA. Polymerase chain reaction primers used were (ϩ)-GTAGCTTCTCATATGTCCCAGCAAC and (Ϫ)-CCCTTGTATGAGACGATCTGAG. These were designed to create an NdeI restriction site to be compatible with the pET expression system (Novagen, Madison, WI). The polymerase chain reaction product was cloned into the pCR2.1 plasmid using the TA Cloning kit (Invitrogen, Carlsbad, CA) and verified by DNA sequencing. Following subcloning into the pET11a vector and transfection into the BL21(DE3)pLysS strain of Escherichia coli (Novagen), protein expression was induced by 1 mM isopropyl-1-thio-␤-D-galactopyranoside for 3 h. Cell mass was pelleted and lysed by freeze-thawing. Particulate matter was removed by centrifugation, and involucrin protein was enriched by heat precipitation as described (22). It was purified to Ͼ97% (by SDS-polyacrylamide gel electrophoresis) by anion exchange chromatography on a HiTrap Q column (Amersham Pharmacia Biotech) using 20 mM Tris-HCl (pH 8.0) and gradient elution with the same buffer containing 1 M NaCl. By circular dichroism, it possessed an estimated ␣-helix content of 68%. As this is similar to native involucrin isolated from keratinocytes (12), it is likely that the recombinant protein had assumed its native configuration.
Preparation of Synthetic Lipid Vesicles (SLV)-Mixtures of dipalmitoyl-phosphatidylcholine, cholesterol, dipalmitoyl-phosphatidylserine (PS), and other lipids where indicated (all from Sigma) were made in chloroform/methanol (95:5). In all cases, the mixtures contained 30% cholesterol. When the amounts of PS or other individual components were varied, phosphatidylcholine was added to make the mixture to 100%. Mixtures were made in 0.5 ml and contained 10 mol of total lipids. The solvent was flushed away under a stream of N 2 , dried further under high vacuum for 4 h, and then resuspended by vortexing in 0.5 ml of a buffer containing 50 mM Tris-Cl (pH 8.0), 100 mM NaCl, 3 mM NaN 3 , 5 mM dithiothreitol, and 200 mM sucrose. The mixture was sonicated on ice five times each for 1 min using a Branson 250 sonifier with micro probe tip, and allowed to stand at 23°C to facilitate assembly of SLV. After 1 h the SLV were diluted with 0.5 ml of the above buffer without sucrose, and 200-l aliquots were centrifuged at 100,000 ϫ g for 30 min in a Beckman Airfuge using the A-10 rotor. The top 175 l was removed, and the pellet was resuspended in another 150 l of sucrose-free buffer. The final stock concentration was 11 mol/ml.
All binding assays were done at 23°C in a final volume of 200 l by mixing SLV with protein amounts empirically found to exceed at least 2-fold the binding capacities of the SLV as follows: for TGase 1, 20 g (0.2 nmol) were mixed with 0.1 mol of SLV lipids; for involucrin, 1.2 nmol were mixed with 1 mol of SLV lipids; for all other proteins, 50 g were used with 1 mol of SLV lipids. In all cases, the buffer contained 50 mM Tris-HCl (pH 8.0), 100 mM NaCl, 1 mM dithiothreitol, 3 mM NaN 3, and other additives where noted. Some mixtures also contained 1 mM CaCl 2 or 2 mM EDTA. After mixing, the samples were centrifuged for 45 min at 100,000 ϫ g, and the protein content from 50 l of supernatant was determined by amino acid analysis after acid hydrolysis. Data were usually not corrected for loss of SLV lipids, as in control experiments, cumulated losses were Ͻ4% as assayed by [ 14 C]phosphatidylcholine tracer, and thus neglected from the calculations.
Preparation of Free Ca 2ϩ Concentrations-The Ca 2ϩ concentrations in the micromolar range were set by buffering free Ca 2ϩ with chelators. The desired Ca 2ϩ /chelator ratios were calculated by the WinMaxC 1.7 computer program (34) and were made by adding the required amounts of 1 M CaCl 2 to 0.2 M stock solutions of the following chelators at pH 8.0. Final chelator concentrations in the samples were 10 mM in all cases. Sodium citrate was the chelator for the 20 -500 M Ca 2ϩ range, sodium nitrilotriacetate for the 1-20 M Ca 2ϩ range, and sodium 1,2-bis(aminophenoxy)ethane-N,N,NЈЈ,NЈ-tetraacetate for the 0.1-1 M Ca 2ϩ range. The actual concentration values were not significantly different from the theoretical calculations (p Ͼ 0.2, n ϭ 5) in the 2-100 M range as assayed by arsenazo III dye spectrophotometry (35).

Identification of [ 14 C]Putrescine-labeled Involucrin Fragments and Measurement of the Reaction
Rate-Three sets of experiments were performed to determine reactive Gln residues in involucrin by TGase 1 and labeling with [ 14 C]putrescine as follows: TGase 1 bound to SLV (2 mol of lipid) containing 15% PS; the TGase 1 enzyme present in crude Sf9 cell particulate fractions; or solubilized TGase 1 in reaction buffer. In each case, the amount of TGase 1 activity was standardized to 0.7 pmol/min using succinylated casein and [ 14 C]putrescine, as described previously (36), and corresponded to about 0.9 pmol. This was set far below saturating amounts on SLV in order to ensure that all enzyme was bound to the SLV. All reactions were done in 250 l and contained 1.2 nmol of involucrin, 1 mM CaCl 2 , 20 mM putrescine with 100 nCi of [ 14 C]putrescine (NEN Life Science Products, 110 Ci/mmol).
After 4 h incubation at 37°C, the reactions were stopped by addition of EDTA to 10 mM and 100 l 20% SDS and vortexed. This mixture was precipitated and washed three times with acetone/triethylamine/acetic acid (90:5:5) (37) to remove the SDS and the lipids. After repeated washings with acetone, the pellet was redissolved in a buffer of 50 mM Tris-Cl (pH 7.5) and digested for 16 h at 37°C with 2% (by weight) of modified trypsin (Boehringer Mannheim). Aliquots of 20 g were resolved on a 250 ϫ 4.6 mm Beckman Ultrasphere C18 HPLC column. Separated peaks were analyzed for radioactivity, and peaks containing activity were attached to a solid support for sequencing as before (38). As the peptide number 40 eluted by 50% acetonitrile (see Fig. 9) was too long for direct sequencing, it was subjected to limited proteolysis by dispase and was sequenced from four peptides isolated by HPLC chromatography as above.
The Gln residues reacted by putrescine by TGase 1 were identified by standard protein sequencing analyses. The PTH-derivative of the ␥-glutamylputrescine eluted as a novel peak at 13.55 min in the HPLC separation step of the Porton Instruments LF 3000 protein sequencer. This was confirmed by measurement of the [ 14 C]putrescine label. This always corresponded to those cycles where a Gln residue was also present.
Determination of Kinetic Parameters of Putrescine Incorporation by TGase 1-Values were measured at 37°C using 0.9 pmol of TGase 1, 20 mM putrescine with 100 nCi of [ 14 C]putrescine, 5 mM CaCl 2 , SLV containing a total of 2 mol total lipid, five concentrations of substrate proteins (0.2, 0.5, 1, 2, and 5 M for SLV-bound TGase 1 with involucrin or 5, 10, 20, 50, and 100 M for solubilized TGase 1 and soluble substrates) as described (25). The calculated K M values pertain to the protein substrates; V max and k cat data are those for putrescine incorporation. The molar mass of succinylated casein was taken to be 25 kDa. The reaction rates were quantified by measuring the incorporated radioactivity as before (35). Kinetic constants were obtained as before (35) using the curve-fitting and regression analysis with Sigmaplot 4.0 software. All data points represent the mean of three measurements, each performed in triplicate. The size of SLV (Ͼ85% below 100 nm, as determined by size-exclusion on Sepharose CL-4B chromatography) was not significantly altered during the binding and labeling reactions.

RESULTS
The present experiments were designed to explore the properties and function of the TGase 1 enzyme on a lipid surface using SLV that mimic the plasma membranes of eukaryotic cells. In preliminary experiments we made the serendipitous discovery that of several known CE substrates, involucrin also binds efficiently to such membranes under the conditions typically used for TGase assays. This observation has important implications for CE assembly, and the details are examined systematically in this paper.
Of Several Known CE Proteins Only Involucrin Attaches to SLV Containing Physiological Concentrations of PS-By using SLV formulations initially taken from established methods for assaying protein kinase C activation (39) that mimic the cytoplasmic surface of plasma membranes of eukaryotic cells, we found that involucrin was readily adsorbed to the SLV. By using saturating amounts of involucrin (0.6 nmol of protein/ mol of lipid) and 1 mM Ca 2ϩ (typical for TGase assays), the soluble involucrin content of the inter-SLV buffer was measured after pelleting of the SLV by ultracentrifugation. Increasing the PS content of SLV from 0 to 30% in 1% increments increased binding of involucrin sigmoidally between 4 and 15 mol % PS (Fig. 1A) and could be fitted with r ϭ 0.93 to a log (y/100 Ϫ y) ϭ 3.82 log [PS%] Ϫ0.904 Hill's equation, where y is the ratio of SLV-bound involucrin to the maximal binding. With increasing concentrations of involucrin, SLV containing 15-30% PS could be saturated by 0.43 Ϯ 0.02 nmol of involucrin/mol of lipid. This corresponds to about one involucrin molecule/500 nm 2 of surface (40). Since an involucrin molecule is 46 ϫ 1.5 nm (70 nm 2 (12), this means that a remarkably high value of about 15% of the SLV surface can be decorated by involucrin. The adsorption of involucrin to the SLV was specific to PS since substitutions by phosphatidic acid, phosphatidylglycerol, or phosphatidylinositol did not enhance binding (Fig. 1B).
We also examined whether a variety of other known CE substrate proteins could bind to SLV. The recombinant proteins available to us included loricrin and the small proline-rich proteins 1 and 2. None of these became attached to SLV of any composition tested (Fig. 2). Similarly, succinylated casein, a commonly used artificial substrate for amine incorporation by TGases, did not associate with SLV.
Ca 2ϩ Ion Dependence of Binding Involucrin to SLV-Next, we determined the optimal free Ca 2ϩ concentration required for binding of saturating amounts of involucrin to SLV containing 15 mol % PS (0.6 nmol/mol lipid) (Fig. 3). Maximal binding occurred with Ͼ20 M Ca 2ϩ , and half-maximal binding was estimated at 4.2 Ϯ 0.7 M, but binding was first detected at 1 M. This Ca 2ϩ -induced binding was completely reversible by excess of the chelators EGTA or EDTA. Magnesium and monovalent ions did not support binding of involucrin to SLV (not shown).
Solubilized Recombinant TGase 1 Binds Spontaneously to SLV-We have shown previously that the recombinant TGase 1 enzyme expressed in the baculovirus system is constitutively N-myristoylated and S-myristoylated or S-palmitoylated on its amino-terminal 10-kDa portion and can be found mostly in the particulate fraction of the insect cell homogenates (33). As for the native enzyme expressed in epidermal keratinocytes, the recombinant TGase 1 can be solubilized from the membranes by extraction with the detergent Triton X-100, and the lipid adducts on the enzyme are retained. Alternatively, the TGase 1 may be solubilized by use of 1 M NH 2 OH-HCl which hydrolyzes the S-acyl adducts off the enzyme (32). When purified from the NH 2 OH-HCl method, no TGase 1 enzyme bound to the SLV, as indicated by amino acid analysis of pelleted SLV mixtures and TGase assays of the resulting supernatants (data not shown). However, when the TGase 1 enzyme purified from Triton X-100 extracts was mixed to SLV, they spontaneously reassociated, as indicated by the disappearance of detectable enzyme activity from the supernatants of pelleted SLV mixtures. Using SLV composed of 15% PS, the saturating amount was Ϸ0.9 nmol of TGase 1/mol of lipid (Fig. 4). The binding of recombinant TGase 1 protein to SLV was not influenced by Ca 2ϩ ions or EGTA or to SLV prepared with varying formulations of ingredients including anionic (20% PS), neutral (only phosphatidylcholine and cholesterol), or cationic (5% stearylamine) lipids (data not shown). Thus the extent of constitutive myristate and palmitate modifications of the recombinant TGase 1 protein by baculovirus is sufficient to permit spontaneous anchorage onto lipid bilayers per se. In contrast, neither the cytosolic TGase 2 nor activated TGase 3 enzymes associated with SLV of any composition tested (Fig. 4). Effect on Kinetic Parameters of TGase 1 Reaction following Attachment to SLV-Kinetic parameters for [ 14 C]putrescine incorporation by SLV carrying 0.94 pmol of recombinant TGase 1/mol of lipid were determined at 37°C using three different protein substrates as follows: succinylated casein, human SPR 2 (28), and human involucrin (Table I), of which only the latter was found to attach to the SLV (Figs. 1A and 2). Kinetic constants for the recombinant SPR2 substrate differed only by about a 2-fold increase in K M if the PS content of SLV was 0 or 15%. As a similar change in K M was observed both with putrescine and succinylated casein, we attribute this K M increase to the reduced diffusional mobility of the enzyme after attachment to the SLV. Interestingly, attachment of TGase 1 to SLV drastically affected its kinetic parameters with respect to the incorporation of putrescine into the standard succinylated casein substrate, so that the V max was reduced by about 5-fold ( Table I). The probable reason for this is the reduced access to many of the Gln residues after anchorage of the enzyme. On the other hand, the V max value for the recombinant SPR2 substrate was almost unchanged. Sequencing analyses revealed that of its several Gln residues, only Gln-6 was reacted by putrescine in all three of the TGase 1 enzyme formulations described in Table I, in agreement with previous findings (28). These data seem consistent with the fact that SPR2 is a much smaller and more flexible substrate with only one reactive Gln residue. The kinetic parameters for involucrin as a substrate were immeasurably low when the SLV were formulated in the absence of PS. In comparison to solubilized TGase 1 enzyme and soluble involucrin, SLV containing 15% PS yielded a 200-fold decreased k cat value and an approximately 40-fold decreased K M . In view of the above observations for succinylated casein, these changes might also reflect drastic changes in the availability of Gln residues on involucrin for reaction with putrescine.
The data of Table I also demonstrate that the kinetic constants of putrescine incorporation into involucrin by TGase 1 alter with the PS content of the SLV. We show further in Fig.  5 that the reaction rate follows a sigmoidal shape with a sharp increase between 6 and 11% with maximal incorporation at 10 -20% PS content. This observation is to be expected from the maximal binding of involucrin to SLV (Fig. 1A) and thus serves as a valid control. Beyond 20% PS content, the efficiency of the reaction declines ( Fig. 5 and Table I), possibly because of inhibition of TGase 1 by excessive charge.
Ca 2ϩ Requirements for TGase 1 Reaction Are Markedly Lowered following Attachment to SLV-To the best of our knowledge, the optimal Ca 2ϩ ion concentration required for the TGase 1 reaction has not been measured for any substrate; standard in vitro solution assays usually contain 1-5 mM Ca 2ϩ (41-43). In the absence of SLV and solubilized recombinant TGase 1 in a mixture containing 1.2 nmol of involucrin, we found that the reaction followed an apparent sigmoidal curve, with half-maximal incorporation occurring at 310 Ϯ 80 M Ca 2ϩ (Fig. 6, yellow squares). However, when an equimolar amount of TGase 1 was bound to SLV containing 15% PS and saturating amounts of involucrin, the Ca 2ϩ activator constant was estimated at 22 Ϯ 1.3 M Ca 2ϩ (Fig. 6, orange triangles). Similar values (19 Ϯ 2.7 M) were obtained for succinylated casein, a substrate that is not significantly bound to SLV (Fig.  2). However, when TGase 1 was attached to neutral SLV containing 0% PS, the half-maximal enzyme activity with succinylated casein was calculated at 370 Ϯ 110 M Ca 2ϩ (data not shown). Together, these control data suggest the changes in Ca 2ϩ sensitivity reflect the altered local Ca 2ϩ micro-environment on PS containing bilayers, rather than of enhanced enzyme affinity toward Ca 2ϩ or its substrates in the membranebound state. The local concentration of Ca 2ϩ ions may be much higher in the intimate proximity of PS-containing membranes as compared with the bulk solution (44), although a wide range of dissociation constants of PS-Ca 2ϩ adducts have been re- ported (for review see Ref. 45). Estimates of the local absolute free Ca 2ϩ concentrations on the surface of SLV are technically not feasible. Nevertheless, our data strongly suggest that crosslinking of involucrin by membrane-bound TGase 1 occurs at a 10-fold lower Ca 2ϩ concentration than that required for the soluble enzyme and substrate. Moreover, comparison with the data of Fig. 3 reveals that this concentration is severalfold higher than that required for efficient involucrin binding to SLV, which thereby suggests a likely temporal order to these processes in vivo.
Solubilized Recombinant TGase 1 Reacts with the Majority of Gln Residues of Involucrin-The Gln residues of involucrin used by recombinant TGase 1 were analyzed in the absence or presence of either natural insect cell membranes or SLV as a catalytic cofactor.
Analysis of [ 14 C]putrescine incorporation into involucrin was done following trypsin digestion and separation of tryptic peptides on a C18 reverse phase HPLC column. Labeled peptide peaks were dried and identified by sequencing. The PTH-de-rivative of ␥-glutamylputrescine isopeptide formed in the TGase reaction was identified as a distinct peak in the Porton Instruments LF 3000 gas-phase sequencer (Fig. 7).
A total of 40 tryptic peptides was reliably separable, of which 29 (Fig. 8A, arrows) were labeled by [ 14 C]putrescine in a reaction with solubilized recombinant TGase 1 and involucrin. Sequencing identified 80 labeled Gln residues Table II (part A), with a total of about 20 mol of putrescine incorporated per mol of involucrin (Table II,

part A). Human involucrin contains 150
Gln residues (20), which means that most were promiscuously labeled under these experimental conditions.
Attachment of TGase 1 to Insect Cellular Membranes or SLV Reveals Highly Specific Utilization of Gln Residues of Involucrin-These experiments were repeated using recombinant TGase 1 and involucrin bound to the crude particulate membranes of Sf9 insect cells. In this case, only five tryptic peptide peaks were labeled (Table II, part B), involving five different Gln residues. Similarly, following binding of recombinant TGase 1 and involucrin to SLV containing 15% PS (Fig. 9B), the same five labeled tryptic peptides were recovered which encompassed the same five Gln residues (Table II, part C). In both cases, only about 1 mol of putrescine was incorporated per mol of involucrin of which Gln-496 was the most brightly labeled.  containing 15% PS at different free Ca 2ϩ concentrations. Activity values are given as percent of the values measured at 1 mM Ca 2ϩ . Also shown for comparison are the data from Fig. 3 showing that involucrin binds to SLV at much lower Ca 2ϩ concentrations (brown circles). Points represent the mean Ϯ S.D. of three independent measurements. Interestingly, these data corroborate an earlier study (11) in which Gln-496 was identified as the most strongly labeled residue in a reaction of involucrin with the TGase 1 enzyme associated with crude keratinocyte membranes. Four other Gln residues (Gln-107, Gln-118, Gln-122, and Gln-133) located in the evolutionarily conserved head domain or ancestral portion of involucrin were also each slightly labeled.

DISCUSSION
Of several enzymes likely to be involved in cross-linking reactions to form the CE barrier during terminal differentiation in stratified squamous epithelia, the TGase 1 enzyme is perhaps the most complex since it exists in multiple forms (32,46). This enzyme is of critical importance in skin barrier function in particular since mutations in its gene resulting in loss of activity cause the devastating life-threatening disease lamellar ichthyosis (47)(48)(49). Most of the TGase 1 enzyme resides on membranes through N-myristoyl and S-myristoyl or S-palmitoyl linkages (30 -32), although minor amounts dissociate to reside in the cytosol (46). To date, almost all studies of this enzyme have involved solution assays of soluble substrates with TGase 1 enzyme stripped and purified from the cellular membranes or soluble recombinant expressed TGase 1 enzymes. In this study, we have systematically developed a more physiological model system to explore the properties of this enzyme. Our data introduce a so far unstudied catalytic cofactor, the membrane surface, in regulating the interaction of membrane-bound TGase 1 enzyme and its substrates. In particular, we have found that of several of its known and proven in vivo substrates, involucrin also binds to SLV membranes of similar PS content to those of the cytoplasmic surface of plasma membranes of eukaryote cells (50). Our data provide evidence that the residue specificity of the TGase 1 reaction is dependent on the attachment of itself and involucrin to the membrane surface (Fig. 9). These observations have important implications for the mechanism of assembly of the CE barrier structure in stratified squamous epithelia.
Optimized involucrin cross-linking by membrane-bound TGase 1 is largely dependent on the ingredients of the membrane and requires Ca 2ϩ ions and PS, an inherent constituent of the cytoplasmic face of membranes in living eukaryote cells (50). PS is required for a number of other physiological pro-cesses, where Ca 2ϩ -dependent binding of proteins to cell membranes is an essential condition for enzyme activity, as exemplified in the case of protein kinase C activation (39, 51) or blood clotting (52). As for these two well studied processes, we found that the activating properties of PS were not substitutable by other natural anionic phospholipids, presumably because both the carboxyl and amino groups are required for sequestration of Ca 2ϩ ions on the membrane surface.
Several studies have examined total Ca 2ϩ concentrations in keratinocytes. Available evidence suggests there is a Ca 2ϩ concentration gradient in the epidermis, for example, from a low level in basal cells which gradually increases toward the granular layer (53). In addition, Ca 2ϩ concentrations are indirectly known to be much higher on cell membrane surfaces (54). In keratinocytes grown in submerged cultures in low Ca 2ϩ medium, under which conditions they do not embark on terminal differentiation, net intracellular Ca 2ϩ concentrations are about 50 -100 nM (55). When cells are grown in higher Ca 2ϩ medium (0.5-1.5 mM), net intracellular Ca 2ϩ levels rise briefly to about 100 -200 nM (55), and the stratification and the terminal differentiation program proceeds (56). Similarly, normalization of a Ca 2ϩ gradient is essential for terminal differentiation and improved barrier function in reconstructed cultured epidermis FIG. 7. Identification of ␥-glutamylputrescine by amino acid sequencing. This assay identifies Gln residues that have been modified by putrescine as a result of the TGase reaction. After Edman degradation, the PTH-derivative of ␥-glutamylputrescine appears as a novel peak eluting at 13.55 min in the sequencing cycles where only a Gln residue would normally be expected. The relative amounts of this peak and that of PTH-Gln provide an estimate of the extent of modification. . Tryptic peptides of involucrin separated by C18 reverse phase HPLC were collected and assayed for isotope incorporation. Forty involucrin tryptic peptides were resolved in this system. Labeled peptides are shown by the numbered arrows. Note that in B, additional minor peptide peaks were contributed by insect proteins, but the same numbering system was retained for clarity. (57). However, in none of these studies has it been possible to measure the micro-environmental concentration at or near the membrane surface.
Typically, involucrin is expressed in mid-late spinous layers in the epidermis (or comparable levels in other stratified squamous epithelia) and is expressed early in cultured keratinocytes as elevated environmental Ca 2ϩ levels initiate terminal differentiation (58). The TGase 1 enzyme is expressed to a minor extent in basal keratinocytes, but its major expression program approximately coincides with that of involucrin in differentiating keratinocytes (59). Our new data from our model SLV system demonstrate that involucrin begins to associate onto SLV above 1 M (Fig. 3). These observations favor the view that involucrin binds to the plasma membranes shortly after its expression. Furthermore, our data demonstrate that the cross-linking of involucrin does not begin until the net Ca 2ϩ concentration rises about 10-fold higher than that required for involucrin binding (Fig. 6). Thus, we propose that the involucrin substrate and TGase 1 enzyme remain in close juxtaposition on cellular membranes until local Ca 2ϩ concentrations rise above a threshold level. Accordingly, together with available in vivo data, our new results offer the possibility that the Ca 2ϩ gradient not only orchestrates the expression of differentiationspecific genes (5,9,56) but also creates the environment required for the initiation of CE barrier formation by juxtaposed attachment of involucrin and TGase 1 to membranes for their subsequent cross-linking together (Fig. 9).
We show here that cross-linking of involucrin by TGase 1 in  Fig. 9 were sequenced. The location and amount of modified Gln residues were identified by the appearance and quantitation of PTH-␥-glutamylputrescine (Fig. 8). Underlined residues denote those seen in in vivo cross-linking (17 solution in vitro is an efficient process (Table I) involving the utilization of more than half of the total Gln residues of involucrin (Fig. 8, Table II). Published data from this laboratory have identified 27 Gln residues that are used for cross-linking in vivo (17), 23 of which were used in the present in vitro experiments (Table II). This suggests that these 80 Gln residues are the most available for reaction on involucrin. Thus it is unlikely that the utilization of multiple Gln residues was due to degradation or denaturation of involucrin (11,22). Conversely, these data also allow the speculation that involucrin may be crosslinked in vivo by soluble TGases as well, including perhaps the minor cytosolic forms of TGase 1. The most striking observation in the present study is the specificity of Gln usage for cross-linking of bound involucrin and TGase 1; more than 50% of TGase 1 reaction involved the single Gln-496. Moreover, this specificity was observed for both crude insect membranes, as well as SLV of several confections, providing the PS content exceeded 5%. Interestingly, this residue was identified as the most reactive in an earlier in vitro study (11). This earlier experiment utilized the TGase 1 activity of a crude keratinocyte membrane fraction, solubilized involucrin and 5 mM Ca 2ϩ . Our present data demonstrate that the added involucrin would have immediately attached to the membranes. As earlier reports showed that shorter CNBr or tryptic peptides of involucrin did not afford such specificity (11,22), together, the previous and our present data indicate that the stereochemistry of binding of intact involucrin and subsequent cross-linking by bound TGase 1 are critical determinants of this specificity.
The kinetic data of Table I shed light on the high degree of specificity of Gln utilization. We propose that the reason for the lowered k cat and K M values of membrane-bound TGase 1 using bound involucrin as substrate is due to restriction of available Gln residues for reaction. Furthermore, the maximal reaction velocity must be limited by the following: (i) the quantity of involucrin molecules attached to a unit of membrane surface; (ii) the lateral diffusion rate of enzyme and substrate along the membrane surface; and (iii) the rate of exchange between sol-uble and membrane-attached involucrin. An alternative hypothesis that association with the SLV membranes might change the conformation and thus specificity of TGase 1 remains to be tested experimentally, but in control experiments documented in Table I, we found no change in substrate specificity of the SPR2 substrate. Furthermore, in the insect membrane or SLV reactions only five Gln residues were used to insert about 1 mol of putrescine/mol of involucrin, of which Gln-496 was the most labeled. In contrast 80 Gln residues were labeled in the soluble reaction to insert about 20 mol/mol of putrescine. Thus the overall kinetic efficiency value for Gln-496 was at least 2-fold higher than that of the average Gln residue in a solution reaction. Therefore, these five residues, and Gln-496 in particular, must be particularly favorably aligned with respect to the active site of the neighboring bound TGase 1 enzyme. It seems obvious that the active site of membranebound TGase 1 must be located at a certain distance from the plane of the membrane, and thus the five utilizable Gln residues in involucrin can only be those that are at a compatible distance (Fig. 9). Clearly, further studies on the three-dimensional structures of involucrin and TGase 1 are warranted.
We have documented that involucrin is cross-linked in vivo to a variety of structural proteins, including in particular desmoplakin at the site of desmosomes, envoplakin, and perhaps periplakin located primarily on plasma membranes between desmosomes, as well as to itself. Furthermore, the predominant cross-linking site with desmoplakin was through Gln-496 of involucrin (17). Together, these and the present data offer a tantalizing snapshot of the earliest stages of CE assembly. It appears that involucrin and TGase 1 associate with the plasma membrane shortly after expression. As the localized Ca 2ϩ concentration rises, the TGase 1 enzyme activates Gln-496 which is favored for transfer to Lys acceptor residues on desmoplakin. Thus we propose that the desmosome is an important site for the initiation of CE assembly. Furthermore, these observations have important implications for the disease lamellar ichthyosis caused by lack of a functional TGase 1 enzyme (33,(47)(48)(49). If the initiation of CE assembly as envisaged above cannot occur, we should expect profound problems with formation of an effective barrier, as it is clear that the TGase 2 and 3 enzymes also expressed in terminally differentiating keratinocytes which do not bind to membranes (Fig. 4) cannot compensate for this critical initial step. FIG. 9. Model for the alignment of involucrin and TGase 1 on SLV on the inner surface of the plasma membrane of keratinocytes. Newly expressed TGase 1 (green sphere) attaches to the membranes by way of the lipid adducts on its amino-terminal portion. As the Ca 2ϩ concentration at the micro-environment of the membrane surface rises above a critical threshold level at the advent of terminal differentiation, the involucrin (red rod) attaches spontaneously. This binding is fostered through ionic interactions of multiple Glu residues of involucrin, Ca 2ϩ , and the anionic PS-rich membrane surface (yellow). We propose that these binding processes align only certain Gln residues of involucrin near the active site of juxtaposed TGase 1 molecules. Crosslinking reactions are initiated as the Ca 2ϩ concentration rises further. The activated Gln residues may then be transferred to other nearby substrates including desmoplakin, envoplakin, etc. (purple spheres) to initiate CE assembly.