Consequences of Seven Novel Mutations on the Expression and Structure of Keratinocyte Transglutaminase*

We report the molecular characterization of seven new keratinocyte transglutaminase mutations (R315C, S358R, V379L, G473S, R687C, deletion Δ679–696, R127Stop) found in lamellar ichthyosis patients. Arg-315, Ser-358, Val-379, and Gly-473 are highly conserved residues in transglutaminases while Arg-687 and Δ679–696 are not. All mutations strongly decreased transglutaminase activity and protein levels. The mutation R127Stop diminished the amount of mRNA. Structural analysis of these mutations based on the factor XIII A-subunit crystal structure demonstrated that Arg-315, Ser-358, Val-379, and Gly-473 are located in the catalytic core domain, and Arg-687 and the deletion are in the β-barrel domains. The side chains of amino acids Arg-315, Ser-358, and Gly-473 make ionic and hydrogen bonds important for folding and structural stability of the enzyme but are not directly involved in catalysis. Val-379 is two amino acids away from the active site cysteine, and its change into leucine disturbs the active site structure. The decreased activity and protein level after expression of the R687C and Δ679–696 TGK cDNA in TGK negative keratinocytes excluded that they are polymorphisms. These results identify important amino acids in the central core domain of transglutaminases and show that the C-terminal end influences the structural and functional integrity of TGK.

Transglutaminases (EC.2.3.2.13, protein-glutamine: amine ␥-glutamyl-transferase) are a superfamily of enzymes which catalyze the formation of intra-and intermolecular ␥-glutamyl-⑀-lysine isodipeptide bonds (1,2). They are calcium-dependent enzymes that contain an active site consisting of a catalytic triad (Cys, His, Asp) (3)(4)(5). The six different classes of transglutaminases are participating in a wide variety of physiological processes (3,6,7). One member of this family, keratinocyte transglutaminase (TGK), 1 is involved in cross-linkage during formation of the cornified cell envelope (CE), a highly insoluble 8 -15-nm wide structure replacing the plasma membrane in terminal differentiating epidermis (8,9). During this process, CE precursor proteins such as loricrin, involucrin, and small proline-rich proteins are sequentially cross-linked on the inner side of the plasma membrane (10 -13). TGK protein is localized mainly to the cell periphery in the granular layer. The enzyme consists of 815 amino acids, and it is post-translationally modified by fatty acid acylation and phosphorylation (14 -17). Several complexes consisting of the full-length protein and polypeptides proteolytically cleaved from it have been identified in the cytosolic and membrane fractions (18,19). Most of the enzyme complexes are attached to the membrane through myristate and palmitate chains (20,21). About 5-10% of TGK activity is found in the cytoplasmic fraction, which might be involved in the final steps of CE assembly. Deletion analysis showed that a molecule in which the first 109 and the last 240 amino acids have been removed retains a specific activity comparable with the full-length enzyme (22). The human TGK gene consists of 15 exons and is located on chromosome 14q11 (23)(24)(25)(26)(27). At least two different allelic variants have been detected in the human population (24).
Autosomal recessive lamellar ichthyosis (LI) (Mendelian Inheritance in Man No. 242100, 242300) is a severe congenital scaling skin disorder with a frequency of about 1:250,000 (28,29). The clinical phenotype is heterogeneous and can range from generalized large brownish plate-like scales with no erythroderma to fine white scales with underlying erythroderma. Moreover, patients may have palmar and plantar hyperkeratosis, scarring alopecia, ectropion, eclabium, and decreased sweating. Patients are often born encased in a shiny, thick parchment-like membrane (collodion baby). By electron microscopy, five types of lamellar ichthyosis (ichthyosis congenita type I-V) have been distinguished (30). Deleterious mutations in the TGK gene have been reported in lamellar ichthyosis patients providing compelling evidence for the importance of the cornified cell envelope for epidermal homeostasis and the barrier function of the skin (31)(32)(33). However, biochemical data clearly showed that about 50 -60% of LI patients have normal TG activity (34). Genetic heterogeneity is further supported by genetic mapping studies identifying two other disease-causing genes, one on chromosome 2q33-35 and another at a currently unknown location (35).
We report seven novel TGK mutations found in LI patients. The consequences of these mutations (five missense mutations, one premature stop codon, and a deletion of 18 amino acids) on the biosynthesis of TGK mRNA and protein are described. The effects of these mutations on folding and structure are analyzed using the three-dimensional structure of factor XIII A-subunit as model. This study identifies structurally and functionally important amino acids of TGK and provides new insight into the structure-function relationship in transglutaminases.

EXPERIMENTAL PROCEDURES
Patients-Families from Switzerland (LI-8), Holland (LI-11), Sweden (LI-20), and the United States (LI-22) were investigated (see Fig. 1). The proband of family LI-8 was born as a collodion baby and died shortly after birth due to bacterial infection. The affected individual of family LI-11 was not born as a collodion baby; his trunk and neck are covered with large plate-like yellow-brown hyperkeratotic scales, and he has very extensive palmar-plantar keratoderma with large fissures. The face is not involved and there is no alopecia or ectropion. The clinical data of the affected members of family LI-20 have been described earlier (36). Patient IV.4 of LI-22 (see Fig. 1) was born as a collodion baby after a normal, full-term pregnancy. He now has generalized, thin white to brown scales that are more plate-like on the scalp and the lower extremities. No blistering or significant erythroderma are apparent. Palms and soles show moderate hyperkeratosis. Patient II.2 of LI-22 ( Fig. 1) was not born as a collodion baby and now has thick scales on the scalp, powdery fine scaling on the back and arms, and hyperlinear palms. The affected members of family LI-22 had normal cholesterol sulfate and cholesterol sulfatase levels.
Cell Culture-Punch biopsies obtained from the probands were used to establish primary cultures on lethally irradiated murine 3T3 fibroblasts as described earlier (37)(38)(39). Secondary cultures were grown in high calcium keratinocyte medium, 10% fetal calf serum until confluency. After an additional 5 days in culture, genomic DNA, RNA, and proteins for measuring transglutaminase activity and immunoblotting were extracted as described below.
Isolation of DNA and RNA-Genomic DNA was purified from cultured cells by phenol/chloroform extraction as described earlier (40) or from blood using NucleoSpin columns (Macherey-Nagel). Total RNA was isolated using the guanidine-thiocyanate method (41).
Transglutaminase Assay-Cells were lysed by sonication in 20 mM sodium phosphate, pH 7.2, 0.5 mM EDTA, 10 mM dithiothreitol, 50 g/ml phenylmethylsulfonyl fluoride. The supernatant, after centrifugation at 25,000 ϫ g at 4°C for 30 min, was used as cytosolic fraction. The cell pellet was re-extracted by sonication with the same buffer supplemented with 1% Triton X-100. After incubation for 10 min at 37°C, the lysate was centrifuged again, and the supernatant (membrane fraction) was collected for measuring the transglutaminase activity (43). Transglutaminase activity is expressed as pmol of 3 H-putrescine incorporated into dimethylcaseine per hour and per mg of protein. Results are indicated as mean Ϯ S.E. in cell extracts from at least two different cell passages, each measured in duplicate.
Western Blot Analysis-Cells were lysed by sonication in 10 mM Tris-HCl, pH 7.4, 5 mM EDTA, 50 g/ml phenylmethylsulfonyl fluoride, 1 g/ml pepstatin, 1 g/ml E-64, 1 g/ml leupeptin. The supernatant, after centrifugation at 25,000 ϫ g for 30 min, was taken as cytosolic fraction. The membrane fraction was obtained after sonication of the cell pellet in the same buffer supplemented with 1% Triton X-100 and centrifugation. 40 g of protein was size-fractionated by SDS-polyacrylamide gel electrophoresis through a 10% separation and 4% concentration gel (containing 4 M urea) and, after partial renaturation, electroblotted to nitrocellulose (9). TGK protein was visualized with antibody B.C1 (8) and the ECL detection kit (Amersham, Switzerland).
Protein Concentrations-Protein content was determined with the Bradford assay (Bio-Rad) using bovine serum albumin as standard (44).
DNA Sequencing and Family Analysis of Mutations-The 15 exons of the TGK gene were amplified by PCR as described (33). Forward primers were biotinylated. PCR products were purified by QIAquick PCR purification kit (Qiagen), and single-stranded DNA was isolated with streptavidin-coated magnetic beads (Dynal) and sequenced with the reverse primers using the Sequenase sequencing kit (Amersham). Nucleotides have been numbered according to Phillips et al. (26). To number amino acids, the first methionine of the open reading frame (15) was designated as number 1. For inheritance analysis in families, DNA was amplified by PCR, digested with restriction enzymes, and separated on agarose or polyacrylamide gels.
Expression of Mutant Proteins-Full-length TGK cDNA was obtained by RT-PCR using patients keratinocyte RNA and primers DH8 5Ј-CATCCATCCTGACCTGTTCCA-3Ј (nt Ϫ79 to Ϫ59 (16)) and DH9 5Ј-GTTTATTAGCATCTGTTCCCCCAGT-3Ј (nt ϩ2580 to ϩ2604 (16)) and cloned into the NotI site of pCI (Promega). The sequence was verified by sequencing. ␤-Galactosidase cDNA was obtained from plasmid H3700-pL2 (45) and cloned into the NotI site of pCI. Plasmids were purified over Qiagen columns and by Triton X-114 extraction (46). Secondary keratinocytes from a TGK negative LI patient cultured on irradiated 3T3 fibroblast feeder layer were transfected at 80 -90% confluency with 4 g of the TGK-expressing plasmid and 2 g of the ␤-galactosidase expressing plasmid (47). Two days later, ␤-galactosidase and transglutaminase activities were determined (43,48).
Modeling of the Protein Defects-The three-dimensional structure of the human factor XIII A-subunit zymogen dimer, experimentally determined by single crystal x-ray diffraction (5), was used as a template for constructing a homology model of the human keratinocyte transglutaminase enzyme using the Biosym InsightII software package. Atomic coordinates for the factor XIII structure were obtained by refining the model against x-ray diffraction data from 10.0 to 2.65 Å resolution using the program X-PLOR (49) to give a crystallographic R factor of 21.7%. The final model exhibits good geometry (root mean square deviation from ideality of 0.012 Å for bond lengths, 1.8°for bond angles, 25.6°for torsion angles, and 1.5°for improper torsion angles); the average value of the individually refined atomic temperature factors is 26.7 Å 2 . Refined coordinates for the factor XIII structure have been deposited with the Protein Data Bank (identifying code: 1ggt). Models of the keratinocyte transglutaminase mutant structures were generated by modifying the homology model using the computer program O (50), and figures were drawn with the program MOLSCRIPT (51).  (Table I). Northern blots showed missing TGK mRNA in proband LI-8 II.2 and aberrant synthesis in LI-22 IV.4 and LI-22 III.3, whereas probands from the other two familes had normal sizes and levels of mRNA (Fig. 2). The banding pattern obtained with the probes DH42 and 3Ј NC in probands III.3 and IV.4 of LI-22 ( Fig. 1) was very similar to the one observed in the previously reported family LI-2, which had a homozygous A to G change in the splice acceptor site of intron 5 (33). TGK protein levels in cytosolic and membrane fractions were strongly decreased in all individuals with low TG activity (Fig. 3a).

Biochemical Characterization of the Patients-The
Sequence Analysis of Patients TGK Gene-The mutations shown in Fig. 4 were detected by direct DNA sequencing of all 15 TGK exons in individuals with low TG activities. LI-8 II.2 had a homozygous C to T mutation at position ϩ1354 in exon 3, changing R127 to a stop codon. This mutation creates a new DdeI site in exon 3 giving rise to a new band of 181 bp. DdeI digestion of amplified exon 3 from the patient and his parents showed that the patient was homozygous for the 181-bp band, whereas the parents were heterozygous for the 181-and 207-bp band (Fig. 4g). Patient LI-11 II.1 was a compound heterozygote for two missense mutations. He carries a G to A change at position ϩ7110, changing Gly-473 to serine (Fig. 4a), and a C to T change at ϩ8479, changing Arg-687 to cysteine (Fig. 4b). In family LI-20, patients III.1 and III.2 carry a heterozygous C to G mutation at ϩ4019 (S358R) inherited from the mother (Fig.  4c) and a heterozygous G to T change at ϩ8509 in the splice donor site of intron 13 inherited from the father (Fig. 4d). Sequencing of 14 cDNA clones obtained by RT-PCR using total RNA from patients LI-20 III.1 and III.2 ( Fig. 1) with primers DH5/DH7 (33) revealed that 10 clones contained a deletion of 54 nucleotides (ϩ8454 to ϩ8509), whereas the remaining clones had normal sequence. This implicates that the splicing machinery uses GT at position ϩ8454 as the new splice donor site, which is consistent with the calculation of the consensus values for putative splice acceptor sites (52, 53) 100 bp up-and downstream of the mutated splice junction (data not shown). This leads to an in-frame deletion of amino acids 679 -696 (⌬679 -696). Since the mother was affected by lamellar ichthyosis and showed low TG activity (Table I), all 15 of her exons were sequenced. This revealed an additional heterozygous mutation replacing a G nucleotide with C at position ϩ4080 changing Val-379 into leucine (Fig. 4e). This mutation was not present in her two children showing that the V379L and S358R mutations are not on the same allele. Patient LI-22 IV.4 is a compound heterozygote for an A to G (ϩ3366) change in the splice acceptor site of intron 5 (data not shown) and a C to T exchange at position ϩ3434 (R315C) (Fig. 4f). The A to G mutation at ϩ3366 creates a new MspI site, allowing us to follow the inheritance of the mutation in the LI-22 pedigree. The mutant allele is present in the patient and was inherited from his mother and maternal grandmother (data not shown).
The second mutation at ϩ3434 destroys the single HaeIII site in exon 6. The restriction enzyme analysis showed that only the patient has this mutant band (data not shown). Whether this mutation was inherited from the father or represents a new mutation could not be tested since DNA from the father was not available.
CpG dinucleotides have on the average a much higher rate of mutations than other dinucleotides (54). In four of the presented 7 mutations (R127Stop, R315C, G473S, and R687C), C nucleotides in CpG are mutated into a T either on the sense or on the antisense DNA strand (G473S). Therefore, these sites could constitute mutational hot spots.
Protein Modeling of the Mutants-The extensive conservation of amino acid residues of 42% between keratinocyte transglutaminase and factor XIII A-subunit indicates that their folding is conserved. Therefore, the factor XIII A-subunit crystal structure served as a reliable scaffold to construct a homology model of keratinocyte transglutaminase (Fig. 5) to better understand the molecular basis for the decreased enzymatic activity caused by the keratinocyte transglutaminase mutations. Factor XIII A-subunit is composed of four domains, which, from the N-terminal end, have been designated as ␤-sandwich, central core domain, and ␤-barrels 1 and 2 (5). For additional indications of the structural and functional importance of the mutation sites, a structure-guided alignment of 19  known transglutaminase sequences was additionally used (data not shown). R315C-Residue Arg-315, in the catalytic core domain, is located in a surface loop between two helices (Fig. 5). The arginine side chain is buried in the structure and forms a salt bridge with Asp-306 as well as a hydrogen bond to the main chain carbonyl group of Met-310 (Fig. 6). These bonds serve to stabilize the conformation of this surface loop. The Arg-315-Asp-306 salt bridge is conserved in a total of 13 of the TG sequences including factor XIII A-subunit and TGK; in one of the other sequences, the equivalent arginine residue interacts with a glutamic acid side chain, and in the remaining molecules, the size of this loop is altered by either amino acid insertions or deletions. The equivalent arginine in factor XIII A-subunit is the site of the deficiency mutation R252I (55), underlining the structural importance of this residue. The TGK FIG. 6. Close-up view of the modelled R315C mutation site. In the factor XIII crystal structure, the side chain of this arginine residue forms a salt bridge with an aspartic acid residue, and both amino acids are conserved in factor XIII and the keratinocyte transglutaminase sequences. Arg-315 also forms hydrogen bonds with the main chain carbonyl of Met-310, which is also conserved. All three residues are located in a surface loop between two helices in the catalytic core domain. Replacement of the large Arg-315 side chain with that of the smaller cysteine residue results in removal of all the wild-type hydrogen-bonding interactions and creates a gap in the molecule. -11 (a and b), LI-20 (c-e), LI-22 (f), and LI-8 (g). a, G to A change in exon 10, resulting in a G473S substitution; b, C to T in exon 13, which alters Arg-687 to cysteine; c, C to G change in exon 7, altering Ser-358 to arginine; d, G to T change in intron 13, destroying its splice donor site and leading to deletion of amino acids 679 to 696; e, an additional mutation was found in the DNA of the mother, replacing a G with a C nucleotide in exon 7, changing Val-379 to leucine; f, C to T change in exon 6, changing Arg-315 to cysteine; g, homozygous C to T mutation in exon 3, changing R127 to a premature stop codon. This mutation creates a new DdeI site (giving rise to a new band of 181 bp), which was used for family analysis. R315C mutation has three effects. First, the described bonds formed by the arginine side chain are removed, rendering the surface loop more mobile and susceptible to proteolytic cleavage. Second, the introduction of the smaller cysteine side chain leaves a void in the molecule that would destabilize the structure; a possible consequence is interference with proper folding and the generation of an altered structure. Third, the introduction of an additional cysteine residue can interfere with proper folding by allowing the formation of an unwanted disulfide bond. Thus, the most likely result of the R315C mutation is the altered conformation of the surface loop yielding a modified structure that is less stable and more susceptible to proteolytic cleavage.

FIG. 4. Sequencing (a-e) and agarose (g) gels demonstrate novel mutations in family LI
S358R-The Ser-358 residue is absolutely conserved in all 19 TG sequences, which suggests structural and/or functional importance. This residue is located in the catalytic core domain (Fig. 5), and its side chain group is buried in the molecule. The Ser-358 side chain hydrogen bonds to the side chains of residues Thr-386 and Trp-288 and to the main-chain carbonyl of Gly-382 (Fig. 7). Since Thr-386 and Trp-288 are also absolutely conserved in the 19 TG and the Gly-382 main-chain torsion angles can easily accommodate other residues (this position is occupied by either a glycine or an alanine in the TG sequences), all three hydrogen bonds involving the Ser-358 are expected to be conserved among the TG structures and are likely to be structurally important. The S358R mutation is expected to have three important consequences. First, the mutation removes the three conserved hydrogen bonds that are likely to be critical for protein folding and stability. Second, to accommodate the much larger arginine side chain, the conformation of the protein at the mutation site must be dramatically altered. Finally, the arginine mutation introduces a buried positive charge in the protein, which will further interfere with proper folding of the protein. The result of the S358R mutation is predicted to be a dramatic misfolding of the catalytic core domain.
V379L-Residue Val-379, which is absolutely conserved, is located 2 positions C-terminal to the catalytic Cys-377 in the The Val-379 residue is located in the active site helix (shown as a coil), two residues C-terminal to the catalytic Cys-377 residue. The conserved Val-379 side chain, shown in dark gray ball and stick, is buried in a closely packed hydrophobic pocket in the catalytic core domain. The V379L mutation (white ball and stick) introduces a larger side chain, resulting in a number of sterically unfavorable short contacts that are relieved by distortion of the protein fold. This in turn leads to a shift of the catalytic Cys-377 residue affecting the enzymatic activity of TGK. active site helix that contains a number of conserved amino acids (Fig. 8). The Val-379 side chain is buried in a tightly packed hydrophobic environment formed by predominantly conserved residues and that cannot accommodate the larger leucine side chain of the V379L mutation without distortion of the local conformation of the protein. As a result of the ordinarily conservative V379L mutation, the position and orientation of the catalytic Cys-377 are likely to be altered, and the catalytic activity of the enzyme compromised.
G473S-Residue Gly-473 is conserved in all TG sequences except for the two band 4.2 proteins. This glycine residue is found on the surface of the catalytic core domain (Fig. 5) and, along with Pro-474 (conserved in all but the band 4.2 se-quences), forms the only cis-peptide bond in the factor XIII A-subunit. The main chain atoms of Gly-473 form hydrogen bonds with the side chains of two residues: Arg-323, which is absolutely conserved among the 19 sequences, and Asp-490, which is conserved in all but the two band 4.2 proteins (Fig. 9). The pattern of conservation of the Gly-473-Pro-474 pair and residues Arg-323 and Asp-490 in all 17 enzymatic TG sequences suggests that the conformation of the protein fold in this region, as determined by the cis-peptide geometry and the conserved hydrogen-bonding interactions, is crucial for catalytic activity. Consistent with this interpretation are the observations that Arg-323 is the site of a previously identified lamellar ichthyosis missense mutation as well as of the factor FIG. 9. Stereo view of the modelled G473S mutation site. Residues Gly-473 and Pro-474 are amino acids conserved among all the enzymatic transglutaminase sequences; in the factor XIII A-subunit structure, the Gly-473-Pro-474 peptide bond is the only one observed to be in the cis conformation. Gly-473 atoms (dark gray) participate in hydrogen-bonding interactions with the side chains of the highly conserved Arg-323 and Asp-490 residues. The main chain conformation of Gly-473 is unsuitable for any other amino acid with its larger side chain; substitution of Gly-473 with a serine residue (white ball and stick) would interfere with proper folding of the protein.
FIG. 10. Stereo view of the R687C mutation site. The Arg-687 residue (dark gray) is buried at the interface between the barrel 1 and catalytic core domains and is variable among the transglutaminase sequences. The side chain of this arginine residue is not predicted to be involved in any interactions critical during the protein folding process. Substitution of the large buried Arg-687 side chain with the much smaller cysteine residue is likely to interfere with the domaindomain interface. The result is a folded protein with a modified quaternary structure that has altered substrate binding and specificity or that is less stable or more easily degraded by proteases. XIII A-subunit deficiency mutant R260C (33,56). The G473S mutation will lead to a misfolded structure that is conformationally distorted since the glycine main-chain torsion angles cannot accommodate any other amino acid with its larger side chain; the serine side chain would introduce sterically unfavorable short contacts with main-chain atoms. The serine substitution will interfere with the folding process and yield an altered structure that is less stable or more susceptible to proteolytic cleavage.
R687C-The Arg-687 residue is buried at the interface between the C-terminal ␤-barrels and the catalytic core domains (Figs. 5 and 10) and is found only in the keratinocyte sequences. The consequence of the R687C mutation is not as evident as for the other four missense mutations. The variability of this residue among the transglutaminases indicates that this residue is not likely to be important during the protein folding process. Although there are few conserved residues at the barrel-core domain interface, the barrel domains are expected to be packed well against the catalytic core domains in all transglutaminase structures. Substitution of the large, positively-charged buried Arg-687 side chain with the much smaller cysteine residue is likely to interfere with the interdomain interface, and to yield a modified quaternary structure. The result is a globally altered molecule that is less stable or more easily degraded by proteases (see Fig. 3a, lanes 4 and 5 and Fig. 3b, lanes 5 and 6) but also might have altered substrate binding and specificity.
⌬679 -696 -The 18 residue deletion in the region corresponding to residues 618 -635 in factor XIII A-subunit forms one long strand that starts in barrel 1 and continues to barrel 2 (Fig. 5). The first part of the peptide forms most of a ␤-strand in barrel 1 at its interface with the core domain. The second part of the peptide forms half of a ␤-strand in barrel 2. Deletion of these 18 residues has dramatic structural effects. In a worstcase scenario, the C-terminal portion of the protein is unable to fold into a globular structure, and the entire molecule is unstable and degraded as shown by Western blot analysis (Fig.  3a, lanes 8 -11 and Fig. 3b, lanes 7 and 8). In a best-case scenario, the C-terminal portion (barrels 1 and 2) folds into an altered globular structure, and the modified protein is stable. However, in this case, the new C-terminal domain will not only have an altered structure but will also not be packed against the catalytic core in the same manner, thus any putative function served by the barrel domains (substrate binding and specificity, specificity of enzyme cleavage, and activation) will be lost.
Transient Expression of the Mutants R687C and ⌬679 -696 -Since the mutations R687C and ⌬679 -696 do not concern highly conserved residues, it was less evident if they would influence TG activity. Therefore, these mutant cDNAs were transiently expressed by cotransfections with a ␤-galactosidase expression plasmid into TGK negative keratinocytes derived from a LI patient. The mutation R687C reduces membrane TG activity to about 5% of the normal level (Table II). An even stronger reduction was observed for the ⌬679 -696 protein molecule (Table II). The three mutations S42Y, R142C, and R323Q were reported earlier in a LI family (33) and were included in Table II to demonstrate the ability of this transient expression assay to detect deleterious mutations. The data show also that the S42Y change, located close to the membrane attachment site of the molecule, is not a disease-causing mutation but does lead to increased cytosolic accumulation of TGK as previously postulated (33). Furthermore, Western blot analysis of cell extracts from transfected cells demonstrated an excellent correlation between the levels of TGK protein and TG activity (Fig. 3b). These results prove that the mutations R687C and ⌬679 -696 are indeed disease-causing mutations. DISCUSSION In this study, we have investigated structure-function relationships in TGK by analyzing mutants found in LI patients. Using biochemical techniques and direct sequencing, we have identified 7 novel mutations in the gene of keratinocyte transglutaminase. Five of the mutations were one-nucleotide changes resulting in single amino acid alterations (R315C, S358R, V379L, G473S, R687C), one point mutation led to a premature termination codon (R127Stop), and one mutation affected the splice donor site of intron 13 leading to an in-frame deletion of 18 amino acids (⌬679 -696). One mutation changing the splice acceptor site of intron 5 has already been reported in a family (33) and, in fact, the aberrant RNA banding pattern (Fig. 2) gave an important clue to identify the mutation. In the case of the nonsense mutation, R127Stop, the steady-state transcript level was very low (Fig. 2). The association of stop mutations with reduced mRNA levels has been reported in other genes and is due to low efficiency in transcript processing and/or mRNA transport from the nucleoplasma (57). In contrast, the mRNA levels of all missense mutations are expressed in comparable amounts as in normal probands, in accordance with observations for other genes.
The three-dimensional structure of TGK is currently not known. Thus, the structural effects of the reported missense and deletion mutations were analyzed using the factor XIII A-subunit structure (Fig. 5) (5). The central core domain, containing the active site cysteine, displays the highest homology between factor XIII A-subunit and TGK, whereas the other domains are less conserved. Sequence alignment of the two proteins shows that most of the mutations (R315C, S358R, V379L, G473S) are located in a region corresponding to the central core domain. Amino acid changes can decrease enzyme activity either by interfering directly with the catalytic mechanism, by introducing gross structural alteration, or by blocking the binding of essential cofactors. With the exception of V379L, these mutations do not concern residues close to the catalytic site. Rather, these mutations are predicted to interfere with proper formation of hydrogen bonds and salt bridges and introduce spatial constraints due to differing side chain sizes that alter protein structure. These mutants are unstable and/or more susceptible to proteolytic degradation (58,59). Our predictions are supported by the results of immunoblotting experiments showing strongly decreased levels of TGK proteins in cultured cells from these patients. Thus, these missense mutations lead to protein instability and premature degradation but do not interfere directly with the catalytic mechanism. Congenital factor XIII deficiency, a rare bleeding disorder, can be caused by mutations in the gene for the factor XIII A-subunit. Arg-252, which corresponds to Arg-315 in TGK, was altered to Ile in a patient affected by this disorder (55). In agreement with our results, low TG activity and protein level were Relative TG activities have been normalized for transfection efficiency by cotransfection with a ␤-galactosidase expression plasmid (see "Experimental Procedures") and are presented as percent of the activity of the wild-type molecule set as 100%. Results are presented as mean Ϯ S.E. from two independent experiments measured in duplicate.
reported in this patient. The qualitative agreement between these results underlines that the factor XIII A-subunit subunit can serve as valuable model for predicting the structural and functional effects of TGK mutants.
Val-379 is located two amino acids C-terminal of the active site Cys-377. Although Val-379 does not belong to the catalytic triad (Cys, His, Asp), it is located in the same ␣-helix as Cys-377. Replacing valine with leucine changes the conformation of this ␣-helix and therefore the spatial position of Cys-377 relative Asp and His due to the additional space occupied by the larger leucine side chain in a tightly packed region (Fig. 8). This structural change, ordinarily conservative, in family LI-20 is drastic enough to cause premature proteolytic degradation (Fig. 3a). Site-directed mutagenesis of Val-316 in factor XIII A-subunit (homolog to Val-379 in TGK) to Ala reduced activity to 34% of the unmutated enzyme (60). The difference in reduction between the enzymatic activities of the two mutants is most likely due to the different sizes of the side chain (Ala versus Leu) replacing valine. Furthermore, we found in another LI patient a homozygotic V379M substitution. 2 These results indicate that the TG activity is very sensitive to changes in the valine two amino acids C-terminal of the active site cysteine.
The mutations R687C and ⌬679 -696 are located at the C terminus, which corresponds to the ␤-barrel domains of factor XIII A-subunit. These domains are not highly conserved; Arg-687 is found only in the keratinocyte sequences, and 11 of 18 amino acids from the deletion mutant are different between TGK and factor XIII A-subunit. Since this precluded analysis of these mutations using the factor XIII A-subunit model, the corresponding mutant cDNAs were expressed in TGK negative keratinocytes. This showed that both mutations strongly decreased the enzymatic activities and led to premature degradation of the enzyme in a manner comparable with mutations in the highly conserved central core or ␤-sandwich domains ( Fig. 3b and Table II). Three mutations, two misssense and a premature stop codon, in the ␤-barrel 2 of factor XIII A subunit were also reported to diminish enzyme activity and protein levels (61)(62)(63). Previous experiments in which deletion constructs of TGK were expressed in bacteria showed that removal of amino acids 675-816 resulted in a substantially reduced specific activity (22). Interestingly, further deletion of 100 amino acids restored the activity nearly to the level of the full-length protein (22). In a series of experiments, it was demonstrated that TGK exists in keratinocytes as complexes of polypeptides derived from the full-length enzyme by proteolysis (18,19). Depending on the differentiation status and cellular localization, enzymatically active 67 kDa, 67/33 kDa, and 10/ 67/33 kDa complexes were found in which the 67 kDa, the 33 kDa, and 10 kDa molecules correspond to the ␤-sandwich plus central core domains, C-terminal ␤-barrel domains, and the first 92 amino acids of the N terminus, respectively. Furthermore, elimination of the two ␤-barrels in bacterial-expressed factor XIII A-chain molecules only slightly diminished enzymatic activity, and the shortened molecules conserved the ability to be activated by thrombin and calcium and the binding and cross-linking of fibrin (64). Interestingly, C-terminal deletions of human tissue transglutaminase were reported to enhance its intrinsic GTP/ATPase activity concomittant with a lowering in TG activity (65). In summary, these data indicate that the ␤-barrel domains are not absolutely required for transglutaminase activity, but they augment activity possibly due to better substrate interaction and enzyme activation. However, our results and those from others suggest that mutations within the ␤-barrel domains have in most cases a profound influence on the whole molecule because they promote premature degradation and/or interfere with the proper functioning of the active site. Additional structural investigations are needed to elucidate further the function of the C-terminal domains of transglutaminases.