Transglutaminase Cross-linking Properties of the Small Proline-rich 1 Family of Cornified Cell Envelope Proteins

Small proline-rich 1 (SPR1) proteins are important for barrier function in stratified squamous epithelia. To explore their properties, we expressed in bacteria a recombinant human SPR1 protein and isolated native SPR1 proteins from cultured mouse keratinocytes. By circular dichroism, they possess no α or β structure but have some organized structure associated with their central peptide repeat domain. The transglutaminase (TGase) 1 and 3 enzymes use the SPR1 proteins as complete substrates in vitro but in different ways: head domain A sequences at the amino terminus were used preferentially for cross-linking by TGase 3, whereas those in head domain B sequences were used for cross-linking by TGase 1. The TGase 2 enzyme cross-linked SPR1 proteins poorly. Together with our data base of 141 examples of in vivo cross-links between SPRs and loricrin, this means that both TGase 1 and 3 are required for cross-linking SPR1 proteins in epithelia in vivo. Double in vitro cross-linking experiments suggest that oligomerization of SPR1 into large polymers can occur only by further TGase 1 cross-linking of an initial TGase 3 reaction. Accordingly, we propose that TGase 3 first cross-links loricrin and SPRs together to form small interchain oligomers, which are then permanently affixed to the developing CE by further cross-linking by the TGase 1 enzyme. This is consistent with the known consequences of diminished barrier function in TGase 1 deficiency models.

A large body of recent amino acid sequencing data has demonstrated that members of the three known classes of small proline-rich (SPR) 1 proteins serve as constituents of the cornified cell envelope (CE) of stratified squamous epithelia (1)(2)(3)(4)(5), which is a specialized structure essential for barrier function (6 -9). The SPR1 (two members), SPR2 (8 -11 members) and SPR3 (one member) are assembled from a common plan (10 -14). Their amino (head) and carboxyl (tail) domains are enriched in Gln and Lys residues and consist of sequences that have been conserved between each member of an SPR class but differ between classes. These flank a central domain consisting of a series of Pro-rich peptide repeats of sequences that likewise have been conserved between members of a class but vary between classes. The SPRs become cross-linked to themselves and other CE structural protein constituents by both disulfide bonds and N ⑀ -(␥-glutamyl)lysine or N 1 ,N 8 -bis(␥-glutamyl)spermidine isopeptide bonds formed by transglutaminases (TGases), resulting in an insoluble macromolecular protein complex ideal for barrier function (6 -9). Moreover, for reasons that are not yet understood, individual SPR family members are differentially expressed in highly variable ways in many different types of epithelia (10 -31).
Examination of the way in which the SPRs were cross-linked to protein partners in human and mouse CE preparations revealed several novel features of their properties and functions (1)(2)(3)(4)(5). First, only head and tail domain Gln and Lys residues were used for cross-linking. Indeed, multiple adjacent residues were often used simultaneously on the same protein to form a complex interchain, and perhaps intrachain, crosslinked network (3)(4)(5). Each SPR molecule participated in as many as four cross-links (4). Second, the SPRs were found to be cross-linked to many protein partners, and indeed, they were consistently found to function as cross-bridgers between themselves or between loricrin, involucrin, etc. (4,5). Fourth, as initially suggested from expression studies by earlier investigators, our sequencing data directly revealed that the amounts of SPRs present in the CEs from various tissues varied widely; while human foreskin epidermal keratinocyte CEs contained about 5% SPRs (2), the CEs of the mouse forestomach contained about 22% SPRs (5). Further modeling analyses revealed that the amounts of SPRs in the epidermal CEs from trunk, lip, and footpad correlate well with the degree of physical and mechanical trauma to which the tissues are normally subjected (5). Thus, we have proposed that cross-bridging SPRs play an important role as modulators of the biomechanical properties of the CEs and the entire epithelium in which they are expressed (5).
In order to further explore these functions of the SPRs, we need to be able to express large quantities of the proteins for study. Since their cross-bridging role is mediated to a substantial degree through TGase cross-linking, we also need to know which enzymes are responsible for their cross-linking in vivo and how this is done. In an initial study, we described the preparation and some properties of a member of the human SPR2 class, and showed that it is cross-linked almost entirely by the TGase 3 enzyme in vitro and in vivo (32). In this paper, we have extended this work to the study of a member of the human SPR1 class. We show here that its cross-linking is much more complex, since it requires at least two TGases (3 and 1) operating in a sequential manner using different Gln residues for cross-linking to its partners as seen in vivo.

MATERIALS AND METHODS
Bacterial Expression and Purification of Human SPR1-A fulllength cDNA clone encoding human SPR1 (clone 15B of Ref. 10) was obtained as a generous gift from C. Backendorf. Following the addition of appropriate linkers, it was inserted into the pET-11a bacterial expression vector (Novagen, Madison, WI) and transformed into the host Escherichia coli B strain BL 21/DE3 (Novagen). Protein expression was induced in the presence or absence of L-[ 35 S]cysteine (2 Ci/ml) as described previously (32,33). Fresh or previously frozen pellets of bacteria were lysed and dialyzed against several changes of 100 volumes of 25 mM sodium citrate (pH 3.6), 1 mM dithiothreitol, 1 mM EDTA, and a mixture of protease inhibitors as described (32,33). While most of the bacterial proteins precipitated, the SPR1 protein remained soluble. Purification to homogeneity was achieved using an Amersham Pharmacia Biotech fast protein liquid chromatography system on a 0.5 ϫ 5 cm Mono-S column equilibrated in the citrate buffer with a 0 -1.0 M NaCl gradient and was eluted with 0.18 M salt.
The fractions were analyzed on 4 -20% SDS-polyacrylamide gels (Novex, San Diego, CA) with Coomassie stain or by Western blotting using a polyclonal antibody broadly reactive against both mouse and human SPR1 proteins (14). The enhanced chemiluminescence detection was performed with the Super Signal CL-HRP Substrate System (Pierce). Alternatively, the SPR1 proteins were monitored by autoradiography.
Isolation of Native Mouse SPR1a/b Proteins-Mouse primary keratinocytes were grown to confluency in low calcium (0.05 mM) medium. After 6 days, they were transferred to medium containing high Ca 2ϩ (1.4 mM), 10 nM staurosporin (to induce SPR1 expression (34)), 100 M LTB-2 (transglutaminase inhibitor) (35) (Syntex Research), and 0.5 C/ml [ 35 S]cysteine. After 24 h, the cells were harvested in phosphatebuffered saline, lysed by polytron homogenization, and centrifuged (10,000 ϫ g, 10 min, 4°C). The cytosol was then dialyzed against the citrate buffer used above, under which conditions the bulk of the keratin proteins precipitated, leaving the SPR1a/b proteins highly enriched. Their final purification was done following the method for the bacterial expressed proteins, and the proteins were eluted by 0.2 M NaCl.
Isolation of TGase Forms for Cross-linking-Three TGase enzymes were used. Human full-length TGase 1 was expressed in baculovirus, and the particulate fraction containing the unprocessed full-length but active TGase 1 form was isolated as described (36). This has a specific activity of about 6 pmol of putrescine incorporation into succinylated casein/h/pmol of TGase 1 protein (36), as measured using 14 C-putrescine (Amersham Pharmacia Biotech; specific activity, 118 mCi/mmol) and remains fully active for at least 1 day. A typical yield was 5-10 nmol of TGase 1/liter of insect cell culture medium. We also isolated the several TGase 1 isoforms present in the cytosolic and membrane-bound compartments of normal human epidermal keratinocytes (NHEK) grown in high Ca 2ϩ medium in the presence of [ 35 S]methionine for 4 days after reaching confluency, exactly as described previously (37,38). These included the membrane-bound full-length 106-kDa enzyme with a specific activity of about 5 (as assayed using the above units); the 67/33/10-kDa complex released from membranes by use of Triton X-100 with a specific activity of about 1000; the cytosolic full-length enzyme with a specific activity of about 50; the cytosolic 67/33-kDa complex with a specific activity of about 500; and the cytosolic 67-kDa form with a specific activity of about 250. We have noted that several of these isoforms begin to lose activity at 4°C within about 6 h of isolation (37,38), presumably because of unfolding and/or destabilization (36). The established method of using 14 C-iodoacetamide for measurement of TGase 1 molar amounts recovered in each culture experiment takes about 1 day (39), and thus to perform experiments using constant molar amounts of enzyme isoforms is problematic because of loss of enzyme activity. However, we have found that when NHEK cells are cultured under consistent conditions, the amounts of [ 35 S]methionine specific activity incorporated into the TGase 1 isoforms (as measured after iodoacetamide titrations) are reproducible and are 21 Ϯ 2 dpm/pmol for the cytosolic 67-kDa form and 40 Ϯ 4 dpm/pmol for all other active forms. (This is to be expected based on the methionine contents (37).) Accordingly, we used the amount of 35 S label to estimate the molar amount of each TGase 1 isoform recovered from fast protein liquid chromatography columns. In a typical culture of normal human epidermal keratinocytes in a 100-mm dish, we recovered 100 -200 pmol of TGase 1, or 15-100 pmol of each of the above membrane-bound or cytosolic isoforms.
Guinea Pig Liver TGase 2 Was Obtained Commercially (Sigma). TGase 3 proenzyme, which has no measurable activity, was isolated and purified from guinea pig skin. It was activated by dispase as described previously (40). The amount used for reactions was quantitated by amino acid analysis.
Cross-linking of SPR1 Proteins-For in vitro cross-linking studies using the SPR1 proteins as complete TGase substrates and measurement of isodipeptide cross-link formation, the purified SPRs were equilibrated into a buffer of 50 mM Tris-HCl (pH 7.5), 50 mM NaCl, 1 mM dithiothreitol, and 1 mM EDTA. Reactions contained 10 g of 35 Slabeled recombinant human or native mouse SPR1 proteins (about 7500 dpm) and about 12.5 pmol of TGase enzyme (ϳ500 nM enzyme concentration; 250 -500 dpm) in a 25-l volume. Reactions were initiated by the addition of Ca 2ϩ to the mixture (5 mM final concentration) and incubated at 37°C. Timed aliquots of 1.5 l each were removed and subjected to total enzymic digestion for measurement of the amount of cross-link formed during the reactions (41). In some experiments, the total reactions were analyzed by SDS-polyacrylamide gel electrophoresis on 4 -20% gradient gels, blotted onto nitrocellulose membranes, and analyzed by autoradiography. In preparative experiments, 100 g of unlabeled recombinant human SPR1 were reacted in a volume 250 l. In this case, we used the baculovirus membrane-bound full-length TGase 1 form and activated TGase 3 at concentrations of 700 nM for 18 h. Experiments showed that the cross-linking had proceeded to completion (see Fig. 3). We also performed a preparative experiment with the high specific activity keratinocyte membrane-bound 67/33/10-kDa isoform at 50 nM enzyme concentration, a reaction that was completed in 2 h.
Kinetic constants of the three TGases were determined for the recombinant human SPR1 protein exactly as described before (32,42).
Protein Chemical Methods-The concentrations and purity of the purified SPR1 proteins and TGase enzymes were determined by amino acid analysis. Uncross-linked or cross-linked SPR1 proteins were digested with trypsin (1:30 by weight; Sigma; sequencing grade) for 6 h at 37°C, and the peptides were resolved on a Phenomex ODS reverse phase HPLC column (2.1 ϫ 250 mm) containing 0.08% trifluoroacetic acid and with a gradient of 5-65% acetonitrile over 70 min (32,33). The peaks were collected and sequenced on a Porton LF3000 gas phase sequencer as described previously (32,33,42,43). In the case of the reactions with the TGase 3 enzyme, or TGase 3 followed by TGase 1, peptides were poorly resolved due to extensive cross-linking; in these experiments, 1-min fractions were collected across the peak for sequencing (32). Circular dichroism spectra on the intact SPR1 proteins or synthetic peptides (see below) were performed as before (32,33,43).
Use of Synthetic Peptides-The following peptides based on published human SPR1 sequences (13) 4 , and (PKVPEPCQ) 6 . Kinetic constants using the head and tail domain peptides as substrates for putrescine incorporation were determined exactly as described before (32,33,42).

RESULTS AND DISCUSSION
Recently, we described the preparation of a recombinant human SPR2 protein and demonstrated that it is a favored substrate of the TGase 3 enzyme in vitro and in vivo (32). In the present study, we have repeated these experiments using recombinant human and native mouse SPR1 proteins. We show here that both the TGase 3 and 1 enzymes are likely to be essential for SPR1 cross-linking in vivo, and in this consecutive order of reaction.
Expression and Purification of the Recombinant Human SPR1 Protein-Following expression in bacteria using the pET11a system, the human SPR1 protein was enriched from bacterial lysates by dialysis into 25 mM citrate buffer, pH 3.6 (in which it was very soluble), and purified by chromatography on a Mono-S fast protein liquid chromatography column (Fig.  1). The maximal yields were of the order of 2-5 mg/liter, typical of the CE proteins loricrin (33) and SPR2 (32) studied previously. As expected from its unusually basic pI value, the expressed SPR1 protein migrated on SDS gels with an apparent molecular mass of ϳ15 kDa, somewhat higher than the 10-kDa size calculated from its known sequence (Fig. 1, inset).
Isolation of Native SPR1a and SPR1b from Cultured Mouse Epidermal Keratinocytes-For the purposes of comparison of biochemical and structural properties, we isolated native mouse SPR1 proteins from cultured keratinocytes. Under the conditions used, mouse keratinocytes express high levels of the SPR1 proteins, which consist of two gene products, termed SPR1a and SPR1b, that differ primarily by the presence of 13 or 14 central domain peptide repeats, respectively (14). By dialysis of phosphate-buffered saline extracts of keratinocytes into the pH 3.6 citrate buffer, the vast majority of epidermal proteins precipitated, thereby enriching for the SPRs, which were then purified to homogeneity on the Mono-S column. They were separated by cutting out bands from SDS gels. Their total yield was ϳ1 mg/10 8 confluent cells. These data indicate that a significant amount of the SPR1a/b proteins remain soluble in the keratinocytes cultured under these conditions. Circular Dichroism Spectra of Native and Recombinant SPR1 Proteins-CD spectra were performed to evaluate the secondary structures of the recombinant human and native mouse SPR1 proteins. They have a limited degree of organized structure in phosphate-buffered saline (data not shown) or the TGase enzyme assay buffer at 20°C, but there was essentially no ␣ or ␤ structure present ( Fig. 2A, lines 1 and 2, for recombinant and native proteins, respectively). However, the signal strength of native mouse SPR1a/b was reproducibly greater than that of recombinant human SPR1. Mouse and human SPR1 proteins have highly homologous head, tail, and central eight-residue peptide repeat sequences but differ by containing 13 versus 6 central repeats, respectively. We performed additional experiments with synthetic peptides. First, peptides corresponding exactly to the head or tail domains of human SPR1 proteins generated only very weak CD signals, suggesting little significant structure (Fig. 2B, lines 1 and 2, respectively). On the other hand, peptides containing two (line 3), four (line 4), or six (line 5) peptide repeats characteristic of the eight-residue central domain of SPR1 proteins generated CD spectra that correlated with peptide size. Next, we assessed the overall structural properties of the SPR1 proteins as a function of temperature and guanidine hydrochloride. The recombinant human SPR1 (Fig. 2C, line 1) protein was at least partially unfolded by heating at 40°C (line 2) and 60°C (line 3), but the signals were normalized when returned to 20°C (dotted line), indicating refolding of the protein structures. Similarly, recombinant human (Fig. 2D, solid line 1) or native mouse (dotted line 1) SPR1 proteins could be reversibly denatured upon the addition of (lines 2) and then renatured upon the subsequent removal of (dashed lines) 4 M guanidine hydrochloride. Together, these data suggest that only the central peptide repeat domains of SPR1 proteins contain organized structures and that the signal strength of the CD spectra correlates to some degree with the numbers of peptide repeats. Moreover, the recombinant human protein possesses an organized structure similar to that of the native mouse SPR1a/b proteins of cultured keratinocytes. From this we conclude that the recombinant human SPR1 protein had folded into a native configuration upon isolation from the bacteria and is therefore appropriate for use in further biochemical experiments.

Epidermal TGases Use Recombinant Human and Native Mouse SPR1 Proteins as Complete Substrates but in Different
Ways-Three TGase enzymes that are expressed in stratified squamous epithelia were used to cross-link in vitro the recombinant SPR1. We used the same molar amount (typically 500 nM) of each of the several isoforms of TGase 1, the only known active form of TGase 2, and inactive or activated forms of TGase 3, in order to make direct comparisons of their reactions (44,45).
First, we measured the amount of isopeptide cross-link formed in the reactions and showed (Fig. 3) that most of the enzymes inserted measurable amounts into recombinant human SPR1, indicating that each enzyme uses this protein as a complete substrate. However, the extent and rates of reactions varied widely. The TGase 2 enzyme inserted relatively little (Ͻ0.2 mol/mol) cross-link (Fig. 3B); the activated TGase 3 enzyme inserted the most (3.1 mol/mol) (Fig. 3C); and, as expected, the proenzyme form of TGase 3 was inactive. A maximal amount of 1.1-1.2 mol/mol of cross-link was inserted by each of the five TGase 1 isoforms employed, but at different rates (Fig. 3A). The reaction was completed within 5 min with the highest specific activity form (67/33/10-kDa complex recovered from the membrane fraction of NHEK cells), but 15-18 h was required for the lowest specific activity form (full-length membrane-bound forms expressed in NHEK cells or baculovirus). Previous work has shown that there is about a 200-fold difference in specific activity between the intact and activated forms (38), which is reflected in these observed reaction rates. In general, the times required for reaction completion for each of the five isoforms mirror accurately their specific activities as previously measured against a succinylated casein substrate. These data support the conclusion that the various TGase 1 forms all use the SPR1 and casein substrates the same way.
Second, we resolved the cross-linking reactions by SDS-polyacrylamide gel electrophoresis and then performed autoradiography. By subsequent excision of the 35 S-labeled bands from the SDS gels, we were able to obtain quantitative information.  (14); lanes 3 and 4, Coomassie staining of native mouse SPR1a and SPR1b, respectively. Inset, molecular mass markers in kDa are as shown.
With each of the TGase 1 isoforms, 85-92% of the SPR1 protein remained as a monomer (Fig. 4A for 67/33/10-kDa complex recovered from the NHEK membrane fraction, cytosolic 67-kDa form, and baculovirus intact form; data for other isoforms not shown); only traces of protein were oligomerized into apparent dimers or larger species trapped at the top of the gel. Similarly, only 8% of the SPR1 protein was oligomerized by the TGase 2 enzyme (Fig. 4A). However, Western blotting analyses with a specific TGase 2 antibody revealed that this was not due to SPR-SPR oligomerization but rather mostly due to autocatalytic cross-linking to TGase 2 itself (data not shown), a phenomenon described previously for this (32,47,48) and the related factor XIIIa TGase enzymes (49). On the other hand, the TGase 3 enzyme cross-linked the proteins, readily giving rise to short oligomeric products so that 88 -98% of the SPR1 protein had been oligomerized (Fig. 4A). These data thus reveal fundamentally different reaction processes between the three TGases.
Third, we cross-linked in the same way samples of native mouse SPR1a/b proteins using the baculovirus full-length form of TGase 1, as well as TGases 2 and 3 (Fig. 4B). The pattern of cross-linking was essentially identical to that shown above. However, in these cases, the monomer and dimer bands of mouse SPR1a/b appeared to be broader, and some protein had migrated at a slightly faster rate than in the EDTA controls. Based on Coomassie staining, Western blots, and amino acid analyses (data not shown), this shift was probably due to intrachain cross-linking (i.e. some monomer and dimer SPR1 proteins had become more compact and migrated faster). A similar observation has been made previously in the in vitro cross-linking of loricrin (33). Moreover, these data mean that the ways in which the various TGases use the SPR1 proteins are independent of the numbers of repeating peptide motifs of the central domain.
Kinetics of TGase 1, 2, and 3 Cross-linking of Recombinant SPR1 Proteins-We determined the kinetic constants of the cross-linking of the recombinant human SPR1 protein with the baculovirus-expressed full-length TGase 1, TGase 2, and activated TGase 3 enzymes (Table I). By use of high concentrations of putrescine, we suppressed putrescine oligomerization of SPR1. Estimates of kinetic values are complicated by the fact that the SPR1 protein serves as a complete substrate and that multiple Gln residues are utilized (see below), so that the values obtained represent average data for the multiple Gln residues used. The data show that there are significant differences in the kinetic efficiencies (k cat /K m ) of the reactions. As suggested by the autoradiograms of Fig. 4, the TGase 3 enzyme cross-linked SPR1 2-4 times more efficiently than TGases 1 or 2. These data are very similar to those for recombinant loricrin (33) and SPR2 (32). From the double displacement mechanisms involved, we also calculated the K m values for putrescine in the human SPR1 reaction, which show a much higher value for the TGase 3 enzyme. These data suggest that this enzyme greatly prefers to use the Gln residues of SPRs for cross-linking.
We also determined kinetic parameters for the TGase 1 and 3 enzymes using as substrates synthetic peptides of sequence corresponding to the head and tail domains of human SPR1 (Table I). While the kinetic efficiency values were all less than for the intact protein, the values for the tail domain peptide were only about 2-fold less. However, there were wide variations for the head domain sequences. We note that there are two Gln/Lys regions encompassing residues 1-13 and 14 -25, which we have termed the head A and head B domains delineated by a triple Pro motif (see Fig. 5D). We found that the TGase 1 enzyme used the head B domain sequences Ͼ100 times more efficiently than the head A domain sequences, but the TGase 3 enzyme used the head A domain 18 times more efficiently than the head B domain. Thus, there are apparent sequence-specific preferences for these two TGases.
Amino Acid Sequencing Analyses of Cross-linking Reactions By The TGase 1 and 3 Enzymes in Vitro-Next, we obtained more specific information on the residues used for cross-linking in the recombinant SPR1 by the TGase 1 and 3 enzymes. Due to very low degrees of reaction (Figs. 3 and 4), we could not obtain useful information for TGase 2 cross-linking. Samples of uncross-linked or cross-linked protein were digested to completion with trypsin, and the peptides were separated by HPLC. Amino acid sequencing of shifted, reduced, or new peaks provided quantitative information on the utilization of every Gln and Lys residue of the protein.
In the case of the reaction with the baculovirus full-length TGase 1 (Fig. 5B), the peptide peaks eluted at 23 min (residues 38 -43), 30 min (residues 44 -51), 36 min (residues 52-67), and 37 min (residues 22-36) containing central domain repeats and were not visibly reduced. (Note that Lys 59 was reproducibly only partially cleaved by trypsin). However, the peptide eluted at 22 min containing residues 1-6 was reduced by about 20%; those eluted at 39 min containing residues 68 -84 and at 41 min containing residues 7-21 were reduced by about 60%. Several smaller new peptides appeared late in the chromatogram that contained cross-links involving Gln and Lys residues of these peptides. Also, we observed in these peaks about a 5% utilization of Lys 43 , indicating a minor degree of cross-linking of the central repeating domain (Fig. 5D, upper numbers). Note that tryptic peptides containing the terminal residues 85-86 and 87-88 were not resolved by this HPLC program. However, these residues were encountered in the unresolved peak eluted between 42-48 min, so that it was possible to estimate the relative percentages of these and all other Gln and Lys residues used for cross-linking as summarized in Fig. 5D (upper numbers). The total amounts used corresponded to the appearance of equivalent amounts of isodipeptide. The total loss of these residues suggested that the TGase 1 enzyme had inserted about 1 mol/mol of cross-link, which is consistent with the amount isolated directly. Furthermore, samples that had been cross-linked by the high specific activity 67/33/10 kDa TGase 1 form recovered from the membrane fraction of NHEK cells gave similar data on residue usage (not shown). Together with several foregoing data, it appears that the forms of TGase 1 all treat the SPR1 substrate the same way and validate use of the baculovirus-expressed full-length enzyme, which is more stable and available in quantity.
In the case of cross-linking with the TGase 3 enzyme (Fig.  5C), the peptides eluted at 23, 30, 36, and 37 min containing central domain sequences were reduced by about 10%, suggesting a minor degree of usage of Lys 37 , Lys 43 , and Lys 59 for cross-linking. However, peptides containing head (residues 1-6 at 22 min and residues 7-21 at 41 min) and tail (residues 68 -84 at 39 min) domain sequences were quantitatively shifted to an unresolved peak late in the chromatogram. By sequencing 1-min time aliquots across this peak, it was possible to estimate the extents of usage of each Gln and Lys residue of the end domains (Fig. 5D, lower numbers), although as many as four sequences were running simultaneously. These accounted for Ͼ95% of about 3.1 mol/mol of inserted isodipeptide.
Samples that had been cross-linked by these TGase 1 or TGase 3 enzymes for only 6 h yielded nearly identical data on Gln/Lys residue usage, except that the amounts used were less because of incomplete reaction, indicating that there were no time-dependent preferences for cross-linking.
Correlation of in Vitro and in Vivo Cross-linking of SPR1 Proteins-In Fig. 6, we have summarized the present in vitro data for TGases 1 and 3, and compared them with the in vivo utilization of Gln and Lys residues of SPR1 proteins observed in our previous studies on CEs isolated from either human foreskin (1, 2) and cultured epidermal keratinocytes (Fig. 6B) (4) 2 or mouse forestomach epithelia (Fig. 6C) (5). Comparisons of mouse and human data are relevant, since the numbers, locations, and sequences around the head and tail domain Gln and Lys residues are very similar (14). Several points emerge. First, there were no cases of usage of central domain Gln or Lys residues in the in vivo studies, which compares well with their summed Ͻ5% utilization for cross-linking in vitro. Second, overall there are remarkable similarities in the patterns of utilization of the head A and B and tail domain sequences: for virtually every residue, the relative profiles were retained. The most significant difference appears to be Gln 5 of human SPR1, which comparatively was used about twice more in vitro than in vivo. Third, these data suggest that the recombinant SPR1 protein is treated in a generally very similar way by the TGase 1 and 3 enzymes cumulatively in vitro as occurs in vivo in the foreskin epidermis or forestomach epithelium and thereby offer further evidence that the recombinant protein had adopted its native configuration in vitro. Moreover, fourth, it becomes clear that both the TGase 1 and 3 enzymes are required for the appropriate cross-linking of the SPR1 proteins in vivo. In view of the differences in the usage of some Gln residues of the head domain in vivo compared with the summed in vitro data for TGases 1 and 3, we cannot however exclude the possibility that another TGase enzyme(s) may also contribute to the effective cross-linking of the SPR1 proteins in vivo (50). We conclude that the TGase 2 enzyme is only weakly involved, if at all (Figs.  3 and 4). Similar conclusions were made for another major CE structural protein, loricrin (33), but these data differ significantly from those of SPR2, for which essentially only the TGase 3 enzyme was used (32).
However, we observed some asymmetry in the distribution of usage of head and tail domain sequences in vitro and in vivo (Fig. 6). The majority of the Lys residues used are located on the tail domain, while most Gln residues used are located on the head domains. Since the first step of a TGase cross-linking reaction involves recognition of a donor Gln residue (44 -46), our in vivo and in vitro observations strongly suggest that TGase substrate specificities are governed primarily by the head domain sequences of the SPRs. Consistent with this conclusion, we observed that the patterns of in vitro cross-linking of recombinant human SPR1 and native mouse SPR1a/b proteins, which have highly homologous head domain sequences, were essentially identical (Fig. 4D), although they contain six or 13/14 central domain peptide repeats, respectively. Also, each of the Gln and Lys residues of the tail domain sequences was used relatively approximately equally by the TGase 1 and 3 enzymes, although 3-fold less by TGase 1 (Figs. 4 and 5). TGases 3 and 1, Respectively-However, there were clear differences in utilization of different head domain sequences. Examination of the kinetic data for synthetic peptides of Table I and the cross-linking data of intact recombinant SPR1 of Figs. 5 and 6 reveal that the head A domain sequences were used almost exclusively (Ͼ90%) by TGase 3, and those of head B mostly (ϳ85%) by TGase 1. Furthermore, in the case of the TGase 3 enzyme, more than 3 mol/mol of isopeptide cross-link were inserted, and ϳ90% of the protein was oligomerized into dimers and tetramers (Fig. 4). It is therefore likely that head A domain sequences were used by the TGase 3 enzyme in vitro largely for interchain head-to-head, or head-to-tail cross-linking. However, we cannot rule out the possibility of significant intrachain cross-linking as well. The reason for this uncertainty is that it was necessary to proteolyze the in vitro crosslinked products (Fig. 5) in order to elucidate sequencing information; thus, only fragments of sequences were obtained joined by isopeptide bonds, so that it remains unclear whether the cross-links were interchain or whether some were intrachain as well (4). On the other hand, the TGase 1 enzyme inserted 1.1-1.2 mol of isopeptide into the recombinant human and native mouse SPR1 proteins using almost exclusively the head B domain sequences (see Fig. 5B). Since only ϳ11% of the protein was oligomerized, it appears that most of the crosslinking in vitro involved intrachain links within individual SPR1 protein molecules, as inferred from the mobility shifts seen in Fig. 4. We note that there was a minor degree of overlap in these two enzyme functions, since 5-10% of TGase 3 crosslinking involved head B domain sequences, and ϳ15% of the TGase 1 reaction involved interchain cross-linking as well as head A domain sequences (Fig. 6).

Differential Functional Use of Head A and B Domains by
The present conclusion that the head A and B domains are used in vitro for cross-linking by the TGase 3 and TGase 1 enzymes, respectively, is entirely consistent with our previous data for the recombinant human SPR2 protein (32). Only interchain cross-linking was observed with both enzymes, and moreover, TGase 3 inserted Ͼ10 times more cross-links than TGase 1 (32). In this regard, it is noteworthy that the SPR2 protein contains the equivalent of head A domain sequences only, since a second Gln/Lys-rich region following a triple Pro motif is absent (13,32). These differences imply important structural-functional differences in their utilization in epithelia.
Analysis of in Vivo Cross-links-SPRs were found to be crosslinked in vivo to many other proteins, including to themselves. Our data base of in vivo cross-links is based on extensive proteolysis procedures in order to obtain informative peptides for sequencing and includes 234 occurrences of human (4) and 121 mouse (5) SPR proteins, of which 38 involved direct SPR-SPR linkages. Since the present work on SPR1 and the previous study on SPR2 proteins explored the cross-linking of the SPR proteins to themselves, it is now possible to predict which TGase enzymes may have formed the human and mouse crosslinks in vivo (Table II). Some peptides clearly resulted from interchain cross-linking between separate SPR protein mole-cules, but in many cases, the short piece of SPR sequence recovered after proteolysis precluded assignation of the type of cross-linking. Of the 38 SPR-SPR cross-linked peptides, eight involved head A-head A sequences of which most were interchain, and 11 involved head A-tail linkages. We can now conclude that these were likely inserted by the TGase 3 enzyme. A group of seven involved head A-head B linkages and seven tail-tail linkages might have been inserted by either the TGase 1 or 3 enzymes. We predict that the TGase 1 enzyme could have inserted four head B-tail cross-links and one head B-head B linkage.
However, reconstruction of the potential TGase enzymes used for the 97 single SPR-protein (X) or 220 SPR-X-SPR linkages is less certain. We note that the head A domain Gln residues were used three times more than head B domain Gln residues (4), which based on the present observations, suggests that the TGase 3 and 1 enzymes preferentially use head A and head B domain sequences, respectively. Thus, we propose that 140 could have been inserted by TGase 3, 61 by TGase 1, and 116 by either enzyme (Table II).
SPR Cross-linking with Loricrin-Previously, we have demonstrated that loricrin is likewise cross-linked in vivo by both the TGase 1 and 3 enzymes using different Gln and Lys residues (33), but in this case, there was more overlap between the various residues used and enzymes (about one-third). We have documented 38 SPR1-loricrin and 19 SPR2-loricrin peptides in human epidermal CEs as well as 58 SPR1-loricrin and 26 SPR2-loricrin peptides from mouse forestomach CEs (4). Accordingly, now we have reexamined these 141 data elements to ascertain which Gln and Lys residues were joined in an attempt to correlate the TGase enzymes used (Table III). The analyses show that SPR1 head A domain or SPR2 head domain sequences, predicted to be used by the TGase 3 enzyme, were preferentially cross-linked to loricrin Gln and Lys residues that themselves were predicted to be utilized rather specifically by the TGase 3 enzyme. Likewise, we find a very high correlation of cross-linking of head B domain SPR1 sequences to loricrin at sites predicted to be used preferentially by the TGase 1 enzyme. These data strongly infer a common highly coordinated mechanism for the cross-linking of loricrin and SPRs together in vivo.

Recombinant SPR1 Is Cross-linked Sequentially by the TGase 3 Followed by the TGase 1 Enzymes to Form Very Large
Oligomers-In view of the above considerations that both the TGase 1 and 3 enzymes are required for the correct crosslinking of SPR1 proteins in vivo using different Gln/Lys residues of the head B and A domains, respectively, we wondered whether the two enzymes operate simultaneously or at different times. To test this, we repeated the in vitro reactions using 35 S-labeled recombinant human SPR1 in which we first separately cross-linked to completion with either TGase 1 or TGase 3 and then performed a second cross-linking reaction on the products with TGase 3 or TGase 1. When baculovirus-expressed full-length TGase 1 was used to cross-link the short oligomeric products of an initial complete TGase 3 reaction, there was a major shift to very large material that could not enter the gel: excision of 35 S label revealed that within 12 h, Ͼ90% had been thus converted (Fig. 7A). Similar data were obtained when the high specific activity 67/33/10-kDa complex was used for 1 h (not shown). In both cases, a total of 4.2 mol of isodipeptide cross-link had been inserted per mol of SPR1 protein, or about 1 mol/mol more by TGase 1, thus implying that additional Gln/Lys residues had been recruited. Following trypsin digestion and HPLC fractionation, a very broad unresolved peak of highly cross-linked material was observed, not unlike that for TGase 3 enzyme alone (Fig. 5, compare E with C). Sequencing 1-min aliquots across the peak revealed significant increases in the amounts of head B Gln residues used for cross-linking, but those of head A domain sequences were essentially unchanged (Fig. 5G, upper numbers; compare with Fig. 5D, lower numbers). Thus, the additional cross-linking performed in the secondary TGase 1 reaction involved mainly head B sequences. We also repeated these experiments in the reverse order of cross-linking by TGase 3 following initial maximal TGase 1 cross-linking (Fig. 7B). In this case, however, there was not a major change in the cross-linking pattern. The secondary TGase 3 enzyme reaction transformed about 20% of the monomeric sized TGase 1 products into short oligomers and 5% into very large oligomers that could not enter the gel, but most remained as the monomer. Also, only about an additional 0.5 mol of cross-link/mol was inserted. Sequencing analyses revealed minor changes in the degrees of utilization of head A domain residues by the TGase 3 enzyme (Fig. 5, F and G, lower  numbers).
Together, these in vitro data strongly imply an obligatory temporal order of the cross-linking of the SPR1 proteins in vivo by the TGase 3 enzyme first, followed later by TGase 1. For apparent spatial and/or structural reasons that are not clear, it appears the head A domain sites must be utilized first by the TGase 3 enzyme before TGase 1 can efficiently recognize the head B domain sites, whereas occupancy of head B sites first interferes with a subsequent TGase 3 reaction. Accordingly, further attempts to obtain detailed structural information on the SPR1 proteins seem desirable.
A Model for Cross-linking of SPRs and Loricrin Together in Vivo-Any model must account for several established aspects of SPR and loricrin expression in epithelia. First, when coexpressed in epithelia, SPR expression always precedes that of loricrin (21,51). Second, we note that SPRs are very soluble and are deposited in both the nucleus and cytoplasm (21,51), and antibodies elicited against tail domain epitopes used for cross-linking (Figs. 4 and 5; Table I) yield a bright cytoplasmic staining reaction in tissues such as footpad epidermis or rodent forestomach, where they are abundantly expressed (14,21). Third, on the other hand, loricrin is insoluble in physiological conditions (33). Fourth, a peripheral staining by the SPR antibodies is observed in more differentiated cells coincidentally with initial loricrin expression (21,51). Fifth, antibodies against both SPRs and loricrin no longer react in fully cornified epithelia, indicating lost epitopes due to cross-linking (14,21,41,52). Sixth, the activated TGase 3 enzyme is soluble (cytosolic) (40,53), whereas the most highly active forms of TGase 1 in keratinocytes are membrane-bound (37,38,54,55). Finally, any model must conform to the phenotypic observations of the TGM1 gene Ϫ/Ϫ mouse (56) and the natural human knockout of the TGM1 gene in lamellar ichthyosis.
Our data allow the following model, which is consistent with all of the extant observations. We propose that SPR proteins might be expressed early because more time is required to gather them together by an active (3) or passive transport system near the cell periphery than is required for loricrin. Then juxtaposed loricrin and SPRs at or near the cell periphery are first cross-linked together to form small oligomers by the cytosolic TGase 3 enzyme using most of the head A domain and some tail sequences of the SPR1 and -2 proteins and favored internal Gln and head/tail Gln/Lys residues of loricrin. Subsequently, the highly active TGase 1 enzyme form(s) anchored to the membrane attach this loricrin-SPR1/2 complex to the growing CE structure by further cross-linking using available SPR1 FIG. 6. Comparative utilization of Gln and Lys residues for cross-linking in vitro and in vivo. A, cross-linking of recombinant SPR1 in vitro (present work). B, summarized data on cross-linking of human SPR1 proteins in vivo generated from amino acid sequencing of isolated CEs from several sources of human epidermal keratinocytes (1, 2, 4). 2 C, summary of data for mouse SPR1 proteins in vivo generated from amino acid sequencing of isolated forestomach CEs (5). Red bars, Gln residues; blue bars, Lys residues; open bars, TGase 1 cross-linking; closed bars, TGase 3 cross-linking. Note that unlike human SPR1a/b, mouse SPR1a/b contain different numbers of central domain peptide repeats, which affects the numbering of terminal domain residues. Thus, tail sequence positions in C are numbered from the conserved terminus, denoted as E.
head B domain and remaining tail sequences of SPR1/2 to the additional available head/tail sequences of loricrin. These temporal differences may be explained in part by TGase enzyme availability as well as in part by the particular conformation of the SPR head domain sequences. It is conceivable that the SPR1/2 head A-like domain sequences that are homologous to those of loricrin (57) evolved so as to favor initial cross-linking by TGase 3 because it is a cytosolic enzyme. Utilization of SPR1  (Fig. 7). Because the vast majority of TGase 1 activity is membrane-bound presumably at or very near the site of final protein deposition of the CE, this enzyme is used for the final reinforcement stages of CE assembly. Furthermore, this model is readily consistent with the pathophysiology of lamellar ichthyosis disease caused by absent TGase 1 activity. Because the TGase 3 enzyme cross-links the proteins with high specificity at specific sites different from those of the TGase 1 enzyme, the TGase 3 enzyme cannot adequately compensate or supplement for the lost TGase 1 enzyme activity. Thus, the reinforcement stage of CE assembly occurs only poorly, resulting in the devastating loss of barrier function seen in human lamellar ichthyosis patients. Similarly, this model is entirely consistent with the observed morphological and histological features and severe loss of barrier function in the mouse Ϫ/Ϫ model.
A similar scenario should be applicable to the CEs of those epithelia that express much less or no loricrin and where the SPRs instead are mostly cross-linked to themselves and other scaffold proteins. Furthermore, the second most frequent partner to which SPRs have been found cross-linked in vivo is involucrin (4). Again, both head A and head B domain sequences were commonly used, suggesting the involvement of both TGases 1 and 3 (4). However, comparable analyses of SPRs and involucrin cross-linking are not yet possible, since only limited data on favored residues used by TGases are available for involucrin (58).
Conclusions-Altogether, these studies indicate that the cross-linking of SPR proteins to the CE for effective barrier function is a complex process, involving multiple TGases operating through numerous Gln and Lys residues, and in multiple well coordinated steps. The availability of recombinant loricrin and SPR proteins now will permit further experiments to explore how they function together to affect the biomechanical properties of the CE and the epithelium in which they are expressed (5). Such experiments may also shed light on the reason why there are multiple apparently functionally very similar yet differentially expressed SPR1 and SPR2 gene products. FIG. 7. Double reaction of recombinant SPR1 reveals an obligatory temporal order of cross-linking by TGase 3 followed by TGase 1. Reactions as in Figs. 3 and 4 were performed with ϳ700 nM of recombinant baculovirus TGase 1 and activated TGase 3. In A, the products of an 18-h TGase 3 reaction were reacted again with TGase 1 for the times shown. In B, an 18-h TGase 1 reaction was reacted again with TGase 3 for the times shown.