Intramolecular Cross-linking of Domains at the Active Site Links AI and B Subfragments of the Ca 2 ’-ATPase of Sarcoplasmic Reticulum *

Glutaraldehyde treatment of sarcoplasmic reticulum vesicles results in formation of cross-linked CaZ’-ATPase oligomers. Under limiting  reaction conditions, where minimal interpolypeptide cross-linking occurs, hydrodynamic properties of the monomer are altered, such that, on sodium dodecyl sulfate-polyacrylamide electrophoresis, the enzyme migrates with an apparent molecular weight of 125,000 (E(125)), as compared to the native enzyme (E(110)). The E(125) species was also formed following reaction with other cross-linking bis-aldehydes, with formaldehyde and with a bissuccinimidyl ester. Derivitization resulted in inactivation of ATPase activity and of phosphoprotein formation from Pi. E(125) formation was inhibited by ATP, ADP, AMPPCP, and orthovanadate, and by specific modification of active site Lys-514 with fluorescein-5‘-isothiocyanate. Tryptic cleavage patterns of the glutaraldehyde-modified enzyme were consistent with covalent linkage of AI and B fragments that have been postulated to comprise the phosphorylation and nucleotide-binding domains (MacLennan, D. H., Brandt, C. J., Korczak, B., and Green, N. M. (1985) Nature 316,696-700). The denaturing detergent, sodium dodecyl sulfate, prevented cross-link formation. Interdomain cross-linking was inhibited by prior modification with either 2,4,6-trinitrobenzene sulfonate, phenylglyoxal, or pyridoxaL5’-phosphate but was unaffected by thiol group modification with iodoacetate or N-ethylmaleimide, suggesting involvement of lysine residues. These findings indicate that intramolecular cross-linking at the active site of the Ca2*-ATPase involves phosphorylationand ATP-binding domains that are widely separated in the linear sequence.

* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The particularly sensitive tryptic cleavage sites, T1 and Tz, which fragment the polypeptide, initially to A (Mr = 55,000) and B (Mr = 54,000) peptides, and subsequently into A, (M, = 33,000) and A2 (M, = 22,000). The cleavage sites approximately delimit the hypothetical cytoplasmic domains, so that A1 contains the phosphorylation site, and B the ATP-binding domain. It has been suggested that A2 constitutes an energy transduction domain (1). The nucleotide domain has been identified, in part, from localization of a lysyl residue (Lys-514), which is specifically modified by fluorescein-5'-isothiocyanate (FITC), resulting in inhibition of ATP-dependent catalytic function (2,3). Another "essential" lysyl residue has been located on the A1 fragment, which contains the aspartate residue that is phosphorylated (Asp-351), through its specific reaction with pyridoxal-5'-phosphate and inhibition of ATPase activity (4).
It has been proposed that catalysis in dehydrogenases and kinases involves conformational changes that result in movements of separate domains with respect to each other about a "hinge region." This movement, triggered by substrate binding, brings the substrates together, enclosing them in an isolated active site where the reaction takes place. Reversal of domain movement releases the products, and the enzyme is ready for another catalytic event (for review see Ref. 5). In the case of the Ca2+-ATPase, it can be predicted that the ATP-binding site on the nucleotide-binding domain, and Asp-351 on the phosphorylation domain, must be in close proximity. Segments of these two domains presumably contribute to formation of the active site.
Intramolecular cross-linking has been used as a method for investigating protein folding and tertiary structure. Introduction of such cross-links into ribonuclease A (6) and lysozyme (7) results in altered physical properties and increased thermal stability. Active site residues have been cross-linked in adenosine cyclic 3',5'-monophosphate-and guanosine cyclic 3',5'-monophosphate-dependentprotein kinases (8,9). In this paper we report on the effects of bifunctional agents on the Ca2+-ATPase that result in intramolecular cross-linking, under conditions where minimal cross-linked oligomers are formed. The modified protein can be conveniently detected by its altered mobility on sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). Cross-linking appears to occur at the active site and to involve residues situated on domains widely separated in the linear sequence.

2042
This is an Open Access article under the CC BY license.
SR vesicles were prepared from rabbit back and hind limb white muscle by the method of Eletr and Inesi (11). These were stored as a suspension (10-20 mg of protein/ml) in 10 mM imidazole, pH 7.4, and 0.3 M sucrose at 0 "C or -60 "C and used within 4 days or two months, respectively. Protein concentrations were estimated by the Lowry method, using bovine serum albumin as standard and sodium deoxycholate as a solubilizing agent.
Cross-linking-Reaction of intact or detergent-solubilized SR vesicles (0.4 mg protein/ml) with various cross-linkers was carried out at 25 'C in the media specified in the legends to the figures. The reaction times with glutaraldehyde and the glutaraldehyde concentration are also indicated in the legends to the figures. Molar concentrations of glutaraldehyde were calculated assuming a molecular weight of 100 and that the 25% aqueous solution is 25% w/v. For the other cross-linkers, the concentrations and times were as follows: glyoxal (17.6 mM) for 10 min, formaldehyde (26.6 mM) for 10 min, dimethyl suberimidate (5 mM) for 30 min, ethylene glycol bis(succinimidy1 succinate) (1 mM) for 20 min, 1, (300 pM) for 1 min, p-phenylene diglyoxal (0.2 mM) for 30 min, and N,N"p-phenylenedimaleimide (0.2 mM) for 30 min. All reagent solutions were made up fresh prior to use. Glutaraldehyde, glyoxal, and formaldehyde were quenched with a 5-to 10-fold molar excess of hydrazine. Dimethyl suberimidate, ethylene glycol bis(succinimidy1 succinate),p-phenylenediglyoxal and N,N'-phenylenedimaleimide were quenched with a 10-fold molar excess of ethylamine, ammonium acetate, arginine, and f3-mercaptoethanol, respectively.
Chemical Modifications-SR vesicles (2 mg SR protein/ml) were chemically modified with various monofunctional reagents in the media indicated below for the specified times at 25 "C. All reagent solutions were made up fresh prior to use, and an aliquot of the stock was added to the SR protein to the final concentration indicated. The glutaraldehyde reaction with the modified SR protein was carried out as above, without prior quenching of the modifying reagents used. Gel Electrophoresis-SDS-PAGE was performed on 4-16% gradient acrylamide according to Laemmli (14). Protein samples for electrophoresis were mixed with an equal volume of solubilization medium (0.1 M TRIS/HCI, pH 6.8,3% (w/v) SDS, 8 M urea, and 4% (w/v) @-mercaptoethanol (if the sample was to be reduced), and 8 pg of total protein aliquots were applied to each well. Approximate molecular weights were assigned by comparison with protein standards. Gel scans of Coomassie Blue-stained, dried gels were performed on a Vitatron densitometer (Manual TLD 100).
ATPase Activity-ATPase activity was measured by the pyruvate kinase-lactate dehydrogenase coupled assay system with 5 mM ATP (15). The activity shown represents the Ca'+-dependent component, obtained from the difference of the activity measured in the presence of calcium ions and that in excess EGTA. The calcium ionophore, A23187 (0.05 pmollmg of protein), was included to make the vesicles "leaky." E-P Measurements-SR vesicle phosphorylation from Pi was de-termined with [3'P]Pi and a filtration procedure (16).
Purification of ea'+-ATPase by Ion Exchange Chramtography-Purification of Ca'+-ATPase, following solubilization in CI2E8, passage through a column of DE52 ion exchange resin and activation with exogenous soybean phospholipid, was carried out as described by Barrabin et al. (17).
EGTA Inactivation of ATPase Activity-The Ca'+-ATPase was inactivated by incubating SR vesicles (0.4 mg of protein/ml) in 100 mM MOPS/TRIS/CHES, pH 8.5, 30 mM NaC1, 1 mM EGTA, and 0.3% Triton X-100 for 30 min at 37 'C. Thereafter, the samples were cooled to 25 "C and cross-linked. The enzyme exhibited no Ca'+dependent ATPase activity after the EGTA treatment.

RESULTS
The protein content of SR vesicle preparations routinely contain about 80% Ca2+-ATPase, with variable amounts of phosphorylase (M, = 93,000), calsequestrin ( M , = 60,000), and other unidentified but lower molecular weight proteins (Fig. L4). The Ca2+-ATPase has a calculated molecular mass of 109,771 daltons, based on the deduced primary sequence (1) and it bands close to this value, according to standard protein markers, in gradient gels. Reaction of SR vesicles with glutaraldehyde resulted in the formation of a protein species with an apparent molecular weight of 125,000, with concomitant loss of the 110,000 ATPase band. This occurred before extensive cross-linking to ATPase dimers or cross-linking of other protein components took place. In this paper, Ca2+-ATPase species as identified by SDS-PAGE will be referred to as E(M,), were M, is the apparent molecular weight, as determined by SDS-PAGE. The pattern of cross-linking and increasing ratio of E(125) to E(110), as the extent of reaction was increased, indicates that the majority of the ATPase cross-links first to E( 125), before cross-linking to higher order oligomers.
The pH dependence of the formation of E( 125) is shown in Fig. 1B. This behavior is consistent with reaction of an ionizable group with a pKo of approximately 7.5, and is maximum above pH 8.0. The observed pH dependence is not due to the lower reactivity of glutaraldehyde at acid pH values, since higher concentrations of glutaraldehyde, or longer reaction times at pH 6.0, resulted in increased amounts of crosslinked oligomers of ATPase but not in the amount of E(125) (not shown). The effects of solubilization of SR vesicles in Triton X-100 (0.2%) and of a longer glutaraldehyde reaction time, together with the ability of various other cross-linking agents to form E(125), are shown in Fig. 2. The concentrations of monovalent and divalent cations are low in order to minimize proteinprotein interactions.2 Reaction with glutaraldehyde under these conditions resulted in close to 100% loss of the native Ca2+-ATPase and appearance of almost stoichiometric amounts of E(125). Other aldehyde-based cross-linkers, such as formaldehyde, glyoxal, and p-phenylenediglyoxal, as well as the bis-succinimidyl ester reagent, ethylene glycol bis(succinimidy1 succinate), also formed E(125) but were not as effective as glutaraldehyde. The bis-imidoester, dimethyl suberimidate, did not cause conversion of E(110) to E(125), although in the absence of detergent this reagent produced cross-linked ATPase dimers, trimers, and other large oligomers, indicating that the concentration of the reagent was not limiting. Oxidation of sulfhydryls with 12, or reaction with the bifunctional sulfhydryl reagent, N,N"p-phenylenedimaleimide, resulted in rather broad-stained protein bands, which made it difficult to decide whether the E(125) had formed. ATP, however, had no effect on the position or size of these broad bands (results not shown), contrary to what was ob-  -93 125 The effects of various nucleotides, p-nitrophenyl phosphate (pNPP) and orthovanadate on the formation of E(125) are shown in Fig. 4. In preliminary experiments, using thin layer chromatography to monitor the identity of the nucleotides, we determined that glutaraldehyde did not react with adenine nucleotides under these conditions. ATP, AMPPCP, ADP, ITP (each 0.5 mM) and orthovanadate (0.1 mM) substantially or completely prevented formation of E( 125). AMP andpNPP (each 0.5 mM) caused partial inhibition. Phosphate (10 mM with 10 mM MgCI,) had very little effect (not shown).

93
Reaction of glutaraldehyde with the Ca2+-ATPase resulted in a time-and pH-dependent inactivation of ATPase activity and of phosphoenzyme formation from Pi (Fig. 5, A and C ) .
Although there was a linear relationship between the extent of inactivation and amount of E(125) species formed, the former was greater than the latter. This effect was more marked at pH 6.5 than at 8 (Fig. 5 B ) . The effect of ATP on ATPase inactivation by glutaraldehyde is shown in Table I. ATP had little protective effect on hydrolytic activity but completely inhibited E(125) formation at both concentrations used.
Results of tryptic digestion experiments are shown in Fig.  6. Mild tryptic digestion of the Ca2+-ATPase resulted in the formation of fragments A (55,000) and B (54,000), followed by the cleavage of A into A, (33,000) and A2 (22,000) (see lanes 6 and 8). Tryptic digestion of glutaraldehyde-treated SR vesicles produced a new protein band, migrating with an apparent molecular weight of 135,000, with loss of E(125) ( l a n e 4 ) . The residual unmodified protein (E(110)) was    weight on the gel of 108,000, close to the position of the 110,000 native protein (lune 5). Very little of the AI fragment was formed. This pattern of cleavage and altered apparent molecular weights was also shown by experiments in which the native enzyme was first cleaved to A and B or AI, Az, and B fragments and then treated with glutaraldehyde (lanes 7,9, respectively). In the first case fragments A and B were converted into E(135), and in the second case AI and B were converted into E(108), while AP was relatively unaffected. Confirmation of this cleavage pathway is provided by the banding pattern which develops with time a t constant trypsin concentration (Fig. 7). It is clear that the cross-linked enzyme, E(125), becomes A-B (E(135)) with the first cleavage, then A,-B (E(108)) and A, on the second. Very little A, was formed. It also appears that A,-B is converted to E(102) on further digestion and that this is most likely due to cleavage a t a site on the A, fragment (18). An interesting observation is that solubilization of the tryptic fragmented Ca2+-ATPase in Triton X-100 (0.5%) did not prevent the cross-linking of A and B or AI and B fragments, indicating a tight association between the fragments (not shown). It has been reported that the tryptic fragments cannot be dissociated in non-ionic detergents (19).
The effects of various treatments of the SR preparations on cross-linking with glutaraldehyde is shown in Fig. 8. Barrabin et al. (17) have reported that passage of SR protein through DEAE-cellulose in CI2EB, and elution of the Ca2+-ATPase with salt, results in a purification and selective enrichment of active over inactive ATPases. We found that most contaminating proteins, with a molecular weight lower than that of the ATPase, are lost with this procedure, but that some higher molecular weight species become evident (lune B ) . The C, E-G) were preincubated for 20 min at 25 "C in 100 mM MOPS/TRIS/CHES, pH 8.5, 100 p~ CaC12, 30 mM NaCI, and either 0.3% C12EB ( l a n e C), 1% Cl2E8 ( l a n e E ) , 1% Triton X-100 ( l a n e F), or 0.2% SDS ( l a n e C), and then cross-linked.
The DEAE-treated SR preparation shown in lane B and a standard SR preparation aged for 7 days at room temperature with 0.02% NaN2, were preincubated as above with 0.3% CI2EB and then cross- D and Z, respectively). An EGTA-inactivated SR preparation (see "Experimental Procedures") was cross-linked ( l a n e H).
. l indicate that a cross-link is formed between regions delimited by tryptic fragments A1 and B.
The evidence is consistent with formation of a cross-link between the Al and B fragments, due to glutaraldehyde reacting a t residues in the active site. Nucleotide binding to the active site, at concentrations known to saturate this site (0.5 mM ATP, ADP, AMPPCP, and ITP), prevents cross-linking, AMP and pNPP (0.5 mM), which are known to bind with low affinity, caused some inhibition. Effects of orthovanadate binding, which results in a tight complex with the enzyme and mimics the transition state of the phosphorylation reaction with Pi (21)(22)(23)(24), suggests that phosphorylation or occupation of the active site by phosphate could prevent the crosslink. Pi (10 mM) binding, however, had little effect and this may be due to the unfavorable pH for the phosphorylation reaction (25). Modification of Lys-514 on the B fragment by FITC, which is considered to be at the nucleotide-binding site (1-3), inhibited the cross-link.
Pyridoxal-5'-phosphate, which reacts specifically with a lysyl residue on the AI fragment, and which is also considered to be at the active site (4), also prevented cross-linking. The greater degree of functional inactivation of the enzyme (Fig.  5B) than expected from the extent of E(125) formation may be explained by a mechanism in which 2 residues at the active site need to be linked in E(125) but that modification of only 1 residue leads to an inactive enzyme. The marked pH dependence of derivitization and the disparity at more acid pH values suggests that the 2 residues involved may have different pK, values. It is interesting that ATP binding affords little protection against inactivation of the enzyme by glutaraldehyde and yet prevents the cross-link. Possibly, ATP binding only blocks reaction with one of the reactive residues and reaction with the other inactivates the enzyme.
The reaction of glutaraldehyde with proteins is complex and appears to involve the side groups of lysine, cysteine, histidine, and tyrosine (12). The best characterized reaction is Schiff base formation between carbonyls and primary amines (12). Carbonyl reactions with other functional groups f ... have not been fully substantiated. Involvement of lysyl residues is suggested by the observation that bis-aldehydes formed the cross-link, as did the bis-succinimidyl ester, ethylene glycol bis(succinimidy1 succinate). These reagents react almost exclusively with the €-amino group of lysine. Glyoxal, although considered to be fairly specific for arginyl residues, also reacts with lysyl residues (12). The lysine specific bisimidoester cross-linking reagent, dimethyl suberimidate, did not, however, react with the Ca2+-ATPase to form any E(125), although in the absence of Triton X-100, higher order crosslinked oligomers were evident. The imidoester functional group is positively charged and quite possibly could have limited access to a hydrophobic pocket, or could be repelled by positively charged regions of the protein, where the crosslink forms. Formaldehyde reacts with a wide range of functional groups in an undefined manner (12). It was difficult to conclude if any E(125) species was formed by sulfhydryl crosslinking with I, oxidation or reaction with N,N-p-phenylene dimaleimide.
The finding that prior chemical modification with ( a ) TNBS, which is specific for lysyl and cysteinyl residues, or with phenylglyoxal, which reacts with both arginyl and lysyl residues, and ( b ) , the specific modifications of lysyl residues by FITC (Lys-514) and pyridoxal-5'-phosphate (Lys-? on the AI tryptic fragment), prevent the glutaraldehyde cross-link, also implicate lysyl residue involvement in the cross-link. Prior modification with iodoacetic acid and N-ethylmaleimide, both specific for sulfhydryl groups, which did not prevent E(125) formation, indicate that cysteinyl residues are not involved. Several buried amino groups on the Ca2+-ATPase have been shown to be unreactive to methyl acetimidate (26) and may explain the ineffectiveness of the ethyl derivative in blocking E( 125) formation. It is therefore suggested that lysyl residues, one close to the aspartyl residue on the A, fragment that is phosphorylated by ATP or Pi (Asp-351), and one close to Lys-514, are involved in the cross-link. It is also possible that more than one type of cross-link could produce E(125).
Altered hydrodynamic properties of a protein on SDS-PAGE with resulting increase in apparent molecular size with the introduction of an intramolecular cross-link and further increase in apparent size with a single cleavage by trypsin has not, as far as we are able to ascertain, been described previously. Proteins that contain disulfide linkages generally migrate on SDS-PAGE, under nonreducing conditions, with lower apparent molecular size, compared with the reduced protein (27). This occurs despite decreased SDS binding in the unreduced, compared with the reduced state (28). Decreased frictional resistance is evidently the dominant factor affecting migration behavior. Reduced proteins assume an extended shape in SDS, the length of which varies in direct proportion to their molecular weight (29). Most intrapolypeptide disulfides form loops of about 50 residues or less, and the observed increase in electrophoretic mobility of unreduced proteins with intrapolypeptide disulfide linkages is simply explained by the decrease in extended length. Much longer loops are possible with artificial cross-links in large proteins, which would cause the shape of the polypeptide in SDS to deviate from the usual rod shape. This may compensate, in terms of frictional resistance, for the decrease in extended size and coupled with decreased SDS binding, could produce a species with an increase in apparent molecular weight on The relationship between the formation of the cross-link, tryptic cleavage, and apparent and actual molecular weights is shown diagrammatically in Fig. 9. A hypothetical cross-link is introduced at Lys-352 and Lys-514, for illustrative purposes and a relatively large loop of 162 residues would be formed by this connection. The reasons for choosing these particular residues are that they are on the correct trypsin subfragments and Asp-351 and Lys-514 are the only residues, thus far, which have been identified as being at the active site. The native polypeptide migrates in the expected position on SDS-PAGE according to its actual molecular weight of approximately 110,000 and evidently assumes the usual extended rod shape in SDS. The cross-link, and resulting loop, cause the polypeptide to exhibit an apparent molecular weight of SDS-PAGE. 125,000, According to this model, TI cleavage produces a bifurcated molecule with an apparent molecular weight of 135,000. The "arms" produced by the cleavage evidently increase the frictional resistance of the polypeptide in SDS. T z cleavage produces an unaltered Az fragment and a polypeptide with an apparent molecular weight of approximately 108,000 and actual molecular weight of 88,000. This deviation is the same as that found following T1 cleavage, and the structure shown in Fig. 9 has a distinct "Y" shape.
Formation of a loop by intramolecular cross-linking, of the size indicated in Fig. 9, is rather unlikely within a single protein domain. However, it is possible if cross-linking occurs between domains . MacLennan et al. (1) have recently proposed, from the predicted secondary structure of the enzyme, that the A1 and B fragments contain the separate phosphorylation-and nucleotide-binding domains, respectively. They are invisaged to consist of two parallel /3-domains, connected by a hinge region, similar to that of phosphoglycerate kinase and hexokinase. In this model ATP, bound to the nucleotidebinding domain in fragment B, would phosphorylate the aspartic acid residue in fragment AI in the adjoining phosphorylation domain. Our proposal that the alteration in apparent molecular weight is due to the formation of a rather large loop is consistent with the phosphorylation-and ATP bindingsites being located in separate domains. In addition, the effectiveness of short cross-linkers, such as formaldehyde, suggests that the cross-link occurs close to the proposed hinge region or another region where the domains have close contact. Significantly, binding of Ca2+, which promotes the Ez to El conformational transition, had no effect on cross-link formation (see Fig. 4). This suggests that active site geometry where the cross-link occurs, is not markedly affected by this transition, and it can be speculated that other factors must be involved in the changes in ligand specificity, since El can be phosphorylated by ATP and not by Pi, while E2 is phosphorylated by Pi and not by ATP (30).
The glutaraldehyde concentration dependence of crosslinking and the pH dependence of the reaction suggest that the majority of the ATPase molecules in SR vesicles behave as a single class with respect to the intramolecular cross-link. This is supported by the finding that, under optimum conditions in detergent, about 80% of the ATPase is cross-linked to E(125) (Figs. 2 and 8). The remainder evidently represents inactive ATPase. Significant amounts of inactive enzyme have been found using other procedures (17, 31). Detergent does not increase the reactivity of the participating residues. The amount of E(125) formed with detergent-solubilized and with intact vesicles is the same if the glutaraldehyde concentration and reaction times are similar (results not shown). This also indicates that detergent does not uncover ATPase that is resistant to the specific cross-link, further supporting the conclusion that the ATPase behaves as a homogeneous class. Consistently, solubilization does not uncover additional phosphorylation sites (17).
The intramolecular cross-link is inhibited by denaturation or inactivation of the enzyme, and hence formation of the cross-link provides a measure of the proportion of ATPases with an intact active site. It is interesting that TI and T, cleavage does not inhibit cross-link formation, and this agrees with the relative insensitivity of ATPase activity to tryptic digestion (32). We will show in a further paper that the inhibition of the cross-link by nucleotides, and by phosphorylation, provides a simple and sensitive assay for nucleotide and analogue binding and of occupancy of the active site during turnover and nonturnover condition^.^