Physiological Reviews

Transglutaminases and Disease: Lessons From Genetically Engineered Mouse Models and Inherited Disorders

Siiri E. Iismaa, Bryony M. Mearns, Laszlo Lorand, Robert M. Graham


The human transglutaminase (TG) family consists of a structural protein, protein 4.2, that lacks catalytic activity, and eight zymogens/enzymes, designated factor XIII-A (FXIII-A) and TG1-7, that catalyze three types of posttranslational modification reactions: transamidation, esterification, and hydrolysis. These reactions are essential for biological processes such as blood coagulation, skin barrier formation, and extracellular matrix assembly but can also contribute to the pathophysiology of various inflammatory, autoimmune, and degenerative conditions. Some members of the TG family, for example, TG2, can participate in biological processes through actions unrelated to transamidase catalytic activity. We present here a comprehensive review of recent insights into the physiology and pathophysiology of TG family members that have come from studies of genetically engineered mouse models and/or inherited disorders. The review focuses on FXIII-A, TG1, TG2, TG5, and protein 4.2, as mice deficient in TG3, TG4, TG6, or TG7 have not yet been reported, nor have mutations in these proteins been linked to human disease.


The human transglutaminases (TGs) are a protein family consisting of a structural protein, protein 4.2, that lacks catalytic activity, and eight zymogens/enzymes designated factor XIII-A (FXIII-A), transglutaminase (TG) 1, TG2, TG3, TG4, TG5, TG6, and TG7, that catalyze a variety of Ca2+- and thiol-dependent posttranslational protein-modifying reactions. All TG family members are comprised of four sequential and structurally distinct domains (an NH2-terminal β-sandwich, an α/β catalytic core, and two COOH-terminal β-barrel domains) that assume a compact conformation in the absence of calcium (Fig. 1A). Two family members, FXIII-A and TG1, have an additional NH2-terminal pro-peptide sequence that may be cleaved to generate the active enzyme (Fig. 1B). A remarkably large conformational change, from the compact form to an extended ellipsoid structure that exposes the TG active site, accompanies activation of TG2 (37, 224) and likely other TG family members (Fig. 1A). Central to TG enzymatic activity is the active site catalytic triad consisting of a cysteine, histidine, and aspartate residue, and a conserved tryptophan that stabilizes the transition state: all four residues are essential for catalysis (Fig. 1B) (92, 116, 158, 187, 197, 218). The catalytically inactive member, protein 4.2, is evolutionarily related to the other TGs, as is evident from its high degree of sequence conservation (37–51% overall sequence similarity). However, none of the four residues critical for catalysis is conserved in protein 4.2.

FIG. 1.

Transglutaminase tertiary structure, protein domains, and genomic organization. A: three-dimensional ribbon structure of the compact, inactive GDP-bound form (37, 162) and the expanded irreversible peptide inhibitor-bound form of TG2 (224). The four structural domains of the protein (β-sandwich, core, and barrels 1 and 2) are color-coded as in Fig. 1B. NH2 and COOH termini are labeled. GDP and the reactive gluten-peptide mimic inhibitor Ac-P(DON)LPF-NH2 are shown as black lines. Although only TG2 (38, 118, 121, 122), TG4 (262), and TG5 (54) have been shown to bind GTP, the crystal structures of all human TGs solved in the absence of calcium [FXIII-A (305), TG2 (37, 162), TG3 (11)] assume the same compact conformation, in which the TG active site is buried. B: dotted lines indicate exons encoding structural domains of the protein, with the relative location of the transition-state-stabilizing Trp, and active site Cys, His, and Asp shown. Genes encoding FXIII-A (F13A1) and TG1 (TGM1) consist of 15 exons (numbered) and 14 introns; genes encoding TG2-7 (TGM2-7) and protein 4.2 (EPB4.2) consist of 13 exons (numbered) and 12 introns. In F13A1 and TGM1, exon 1 is noncoding, exon 2 encodes an NH2-terminal pro-peptide that is cleaved upon activation of TG activity, and exons 10 and 11 are separated by an intron that is not present in TGM2-7 and EPB4.2.

Although protein cross-linking is the best-known catalytic reaction of these enzymes, they mediate at least five posttranslational modification reactions that can be divided into three types: transamidation, esterification, and hydrolysis. All five reactions involve a glutamine-containing protein or peptide as the first, acyl or acceptor substrate, which reacts with the active site cysteine to form a covalently linked γ-glutamylthioester, the acylenzyme intermediate, with ammonia being released as a by-product (Fig. 2). This initial acylation step is followed either by a direct attack by water (hydrolysis) or by the binding of a second substrate, such as an amine (transamidation reaction) or an alcohol (esterification), to cleave the thioester bond (the deacylation step) with generation of the hydrolyzed, cross-linked, or esterified product, respectively (for review, see Ref. 168). Transamidation reactions result in either protein cross-linking, in which a glutamine residue is cross-linked via an Nε-(γ-glutamyl)lysine isopeptide bond to a lysyl residue; amine incorporation, in which a primary amine is incorporated onto a glutamine side chain of the acceptor protein; or acylation, in which a lysine side chain in a donor protein is acylated. Esterification, on the other hand, results in fatty acid modification of an acceptor-protein glutamine side chain, whereas hydrolysis results in either deamidation, in which the donor is H2O and the -NH2 group on the acceptor protein glutamine side chain is replaced with an -OH group to form a glutamate, or isopeptide cleavage, which involves a H2O molecule acting as the donor, leading to scission of the isopeptide bond cross-linking two proteins.

FIG. 2.

Reaction cycle for TG-catalyzed posttranslational modifications. The active site Cys and His form a thiolate-imidazolium ion pair. Binding of the Gln-containing substrate 1 enables attack of the γ-carboxamide group in the substrate by the thiolate. The reaction proceeds through oxyanion intermediate 1, which is stabilized by H-bonding both to the backbone nitrogen of the active site Cys and to the Nε1 nitrogen of the transition-state-stabilizing Trp. In the ensuing acylation step, NH3 is released as the acylenzyme intermediate is formed. A high specificity for the second substrate is manifested in a saturable complex between substrate 2 and the acylenzyme intermediate. Nucleophilic attack by the amino group of substrate 2 leads to formation of oxyanion intermediate 2, which is again stabilized by H-bonding interactions to both the active site Cys and the transition-state-stabilizing Trp. In the final deacylation step, cross-linked product is released and the TG is regenerated. [Adapted from Iismaa et al. (116).]

The different TGs generally exhibit remarkable substrate specificity. Although the same protein substrate can be recognized by distinct TGs, the enzymes have different affinity and/or specificity toward the reactive glutamines/lysines. The covalent isopeptide cross-link is stable and resistant to proteolysis, thereby often increasing the resistance of the resulting supramolecular structure to proteolytic degradation. The reactions catalyzed by TGs are thus essential for biological processes such as blood coagulation, skin barrier formation, and extracellular matrix assembly. TGs also contribute to the pathophysiology of blood and skin disorders and various inflammatory, autoimmune, and degenerative conditions.

In addition to these well-studied enzymatic actions, some members of the TG family participate in a plethora of other biological processes through actions unrelated to their transamidase catalytic activity (for review, see Ref. 168). For example, TG2 (38, 118, 121, 122), TG4 (262), and TG5 (54) bind and hydrolyze GTP: GTP binding causes a transition to the compact, inactive conformation and inhibition of TG activity (37, 162). To date, however, only TG2 has been shown to utilize this function in its role as a high-molecular-weight G protein (Gh) involved in signaling by certain G protein-coupled receptors (59, 202, 216, 300). FXIII-A (293) and TG2 (16, 167, 283, 311) also function as adaptor proteins to facilitate matrix interactions during cell adhesion. Finally, TG2 has also been reported to act as a serine/threonine kinase (190–192), a protein disulfide isomerase (110), and an extracellular ligand for the orphan GPCR, GPR56, which is involved in tumor suppression (117, 303), although involvement of these latter functions in physiology and/or disease remains unclear.

Although the biological actions and enzymology of the mammalian TG family have been considered in previous reviews (e.g., Refs. 103, 168), much has been learned recently, particularly about TG2, from studies of genetically engineered animal models. In contrast to in vitro experiments and cell culture models that are useful for dissecting protein functions, such studies allow the integrated actions of proteins to be evaluated in the setting of diverse cellular contexts and inputs and, while complicated, are frequently the only avenue for elucidating the full array of activities of a protein of interest. Moreover, if they model human diseases caused by an apparent inherited gene mutation, they can be a powerful tool for confirming disease causation, as well as for allowing the evaluation of disease mechanisms and the therapeutic potential of novel drugs and therapeutic strategies. Furthermore, where no inherited disorder has been identified, due, for example, to redundancy or to mutations causing an embryonic lethal phenotype, animal models involving conditional gene inactivation may be the only approach for determining the physiological role of a protein of interest.

Caution, however, is required in performing animal studies, as poor animal husbandry practices and/or the use of nonlittermate or inappropriate background strain controls can confound the interpretation of experimental results. Thus inherent genetic differences that exist among the commonly used mouse strains can influence physiological responses [e.g., sensitivity to Fas-mediated cell death (212), insulin secretory function in response to a high-fat diet (19), inflammatory cell recruitment (302), and postinfarct ventricular rupture (94)]. Moreover, a sequential series of sibling matings within lines that have become physically separated (e.g., homozygous knockout crossed with homozygous knockout, and wild-type crossed with wild-type) leads to changes between the lines after five generations (∼18 mo of breeding) due to genetic drift. Furthermore, continued sequential sibling matings lead to the generation of substrains with substrain-specific spontaneous mutations then becoming fixed within the lines (268, 278). To distinguish subtle effects of a gene deletion/transgene from effects of other genes within the background of the strain, it is, therefore, essential to compare animals in which differences in the genetic background have been eliminated as a variable in the experiment (253). This is best done in transgenic mice by back-crossing to the inbred strain used to create the mouse, and in gene deletion (knockout) mice by back-crossing to an inbred strain, for 10 or more generations. Homozygous knockout and wild-type littermate controls should only be generated from heterozygous crosses.

Based on the above considerations, we review here recent insights into the physiology and pathophysiology of members of the TG family that have come from studies of genetically engineered animal models and/or inherited disorders. The review focuses on FXIII-A, TG1, TG2, TG5, and protein 4.2, as mice deficient in TG3, TG4, TG6, or TG7 have not yet been reported, nor have mutations in these proteins been linked to human disease. The major conclusions that can be drawn from consideration of the available studies are as follows: the clotting disorders in FXIII-deficient patients are phenocopied in FXIII-A-deficient mice, demonstrating the integral and unique extracellular cross-linking role of plasma FXIIIa in blood coagulation (see sect. ii). The intracellular cross-linking roles unique to TG1 or TG5 in skin barrier formation are evident from TG1- or TG5-deficient patients presenting with lamellar ichthyosis or acral peeling skin syndrome, respectively, and from TG1-deficient mice (see sect. iii). Studies with TG2-deficient mouse models indicate a pleiotropic role for TG2 in pathophysiology, with demonstrated involvement in cataract development, gluten sensitivity diseases, neurodegeneration, and tissue remodeling/repair associated with heart, liver, and kidney disease, cancer, and bone development (see sect. iv). The only TG other than TG2 with manifest pleiotropy is FXIII-A; the roles of FXIII-A and TG2 are often complementary, as observed for bone development, but can also be opposing, as evidenced by cancer progression in knock-out mouse models (see sects. ii and iv). Finally, studies with protein 4.2-deficient mice and inherited mutations associated with hereditary spherocytosis indicate a noncatalytic, intracellular structural role for protein 4.2 in ion transport across the RBC membrane (see sect. v). The key concepts that have emerged from these studies of genetically engineered animal models and clinical inherited disorders, and areas of future investigation, are discussed (see sect. vi).


A. General Overview and Protein Expression

FXIII, as a zymogen, has both extracellular and intracellular functions. In plasma, it is thought to circulate as a heterotetrameric ensemble (plasma coagulation factor XIII, pFXIII-A2B2 or fibrin stabilizing factor), where the A subunit (∼83 kDa, encoded by the F13A1 gene on human chromosome 6p24–25; Fig. 1B) belongs to the TG family of proteins and the regulatory/carrier B subunit (∼80 kDa, encoded by the F13B gene on chromosome 1q31-32.1 and synthesized in the liver and kidney), is related to the “small consensus” or “sushi” repeat protein family (159). The richly sialylated B subunit protects the A subunit in the circulation. This is evident from the fact that a human null mutation of the B subunit was associated with A subunit deficiency (237). Cellular factor XIII, cFXIII, is thought to exist as an A2 homodimer (cFXIII-A2) and is found in the cytoplasm of megakaryocytes/platelets (up to 50% of the total FXIII-A activity in blood is in platelets; Ref. 5). Platelets have been reported to also contain low levels of TG2 (228). FXIII-A can be detected in macrophage populations associated with wound healing and tumor progression (288) and in other cell types such as chondrocytes and osteoblasts (209–211). The role of FXIII-A in bone development is detailed in section ivF.

B. Factor XIII Mouse Models

Mixed strain [SVJ129-C57BL/6-Swiss albino-129O1a-CBACa (156) or SVJ129-C57BL/6J-NIH Black Swiss (260)] FXIII-A-deficient mice were developed by replacement of exon 7 of the F13A1 gene (which encompasses the catalytic active site of FXIII-A) with a neomycin resistance gene (Fig. 1B). Plasma FXIII-A activity was reduced to 50% in heterozygous mice and was absent in homozygous mice (156, 260). Plasma levels of FXIII-B were also decreased (to <20% of wild-type levels) in FXIII-A−/− mice, indicating that both FXIII-A and -B are protected in the A2B2 complex in the circulation (260). Spontaneous hemorrhage, including hemothorax, hemoperitoneum, subcutaneous bleeding, and early abortion contributed to reduced survival of pregnant female FXIII-A−/− mice (156, 260). Although heart structure and function were normal in FXIII-A−/− mice, a gender-specific effect of widespread fibrosis with hemosiderin deposition and reduced survival of male FXIII-A−/− mice was observed (261).

Mixed strain (SVJ129-C57BL/6J) FXIII-B-deficient mice (260) have been developed by replacement of exons 1 and 2 of the F13B gene (which encompass the translation initiation codon and signal peptide) with a neomycin resistance gene. Plasma FXIII-B was absent in homozygous mice and, although steady-state FXIII-A mRNA levels were unchanged, FXIII-A protein and activity were only ∼1–20% of wild-type levels. Unlike FXIII-A−/− mice, survival of FXIII-B−/− mice did not differ from that of wild-type controls (261), and no female FXIII-B−/− mice died during pregnancy (260).

C. Hemostasis and Thrombosis

The pivotal extracellular role of pFXIII in hemostasis is well established. In human blood coagulation, the activated form of pFXIII (designated pFXIIIa or pFXIII-A2*) stabilizes fibrin matrix assembly during clot formation by catalyzing the covalent linkage of fibrin monomers at selected locations [a process akin to spot-welding (164); Fig. 3]. In addition, pFXIIIa enhances incorporation of α2-plasmin inhibitor into the clot network (238). Though it does not influence clotting and primary bleeding times, the pFXIII system is recognized as an integral part of the normal coagulation process, essential for establishing a properly functioning fibrin matrix that provides the clot with the necessary structural stiffness and plasticity for wound closure, and also with sufficient resistance against lytic enzymes (Fig. 3). The FXIII-B subunits help with binding the pFXIII-A2B2 zymogen to fibrinogen and, importantly, also act as a brake in the thrombin-catalyzed cleavage of the zymogen and hence control the conversion of pFXIII to pFXIIIa. Timing of the operation of the pFXIII system is regulated by unique “feed-forward” control mechanisms. Fibrin accelerates thrombin-catalyzed activation of the pFXIII zymogen by about two orders of magnitude; binding to fibrin allows for the rapid dissociation of subunits to take place at the physiological Ca2+ concentration of plasma. Also, the prior end-to-end assembly of fibrin units boosts the reactivity of cross-linking sites at αCOOH-terminal sites, by ∼10-fold (164).

FIG. 3.

Pathways of FXIII-A activation and the clotting of fibrinogen in blood plasma. The FXIII zymogen exists in plasma as the pFXIII-A2B2 heterotetramer and as the cFXIII-A2 homodimer in cells. In the physiological pathway of clotting in plasma, thrombin has the dual role of controlling the rate of fibrin production from fibrinogen and the rate of activation of the plasma FXIII zymogen by cleavage of ArgGly peptide bonds in the NH2-terminal domains of fibrinogen and the A subunits of pFXIII. Removal of the activation peptide from the A subunits converts pFXIII-A2B2 to pFXIII-A2′B2, and active pFXIIIa (pFXIII-A2*) is generated by a conformational change following dissociation from the carrier subunits. These reactions occur in the plasma milieu of 1.5 mM Ca2+. The fibrin substrate exerts major regulatory influences on the formation of the A2* enzyme at every step on the way. In the absence of pFXIIIa, the fibrin clot is soluble in 30% urea or in 1% monochloroacetic acid and can be readily lysed by plasmin. However, the pFXIIIa-catalyzed covalent ligation of the clot by Nε(γ-glutamyl)lysine bridges renders it insoluble in these reagents and, especially with the concomitant incorporation of the α2 plasmin inhibitor (α2PI) into the network, more resistant to lysis. Activation of platelets and monocytes, with concomitant influx of Ca2+, brings about the conversion of cFXIII-A2 to cFXIII-A2°, causing cross-linking of cell proteins. Platelets and other cells have also been shown to express FXIII-A2 or A2° on their surfaces.

In the explosive process of blood clotting, activation of pFXIII is a multistep process: thrombin-catalyzed release of a 34-residue activation peptide from the NH2 termini of the A subunits by limited proteolysis is critical for loosening their attachment from the B subunits, but the catalytic Cys in the active center of the pFXIII-A2 enzyme (shown as pFXIII-A2* in Fig. 3) becomes accessible only as a consequence of the Ca2+ and fibrin-assisted separation from the B subunits. The cross-linking activity of pFXIII can also be activated without prior thrombin cleavage, but the A2B2→A2′+B2 dissociation inherent in this type of activation requires a much higher than physiological concentration of Ca2+. The next step in the process can take place at low concentrations of Ca2+ and is a facile conformational change from the free pFXIII-A2′ to an enzymatic form that has the same accessibility of its active center Cys (determined by titration with the alkylating agent, iodoacetamide) as that of pFXIII-A2*. However, since its tertiary structure differs from pFXIII-A2* by the presence of the NH2-terminal activation peptide, the enzyme generated by the thrombin-independent pathway is denoted as FXIII-A2°. This also represents the active form of the enzyme found in platelets, monocytes, and other cells (cFXIIIa), where the zymogen exists as free A2 homodimers without the carrier B subunits; A2 becomes activated to A2° by a relatively small elevation of intracellular Ca2+ concentration (Fig. 3). Although both A2* and A2° have the same steady-state kinetic constants on small test substrates, they seem to behave differently in reactions with larger protein substrates.

A relatively common single nucleotide polymorphism in F13A, leading to replacement of Val-34 with Leu in the activation peptide, results in faster FXIII-A activation due to accelerated thrombin cleavage of the activation peptide (22). This Val34Leu polymorphism has been associated with improved wound healing (99) and survival after myocardial infarction (98), but also an increased risk of oral cancer (294).

Several hundred autosomal, recessively inherited cases of human FXIII deficiency are known worldwide (for a comprehensive list of FXIII-A mutations, see Online Mendelian Inheritance in Man Most are due to loss-of-function missense mutations in F13A that, on account of transcriptional or folding problems, prevent the production of the functionally competent A subunits of the zymogen. However, as indicated above, in a small number of cases, the absence of the factor in the circulation is due to mutation in F13B that results in a failure to produce the carrier B subunits, without which the A subunits do not survive in the circulation (126, 147, 237). In either case, in the absence of FXIIIa-mediated cross-linking activity being generated during the coagulation process, the hemostatic plug fails on account of its mechanical weakness and increased susceptibility to lysis. Both FXIII-A (156, 260) and FXIII-B knockout mice (260) phenocopy human FXIII deficiency, showing decreased fibrin cross-linking in plasma and, consequently, plasma clots that are similarly less stiff and more easily lysed than controls. Administration of recombinant FXIII-B into FXIII-B−/− mice accelerated fibrin cross-linking in plasma and assisted maintenance of plasma FXIII-A levels (260).

Most FXIII-deficient patients are diagnosed soon after birth as they display prolonged umbilical cord bleeding that continues for a day or more, although in a few cases the genetic defect in some individuals may remain undetected until quite late in life. Bleeding phenotypes vary greatly from very mild to life-threatening (intracranial hemorrhage in ∼30%; Ref. 170). Judicious replacement therapy with FXIII concentrates administered once a month can mitigate and even prevent the bleeding episodes. FXIII-A−/− mice (156, 260) exhibit a relatively mild bleeding predisposition with somewhat prolonged tail-tip bleeding times, whereas other laboratory values of hemostasis are normal.

Platelets activated simultaneously with collagen and thrombin comprise a subpopulation of activated platelets that likely contribute significantly to thrombotic processes because they become “coated” with a number of procoagulant proteins, including factor V, fibrinogen, fibronectin, thrombospondin, and von Willebrand factor (64). Many of these procoagulant proteins are posttranslationally modified by transamidase-mediated conjugation with serotonin (65), a process that enhances their procoagulant activity (275). Although the formation of a polymer or of γ:ε isopeptides in these platelet reactions has not been shown, inhibition of coated-platelet formation, by an anti-FXIII-A antibody and by competitive substrate inhibition with dansylcadaverine, suggested a role for FXIII-A in coated-platelet formation (65). However, platelets from mixed strain (SVJ129-C57BL/6-Swiss albino-129O1a-CBACa) FXIII-A−/− mice showed unaltered coated-platelet formation relative to FXIII-A+/− littermates (128). The reason for these discrepant findings is unclear and may reflect differences between human and mouse platelets in some other component, such as TG2. Thus murine FXIII-A−/− platelet lysates had higher transamidase activity than platelet lysates from wild-type C57BL/6 mice, despite the absence of FXIII-A expression (128), suggesting that TG2 (or some other TG) may compensate for the absence of FXIII-A, akin to bone development (see sect. ivH). Studies with both TG2−/− and TG2−/−/FXIII-A−/− double knockout mice, if viable, would help further delineate the respective roles of these transglutaminases in coated-platelet formation.

Serotonin has also been demonstrated to posttranslationally modify small GTPases by transglutaminase-mediated cross-linking, during platelet activation. As a result, the small GTPases are constitutively activated, leading to irreversible aggregation and platelet α-granule release (301). These findings stem from experiments with mixed strain (SVJ129-C57BL/6) tryptophan hydroxylase-deficient (TPH) mice, in which the concentration of peripheral serotonin, as found in platelets, is greatly reduced. TPH−/− mice exhibit impaired hemostasis with prolonged bleeding time. Although neither the ultrastructure nor the in vitro aggregation properties of serotonin-deficient platelets differ from normal, their adhesion in vivo is greatly reduced due to a blunted secretion of adhesive α-granular proteins, such as the multimeric form of the von Willebrand factor which, under the high shear rates existing in the circulation, is essential for platelet adhesion. Thus normal secretion of α-granules depends on the functioning of a transglutaminase-regulated signaling process requiring cytoplasmic Ca2+ and serotonin. Again, it is unclear if FXIII-A is involved or if other TGs, such as TG2, also play a role. Further studies with TG2−/−, FXIII-A−/−, and TG2−/−/FXIII-A−/− double-knockout mice are clearly needed.

In women, habitual abortion, as in afibrinogenemia and in some cases of hypodysfibrinogenemia, is a frequent manifestation of FXIII deficiency (170). Similarly, mixed strain (SVJ129-C57BL/6-Swiss albino-129O1a-CBACa) FXIII-A−/− mice are fertile, but reproduction is greatly impaired (148), with female knockout animals frequently developing overt genital bleeding around the 10th day of gestation (regardless of the genotype of the male mating partners), resulting in fetal abortion. About 50% of pregnancies result in maternal death. Because the occurrence and severity of bleeding events, and the number of abortions, varies among the pregnant females, it is likely that hemorrhage in these animals is due to chance rupture of blood vessels in the placenta. Thus FXIII-A also functions in the maintenance of tissue integrity to allow a normal pregnancy, although its mechanism of action in this regard is yet to be delineated.

D. Tissue Remodeling in Cancer and Wound Repair

Tissue remodeling is a complex process involving temporarily overlapping phases that include inflammation, tissue formation and remodeling, and the interaction of platelets, macrophages, fibroblasts, and endothelial cells. FXIIIa, like TG2 (see sect. iv, F and G), is likely involved in several steps of the remodeling processes involved in cancer and tissue repair, including cell adhesion and migration, clot formation, and angiogenesis.

FXIII-A-deficient mice have recently been used to investigate the role of FXIIIa in tumor growth and metastasis. In contrast to TG2−/− mice (see sect. ivG), tumor size was comparable between mixed strain (97% C57BL/6J-3% SVJ129) FXIII-A-deficient and wild-type mice (215). Significantly, hematogenous metastasis to lung and liver of either intravenously or intradermally injected tumor cells was diminished in FXIII-A−/− relative to wild-type mice, whereas lymphatic metastasis to lymph nodes of intradermally injected tumor cells was unaffected. Analysis of the early fate of circulating tumor cells indicated equivalent initial localization of tumor cells to the lungs within 30 min of injection, but reduced survival of these cells after 26 h in FXIII-A−/− mice due to increased clearance by natural killer cells (215). Thus, although FXIII is not required for the growth of established tumors or the initial adhesion of circulating tumor cells, it supports metastatic spread by limiting natural killer cell function. The mechanism involved has not yet been established; however, a plausible hypothesis may involve tumor cell evasion of natural killer cells through FXIII-mediated tumor cell-associated matrix cross-linking.

Impaired wound healing is a fairly common problem in individuals with hereditary FXIII deficiency (see Refs. 69, 170 for review) and may impact on recovery following myocardial infarction (200). FXIIIa stimulates endothelial, monocyte, and fibroblast cell proliferation and migration, while inhibiting apoptosis (66, 70). In models of blood vessel tube formation developed using human umbilical vein and microvascular endothelial cells, FXIIIa produced dose-dependent enhancement of endothelial cell arrays and, in a rabbit cornea model, injection of FXIII-A caused neovascularization (70). In murine models with neonatal heterotopic cardiac allograft transplant and Matrigel implant, the number of new vessels and percentage of heart beating areas were significantly greater following FXIII-A injection into the allografts than in controls, and by incorporating FXIII-A into the Matrigel, the number of blood vessels was almost the same as obtained by incorporation of the well-known proangiogenic factor basic fibroblast growth factor (bFGF) (67). The angiogenic activity of FXIII-A also promotes the healing process with biomaterials, such as in experimental rat tibial bone defects filled with a nanoparticulate hydroxyapatite paste (137). This angiogenic effect is mediated by phosphorylation and activation of the endothelial cell-specific receptor for vascular endothelial growth factor, VEGFR-2, leading to upregulation of the transcription factors Egr-1 and c-Jun (66, 68). c-Jun downregulates mRNA expression of the platelet-secreted angiogenesis inhibitor thrombospondin 1 (66, 70) and secretion of the protein (70) via Wilm's tumor-1 (68). Remarkably, key to the activation of VEGFR-2 seems to be FXIIIa-catalyzed extracellular cross-linking of VEGFR-2 to the β3 subunit of integrin αvβ3 (68), an integrin that is highly expressed by activated endothelial and tumor cells and plays a critical role in angiogenesis (296) (Fig. 4). Monocyte/macrophage FXIIIa may also contribute to inward remodeling of small arteries in response to a reduction in blood flow (31).

FIG. 4.

FXIIIa cross-link-activated receptor signaling. FXIIIa-catalyzed extracellular cross-linking of β3 integrin to VEGFR-2 results in VEGFR-2 activation and angiogenesis. FXIIIao-catalyzed intracellular dimerization of AT1 receptors enhances AT1 receptor signaling in hypertensive patients and results in increased adhesiveness of monocytes to endothelial cells.

Mixed strain (SVJ129-C57BL/6-Swiss albino-129O1a-CBACa) FXIII-A-deficient mice subjected to 3-mm skin-punch biopsies exhibited a delay in wound healing relative to wild-type CBA mice, which was restored by intraperitoneal injection of FXIII concentrate every 3 days (124). FXIII-A-deficient mice (background strain not specified) also had increased incidence of fatal left ventricular rupture, compared with wild-type CBA mice, following myocardial infarction (199). This increased rupture was associated with an attenuated inflammatory response, impaired collagen synthesis, and enhanced ECM degradation. Daily intravenous FXIII-A replacement therapy improved survival, increased neutrophil migration, and reduced ECM degradation, but did not restore collagen synthesis and resulted in thinner infarct scars and larger left ventricles than in wild-type or untreated FXIII-A-deficient mice (199). This inability to completely rescue FXIII-A deficiency by extracellular reconstitution may reflect an important intracellular role for platelet or monocyte/macrophage cFXIII.

Calcium ionophore activation of the cFXIII-A2 zymogen in normal human monocytes, in conjunction with stimulation by angiotensin II, caused covalent dimerization of the angiotensin II type 1 (AT1) receptor (3). The enzymatic cross-linking, which was inhibited by the TG inhibitors dansylcadaverine or cystamine, or by the AT1 receptor antagonist losartan, involved residue Gln-315 of the COOH-terminal intracellular tail of the receptor and led to its sustained upregulation, while enhancing also the adhesion of monocytes to endothelial cells (Fig. 4). Cross-linked AT1 receptor dimers are characteristically found in monocytes of patients with essential hypertension, but are absent in monocytes of FXIII-deficient individuals. Treatment with the angiotensin converting enzyme inhibitor captopril, (3) or the AT1 receptor antagonist eprosartan (80), diminished the adhesiveness of monocytes of hypertensive patients. Cross-linked AT1 receptor dimers were also found in monocytes of dyslipidemic pure strain (C57BL/6J) ApoE−/− mice, which are prone to excessive aortic root plaque formation due to enhanced monocyte adhesion (3). Captopril treatment significantly decreased the amount of cross-linked AT1 receptor dimers and suppressed monocyte cFXIII-A2 activity (3). These findings provide the molecular framework for explaining the connection between hypertension and vascular diseases. In a general sense, they also highlight the role played by FXIIIa in amplifying some receptor-mediated signaling processes (Fig. 4). As presented above, we have examples for two different modes of amplification of such “cross-link-activated receptor signaling” (CLARS): the cross-linking of VEGFR-2 is likely catalyzed by FXIIIa from outside of the cell (CLARS1), whereas that of AT1 receptor is by the intracellular enzyme (CLARS2).


A. General Overview and Protein Expression

The multilayered epidermal surface of the skin provides a flexible, physical, and chemical barrier to the external environment. It is continually regenerated by the differentiation of keratinocytes into corneocytes, which occurs during their passage from the innermost proliferative (basal epidermal) layer through the spinous and granular layers. Here, daughter cells differentiate to synthesize and then cross-link cornified cell envelope (CE) structural proteins. Corneocytes deposit beneath their plasma membrane a 15-nm-thick CE consisting of a 5-nm-thick cross-linked lipid layer that coats the exterior and is covalently attached to a 10-nm-thick highly cross-linked protein scaffold. TG2 is expressed primarily in the basal layer of the epidermis (7), TG1 and TG5 are present in the spinous and granular layers, and TG3 seems to be confined to the upper granular layer (52).

In hair follicle development, elongated keratinocytes of the hair germinal follicle, visible as an epidermal thickening, grow into the dermis to form a hair peg. The base of the hair peg undergoes bulblike enlargement to form the dermal papilla, concomitant with the differentiation of the hair shaft, development and elongation of the inner root sheath, and differentiation of the outer root sheath and sebaceous gland. The hair follicle elongates through the deep subcutis at an angle of 40° to the epidermis to reach the subcutaneous muscle layer, and the hair shaft emerges through the epidermis (217). The only TG present in the hair germinal follicle is TG2. During its development, TG1, TG2, and TG3 are present in the inner cells of the hair peg. TG1 remains restricted to the inner root sheath and the inner cells of the dermal outer root sheath; TG3 becomes restricted to the hair shaft, while TG5 expression overlaps with that of involucrin in the inner root sheath, the outer root sheath, and the hair shaft (284). CE formation is observed throughout the inner root sheath, in the inner cells of the dermal outer root sheath, and in the hair canal (17).

The molecular events of CE development are believed to be almost identical in the epidermis and hair follicle (17). TG2 plays a role in stabilization of the dermoepidermal junction (7) but is not involved in cornification. TG1, TG3, and TG5 have distinct, yet complementary, roles in cross-link formation during CE assembly. For example, in the spinous layer of the epidermis, CE formation is likely initiated by TG1- and TG5-mediated cross-linking of envoplakin and periplakin to the desmosome (55). In the granular layer, loricrin, the major CE protein, and small proline-rich proteins are cross-linked by TG3 to form oligomers and in the final reinforcement stages are cross-linked to the CE by TG1 (56). TG1 and TG3 cross-link at different sites (56). Formation of the lipid envelope occurs by TG1-mediated cross-linking of lipids (ω-hydroxyceramides) to already cross-linked proteins such as envoplakin, periplakin, and involucrin (206).

1. TG1

TG1 (encoded by the gene TGM1 on human chromosome 14q11.2) is synthesized as an inactive zymogen of 106 kDa (Fig. 1B). It is expressed at low levels in proliferating keratinocytes and is progressively induced during keratinocyte differentiation (265), concomitant with the expression of keratinocyte differentiation regulator, tarazotene-induced protein 3 (TIG3), which regulates terminal differentiation by activating TG1 (271, 272). TG1 is also expressed in the adherens junctions of epithelial cells in lung, kidney, and liver (113) and in myocardial microvascular endothelial cells (35). The majority of keratinocyte TG1 is bound to the plasma membrane in proliferating or differentiating cells via NH2-terminal myristate or palmitate linkages, respectively (266). During terminal differentiation, a significant fraction of the membrane-bound TG1 zymogen is proteolytically processed into a highly active 10/67/33-kDa complex (265) responsible for most of the TG1 enzyme activity that not only cross-links CE proteins, but also links the terminal (ω) hydroxyl group of long-chain ω-hydroxyceramides onto CE proteins (206).

2. TG3

TG3 (encoded by TGM3 on human chromosome 20q11-12; Fig. 1B) is synthesized as an inactive 77-kDa zymogen that is proteolytically processed by cathepsin L (60) into an active 50/27-kDa complex (139). Trichohyalin, a major structural protein of the hair follicle, is cross-linked predominantly by TG3 (280). To date, no mutations in TGM3 have been linked to disease. A preliminary report at the 9th International Conference on Transglutaminases and Protein Cross-linking (Morocco, Sept 1–4, 2007) detailed a curled structure of the fur and whiskers of TG3−/− mice, with irregularities in ultrastructure and increased protein extractability reflecting altered cross-linking of hair structural proteins (285).

3. TG5

TG5 (encoded by TGM5 on human chromosome 15q15.2; Fig. 1B) has a molecular mass of 81 kDa (6). Like TG1 and TG3, activation of TG5 cross-linking activity may require proteolytic processing (223). TG5 appears to associate with the vimentin intermediate filament network in cultured keratinocytes that are undergoing epithelial-mesenchymal transition (53). TG5 is also found in several fetal and adult tissues; however, the exact cell type that expresses this protein in these tissues is not yet known (54). A loss-of-function mutation in TGM5 has been reported to cause acral peeling skin syndrome (57).

B. TG1 Mouse Models

To gain insights into the regulation of TG1 expression, the TGM1 promoter has been analyzed in vivo using transgenic mouse lines expressing reporter genes under the control of various lengths of the human [2.5 kb (304), 2.2, 1.6, 1.1, and 0.3 kb (221)] or rabbit [2.9 kb (184)] promoters or under the control of a 2.2-kb human TGM1 promoter fragment with mutated AP1 or CRE binding sites (221). Correlation of transgene expression with that of endogenous TG1 demonstrated that the 2.5-kb human (304) and 2.9-kb rabbit TGM1 promoter (184) were sufficient to confer tissue- and terminal differentiation-specific expression. Like TG1 expression, the reporter was upregulated by topical 12-O-tetradecanoylphorbol-13-acetate (TPA) treatment of adult tail skin (304). The minimal length of human TGM1 promoter required for appropriate expression was narrowed to slightly less than 1.6 kb and shown to contain binding sites for the transcription factors AP1 and Sp1 (221). Transgenic reporter lines generated with the functional TGM1 promoter regions may be useful to assess the in vivo effects of topical application of chemicals or drugs on human TG1 promoter activation. Functional TGM1 promoter regions may also be useful for gene therapy or generation of conditional (tissue- and terminal differentiation-specific) knockout mice via Cre/loxP site-specific recombination.

Mixed-strain (SVJ129-C5BL/6) TG1-deficient mice (182) were developed by replacement of exons 1–3 (encoding the unique NH2-terminal membrane-anchoring domain of TG1) with a neomycin resistance gene (Fig. 1B). TG1+/− mice were phenotypically indistinguishable from wild-type animals, whereas TG1−/− neonates were smaller, often encased in an inelastic translucent membrane reminiscent of that of “collodion babies,” did not feed, became progressively dehydrated, and died within 4–5 h after birth (182). Normal development of the skin permeability barrier, which occurs around 16 days post coitum (dpc), was impaired in TG1−/− mice due to defects in CE formation (154). Similar defective barrier functions were noted in stratified squamous epithelia of the tongue and forestomach, whereas epithelial cells and adherens junctions in lung, liver, and kidney were unaffected (154). Compared with control grafts from wild-type mice, skin from TG1−/− neonates grafted onto nude mice showed a substantial delay in epidermal regeneration following a full-thickness 10-mm linear incision, thereby demonstrating the importance of TG1 in skin wound healing (123).

C. Lamellar Ichthyosis

Keratinization disorders or ichthyoses are characterized by thickened and scaly skin due to several distinct defects. Lamellar ichthyosis is caused by defects in TG1 activity (114) that result in an inability to assemble both the lipid envelope by ester bond formation as well as the protein envelope by cross-linking of CE proteins (206). Mutations in TGM1 have been identified due variously to nonsense mutations, amino acid substitutions, or splice site changes, which affect either enzyme activity or posttranslational proteolytic processing (reviewed in Ref. 140; for a comprehensive list of TG1 mutations, see Online Mendelian Inheritance in Man).

In X-linked ichthyosis, accumulation of cholesterol sulfate, due to a deficiency in steroid sulfatase, inhibits the capacity of TG1 to perform protein cross-linking and ester linkage of ω-ceramides, resulting in inhibition of lipid-bound envelope formation. As a result, mechanical stability of the intercorneocyte lipid layers is impaired, leading to a lamellar ichthyosis-like phenotype (205).

Consistent with an exclusive role for TG1 in the final reinforcement stages of CE assembly, TG1−/− mice show defective epidermal maturation late in embryonic development (16.5–17.5 dpc) that results in a drastic increase in skin permeability and death from dehydration within a few hours of birth (154, 182). Transplantation of TG1−/− mouse skin to nude mice resulted in an ichthyosiform phenotype with epidermal hyperplasia and marked hyperkeratosis. However, transepidermal water loss was comparable to that of grafted control skin, as was evident when ichthyosiform thick scales were removed. Thus hyperkeratosis is an apparent compensatory response to limit transepidermal water loss from the surface of the TG1−/− mouse skin (154).


A. General Overview and Protein Expression

TG2 (encoded by TGM2 on human chromosome 20q11-12; Fig. 1B), a 74- to 80-kDa protein, is perhaps the most intriguing member of the TG superfamily. This protein is found throughout the body due to its constitutive expression in endothelial cells, smooth muscle cells, and fibroblasts as well as in a number of organ-specific cell types (9, 87, 89, 286, 287). TG2 is located in the ECM or at the cell surface in association with the ECM (97) and also intracellularly, where it is mostly cytosolic, but is also found associated with the inner face of the plasma (38, 118, 121, 122) or nuclear membrane (254). TG2 is not present in the mitochondria (151). The pathway(s) and molecular mechanisms by which TG2 is externalized are largely unknown.

The transamidase activity of TG2 plays an extracellular role in matrix stabilization, which is important in wound healing, angiogenesis, and bone remodeling, and an intracellular role, predominantly in the cross-linking of proteins during apoptosis. TG2 has five reported functions in addition to its transamidase activity. These include GTPase activity and intracellular G protein signaling via the α1B1D adrenergic receptors (59, 202), the TPα thromboxane A2 receptor (300), and the oxytocin receptor (216); an extracellular stabilizing function by being able to form tight ternary TG2:fibronectin:collagen complexes (231, 292), and an adapter function to facilitate cell adhesion to fibronectin by interacting directly with β135 integrins (16, 311), with heparan sulfate chains of the heparan sulfate proteoglycan receptor, syndecan 4 (283), or with the orphan G protein-coupled cell-adhesion receptor GPR56 (117, 303). TG2 has also been reported to possess intracellular serine/threonine kinase activity, with insulin-like growth factor binding protein (IGFBP) 3, p53 tumor suppressor protein, or histones (190–192) as substrates, as well as extracellular protein disulfide isomerase (PDI) activity (110). The TG and G protein signaling activities of this protein are reciprocally regulated by the binding of Ca2+ and GTP, respectively (38, 120, 202): calcium-activated TG2 assumes an expanded ellipsoid structure (37, 224), while GTP binding promotes transition to the compact, transamidase-inactive conformation (37, 162) (Fig. 1A).

B. TG2 Mouse Models

Like TG1, the Tgm2 promoter has been analyzed to gain insights into control of murine TG2 expression using transgenic mouse lines expressing the Escherichia coli β-galactosidase (lacZ) reporter gene under the control of a 3.8-kb 5′-flanking promoter region of Tgm2 (pmTG3.8LacF) (198). Correlation of transgene expression in embryos at 13.5 dpc with that of endogenous TG2 protein demonstrated that the 3.8-kb Tgm2 promoter region was sufficient to confer specific patterns of expression during limb development. However, in contrast to expression of endogenous TG2, no β-galactosidase activity was detected in the liver or heart of transgenic embryos. Thus, unlike TGM1, where a 1.6-kb promoter fragment is sufficient to direct appropriate expression, the 3.8-kb 5′-flanking promoter region of the Tgm2 gene regulates some, but not all, of the physiological expression of TG2.

Transgenic mice that specifically overexpress rat TG2 cDNA in the heart under the control of the murine α-myosin heavy chain promoter (256, 300, 312) have been generated by two independent groups and have been used to investigate the role of TG2 in intracellular signaling pathways relevant to cardiac hypertrophy (see sect. ivH). A transgenic mouse model overexpressing human TG2 in neurons and heart has recently been generated using the murine prion promoter (MoPrP) from the MoPrP.XhoI expression transgenic vector (291). This mouse model was created to study the role of TG2 in neurodegenerative conditions (see sect. ivJ).

Two mixed strain (SVJ129-C57BL/6) knockout mouse models for TG2 were developed simultaneously by different groups to evaluate the in vivo function of TG2. One model was developed by replacement of part of exon 5 and all of exon 6 (exons that encode part of the catalytic core domain) with a neomycin resistance gene (73), while the other model was developed using the Cre/loxP system with LoxP sites inserted into introns 5 and 8 for deletion of exons 6–8 of the catalytic core (203) (Fig. 1B). Both models showed absence of TG2 protein in homozygote progeny. Although TG2 has been reported to either facilitate or attenuate apoptosis (reviewed in Refs. 88, 235), TG2−/− mice showed no obvious abnormal phenotype, with no abnormalities of developmental apoptosis and no difference in induced apoptosis of TG2−/− thymocytes and fibroblasts (73, 203). This, in itself, was surprising given the ubiquitous expression of TG2. Nevertheless, these mice have been useful in defining the role of TG2 in pathology and under conditions of stress.

C. TG2 and Erythrocytes

The paradigm that an increase in intracellular Ca2+ concentration activates latent transglutaminases in many cell types was first established for TG2 in human red blood cells (RBCs) by the demonstration that TG2-catalyzed remodeling of membrane skeletal proteins via Nε-(γ-glutamyl)lysine linkages was the immediate cause for the apoptosis-like morphological and physical alterations in these cells (175, 252). Competitive or direct inhibitors of cross-linking activity offered effective protection against the permanent abnormal shape changes and loss of membrane deformability (257). SDS-PAGE of ghosts from Ca2+-treated RBCs revealed large polymeric clusters ranging in molecular weight from ∼1 × 106 to ∼6 × 106 Da; some polymers appeared to be situated peripherally to the inner leaflet (i.e., extractable by mild alkali), but a major fraction was firmly anchored into the membrane and could not be stripped away by alkali treatment. Since neither chemical nor biochemical procedures are available at present to separate the Nε-(γ-glutamyl)lysine-bonded structures into their polypeptide building blocks, indirect methods have been employed for assessing polymer composition, with good agreement between different modes of analyses. For example, immunological approaches have utilized cross-reactive antibodies for recognizing accessible epitopes both in monomers and polymers (46), whereas the identification of tryptic fragments from polymers has been accomplished by proteomics (L. Lorand, personal communication). Membrane skeletal as well as cytoplasmic proteins are incorporated into the polymers (e.g., spectrins, ankyrin, band 4.1, hemoglobin, among several others), and the anion transporter (band 3) and possibly other transmembrane proteins serve to anchor the polymers firmly into the membrane.

High-molecular-weight polymers cross-linked by Nε-(γ-glutamyl)lysine bonds have also been isolated from RBCs of a Hb-Koln and a sickle cell patient (172, 173). A causal relationship is suggested between polymer formation and the shortened life span of these cells, whereby elevated Ca2+ levels in the pathological RBCs activate TG2 transamidase activity and the cells become susceptible to survival hazards in the circulation through changes in their physical properties (i.e., cell membrane stiffening and irreversible structural fixation of abnormal cell shape, e.g., echinocyte, sickle cell), in addition to the known translocation of phosphatidylserine to the cell surface, which serves to attract phagocytes.

Elevated intracellular Ca2+ also activates membrane proteases, which attack mainly glycophorin and the band 3 anion transporter (166). TG2-mediated cross-linking can be uncoupled from proteolysis by differential inhibitors (165), or by a few days of blood bank storage, which greatly diminishes Ca2+-induced proteolysis without impairing TG2-catalyzed modification of membrane skeleton (171).

In contrast to human RBCs, the prime response to Ca2+ overload of rat RBCs is proteolysis of membrane proteins (mostly band 4.1, band 3, and band 2.1 proteins) by cytosolic proteases (mainly calpain), rather than TG2-dependent cross-linking. This, incidentally, seems to explain the greater ease of fusion of rodent compared with human RBC membranes and is attributed to the higher ratio of protease to protease inhibitor (calpain to calpastatin) in the rodent RBC (102, 149). Given this and other fundamental species differences (see sect. vC), it is appropriate to question the relevance to human RBCs of observations made with rodent RBCs.

RBCs from mixed strain (SVJ129-C57BL/6) TG2−/− mice had no transamidating activity, as expected, and RBC and reticulocyte counts, hemoglobin concentration, hematocrit, mean corpuscular volume, and mean corpuscular hemoglobin contents showed only minor abnormalities relative to wild-type mice, with a trend toward a reduced hematocrit (possibly due to anemia) in the TG2−/− mice (41). Although no significant differences were noted by SDS-PAGE between TG2−/− and TG2+/+ RBC membrane fractions, high-density (“old”) but not low-density (“young”), TG2−/− RBCs were more resistant to osmotic lysis than wild-type RBCs, both in the absence of, or following, pretreatment with Ca2+/ionophore to induce morphological changes (41).

As with the rat, the prime response of mouse RBCs to Ca2+ influx is activation of calpain and membrane protein degradation, with TG2-mediated cross-linking of the small to the large subunits of μ-calpain enhancing protease activity (239). Following ionomycin addition to induce cell death, cell-surface phosphatidylserine exposure was delayed in mixed strain (SVJ129-C57BL/6) TG2−/− relative to wild-type RBCs, which accords with the notion that TG2, through calpain activation, promotes cell surface migration of phosphatidylserine (239). This notion was strengthened by the observation that biotinylcadaverine, a competitive inhibitor of TG2 transamidase activity, delayed phosphatidylserine exposure in wild-type, but not TG2−/− RBCs. To investigate whether TG2 affects RBC survival, wild-type and TG2−/− RBCs were stained with a fluorescent dye, injected intraperitoneally into wild-type mice (rather than directly reinjecting them into the circulations of the individual donor mice), and their time-dependent disappearance from the circulation was followed. An unusual, early increase in life-span profile obtained for the TG2−/− cells was interpreted to indicate a delay, relative to wild-type cells, in TG2−/− cell uptake from the peritoneal cavity into the bloodstream, but otherwise, RBC survival rate was judged not to be significantly different (239).

D. Gluten Sensitivity Diseases

Gluten sensitivity diseases, which affect 1–2% of the general population, are autoimmune disorders triggered by dietary exposure to gluten present in wheat, barley, and rye. Symptoms resolve with a gluten-free diet and reoccur with resumption of gluten ingestion. Gluten proteins, which are unusually rich in proline (∼15%) and glutamine (∼35%), are largely resistant to gastrointestinal proteases due to a lack of proline- and glutamine-specific endoprotease activity. As a result, immunotoxic peptides accumulate and are transported by retrotranscytosis (183) across the intestinal epithelium, where they initiate a deleterious adaptive and innate immune response in a proportion of genetically susceptible individuals. Common manifestations include enteropathy (celiac disease or celiac sprue) that results from T cell-mediated flattening of the small intestinal mucosa and is characterized by malabsorption and chronic diarrhea; dermatopathy (dermatitis herpetiformis), a blistering skin disease of the extensor surfaces of the major joints; and neuropathy (gluten axonal neuropathy) as well as cerebellar ataxia due to the loss of Purkinje cells. Patients develop IgA antibodies against gluten epitopes and predominantly IgA autoantibodies against TG2 in celiac disease (76), against TG3 in dermatitis herpetiformis (79, 227, 242), or against TG6 in gluten neuropathy (106). Furthermore, tissue analyses show small intestinal IgA-TG2 deposits in celiac disease (142), dermal IgA-TG3 deposits in dermatitis herpetiformis (242), and cerebellar IgA-TG6 deposits in gluten ataxia (106).

Recent immunological and biochemical evidence indicates a central role for TG2 in the adaptive immune response elicited in gluten sensitivity diseases (Fig. 5), although, to date, little attention has been paid to the role of TG3 or TG6 in the immunopathophysiology of gluten-induced dermatopathy and neuropathy. TG2 contributes to the T cell-mediated adaptive immune response by the selective deamidation of gluten peptides (21, 193, 230, 255, 295). These are internalized by antigen-presenting cells and loaded onto HLA DQ2 (or DQ8) molecules on the surface of the antigen-presenting cells. This results in activation and clonal expansion of gluten-specific, DQ2- and DQ8-restricted CD4+ T cells in the lamina propria, which initiates a cell-mediated Th1 response causing crypt cell hyperplasia and villous flattening (see Ref. 258 for review). Activated T cells also mediate the humoral adaptive response by stimulating gluten-specific B cells to produce gluten-specific antibodies. The transamidase activity of TG2 may also contribute to the humoral response by forming hapten-carrier complexes in the form of covalent TG2-gluten peptide-complexes (linked via a thioester bond to the active site Cys of TG2 or via isopeptide bonds to particular Lys residues of the enzyme) (91) and collagen-gluten peptide-complexes (77). These complexes result in activated T cells stimulating B cells that express TG2-specific (259) or collagen-specific (77) antibodies, respectively. TG2 autoantibodies only partially inhibit the transamidating activity of TG2 (78), but may play an important role in disease initiation/progression through molecular mimicry. Thus a subset of TG2 autoantibodies cross-react with self-antigens such as heat shock protein 60, desmoglein 1, or Toll-like receptor 4, as well as viral antigens (309) or deamidated gliadin-derived peptides (144). This increases epithelial cell permeability (309) and proliferation (33, 107), which, in turn, induces monocyte activation (309). Cell-surface TG2 has been reported to play a role in the innate immune response induced in gluten sensitivity diseases by a class of immunotoxic gluten peptides that does not stimulate gluten-specific CD4+ T cells (177). However, this work needs to be reevaluated, as the antibody used in these experiments (monoclonal antibody 6B9) has recently been shown to recognize CD44 and not cell-surface TG2 (263).

FIG. 5.

Adaptive and innate immune responses to gluten in the gut. Ingested gluten is digested by gastrointestinal proteases in the small intestine to innocuous peptides or proteolytically resistant, glutamine (Q)-containing immunotoxic gluten peptides. Immunogenic gluten peptides access the lamina propria by unknown mechanisms. Their Q residues are deamidated to glutamic acid (E) by TG2, internalized by antigen-presenting cells (APC), and presented via HLA DQ2 (or DQ8) molecules on the surface of APC to gluten-specific, DQ2- and DQ8-restricted CD4+ T cells in the lamina propria. The subsequent activation and clonal expansion of the T cells results in a cell-mediated Th1 response that initiates destruction of the epithelial layer and villous flattening as well as a humoral response that stimulates gluten-specific and TG2-specific B cells to produce gluten- and TG2-specific antibodies. This renders the epithelium more permeable, facilitating increased access of immunogenic gluten peptides and propagation of disease. [Adapted from Bethune and Khosla (44).]

A microbial transglutaminase from Streptoverticillium mobaraense that is commonly used in the food industry, for example, in seafood, meat products, noodle pastas, dairy products, and baked goods, has recently been shown to deamidate the same glutamine residues of gluten peptides as TG2, thereby enhancing peptide immunogenicity and exacerbating the activation of gluten-specific T cells from celiac disease patients (74). Food industry use of microbial transglutaminases may thus be a significant contributing factor to the development of celiac disease in susceptible individuals. The potential effect of other forms of food processing on epitope formation is unknown.

C57BL/6 TG2+/+ and TG2−/− mice have been used to verify that TG2 is an autoantigen in celiac disease patient sera (143), and serological testing for anti-TG2 autoantibodies is now a primary tool for diagnosing individuals with celiac disease. Attempts to establish models of gluten sensitivity in rabbits, dogs, and mice have been unsuccessful (see Ref. 44 for review). A gluten-sensitive rhesus macaque model exhibited histological signs and symptoms of gluten-induced enteropathy accompanied by anti-gliadin antibody production; however, no anti-TG2 antibodies were observed and an association with MHC class II haplotypes has not yet been confirmed (43). Coadministration of gluten and an orally active glutamine-specific endoprotease to hydrolyze glutamine-rich gluten peptides in vivo prevented gluten-induced clinical relapse, indicating a marked loss of T-cell reactivity towards the shorter endoprotease-digested gluten peptides (45). Unexpectedly, however, this treatment increased the production of anti-gliadin antibodies and initiated the production of anti-TG2 antibodies. This not only suggests more efficient penetration of the enterocyte barrier by the shorter peptides, but more importantly, supports a noncausative role for anti-TG2 antibodies in disease progression (45). Combination enzyme therapy to digest glutamine-rich gluten peptides into even smaller (potentially nonimmunogenic) fragments, using a glutamine-specific endoprotease from barley as well as a bacterial prolyl endopeptidase to detoxify prolyl-rich gluten peptides (96), may further limit activation of the immunological responses associated with gluten sensitivity diseases.

E. Cataracts

Age-related cataracts are characterized by aggregation of lens proteins with subsequent lens clouding. It has long been accepted that TG2-mediated cross-linking of some of the β-crystallins (169, 195, 297), αB-crystallin (104, 174), and the intermediate filament protein vimentin (62) is likely involved. Very high (unphysiological) levels of oxidative stress were shown to stimulate the in situ transamidase activity of TG2 in human lens epithelial cells (249) via transforming growth factor (TGF)-β2 activation of the Smad3 signaling pathway, resulting in TG2-mediated cross-linking and aggregation of lens proteins (250) (Fig. 6). Lenses from mixed-strain (C57BL/6-SV129) TG2-deficient mice cultured for 10 days with TGF-β2 showed no signs of cortical opacification, whereas lenses from wild-type C57BL/6 mice opacified within 5 days of TGF-β2 treatment and were completely opaque by day 10 (250). Thus the transamidating activity of TG2 likely plays a critical role in the formation of lens protein aggregates induced by oxidative stress and may contribute to the pathogenesis of other diseases involving TGF-β-mediated responses.

FIG. 6.

Stress, injury, and inflammatory response pathways regulating TG2 expression and/or activity. Oxidative stress triggers either an increase in intracellular calcium levels, which stimulates TG2 transamidase activity, or TGF-β-mediated activation of NFκB [as a result of its release from its IκBα regulatory subunit that follows phosphorylation of IκBα by IκB kinase (IKK)], which induces TG2 expression, translocation, and activation and results in an increased inflammatory response. Lipopolysaccharide (LPS) or cytokines also trigger activation of NFκB, while ethanol induces nuclear translocation of TG2. Enhanced transamidation by TG2 leads to polymerization of various substrate proteins, including crystallin, IκBα, PPAR-γ, or Sp1, thereby promoting an inflammatory response or apoptosis. In cardiomyocytes overexpressing TG2, enhanced signaling through the thromboxane A2 (TXA2) receptor (TPα) leads to sustained activation of phospholipase A2 (PLA2) and cyclooxygenase 2 (COX-2), with consequent hypertrophy and apoptosis. AA, arachidonic acid; iNOS, inducible isoform of nitric oxide synthase; iPs, isoprostanes; PGH2, prostaglandin H2; PL, phospholipid; PPARγ, peroxisome proliferator-activated receptor-γ; TGF-β, transforming growth factor-β; TLR4, Toll-like receptor 4.

F. Inflammation and Tissue Remodeling/Repair

Tissue remodeling/repair is a dynamic process involving an initial inflammatory phase followed by tissue formation/stabilization and remodeling. Many different cell types, such as keratinocytes, platelets, macrophages, fibroblasts, and endothelial cells, interact to facilitate remodeling (reviewed in Ref. 282). TG1 is important for cutaneous regeneration (123). Both TG2 and FXIII-A appear to be involved in tissue injury/remodeling, although their precise contributions to the various different stages of injury/remodeling have not yet been fully characterized.

1. Expression of TG2 and inflammatory mediators

Increased TG2 expression and transamidation activity is a common feature of many inflammatory diseases and events. Cytokines and growth factors secreted during the initial phase of cell injury regulate TG2 expression (Fig. 6). For example, TGF-β acts to increase TG2 in keratinocytes (101) and on the surface of dermal fibroblasts (229) via a TGF-β1 response element in the Tgm2 gene promoter (233). TNF-α enhances TG2 synthesis in liver cells via activation of IκBα phosphorylation. This causes the dissociation of IκB from NFκB, allowing the nuclear translocation of NFκB, which then binds to the Tgm2 promoter (153). TG2 enhances NFκB activation and, thus, inflammation in microglial cells, by cross-linking IκBα, which also results in NFκB dissociation and nuclear translocation (157). Increased TG2 expression, resulting from increased binding of NFκB to the Tgm2 promoter, enhances extracellular TG2 cross-linking activity, a response observed in experimentally induced hepatic fibrogenesis in rats (189) and in patients with hepatic disease, who have increased ECM TG2 levels (204, 222). TG2 expression is also enhanced in cartilage tissue by interleukin (IL)-1 (129) and in liver cells by IL-6 (274). Mallory body inclusions consisting primarily of transglutaminase cross-linked keratins 8 and 18 are characteristic of several liver disorders, although it is not yet clear whether these bodies protect from, or promote, injury (310). In response to cutaneous injury, TG2 expression and activity is increased at sites of neovascularization in the provisional fibrin matrix, endothelial cells, skeletal muscle cells, and macrophages infiltrating wounds in the border zone between normal and injured tissue, and, later, in the granulation tissue matrix where it is thought to cross-link ECM substrates (49, 109).

Recent studies using lipopolysaccharide to induce murine endotoxic shock demonstrated increased survival of C57BL/6 TG2-deficient mice relative to their wild-type counterparts (84). This was associated with decreased NFκB activation, decreased neutrophil recruitment into kidneys and the peritoneum, and reduced renal and myocardial tissue damage. This phenotype indicates that TG2 promotes the pathogenesis of endotoxic shock. Further delineation of the mechanisms involved is required.

Consistent with a stimulatory role for TG2 in inflammation is recent work using airway epithelial cells from cystic fibrosis patients defective in the cystic fibrosis transmembrane conductance regulator (CFTR) gene. A marked upregulation of TG2 in these cells leads to cross-linking and subsequent proteasomal degradation of peroxisome proliferator-activated receptor (PPAR)-γ, a nuclear hormone receptor that negatively regulates inflammatory gene expression (Fig. 6). As a result, these patients display an enhanced inflammatory response (178).

2. Hepatic and renal injury

TG2-deficient mice have been used in an attempt to investigate if accumulation of TG2 during hepatic or renal injury plays a protective or pathological role. Consistent with a protective role for TG2 in liver damage, treatment of C57BL/6 TG2-deficient mice with the hepatotoxin carbon tetrachloride resulted in an increased incidence of death postinjury, relative to control C57BL/6 mice, an effect associated with an increased inflammatory response and increased ECM accumulation (204). In contrast, consistent with a proapoptotic role for TG2 is the finding that compared with their wild-type mixed-strain (SVJ129-C57BL/6) littermates, apoptosis of hepatocytes from TG2−/− mice was markedly reduced following treatment of isolated hepatocytes with ethanol or treatment of animals with a low dose (0.1 μg/g body wt) of the anti-Fas antibody Jo2, which is a known, potent inducer of endothelial cell and hepatocyte apoptosis that contributes to alcohol-induced liver damage (281). This finding contrasts with that of an earlier study reporting increased death of mixed-strain (SVJ129-C57BL/6) TG2-deficient mice relative to wild-type FVB mice, when treated with a high dose (1 μg/g body wt) of Jo2 (240). The problem with this earlier study is that a Jo2 dose of 1 μg/g body wt is sublethal in FVB mice, whereas C57BL/6 mice are more sensitive to Jo2, with lethality being observed at doses as low as 0.3 μg/g body wt (240); thus it is not clear if the increased sensitivity of the TG2-deficient mice was due to the influence of their C57BL/6 genetic background or the lack of TG2. Indeed, when using wild-type SVJ129-C57BL/6 littermates from heterozygous TG2+/− crosses, the high dose of Jo2 used by Sarang et al. (240) caused massive hepatic necrosis, both in TG2+/+ and TG2−/− mice (281), indicating that the different genetic backgrounds of the wild-type and TG2-deficient mice used by Sarang et al. (240) likely accounts for the discrepancy between the two studies. The study of Tatsukawa et al. (281) also highlights a role for TG2 in promoting hepatic apoptosis caused by relatively low doses of Jo2, but not by higher doses. The proapoptotic effect of TG2 in the pathophysiology of liver cell death involves inactivation of the transcription factor Sp1 by TG2-mediated cross-linking, with resultant inhibition of the expression of genes required for cell viability (281) (Fig. 6). In a third type of liver injury model, C57BL/6 TG2−/− mice were found to be highly resistant to drug-induced Mallory body formation and associated liver hypertrophy relative to C57BL/6 TG2+/+ mice (270). Hepatocellular damage, however, was similar between wild-type and TG2−/− mice, although TG2−/− mice had marked hepatic cholestasis, characterized by more gallstones, jaundice, and ductal proliferation than wild-type mice (270). Thus the amelioration of Mallory body formation and injury-mediated hypertrophy attributable to TG2 is associated not with an improvement in liver cell damage but, rather, with increased cholestasis.

Studies using C57BL/6 TG2−/− tubular epithelial cells, in conjunction with transfected opossum kidney proximal tubular epithelial cells, have shown that TG2 contributes to renal extracellular matrix accumulation primarily by accelerating the deposition of soluble collagens III and IV into the matrix (90). Involvement of TG2 in interstitial renal fibrosis has also been demonstrated in C57BL/6 TG2-deficient mice subjected to unilateral ureteral obstruction (251). Although apoptosis was similar, fibrosis development was substantially reduced in TG2−/− mice relative to TG2+/+ mice. This was associated with reduced macrophage and myofibroblast infiltration and decreased collagen I synthesis due to decreased TGF-β activation.

3. Phagocytosis

Macrophages that infiltrate a wound after injury contribute to the resolution of inflammation by generating TGF-β. In turn, activation of TGF-β stimulates phagocytosis of apoptotic cells and suppresses further proinflammatory mediator production (82). During monocyte differentiation into macrophages, high levels of integrin-bound surface TG2 are induced (14, 196) concomitant with a decrease in FXIIIA expression (248). This correlates with macrophage phagocytotic capacity (246, 248) as well as macrophage adhesion and migration (14). TG2−/− macrophage adhesion and migration have not been investigated directly. However, TG2−/− mice have been reported to exhibit a defect in macrophage clearance of necrotic or apoptotic cells. Systems that have been studied include clearance of necrotic (204) or apoptotic (85) C57BL/6 TG2+/+ and TG2−/− cells after hepatotoxin exposure, clearance of apoptotic cells (background strains of TG2+/+ and TG2−/− mice not stated) during induced thymic involution (277), and clearance of apoptotic 94% C57BL/6–6% SVJ129 TG2+/+ and TG2−/− neutrophils during goutlike peritoneal inflammation (234). This defect was associated with the development of autoimmunity in aged mice (277). However, ingestion of bacteria, yeast, or opsonized nonapoptotic thymocytes by TG2−/− macrophages was normal (277). This finding is in contrast to monocytes of patients with FXIII deficiency, which have been reported to display impaired phagocytosis of sensitized red blood cells (FcγR-mediated) and of complement-coated (complement receptor-mediated) and uncoated (lectin-like receptor-mediated) yeast particles (243). The clinical significance of these findings, however, is unclear, since patients with FXIII deficiency do not have an increased susceptibility to infectious diseases resulting from impaired phagocytic elimination of microbes (288). The impaired engulfment of apoptotic cells (85, 234) by TG2−/− macrophages appears to be due to a deficiency in activation of latent TGF-β1 (234, 277). Although the transamidase activity of TG2 has been implicated in TGF-β release from latent matrix storage complexes (141, 207), rescue of the defective phagocytosis by TG2−/− macrophages appears not to require the transamidase activity of TG2, since it was observed even with exogenous addition of a catalytically inactive mutant form of TG2 (234). Furthermore, rescue of phagocytosis appeared to be dependent on exogenous TG2 binding to GDP or adenine nucleotide, but not to GTP. Indeed, exogenous GTP-bound TG2 inhibited phagocytosis of apoptotic cells by wild-type macrophages (234). Together, these results indicate that a particular conformation of TG2 is required for TGF-β activation during the resolution phase of inflammation. The mechanism by which exogenous TG2 activates TGF-β is not yet clear and requires further investigation.

The phagocytic capacity of neutrophils stimulated by intraperitoneal administration of yeast extract was not compromised in mixed strain (SVJ129-C57BL/6) TG2−/− mice relative to those from wild-type mixed strain (SVJ129-C57BL/6) animals (32). However, TG2−/− neutrophils were deficient in superoxide generation (32) due to downregulation of GP91PHOX (32), the major subunit of the NADPH oxidase system that generates reactive oxygen species in response to invading microorganisms as part of a phagosomal killing mechanism. TG2−/− mice have not yet been tested for their ability to resolve microbial infections. With the use of adhesion-dependent assays, migration of neutrophils into the peritoneum in response to yeast extract and chemotaxis of neutrophils towards fMLP have also been reported to be defective in TG2−/− mice (32). However, this study did not rule out potential impairment of TG2−/− neutrophil adhesion as a contributing factor. Given that cell surface TG2 is involved in integrin-dependent fibroblast (16) and monocyte (14) adhesion and migration (see below), and that neutrophil migration is also integrin dependent (160), the observed neutrophil migration defect may in part be related to an adhesion defect.

4. Cell adhesion and migration

Interaction of cells with the surrounding ECM is critical for cell adhesion and migration. The role of TG2 in this process is being delineated primarily from cultured cell studies. TG2 overexpression in fibroblasts increased cell attachment (100, 298) and spreading (100). In contrast, fibroblasts deficient in TG2, that were either isolated from mixed-strain (SVJ129-C57BL/6) mice (203) or generated by stable transfection of antisense constructs (134, 267, 298), exhibited decreased adhesion (134, 203, 298) and spreading (134, 267). Available evidence indicates TG2 contributes to cell adhesion in two ways (Fig. 7). First, TG2 stabilizes ECM proteins by enhancing the initial phase of fibronectin matrix formation (15) and by cross-linking ECM proteins (8, 93, 109). Second, cell-surface TG2 promotes fibroblast adhesion to fibronectin (10, 16, 283, 299, 311) by interacting directly with β1/β3/β5 integrins (16, 97, 311), with heparan sulfate chains of the heparan sulfate proteoglycan receptor syndecan-4 (283, 299), and with the orphan G protein-coupled cell-adhesion receptor GPR56 (303). Enhancement of integrin- (16) or syndecan-4-mediated (283, 299) fibroblast adhesion and spreading on fibronectin, by cell surface TG2, is not mediated by its transamidase activity. Rather, it involves modulation of common targets of receptor-dependent outside-in signaling, such as PKC-α activation (283, 299). During tissue injury and/or remodeling, activated matrix metalloproteinases degrade fibronectin and other matrix proteins to generate Arg-Gly-Asp (RGD)-containing peptides. These peptides compete with, and thus disrupt, integrin binding to the RGD-containing domains of matrix components, such as fibronectin, and, consequently, impair adhesion. Independent of its transamidase activity, extracellular TG2 restores fibroblast adhesion in the presence of RGD-containing peptides (299) by direct interaction, not with integrin, but with the heparan sulfate chains of sydnecan-4. This results in intracellular cross-talk between syndecan-4 and β1 integrin via syndecan-4-driven activation of PKC-α and PKC-α-induced β1-integrin activation (283). Thus TG2 may have a physiological role in protecting cells from anoikis (detachment-induced apoptosis) induced by inhibition of integrin-dependent adhesion following injury. Again, independent of its transamidase activity, intracellular TG2 may also regulate fibroblast spreading on fibronectin through PKC-α activation and FAK phosphorylation (267), as well as inhibiting smooth muscle cell motility by direct association with the cytoplasmic tail of α-integrin subunits (136). Studies with TG2−/− cells should confirm and extend these findings.

FIG. 7.

Contribution of TG2 to cell adhesion. TG2-mediated cell adhesion involves binding and cross-linking of fibronectin (shown) and other ECM proteins, as well as promotion of cell adhesion by direct interaction of TG2 with α or β1/β3/β5 integrins, the G protein-coupled cell-adhesion receptor GPR56, or the syndecan-4 heparan sulfate proteoglycan receptor, resulting in activation of protein kinase C-α (PKC-α) signaling.

5. Arterial remodeling

Preliminary studies using TG2-deficient mice have investigated the role of TG2 in arterial remodeling. Compared with wild-type littermates, arteries of mixed strain (SVJ129-C57BL/6) TG2−/− mice showed a delayed, but not suppressed, capacity to undergo inward remodeling (that is, a decrease in lumen diameter) in response to surgical reduction in blood flow (31) or nitric oxide inhibitor (Nω-nitro-l-arginine methyl ester hydrochloride, l-NAME)-induced hypertension (225). Remodeling was accompanied by increased arterial expression of FXIII-A mRNA in TG2−/− mice and of TG2 and FXIII-A mRNA in wild-type mice upon surgical reduction of blood flow (31), but no differences in FXIII-A or TG2 mRNA levels were observed in wild-type or TG2−/− mice with l-NAME-induced hypertension (225). Depletion of accumulated adventitial monocytes/macrophages by phagocytosis of the liposome-encapsulated toxic drug clodronate inhibited both arterial FXIIIA expression and surgically induced inward remodeling in TG2−/− mice (31), indicating potential compensation for TG2 in low flow artery remodeling by monocyte/macrophage-derived FXIIIA. Interestingly, l-NAME treatment resulted in stiffening of TG2+/+ but not TG2−/− erythrocytes, indicating a role for TG2 in erythrocyte deformability (225). The reduction in inward remodeling observed in TG2−/− mice in response to reduced blood flow is phenocopied in pure strain (SV129) vimentin−/− mice (245). Given that vimentin is a major TG cross-linking substrate in carotid atherosclerotic plaque (105), TG2-mediated dimerization of this integral cytoskeletal protein may regulate inward remodeling. Consistent with TG2 involvement in this process, inward remodeling in rats was inhibited by cystamine, a broad-spectrum inhibitor of sulfhydryl enzymes (81). Outward remodeling (increase in vessel size), on the other hand, involves vimentin (245), but does not appear to involve TG2, as this was unaffected in TG2−/− mice (31), despite cystamine inhibition of outward remodeling in rats (81). Clearly, further work is required to clarify the mechanisms responsible for vascular remodeling in these various models.

G. Cancer

Cancer progression shares many similarities with inflammatory responses (179) and tissue injury/remodeling (244). Numerous conflicting reports implicate either TG2 up- or downregulation in this process (185). TG2 expression is often downregulated in primary tumors and during tumor progression, whereas upregulation of TG2 is often associated with secondary metastatic tumors and resistance to chemotherapy (reviewed in Ref. 150). Recent studies demonstrating hypermethylation of a CpG island that overlaps both the transcriptional and translational start site of the TGM2 gene suggest that epigenetic gene silencing may mediate the reduction in TG2 expression in primary tumors (12). In N-myc-overexpressing neuroblastomas, N-myc has been shown to repress TG2 expression by recruiting the transcription factor Sp1 and histone deacetylase (HDAC) 1 to Sp1 binding sites in the TGM2 core promoter (163). Moreover, in contrast to the action of FXIII-A (see sect. iiD), upregulation of TG2 by N-myc siRNA or HDAC inhibitors correlated with marked reduction in tumor growth (163). Similarly, intratumor TG2 injections inhibited CT26 colon carcinoma tumor growth in mice (135).

The host ECM plays an important role in the regulation of tumor growth, in metastatic spread, and in tumor angiogenesis. Thus degradation of cell-surface TG2 by membrane type 1 matrix metalloproteinase and matrix metalloproteinase-2 at the boundary between normal and tumor tissue suppressed cell adhesion and migration on fibronectin, but stimulated migration on collagen (39, 40). Exogenous addition of TG2 increased ECM accumulation and reduced ECM turnover, conferring resistance to ECM degradation by matrix metalloproteinase-1 and blocking capillary growth in in vitro angiogenesis culture studies (135). In direct contrast to studies with FXIII-A−/− mice (see sect. iiD), studies with mixed strain (SVJ129-C57BL/6) (135) or C57BL/6 (75) TG2−/− mice have shown that subcutaneous implantation of B16F1 (135) or B16-F10 (75) melanoma cells resulted in increased tumor size in (75, 135), and reduced survival of, TG2−/− mice, relative to wild-type mice (135). Hematogenous metastasis to lung and other organs of intravenously injected tumor cells was increased, although individual metastasis size was smaller, in TG2−/− compared with wild-type mice (75). Inhibition of TG2, by treatment with TG2 inhibitors (306, 307), antisense (108), ribozyme (108) or RNAi (13, 112, 138, 308) technologies, sensitized cancer cells to chemotherapeutic agents by disrupting fibronectin assembly in the ECM and promoting cell death or autophagy (306, 308). Although it is not yet clear which location or biochemical function of TG2 is involved in these effects, it appears that TG2 and FXIII-A have opposing roles in cancer progression.

TG2 has recently been shown to interact in the ECM with the novel G protein-coupled receptor GPR56, a suppressor of melanoma tumor growth and metastasis that is markedly downregulated in highly metastatic cells (303) (Fig. 7). Tumor cell growth and metastasis may thus involve localized interactions of TG2 with GPR56, integrins, and other matrix proteins such as fibronectin. The functional consequences on tumor growth, metastasis, and angiogenesis of TG2 interaction with sydnecan family members, the expression of which are often downregulated in primary tumors and upregulated in metastases (36), have not yet been investigated.

H. TG2, FXIII-A, Bone, and Osteoarthritis

Bones form during vertebrate development by one of two independent, yet coupled, pathways. Those of the skull, part of the clavicle, and the bony collars of long bones are formed by intramembranous ossification. This involves differentiation of mesenchymal cells into osteoblasts, which deposit an extracellular matrix that is subsequently mineralized. Bones of the axial and appendicular skeleton and most of the facial bones are formed by endochondral ossification, whereby a hyaline cartilage blueprint of the future skeleton is laid down in the embryo (chondrogenesis) and then replaced by bone (osteogenesis). Some hyaline cartilage remains at the ends of the bone in the epiphyseal growth plates to allow lengthening of bones during childhood.

Chondrogenesis begins with the localization of undifferentiated mesenchymal cells in the site of the future bone. Cells in the core of this localized cellular mass differentiate into chondrocytes that express and secrete TG2 and FXIII-A, but not TG1, TG3, or TG5 (9, 133, 211), while a thin layer of cells that surrounds the core differentiates into perichondrial cells. The chondrocytes rapidly proliferate and secrete collagen type II and aggrecan, resulting in the formation of cartilage.

Osteogenesis commences with cessation of chondrocyte proliferation and initiation of chondrocyte hypertrophy. Expression of both TG2 and FXIII-A is upregulated in hypertrophic chondrocytes (9, 161, 210). These cells secrete type X collagen, a marker of chondrocyte hypertrophy, differentiate further, and deposit mineral into their surrounding matrix. They also secrete metalloproteinases to degrade nonmineralized matrix, and, possibly, growth factors to induce invasion of the mineralized cartilaginous matrix by vascular tissue, before undergoing apoptosis. The invading vascular tissue delivers osteoclasts to the region. Simultaneously, perichondrial cells differentiate into osteoblasts that express and secrete both TG2 and FXIII-A (111, 209) and form a bony collar. The FXIII-A protein found in bone cells and bone matrix appears to be proteolytically processed during osteoblast differentiation to a 37-kDa form that retains its calcium binding region, but not its thrombin cleavage sites (201). Osteoclasts degrade the mineralized matrix, which is then replaced with bone matrix by osteoblasts. This osteoblast-derived bone matrix contains both collagen types I and II and noncollagenous proteins such as fibronectin, SIBLING (small integrin binding ligand N-linked glycoprotein) proteins, including osteopontin, bone sialoprotein, dentin matrix protein-1, and dentin phosphoprotein, the calcium-binding cysteine-rich glycoprotein, osteonectin, and fibrillin-1 (reviewed in Ref. 208), all of which are in vitro TG substrates. Inhibition of chondrocyte transamidase activity reduces mineralization in preosteoblasts (18) and in cocultures of chondrocytes and preosteoblasts (209), while addition of either exogenous TG2 (132, 209) or FXIII-A increases mineralization/hypertrophy (131). FXIII-A may have a more prominent role than TG2 in early matrix mineralization by intramolecular collagen cross-linking, as well as cross-linking collagen and noncollagenous matrix proteins (201). Consistent with this notion, transamidase activity of lysates prepared from mixed strain (SVJ129-C57BL/6) chondrocytes isolated from knees of TG2−/− mice was reduced by only ∼50% compared with wild-type littermates (133), while transamidase activity in bone extracts showed no difference between mixed strain (SVJ129-C57BL/6) wild-type and TG2−/− mice (201), with no substantial overexpression of FXIII-A, or other TGs in either case. More significantly, independently of transamidase activity, FXIII-A promotes chondrocyte hypertrophy by interacting extracellularly with α1 integrin to stimulate TG2 externalization (131) (Fig. 8). Chondrocyte hypertrophy induced by exogenous TG2 is independent of transamidase activity and fibronectin binding, but is α5/β1 integrin dependent and is enhanced by GTP binding (132, 279). FXIII-A also appears to be required intracellularly as exogenously added FXIII-A did not induce chondrocyte hypertrophy in mixed-strain (129Ola-CBACa) FXIII-A−/− chondrocytes (131). Induction of hypertrophy by FXIII-A and TG2 is associated with FAK and p38MAPK phosphorylation (131, 279). These studies thus indicate important temporal signaling roles for FXIII-A and TG2. This involves extracellular fibronectin-independent interactions between FXIII-A and α1β1 integrin, followed by interactions between GTP-bound TG2 and α5/β1 integrin. These interactions mediate the induction of chondrocyte hypertrophy and consequent mineralization. Despite this evidence for the involvement of FXIII-A and TG2 in bone development, no gross skeletal abnormalities are found in either TG2 (73, 203) or FXIII-A (156) knockout mouse models. Thus, although the functions of TG2 and FXIII-A may be coordinated in developing bone, with both enzymes potentially promoting cell adhesion, matrix assembly and mineralization and matrix-mediated cell differentiation, neither of the individual TGs alone is critical for bone development and one can perhaps compensate for loss of the other.

FIG. 8.

Contribution of TG2 to osteogenesis and osteoarthritis. Chondrocyte hypertrophy is induced temporally by FXIII-A-dependent, followed by TG2-dependent, integrin interactions resulting in MAPK activation. This progresses to osteoarthritis during low-grade, chronic inflammation through interleukin 1β (IL-1β)-stimulated secretion of S100A11 and expression of the receptor for advanced glycation end products (RAGE), TG2-mediated formation of the dimeric S100A11 RAGE ligand, and subsequent activation and propagation of MAPK and NFκB signaling.

Osteoarthritis, a condition that affects more than 20 million people in the United States alone, is associated with regions of articular chondrocyte hypertrophy, formation of osteophytes (bony outgrowths), and degeneration of cartilage. Articular cartilage (hyaline cartilage without a perichondrium), which cushions the ends of bones in joints by elaborating synovial fluid in response to loading, normally contains resting chondrocytes that do not differentiate or mineralize their surrounding matrix. A number of proinflammatory factors known to stimulate chondrocyte hypertrophy are upregulated in osteoarthritic cartilage, such as IL-1β (24) and the CXC chemokine subfamily members IL-8 (CXCL8) and growth-related oncogene α (GROα; CXCL1) (48). IL-8 signals through CXCR1 and CXCR2, whereas GROα activates CXCR2, but not CXCR1 (213). Both IL-8- and GROα-induced hypertrophy of wild-type murine chondrocytes is associated with increased transamidase activity (186). Unlike lysates prepared from murine knee chondrocytes isolated from mixed-strain (SVJ129-C57BL/6) wild-type mice, those from TG2−/− littermates showed no increase in transamidase catalytic activity or matrix calcification in response to stimulation with IL-1β, GROα, or the all-trans form of retionoic acid (ATRA). They also failed to exhibit features characteristic of chondrocyte hypertrophy, such as collagen X secretion (133, 186) and expression of the transcription core binding factor α1 Cbfa1 (133). In contrast, stimulation of both wild-type mixed-strain (SVJ129-C57BL/6) and TG2−/− littermate knee chondrocytes with C-type natriuretic peptide (CNP), an essential factor in endochondral development, resulted in comparable Cbfa1 expression and chondrocyte hypertrophy without associated mineralization or modulation of transamidase activity (133). These studies thus support a direct contribution of TG2 to IL-1β-, GROα-, and ATRA-induced transamidase activity and mineralization and demonstrate that there are distinct TG2-dependent (IL-1β-, GROα-, or ATRA-induced) and TG2-independent (CNP-induced) pathways of chondrocyte hypertrophy.

The transamidase activity of TG2 has recently been implicated in promoting osteoarthritis progression. Studies using murine femoral head articular cartilage explants and knee chondrocytes from C57BL/6 TG2−/− mice have shown that the transamidase activity of TG2 is required to convert S100A11 (a calcium-binding, EF-hand S100/calgranulin protein secreted in response to IL-1β during low-grade chronic inflammation) into a covalent homodimer (Fig. 8). This accelerates chondrocyte hypertrophy and matrix catabolism via receptor for advanced glycation end products (RAGE) signaling (58). In a surgically induced knee joint instability model of osteoarthritis, mixed strain (SVJ129-C57BL/6) TG2−/− mice had reduced cartilage destruction and increased osteophyte formation relative to wild-type littermates (214). This phenotype was associated with increased expression of TGF-β1, which promotes osteophyte formation (30), in the osteoarthritic tissues of the TG2−/− mice compared with TG2+/+ mice (214). Together, these studies point to a role for TG2 in modulating cartilage repair and stability in response to certain pathological stressors and identify TG2 as a potential target for therapeutic intervention in osteoarthritis.

I. Cardiovascular Disease

Various parameters of cardiac function (heart rate, systolic and diastolic blood pressure, maximum rates of contractility and relaxation) were measured in mixed-strain (SVJ129-C57BL/6) TG2−/− mice at rest and shown not to be different from those of wild-type littermates (203). In contrast, a recent study reported that TG2−/− mice (the background strain was not indicated) had reduced heart rate, coronary flow, aortic flow, and aortic pressure as well as reduced myocardial ATP levels, compared with C57BL/6 wild-type controls (241, 276). Moreover, in response to an ischemia/reperfusion challenge, infarct size was reported to be increased in TG2−/− hearts, as was the incidence of reperfusion-induced ventricular fibrillation, and ATP levels were even further reduced (276). Nevertheless, given that the preischemic coronary flow was already substantially reduced in these TG2−/− hearts, the validity of determining energy substrates, in this context, is questionable. Surprisingly, however, two earlier studies from the same group reported no significant difference in ATP levels in heart, liver, skeletal muscle (180), or brain (34) in the TG2−/− mice at rest. Only one of these studies (34) specified the background strain of the TG2−/− mice used, which was C57BL/6 (i.e., mixed SVJ129-C57BL/6 backcrossed for 10 generations to C57BL/6). Upon exercise (180) or in response to a toxin challenge (34), ATP content in brain and skeletal and cardiac muscle appeared to be depleted due to reduced activity of mitochondrial complex I, and increased activity of mitochondrial complex II. Given the absence of precise animal husbandry details, the discrepancies in cardiac function and ATP levels of TG2−/− mice at rest (Ref. 276 versus Refs. 34, 180, 203) may merely reflect different husbandry practices and/or the use of nonlittermate controls.

Cardiac hypertrophy is an adaptive response to increased work load induced by pressure- or volume-overload and, in response to chronic overload, cardiac function deteriorates, resulting in decompensated heart failure. Steady-state TG2 mRNA levels are increased in both volume- and pressure-overload-induced hypertrophied rat ventricles and are further increased during the transition to heart failure (125). Increased TG2 expression has also been documented in failing hearts from patients with dilated but not ischemic cardiomyopathy, although both the GTP-binding and transamidase activities of TG2 were decreased in both (115). The hypertrophic effect of transgenic cardiac-specific TG2 overexpression has been investigated in the mouse by two independent groups (256, 312). Animals between 3 and 4 mo of age developed a unique type of cardiac hypertrophy with a mild elevation in some markers of hypertrophy (α-myosin heavy chain and α-skeletal actin) but not others (atrial naturietic factor) (256). Older animals (7–10 mo old) developed cardiac decompensation characterized by cardiomyocyte apoptosis and fibrosis, with an increase in mortality evident by 15 mo (312). The primary and initiating event underlying the cardiac failure observed in these mice was increased biosynthesis of the prostanoid thromboxane (TxA2), and enhanced signaling from its G protein-coupled receptor, TPα (312) (Fig. 6). This response leads to ERK1/2 activation, which results in a feed-forward loop that accounts for the hypertrophy, as well as for sustained activation of phospholipase A2 and induction of the prostanoid biosynthetic enzyme COX-2 and, as a consequence of the latter, enhanced TxA2 production (119, 312) Coincident increased oxidant stress leads to oxidative modification of arachidonic acid and formation of free radical-catalyzed products called isoprostanes (iPs). Formed initially in situ in the phospholipid domain of cell membranes, they are then cleaved by phospholipases, circulate in esterified and unesterified forms, and importantly, act as incidental ligands for prostanoid receptors (152). Oxidant stress thus contributes to the hypertrophy phenotype via TP stimulation (119, 312). Apoptosis may be a direct result of enhanced TPα signaling (95) or may be secondary to TP-mediated vasoconstriction causing myocardial ischemia. Although intracellular TG2 levels were elevated in the cardiac TG2-overexpressors relative to control mice, intracellular transamidase activity was not measured in vivo or in intact cardiomyocytes. It is thus not clear if the transamidase activity of TG2 contributes to the regulation of COX-2 expression. Nevertheless, these studies provide compelling evidence for a potential link between TG2 and prostanoids in cardiac dysfunction due to hypertrophy disorders in humans.

In atherosclerosis, fatty material is deposited in arterial walls, and subsequent migration of inflammatory cells (macrophages and T lymphocytes) and vascular smooth muscle cells into the lesion forms a plaque that results in a narrowing of the lumen that may eventually impair blood flow, either directly or as a result of plaque rupture, which leads to the rapid formation of an occlusive thrombus. TG2 expression in intimal and medial smooth muscle cells, and in the extracellular matrix, increases concomitantly with the progression of atherosclerosis (25, 273), with leukocytes being the major source of TG2 in atherosclerotic lesions (47). Apoptosis is observed in developing plaques, with endothelial cell and macrophage apoptosis increasing early in atherogenesis, and vascular smooth muscle cell and macrophage apoptosis contributing to plaque rupture (269). Recent work indicates macrophage-expressed TG2 limits expansion of atherosclerotic plaques (47). Thus, following a high-fat diet to induce atherosclerosis in irradiated C57BL/6 low-density lipoprotein receptor knockout (LDLR−/−) mice that had been transplanted with mixed-strain (94% C57BL/6–6% SVJ129) wild-type or TG2−/− bone marrow, significantly larger aortic valve atherosclerotic lesions were observed in those animals that received the TG2−/− marrow (47). Lesions in recipients of TG2−/− bone marrow appeared to have more expanded necrotic cores, indicative of plaques that are vulnerable to rupture, with macrophages localized deeper in the intima than in those that received the wild-type bone marrow. The function of TG2 as an integrin-associated adhesion receptor involved in monocyte migration (14) may account for the lack of macrophages at or near the endothelium of knockout bone marrow recipients, while decreased extracellular matrix cross-links (47) and a decrease in the efficiency of apoptotic cell clearance (47, 277) by TG2−/− macrophages may contribute to atherosclerotic expansion in the same animals (47). In addition, decreased macrophage activation of TGF-β (277), which is thought to stabilize plaques by stimulation of collagen synthesis (47), may promote plaque instability in knockout recipients. Interestingly, irradiated LDLR knockout mice transplanted with TG2−/− bone marrow had significantly lower triglyceride levels after 4 and 8 wk on a high-fat diet (47). However, the cause and potential implications of this finding remain unclear.

Atherosclerotic lesions calcify nonlinearly over time as a result of osteogenic and chondrogenic differentiation of vascular smooth muscle cells in a process akin to bone formation (see Ref. 4 for review). Although minor differences were observed between freshly isolated 99.6% C57BL/6–0.4% SVJ129 TG2+/+ and TG2−/− mouse aortic smooth muscle cells and aortas, stimulation of smooth muscle cells (by an inorganic phosphate donor or by bone morphogenetic protein-2) to drive intra-arterial chondro-osseous differentiation and calcification, failed to induce either differentiation or calcification of TG2−/− smooth muscle cells (130). This indicates a central role for TG2 in these processes. Exogenous addition of TG2 to mouse vascular smooth muscle cells has been shown to enhance calcification via binding to low-density lipoprotein receptor-related protein 5 (LRP5) and subsequent activation of β-catenin signaling (86). TG2 released during artery wall repair may thus promote calcification.

J. Neurodegenerative Disorders

Increased levels and activity of TG2 have been observed in many neurodegenerative diseases. Studies with TG2−/− mice have demonstrated that TG2 accounts for at least two-thirds of the transamidase activity in mouse forebrain lysates, is localized predominantly in neurons (27, 151), and contributes to the pathogenesis of Huntington's disease. Huntington's disease is an autosomal dominant neurodegenerative disorder characterized by extensive neuronal cell apoptosis in the striatum and cerebral cortex, leading to progressive motor dysfunction and psychiatric disturbances with gradual dementia. It is caused by an expansion of ∼40 or more CAG repeats in the gene encoding the cytosolic protein, huntingtin (htt) (176). Increased length of CAG repeats correlates with earlier age of disease onset (176). Since CAG encodes glutamine (Q), CAG repeats result in mutant htt proteins with long NH2-terminal polyQ tracts (176). Given that Q residues are potential substrates for TG-mediated cross-linking, it was initially postulated that TGs contribute to disease pathogenesis by catalyzing the covalent formation of large Q-Q linked insoluble htt aggregates. However, it is now clear that such aggregates are due to noncovalent hydrogen bonding interactions or polar zippers between adjacent Q residues (176). Similarly, although insoluble htt aggregates were originally proposed to cause neurotoxicity, recent evidence suggests that it is the soluble forms of mutant htt that are neurotoxic (reviewed in Refs. 188, 235). TG2 ablation in R6/1 (181) and R6/2 (29) mouse models of Huntington's disease is associated with increased life span and improved motor function. Nonetheless, in keeping with a role of the soluble rather than aggregated forms of htt being neurotoxic, TG2 ablation in the R6/1 and R6/2 animals resulted in increased aggregate formation (29, 181). Thus TG2 apparently inhibits aggregate formation, an effect that in vitro appears to result from TG2-mediated cross-links increasing the solubility of polyQ-expanded htt aggregates (155). It is possible, therefore, that TG2 contributes to the pathogenesis of Huntington's disease, not by promoting htt aggregate formation, but somewhat counterintuitively by increasing the formation of soluble high-molecular-weight htt complexes.

The well-established therapeutic benefit of treating transgenic mouse models of Huntington's disease with cystamine, a broad-spectrum inhibitor of sulfhydryl enzymes (reviewed in Ref. 235), has been shown to be independent of its ability to inhibit TG2 activity, since the delayed onset of motor abnormalities and extended life span following cystamine administration was similar in R6/2 TG2+/+ and R6/2 TG2−/− mice (28).

Alzheimer's disease, characterized by progressive neurodegeneration and a decline of cognitive function, is associated with the accumulation of fibrillar β amyloid (Aβ) in extracellular plaques. Disease severity correlates with impaired Gαq/11-coupling of the M1 muscarinic receptor (290). Recent studies, using TG2-encoding lentivirus injection in a transgenic mouse model of Alzheimer's disease (APPSw), indicate this is due to the formation of angiotensin II type 2 (AT2) receptor oligomers in the brain that sequester Gαq/11 (1). AT2 oligomers are triggered dose-dependently by aggregated Aβ in a two-step process involving oxidative cross-linking to form dissociable AT2 receptor dimers that do not sequester Gαq/11 (1), followed by TG2-dependent cross-linking of residues in the receptor COOH terminus to form nondissociable AT2 receptor oligomers (2).

Excitotoxic neuronal cell death, caused by calcium overload following overactivation of ionotropic glutamate receptors, is an important component of acute neuronal injury and chronic neurodegenerative disorders (23). Transgenic mice that overexpress human TG2 primarily in neurons, which resulted in ∼70-fold increase in in vitro cross-linking activity, displayed significantly more hippocampal neuron apoptosis than control mice following kainic acid-induced overstimulation of glutamate receptors (291). Thus TG2 may contribute to neurodegenerative diseases by facilitating excitotoxic-induced neuronal cell death.

Although elevated TG2 expression and activity have also been linked to the pathogenesis of Parkinson's disease, amyotrophic lateral sclerosis, and other neurodegenerative disorders (reviewed in Ref. 235), mechanistic studies in TG2-modified animals, aimed at evaluating the role of TG2 in these diseases, have not yet been undertaken. Recent studies suggest that TG2-mediated intra- and intermolecular cross-linked α-synuclein is unable to form the cytotoxic, structured fibrillar aggregates responsible for neural membrane disruption in Parkinson's disease (247). TG2-mediated cross-linking may, thus, prevent progression of Parkinson's disease. A short isoform of TG2 that was originally identified in Alzheimer's disease brains (61) and is unable to bind GTP, appears to be cytotoxic to NIH-3T3 and SKBR3 breast cancer cells following transamidation activity-independent self-aggregation (20). Double mutants of TG2, that are both GTP binding and transamidation defective, were also proapoptotic in NIH-3T3 cells, highlighting the importance of GTP binding for the antiapoptotic function of TG2 (72). It remains to be determined if the proapoptotic function of the short TG2 isoform plays a role in TG2-associated neuronal cell death in neurodegenerative disease progression.

K. Diabetes

Maturity-onset diabetes of the young (MODY) is a heritable variant of type 2 or non-insulin-dependent diabetes mellitus (NIDDM) that occurs in nonobese individuals and presents in adolescence or when subjects are in their early 20s. Mutations in several genes (e.g., glucokinase, transcription factors) that impair insulin secretion have been implicated in its pathogenesis (83). Another potential candidate gene for MODY is TGM2, with heterozygous missense mutations that impair transamidase activity (M330R, I331N, or N333S) having been identified in one or more affected individuals of three different families (42, 226). TG2 is the predominant TG expressed in pancreatic islets (226), and its transamidase activity has been suggested to modulate the exocytosis of glucose-stimulated insulin secretion that accompanies calcium influx into pancreatic β-cells (51). Consistent with an involvement in insulin release but not synthesis, pancreatic islets from mixed strain (SVJ129-C57BL/6) TG2−/− mice were unchanged from wild-type islets with respect to total insulin content, islet size, or morphology (42). However, less insulin was secreted relative to wild-type islets in response to a high glucose challenge. This was reflected in lower blood insulin and higher blood glucose levels in TG2−/− mice following intraperitoneal glucose loading (42). Paradoxically, TG2−/− mice showed a hypoglycemic response to exogenous insulin, indicating increased insulin sensitivity (42), a phenomenon that has not been reported in MODY patients. The defect in TG2−/− glucose-stimulated insulin release has been attributed to mitochondrial TG2 functioning as a protein disulfide isomerase to reduce mitochondrial respiratory chain activity and, consequently, ATP production, thereby resulting in a dramatic decrease in the general motility of TG2−/− mice (180). However, such a pathogenetic mechanism is not generally associated with MODY. Moreover, a detailed study failed to find TG2 in mitochondria (151), and reduced motility due to low ATP levels (the basis for postulating the above pathogenetic mechanism) is not observed in all TG2−/− colonies (as discussed in sect. i above). Thus the precise role, if any, of TG2 and TGM2 mutations in MODY pathogenesis remains unclear.


A. General Overview and Protein Expression

Protein 4.2 (also known as band 4.2 or pallidin) is the only catalytically inactive member of the TG superfamily. Encoded by the EPB4.2 gene on human chromosome 15q15.2 (Fig. 1B), protein 4.2 is a 72-kDa peripheral membrane protein that is restricted to cells of the erythroid lineage (313) and accounts for ∼5% of the total RBC membrane protein (264). Protein 4.2 is myristylated (232) and palmitoylated (71). It binds ankyrin (145) as well as the cytoplasmic domain of anion exchanger 1 (AE1, also known as band 3) (146), both of which, in turn, link the erythrocyte membrane to the underlying cytoskeleton. It thereby contributes to membrane shape and stability. Protein 4.2 also associates with the Rhesus (Rh) protein complex (Rh-associated glycoprotein, Rh polypeptides, glycophorin B, CD47, LW) in RBC membranes (194).

In contrast to many other TGs, protein 4.2 binds ATP but not GTP (26), although the significance of this is as yet unknown.

B. Protein 4.2 Mouse Model

Mixed strain (SVJ129-C57BL/6) protein 4.2-deficient mice (219) were developed by replacement of exons 4–7 and part of exon 8 (which encode most of the catalytic core domain) with a neomycin resistance gene (Fig. 1B). Homozygous 4.2−/− mice were phenotypically indistinguishable from wild-type mice and were born at the expected Mendelian frequency. RBCs from 4.2−/− mice were smaller than wild-type RBCs due to membrane loss, were dehydrated due to altered cation concentrations, had reduced levels of band 3, and were mildly spherocytic instead of biconcave due to loss of lipid anchoring and increased fragility of the RBC membrane.

C. Hereditary Spherocytosis

Hereditary spherocytosis, the most common inherited disorder of the RBC membrane, is characterized by ion transport abnormalities accompanied by osmotic changes that lead to hemolytic anemia, jaundice, and enlargement of the spleen. Numerous mutations underlie hereditary spherocytosis, including those in genes encoding the proteins, spectrin, ankyrin, band 3, and protein 4.2, which connect the membrane skeleton to the lipid bilayer. However, mutations in the gene encoding protein 4.2 are rare; for a comprehensive list of EPB4.2 mutations, see Online Mendelian Inheritance in Man. In humans, expression of band 4.2 is reduced as a result of band 3 mutation (127, 236) and is totally absent in band 3−/− mice (220). Taken together with data from 4.2−/− mice, which display a reciprocal reduction in band 3 anion exchanger activity due to reduced levels of band 3 (219), this indicates that binding of protein 4.2 to band 3 enhances band 3-mediated anion transport by enhancing membrane incorporation of band 3, a finding that has been confirmed in Xenopus oocyte expression studies using wild-type protein 4.2 (289).

Human erythrocyte membranes completely lacking protein 4.2 are severely depleted in the CD47 component of the Rh protein complex (50), whereas RBC membranes from 4.2−/− mice displayed normal CD47 expression (194). Thus, although human protein 4.2 influences CD47 levels and is critical for linkage of CD47 to the membrane skeleton (63), this function does not appear to be conserved across species. This again questions the relevance to human RBCs of studies done in the mouse (see also sect. ivC).


It is clear from TG-deficient patients and genetically engineered TG-deficient mouse models that four of the nine human TGs, i.e., FXIII-A, skin and hair TGs 1 and 5, and protein 4.2, have distinct biological roles: the clotting disorders in FXIII-deficient patients that are phenocopied in FXIII-A-deficient mice demonstrate the integral extracellular cross-linking role unique to pFXIIIa in blood coagulation; the intracellular cross-linking roles unique to TG1 or TG5 in skin barrier formation are evident from TG1- or TG5-deficient patients and from TG1-deficient mice; and despite species differences between mouse and human RBCs, the noncatalytic, intracellular structural role of protein 4.2 in ion transport across the RBC membrane is apparent from protein 4.2-deficient mice and various human mutations that reduce intracellular protein 4.2 levels. Mice deficient in TG3 appear to have altered hair ultrastructure, although detailed phenotypic analysis of this model has yet to be published. Mice deficient in TG4, TG6, or TG7 have not yet been reported, nor have mutations in these proteins been linked to human disease.

Mutations in human TG2 have been linked to diabetes; however, available evidence from biochemical studies and TG2-deficient mouse models indicates a pleiotropic role for TG2 in pathophysiological conditions, or in amplifying or moderating pathophysiological influences. TG2 has been shown to be involved in cataract development, gluten sensitivity diseases, neurodegeneration, and tissue remodeling associated with cancer, wound repair, and bone development. The transamidase activity of TG2 plays a critical role in the formation of lens protein aggregates induced by oxidative stress. The extracellular transamidase and deamidase activities of TG2 are central to the adaptive immune response that is elicited in gluten sensitivity diseases. Whereas TG2-mediated cross-linking of amyloid-forming proteins may prevent Parkinson's disease progression, TG2 may contribute to the pathogenesis of Alzheimer's disease and Huntington's disease by mediating the formation of AT2 receptor oligomers and the formation of soluble high-molecular-weight htt complexes, respectively. The role of the short, proapoptotic, neural TG2 isoform in neurodegenerative disease progression remains to be investigated. The transamidase activity of TG2 contributes to extracellular matrix mineralization during bone development, to the excessive matrix mineralization characteristic of bone pathologies, such as osteoarthritis, and to stabilization of host matrix proteins that may retard tumor progression. Extracellularly, and independently of its cross-linking activity, TG2 promotes cell adhesion during tissue remodeling by stabilizing the ECM and by interacting with cell-surface adhesion receptors such as β1/β3/β5 integrins and the heparan sulfate proteoglycan receptor syndecan-4. More specifically, and perhaps more relevant to wound healing or cancers than to bone development, extracellular TG2 may function in syndecan adhesion-mediated signaling cascades that alter or amplify integrin-mediated signaling pathways. TG2 also functions intracellularly, to regulate cell spreading and motility independent of its cross-linking activity and, upon overexpression in cardiac cells, to enhance signaling through the prostanoid biosynthetic pathway leading to cardiac hypertrophy and contractile dysfunction. However, it is not yet known if it is the transamidase activity of TG2 or its other functions that contribute to cardiac hypertrophy. Intracellularly, as a transamidase, TG2 enhances inflammation through cytosolic cross-linking of IκBα, which enhances nuclear translocation and activation of the transcription factor NFκB, and TG2 promotes injury-induced hepatic apoptosis by translocating to the nucleus, where it cross-links and inactivates the transcription factor Sp1, which is required for the expression of cell viability genes. TG2 has also been implicated in the promotion of renal fibrosis, inflammation associated with endotoxic shock and cystic fibrosis, activation of latent TGF-β1 to trigger apoptotic cell engulfment by macrophages during the resolution phase of inflammation, and arterial remodeling and atherosclerosis. However, the specific biochemical function(s) and mechanism(s) involved in modulating these responses require further characterization.

To date, the only other TG with manifest pleiotropy is FXIII-A, which, like TG2, is involved in tissue remodeling associated with cancer, wound repair, and bone development. Activated pFXIII-A2* may have an additional cross-linking role in coagulation-related events, such as coated platelet formation, and cFXIII-A2° may be involved intracellularly in the serotonylation of small GTPases, although this requires further studies. The cellular localization of FXIII-A is often coincident with TG2, and both proteins participate in intracellular as well as extracellular cross-linking. The proangiogenic activity of FXIII-A in tissue remodeling, for example, is mediated by extracellular cross-linking of VEGFR-2 to β-integrin subunits, whereas the dimerization of AT1 receptors by intracellular cFXIII-A2°-mediated cross-linking contributes to the development of essential hypertension. Extracellularly, both FXIII-A and TG2 can facilitate integrin-mediated cell adhesion, independent of cross-linking activity. TG2 also interacts with other cell surface receptors such as GPR56 and syndecans; however, similar receptor interactions with FXIII-A have yet to be investigated. The roles of FXIII-A and TG2 are often complementary, as observed for bone development. Thus the temporal expression and localization of FXIII-A and TG2, but not the other TGs, in developing bones are consistent with a role for their transamidase activity in extracellular matrix mineralization and their adaptor function in the regulation of cell differentiation, although neither TG, alone, is critical for bone development. FXIII-A and TG2 can also have opposing actions, as evidenced by cancer progression in knock-out mouse models, where, unlike FXIII-A, which is not required for tumor growth but supports hematogenous metastasis, TG2 expression correlates with a reduction in both tumor growth and hematogenous metastasis. The mechanism(s) responsible for these disparate effects remains unclear.

The diverse intracellular and extracellular roles of TG2 and FXIII-A, and the many proteins with which they interact, indicate a complex “dance” between TGs and matrix, receptor, cytosolic, and nuclear proteins. Future studies delineating signaling pathways are needed to link specific effects with a specific TG location(s) and biochemical function(s). Indeed, a key area of research, about which little is currently known, is how TG2 is secreted from the cell. Further dissection of the overlapping or opposing roles of TG2 and FXIII-A would benefit from the use of tissue-specific, temporal-specific, or double-knockout mouse models. However, such studies must ensure the use of appropriate littermate controls to avoid potentially confounding background genetic variations. Finally, a major challenge will be the development, and temporal analysis in animal models, of specific pharmacological inhibitors of transamidase activity for efficacious treatment of cataracts, gluten sensitivity and neurodegenerative diseases, osteoarthritis, hypertension, and cancer.


This review was made possible through research funding from National Health and Medical Research Council Grants 459406, 526622, and 568753 and from Australian Research Council Grant DP0665422.


Address for reprint requests and other correspondence: S. E. Iismaa or R. M. Graham, Molecular Cardiology and Biophysics Division, Victor Chang Cardiac Research Institute, 405 Liverpool St., Darlinghurst, Sydney, NSW 2010, Australia (e-mails: s.iismaa{at} or b.graham{at}


View Abstract