Paradoxical Inhibition of Protein Aggregation and Precipitation by Transglutaminase-catalyzed Intermolecular Cross-linking*

Cross-linking of proteins catalyzed by tissue transglutaminase has been suggested to play key roles in a variety of cellular events, including cell apoptosis and human pathogenesis (e.g. polyglutamine and Alzheimer diseases). It has often been suggested that tissue transglutaminase enhances aggregation and precipitation of damaged or pathogenic proteins. To ascertain whether this is accurate, we investigated the effects of tissue transglutaminase-catalyzed modulation on the aggregation of structurally damaged and unfolded proteins. Our results indicated that the aggregation and precipitation of some unfolded proteins were inhibited by transglutaminasecatalyzed reaction, although the effect was strongly dependent upon the target protein species. To elucidate the molecular events underlying the inhibitory effect, extensive analysis was performed with regard to reduced β-lactoglobulin using a number of techniques, including chromatography and spectroscopy. The results indicated that cross-linking yields high molecular weight soluble polymers but inhibits the growth of insoluble aggregates. The cross-linked β-lactoglobulin retained stable secondary structures with a hydrophobic core. We concluded that the transglutaminase-catalyzed intermolecular cross-linking did not necessarily enhance protein aggregation but could sometimes have a suppressive effect. The results of the present study suggested that tissue transglutaminase modifies aggregation and deposition of damaged or pathogenic proteins in vivo in a wide variety of manners depending on the target protein species and solution conditions.

Protein clustering occurs in the highly concentrated intra-and extracellular solutions of the living body (1,2). Fine balancing of the intermolecular forces of proteins in biological fluids may realize a variety of physical states, such as dispersed, equilibrium clustered, glass-like, or crystal states, as in pure solutions (3)(4)(5)(6)(7), because biological environments are highly diverse. Investigation of the factors that affect the intermolecular interactions of proteins are of primary importance from a biological viewpoint. Such information is also essential for clinical control of abnormal protein aggregation, such as amyloid formation in patients with neurodegenerative diseases (8 -12).
Biological protein assembly is mediated not only by noncovalent interactions but also by covalent cross-linking catalyzed by enzymes. The most important group of cross-linking enzymes is the transglutaminase (TG) 1 family (13)(14)(15). TG are thiol-and Ca 2ϩ -dependent acyl transferases that catalyze formation of amide bonds between the ␥-carboxamide groups of peptide-bound glutamine residues and the primary amino groups of various compounds, including the ⑀-amino group of lysine in proteins. It is commonly held that covalent crosslinking of protein molecules always enhances protein aggregation and precipitation. In fact, cross-linking by TG stabilizes blood clots and skin keratin under physiological conditions (13,15). Tissue-type transglutaminase (tTG) is also a candidate for the factor that induces formation of neuronal inclusions in the brains of patients with human diseases, such as the polyglutamine and Alzheimer diseases (16 -19).
However, we have recently demonstrated that the cross-linking reaction catalyzed by tTG does not always enhance protein aggregation; the cross-linking reaction in the intramolecular mode was shown to strongly suppress the aggregation of diseaserelated proteins (20). This led to the question of whether the intermolecular mode of cross-linking would also suppress protein aggregation. In fact, Lai et al. (21) recently reported that tTGcatalyzed intermolecular cross-linking of a polyglutamine-containing protein formed high molecular weight soluble polymers, but inhibited precipitation. An accurate answer to this question must be pursued, as the tTG-catalyzed cross-linking has been suggested to play key roles in a variety of cellular events, including regulation of cellular growth, differentiation, apoptosis, and human pathogenesis (14,22).
In the present study, we examined whether the intermolecular cross-linking of damaged and structurally unfolded proteins enhances or inhibits their aggregation and precipitation. To assess this problem from a general viewpoint, we examined four structurally unrelated proteins as substrates of tTG: bovine ␤-lactoglobulin A (␤LG; ␤-sheet-type), bovine serum albumin (BSA; ␣-helix-type), human ␣-lactalbumin (␣LA; ␣/␤-type), and bovine ␣-casein (natively unfolded type). ␤LG, ␣LA, and BSA were unfolded by reducing the disulfide bonds, which might be the simplest model for biological protein damage. First, we performed simple aggregation and cross-linking analysis of all the proteins. This was followed by extensive analysis for ␤LG as a typical case using biochemical and biophysical techniques.

EXPERIMENTAL PROCEDURES
Materials-Guinea pig liver transglutaminase was purified on an immunoadsorbent column as described previously (23). ␤LG, ␣LA, BSA, bovine ␣-casein, and bovine ␣-chymotrypsin were purchased from Sigma-Aldrich and were used without further purification. Other chemicals of reagent grade were purchased from Nacalai Tesque Co. (Kyoto, Japan).
Aggregation and Cross-linking Reactions-All of the sample solutions were prepared in 10 mM Tris (pH 7.5), and the pH was adjusted carefully with small amounts of HCl. Non-covalent aggregation of ␤LG, BSA, and ␣LA was initiated by adding 10 mM dithiothreitol (DTT) and 5 mM CaCl 2 at 37°C. The tTG-catalyzed cross-linking reaction was initiated by adding 130, 260, or 650 nM tTG, 10 mM DTT, and 5 mM CaCl 2 at 37°C. Irreversibly inactivated tTG used for control experiments was prepared by heat treatment of tTG at 70°C for 5 min. A reference experiment under reducing conditions without aggregation and cross-linking of ␤LG was performed in a solution containing 10 mM DTT and 1 mM EDTA. SDS-PAGE analysis was performed by the standard method of Laemmli (24) with densitometric measurement of the Coomassie Brilliant Blue-stained gel. Solution turbidity was measured by absorbance at 320 nm using a U-3300 spectrometer (Hitachi, Tokyo).
In the experiments using ␣-casein, the protein was dissolved in 10 mM Tris (pH 7.5), 5 mM CaCl 2 , and 10 mM DTT and incubated at 37°C for 24 h. At this time point, tTG was added to the solution, and then the turbidity was monitored.
Size-exclusion Gel Chromatography and Microfiltration-Analytical size-exclusion gel chromatography (SEC) was performed using the Class M10A HPLC system (Shimadzu, Kyoto, Japan) equipped with a G3000SW column (Tosoh, Tokyo). The equilibrium and elution buffers were 0.1 M sodium phosphate buffer (pH 7.5), and the flow rate was 1 ml/min. Absorbance at 280 nm was monitored. The samples loaded onto SEC (20 l) were pretreated by microfiltration with an Ultrafree-MC filter (pore size, 0.22 m; Millipore Co., Bedford, MA).
Spectroscopic Measurements-Circular dichroism (CD) spectra were measured with a Jasco J-820 spectropolarimeter (Jasco, Tokyo) using a quartz cell with a path length of 0.5 or 1 mm. The protein concentrations were 6 and 150 M for far-and near-UV measurements, respectively. The temperature of the cell was controlled using a Peltier-type PTC-423S apparatus (Jasco).
Phase Separation of ␤LG Solutions in the Presence of CaCl 2 -The cross-linked samples were prepared in advance by adding 650 nM tTG and 5 mM CaCl 2 to the reduced reference sample followed by incubation at 37°C for 24 h. The reduced reference and cross-linked samples containing 100 M ␤LG and various concentrations of CaCl 2 were incubated at 37°C for 24 h. The samples after incubation were subjected to microfiltration (pore size, 0.22 m), and the protein concentration of the supernatant was determined by averaging estimates obtained from absorbance at 280 nm and the BCA method. The concentrations of CaCl 2 in the presence of 1 mM EDTA were determined by the method of Goldstein (25).
Proteolytic Digestion Experiments-The reaction mixtures containing 270 M ␤LG and 10 mM DTT were used for the digestion experiments monitored by SDS-PAGE. Depending on the sample species, 650 nM tTG, 5 mM CaCl 2 , or 1 mM EDTA was added to the solutions. The mixtures were incubated at 37°C for 24 h. Then, ␣-chymotrypsin was added at a final concentration of 10 g/ml, and the mixtures were incubated at 25°C with stirring.

tTG-catalyzed Suppression of Non-covalent Protein
Aggregation and Precipitation-Structurally damaged and unfolded protein molecules have a strong tendency to form aggregates because exposure of hydrophobic parts of the protein interior to the aqueous environment enhances the intermolecular hydrophobic interactions that overcome electrostatic or other types of intermolecular repulsion. Accordingly, we expected that reduction of disulfide bonds and the subsequent unfolding of proteins would also induce their non-covalent aggregation in a medium of relatively high ionic strength that partially shields electrostatic repulsion. This type of aggregation and precipitation was actually observed for ␤LG, ␣LA, and BSA at neutral pH at 37°C in the presence of DTT and Ca 2ϩ (Fig. 1, A-C, star symbols). Aggregation of the reduced ␤LG was suppressed completely by chelation of trace ions by EDTA (Fig. 1A, filled circles). This sample containing highly reduced ␤LG without any aggregates in the presence of 1 mM EDTA is referred to as the "reduced reference" sample throughout this report. Addition of an excess amount of CaCl 2 (e.g. 5 mM) to this reduced reference solution strongly induced non-covalent aggregation and precipitation of ␤LG (Fig. 1D, star symbols).
The addition of tTG to the aggregating solutions described above inhibited the increase in turbidity of the proteins depending on the concentration of tTG (Fig. 1). Notably, the effect of tTG was much weaker for ␣LA than for ␤LG and BSA. Heat-inactivated tTG did not show this inhibitory effect (data not shown). Ca 2ϩ -induced aggregation of the reduced reference ␤LG shown in Fig. 1D could also be suppressed very efficiently by tTG (Fig. 1D, open symbols). Furthermore, disassembly of preformed aggregates by tTG was observed for ␣-casein. A solution of ␣-casein became turbid in the presence of CaCl 2 , and the turbidity decreased sharply by adding tTG at 37°C (Fig. 2). These results indicated that the tTG-catalyzed reaction inhibited the non-covalent aggregation of unfolded proteins for at least some protein species.
The solubility of tTG-modified ␤LG as a function of ionic strength was compared with that of the reduced reference ␤LG by changing the concentration of CaCl 2 and measuring the residual protein concentration in the solution phase at 37°C. The tTG-modified ␤LG exhibited phase separation and precipitation at [CaCl 2 ] м 20 mM, whereas the reduced reference ␤LG was precipitated at [CaCl 2 ] м 0.5 mM (Fig. 3). These results confirmed that tTG-catalyzed modification of the unfolded protein increased its solubility.
The proteins used here were cross-linked very efficiently by tTG in the presence of Ca 2ϩ and DTT ( Fig. 4 for ␤LG; Refs. 26 and 27). Therefore, the cross-linking reaction was the most plausible mechanism for the inhibition of precipitation and increase in protein solubility. To determine the detailed molecular events underlying this paradoxical phenomenon, further analyses were performed for the case of ␤LG shown in residues located on flexible protein chains is necessary for efficient cross-linking.
Size-exclusion Gel Chromatography Combined with Microfiltration-Microfiltration (pore size, 0.22 m) and SEC were combined to monitor the aggregation and cross-linking processes of ␤LG (Fig. 5). This method separated three different populations of ␤LG, although some errors caused by dissociation of rapidly dissociating oligomers during the time of chromatographic measurement could not be excluded. The three groups were large aggregates of Ͼ0.22 m (group L; star symbols in Fig. 5, B and C), small aggregates of Ͻ0.22 m (group S; filled symbols in Fig. 5, B and C), and non-aggregated ␤LG (group N; open symbols). Fractional changes in each population were monitored by measuring peak areas on the chromatograms (Fig. 5, B and C). The peak labeled "N" in Fig. 5A corresponds to the N group. The area of the S group was estimated by integrating the chromatogram at the elution time of 5 to 8.7 min including the void fraction (labeled "S" in Fig.  5A). The fractional values for the N and the S groups were calculated by dividing the areas by that of the N group at zero reaction time. The L fraction was estimated by assuming that the sum of the N, S, and L fractions should be unity.
In the case of the reduced reference sample containing 1 mM EDTA, neither the L nor S group appeared even after incubation for 24 h (data not shown). In the sample containing CaCl 2 without tTG, the large aggregates of the L group appeared soon after the beginning of incubation, but the S fraction did not increase significantly (Fig. 5B). All of the molecules were in the L group after incubation for 24 h.
In contrast, in the presence of tTG and CaCl 2 , the L fraction did not increase substantially, whereas the S fraction increased rapidly (Fig. 5C). In this case, essentially all the molecules were trapped in the S group after incubation for 24 h (Fig. 6, open symbols). These results indicated that reduced ␤LG modified by tTG could not grow to form large aggregates. The half-decay time constant () of the N fraction of the tTG-containing sample was 4.2 Ϯ 0.2 h (Figs. 5C and 6, open symbols). These decay kinetics were similar to those observed for the decay of the monomer fraction determined by SDS-PAGE analysis (Fig. 6, filled symbols). Both decays advanced at the same rate in the initial 4 h of incubation, indicating that growth of the S fraction in the presence of tTG could be explained completely by the cross-linking reaction at the early stages. The cross-linked high molecular weight ␤LG was trapped in the solution phase. At later stages of incubation, the decay of the N fraction of SEC was slightly faster than that of the monomer fraction on SDS-PAGE (Fig. 6), suggesting a minor contribution of non-covalent mechanisms for the later assembly process of ␤LG.
Notably, most of the ␤LG molecules in the S group under cross-linking conditions were eluted in the void fraction at all incubation times (Fig. 5A, bottom panel). Oligomers with lower molecular weights were not present to a large extent. These observations were consistent with the results of SDS-PAGE analysis (Fig. 4), which did not exhibit a clear ladder-type pattern of polymerization. These results suggested that the rate-limiting step in polymerization was not the cross-linking step itself. The details of the cascade of events leading to cross-linking will be discussed later.
Conformational Changes of ␤LG Accompanied by Cross-linking-The conformational changes of ␤LG accompanying reduction and cross-linking were analyzed by CD spectroscopy. The spectrum in the near-UV region of the reduced reference ␤LG showed a total loss of tertiary structures by incubation at 37°C for 24 h (Fig. 7A, thick solid line). The decrease in ellipticity at ϳ206 nm of the far-UV CD spectrum also demonstrated unfolding of secondary structures by reduction (Fig. 7A, thick solid line). The near-UV CD spectrum of ␤LG in the presence of tTG and Ca 2ϩ indicated that the cross-linking reaction did not inhibit the breakdown of the tertiary structures (Fig. 7B, broken line). However, unfolding of the secondary structures was partly blocked by cross-linking (compare the broken and thick solid lines in Fig. 7B). The addition of tTG and CaCl 2 to the fully reduced reference sample also resulted in significant recovery of the shape of the far-UV CD spectra (Fig. 7C). The spectral shape of the cross-linked ␤LG shown in Fig. 7C was quite similar to that shown in Fig. 7B.
Kinetic details of the conformational changes of ␤LGs were monitored by ellipticities at [] 206 and [] 293 . Tertiary structures of ␤LG represented by [] 293 were lost in a single-phase manner for both the reduced reference and the cross-linked ␤LGs (Fig.  8A). The decay time constants of the two were similar to each other ( ϭ 4.2 Ϯ 0.3 and 5.0 Ϯ 0.4 h, respectively), indicating that the cross-linking reaction did not substantially perturb the unfolding process of the tertiary structures. In contrast, the crosslinking reaction significantly altered the kinetic trace of [] 206 . The trace for the reduced reference sample decreased in a biphasic manner with 1 ϭ 1.8 Ϯ 0.2 and 2 ϭ 20.7 Ϯ 3.2 h (Fig. 8B,  filled symbols). The trace for the cross-linking sample also decreased monotonically until 5 h of incubation (Fig. 8B, open  symbols). The decay constant of this part was 1.8 Ϯ 0.3 h, exactly the same as that of the faster phase of the reduced reference sample. However, the change in [] 206 of the slower phase was blocked completely by the tTG-catalyzed reaction. The trace exhibited a relatively sharp corner at 4 -6 h of incubation, and the ellipticity even increased marginally at longer incubation times. These results indicated that the cross-linking reaction inhibited unfolding of the secondary structures at later stages. Note that CD analysis was not informative for the non-covalently aggre-gated sample without tTG because of the strong light scattering of the turbid solution.
Stability of Cross-linked ␤LG-The stability of the residual secondary structure of cross-linked ␤LG was compared with that of the reduced reference ␤LG by far-UV CD spectroscopy. By decreasing the temperature from 25 to Ϫ3°C, the residual structure of the reduced reference ␤LG was unfolded as indicated by a strong decrease in ellipticity at ϳ200 nm (Fig. 9A, broken line). This change was completely reversible, indicating that the residual structure still contained a cooperative hydrophobic core (28,29). The spectrum of the cross-linked ␤LG also changed reversibly in the direction of unfolding by cooling, but the magnitude of the change was much smaller (Fig. 9B, broken line). The results indicated that the residual secondary structure with a hydrophobic core of the cross-linked ␤LG was formed more tightly than that of the reduced reference ␤LG. We also compared the stability of the residual secondary structures by guanidine hydrochloride (GdnHCl)-induced denaturation. The denaturation curve monitored by [] 220 was composed of two phases for both the ␤LG species (Fig. 9C). The molecular origin of this biphasic behavior was not pursued in the present study. The results shown in Fig.  9C indicate that the residual secondary structure of the crosslinked sample was more stable than that of the reduced reference against the denaturant.
Finally, stability against ␣-chymotrypsin digestion was tested for the three ␤LG species. The reduced reference ␤LG was digested easily by the protease (Fig. 10, Reference), whereas digestion of the non-covalently formed aggregates was negligible (Fig. 10, Non-covalent). The cross-linked sample was also degraded at a significant rate (Fig. 10, Cross-linked). These results indicated that the susceptibility of cross-linked ␤LG to the proteolysis was much greater than that of the non-covalent aggregates and was even comparable with that of the reduced ␤LG monomer.

Inhibition of Protein Aggregation by tTG-catalyzed Cross-
linking-Clustering behaviors of unfolded protein molecules in solution are finely influenced by changes in the attractive and repulsive intermolecular forces. Here, we studied a covalent mechanism that could shift the balance of the associationdissociation equilibrium. It is natural to suppose that the covalent cross-linking of unfolded protein molecules would always enhance their aggregation and precipitation (Fig. 11A,  Case 1). However, we found that the tTG-catalyzed cross-linking of some unfolded proteins inhibited their non-covalent aggregation and precipitation by producing soluble multimers (Fig. 11A, Case 2). Notably, the efficiency of the tTG effect was strongly dependent upon the species of substrate proteins (Fig.  1, A-C), indicating that the inhibitory effect is active only for some specific protein groups or conformations. It is known that 1 g of guinea pig liver contains ϳ5.8 nmol of tTG (23), which gives a rough estimate of 1ϳ3 M for the intracellular concentration of tTG. The concentrations of tTG used in the present study were comparable with or lower than this estimate. The detailed molecular events underlying the paradoxical inhibition of protein aggregation by tTG were analyzed for ␤LG. Based on all of the results, we propose the scheme shown in Fig. 11B. Three different denatured states of reducing ␤LG (D 1 , D 2 , and D 3 ) were identified using the kinetic data of [] 293 and [] 206 of the reduced reference sample (Fig. 8, filled circles) The kinetic results indicated that ␤LG under reducing conditions initially unfolded to a mildly denatured state (D 1 ), which lacked some secondary structure but still contained tertiary structure. Then, the unfolding advanced to the D 2 state in which the tertiary structures of ␤LG disappeared completely. After a longer incubation time, ␤LG in the D 2 state further lost its residual secondary structure and fell into the more extensively unfolded state, D 3 .
Cross-linking did not affect the decay of [] 293 (Fig. 8A) but inhibited the slower phase of the change in [] 206 (Fig. 8B). This observation indicated that cross-linking blocked the transition from the D 2 state to D 3 and trapped the molecules as soluble multimers (Fig. 11B). The decay constant of the non-aggre-gated fraction by SEC for the cross-linked sample was 4.2 Ϯ 0.2 h (Figs. 5C and 6), which was the same as that observed for the decay of the tertiary structure ( ϭ 4.2 Ϯ 0.3 h). This result indicated that the cross-linking reaction occurred rapidly in the D 2 state of ␤LG. The rate-limiting step of the cross-linking reaction is probably the conformational change of ␤LG. The extensive unfolding from D 2 to D 3 could also be reverted by the action of tTG (Fig. 7C). The decay time constant of the N fraction of the non-covalent aggregation processes (4.4 Ϯ 0.4 h; Fig. 5B) was also identical to that of [] 293 of the reduced reference ␤LG, suggesting that the non-covalent aggregation was also initiated in the D 2 state (Fig. 11B).
Properties of Cross-linked ␤LG Molecules-The cross-linked ␤LG could not form precipitates under low ionic strength. Higher concentrations of CaCl 2 were required for its phase separation than for the reduced ␤LG monomer (Fig. 3). This indicated that cross-linking weakened the intermolecular attractive forces or, alternatively, that it enhanced the electrostatic repulsion. The observation that the effect of tTG on the aggregation of ␣LA was much weaker than that of ␤LG and BSA (Fig. 1, A-C) can also be explained by variations in force balancing; the attractive force enhanced by unfolding of ␣LA was probably too strong to be overcome by the cross-linking effect.
The change in the force balance caused by the tTG activity could have originated from the conformational shift of the unfolded ␤LG by cross-linking because cross-linking substantially suppressed unfolding of the secondary structure (Fig.  8B). Furthermore, the residual structures with a hydrophobic core were more stable for the cross-linked ␤LG than for the reduced reference (Fig. 9). It is plausible that cross-linking trapped ␤LG in the conformation with less exposure of the hydrophobic core, which would reduce the hydrophobic attractive force among the molecules and increase the solubility of  the unfolded proteins. Alternatively, cross-linking may also have altered the charge distribution on the unfolded molecules. All of the protein species used in the present study were acidic and carried net negative charges at pH 7.5. Formation of one isopeptide cross-link removes one positive charge on lysine from the proteins and increases their net negative charge. This could enhance the intermolecular electrostatic repulsion, which might explain the high solubility of the cross-linked molecules.
Biological Implications and Conclusion-Structurally unfolded protein molecules expose their hydrophobic interiors to the aqueous environment and therefore exhibit a strong tendency to form insoluble aggregates. This phenomenon has profound biological significance related to biosynthesis and biological disposal of proteins (30 -34). The present study has suggested that the tTG-catalyzed cross-linking reaction did not always enhance protein aggregation but sometimes showed an inhibitory effect. This view is important, as tTG is believed to play key roles in apoptosis and human degenerative diseases (14,22). In the case where cross-linking increases the solubility of the damaged proteins, these proteins might be removed more efficiently from the intra-and extracellular spaces by fluid flow or proteolysis. Cross-linking could also suppress the pathogenic amyloid-type aggregation. In fact, Lai et al. (21) showed that the cross-linking of polyglutamine-containing proteins increased their solubility and inhibited their fibril formation. Soluble oligomers of pathogenic proteins are not always benign and could be real pathogens for degenerative disease (12).
However, it should also be emphasized that tTG-catalyzed cross-linking does not necessarily inhibit protein aggregation and precipitation. We found that the effect of tTG on the aggregation process was strongly dependent on the species of the substrate proteins (Fig. 1, A-C). Furthermore, when crosslinking occurs with the substrate proteins in an extremely high concentration, it will induce gelation of the proteins. This is probably the case for blood clotting and skin keratin formation. It should also be noted that tTG and N ⑀ -(␥-L-glutamyl)-L-lysine linkage are often co-localized with the intracellular inclusions in patients with the neurodegenerative diseases (36 -39). Cultured cell models over-expressing tTGase and/or amyloidogenic proteins, such as , ␣-synuclein, and huntingtin, have also been shown to exhibit tTGase-induced intracellular cross-linking and deposit formation (35,38,39). These observations suggested that tTG-catalyzed cross-linking enhances the aggregation and deposition of some pathogenic proteins in vivo. It is possible that tTG-catalyzed cross-linking modifies protein aggregation and deposition in vivo in a variety of fashions, including both enhancement and inhibition of aggregation de-pending on the target protein species and solution conditions. Further studies will be required to examine larger numbers of protein species under a variety of solution conditions, to establish the biological significance of the present findings.