Human cardiac troponin I: precise identification of antigenic epitopes and prediction of secondary structure

¹ Centre National de la Recherche Scientifique, UMR 9921, Faculté de Pharmacie, Ave. Charles Flahault, 34060 Montpellier Cedex 2, France.

² ERIA, Rue d'Italie, 69780 Mions, France.

³ Sanofi Recherche, Ave. du Professeur Blayac, 34000 Montpellier, France.

⁴ Sanofi Diagnostics Pasteur, Ave. Raymond Poincaré, 92230 Marnes-la-Coquette, France.
^a Author for correspondence. Fax 33 4 67 54 86 10; e-mail granier{at}pharma.univ-montp1.fr.

	Abstract

Top Abstract Introduction Materials and Methods Results Discussion References

 
The presence of human cardiac troponin I (hcTnI) in serum is considered to be a highly specific biochemical marker of acute myocardial infarction. To better understand the antigenic properties of hcTnI, a set of 68 overlapping peptides covering the complete amino acid sequence of hcTnI was prepared and used in epitope mapping experiments. All 16 anti-hcTnI monoclonal antibodies tested were found to recognize a peptide epitope, indicating that recognition by anti-hcTnI monoclonal antibodies was not dependent on the tertiary structure of the protein. Furthermore, the peptide reactivity with anti-hcTnI polyclonal antibodies indicated that most of the sequence of the protein was antigenic; in particular, the N- and C-terminal extremities were found to be the strongest antigenic regions. By using accurate secondary structure prediction methods, hcTnI was found to be an all-alpha type protein, with five regions predicted as helices. Matching the results of the epitope analysis with the structural prediction led us to the view that hcTnI is not a globular protein but probably adopts an extended conformation, allowing a large part of the amino acid sequence of this molecule to be recognized by the immune system. This improved knowledge of the antigenic and structural properties of hcTnI may help in developing new antibodies and immunoassays for use in diagnosing myocardial infarction.

	Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

 
Increased serum concentrations of biochemical markers of myocardial injury, together with chest pain and electrocardiographic changes, are considered to be convincing elements for acute myocardial injury (AMI)¹ diagnosis (1). The presence of human cardiac troponin I (hcTnI) in serum is now considered as one such highly specific biochemical marker (2)(3)(4)(5). This molecule can be detected in patients' sera by using selected monoclonal antibodies (mAbs) (6)(7). To achieve the required diagnostic specificity for myocardial infarction, the antibodies should distinguish the cardiac isoform from the skeletal isoforms of troponin I (8). Despite the clinical interest of mAbs directed against hcTnI, very little is known about the regions of the troponin molecule these antibodies bind to. An attempt to map peptide epitopes of anti-hcTnI mAbs was reported a few years ago (6). These results explained the cardiac specificity of mAb 11E12 because this antibody was found to recognize residues 27–36 (sequence RAYATEPHAK), part of the N-terminal stretch of 32 residues that is present only in the cardiac isoform of troponin I (9). A better knowledge of the antigenic properties of hcTnI might yield methods for developing antibodies with the desired characteristics of specificity and affinity for use in the clinical diagnosis of AMI. We, therefore, attempted to precisely identify the epitopes on the hcTnI recognized by a panel of 16 anti-hcTnI mAbs. Because it is commonly accepted that peptides 5 to 10 residues long correspond to the size of B-cell epitopes, methods of multiple peptide synthesis appear to be ideally suited for epitope mapping (10). The Spot method (11) was used here, both to prepare a large set of overlapping peptides spanning the complete 210-residue sequence of hcTnI and to evaluate their recognition by antibodies.

Because mAbs to globular proteins very often recognize assembled topographic epitopes (12) and much less frequently peptides, the analysis of antigenic epitopes recognized by anti-troponin I antibodies could also provide the necessary information to determine whether hcTnI is globular. In fact, the hydrodynamic behavior of bovine cardiac troponin I has been shown to be different from that expected for a typical globular protein (13); however, it was also postulated that bovine cardiac troponin I could be a compact globular protein with a well-defined hydrophobic core (14). The actual overall conformational type of cardiac troponin I is, therefore, not known. On the basis of our observations on the peptide reactivity of anti-hcTnI mAbs, we have attempted to clarify this issue. We also report here the detailed secondary structure of hcTnI, as predicted by a powerful algorithm (15), and analyze the relationship between epitope localization and secondary structure elements.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

 
spot multiple peptide synthesis
Sixty-eight overlapping decapeptides frameshifted by three residues representing the complete hcTnI protein sequence (9) were synthesized on a cellulose membrane (Abimed GmbH) by the Spot technique (11). The synthesis was performed several times, either manually (16) or by using an ASP 412 spotter (Abimed GmbH).

antibody assay
The set of membrane-bound peptides was probed by incubation with mAb (1 mg/L), rabbit polyclonal IgG (1 mg/L), or mouse polyclonal serum (diluted 1:250). The binding was revealed by alkaline phosphatase-conjugated anti-mouse or anti-rabbit antibody (Sigma Chemical Co.; diluted 1:1000), as described (11). The membrane was further treated so as to remove precipitated dye and bound antibodies and reused when necessary.

synthesis of biotinylated soluble peptides
Peptides i₁ (biotinyl-GSMMQALLGARAKE) and i₂ (biotinyl-GSNYRAYATEPHAK) corresponding, respectively, to residues 154–165 and 25–36 of the hcTnI sequence extended by a biotinylated two-residue spacer (Gly-Ser) were prepared by Fmoc solid-phase synthesis on a AMS 422 robot (Abimed GmbH) (17). After cleavage from the resin, the peptides were analyzed by analytical reversed-phase HPLC (18) (C₁₈ column; 15-min gradient from 5% to 60% acetonitrile in 1 mL/L aqueous trifluoroacetic acid; 1 mL/min) and found to be >90% pure.

monoclonal and polyclonal antibodies
The mAbs were derived from mice immunized with purified hcTnI. mAbs 10B11, 10F4, 8E10, 2A3, 5F1, and 7F4 were obtained from HyTest; the others correspond to those described previously (6). The specificity of these antibodies for hcTnI has been studied previously (3)(6)(19). Rabbit polyclonal IgG 212, 232, and 245 were prepared from the sera of three rabbits immunized with purified hcTnI. Mouse polyclonal antibodies E and G are pools of sera from three BALB/c and six Biozzi mice, respectively.

real-time analysis of the interaction between mAbs and peptides or hcTnI
Studies were performed at 25 °C by using a BIAcore apparatus (Biacore AB). Purified hcTnI protein (20 mg/L) in 10 mmol/L sodium acetate, pH 4.43, was immobilized on the flow cell of a CM5 sensor chip surface (20), activated with 100 mmol/L N-ethyl-N'-(3-dimethylaminopropyl) carbodiimide hydrochloride and 400 mmol/L N-hydroxysuccinimide. The amount of immobilized hcTnI for kinetic experiments was ~150–200 pg/mm. Soluble biotinylated peptides i₁ and i₂ (10 mg/L) in HEPES-buffered saline (HBS; Biacore AB), pH 7.4, were immobilized on flow cells of a streptavidin-coated sensor chip. Next, sensor chips coupled with peptides were saturated with biotin (50 mg/L) in HBS buffer. The amount of immobilized peptides was ~20 pg/mm. The binding kinetics of analytes (mAbs 11E12 and 8E1) to their immobilized reactants (peptides i₁, i₂, or hcTnI) were determined by injecting 30 mL of analytes (5–50 mg/L) in HBS buffer, pH 7.4, at a flow rate of 10 mL/min. Dissociation was observed in HBS without dissociating agent at 10 µL/min. The results are expressed as variation of resonance units as a function of time. The kinetic variables were measured by using BIAevaluation Software (21). The dissociation rate (off-rate) constants, k_d, were determined from a plot of ln(R_o/R) vs time (t), where R is the surface plasmon resonance signal at time t. The association rate (on-rate) constants, k_a, were determined from a plot of ln[abs(dR/dt)] vs time, and the dissociation constants, K_D, were determined from the ratio k_d/k_a.

prediction of secondary structure
The Rost and Sander (22) procedure, which combines sequence homology information and neural networks, was used to predict the secondary structure of hcTnI on the basis of the amino acid sequence of hcTnI (9) sent to the Predict Protein EMBL mail server (23). The method achieves >70% accuracy in identifying -helices and ß-sheets. For the sake of clarity, the detailed analysis is not provided (available on request), and only the final prediction is given herein.

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

 
identification of peptide epitopes recognized by anti-hcTnI mAbs
The amino acid sequence of hcTnI (9) was presented in the form of an array of 68 overlapping decapeptides (7- residue overlap) synthesized on a cellulose membrane (11)(16). The 16 different anti-hcTnI mAbs were evaluated individually for their capacity to react with this set of immobilized peptides. As an example, the recognition pattern observed for two different mAbs is shown in Fig. 1 . Antibody 10B11 bound to spots 5 and 6 and not to the preceding or following spots (Fig. 1A ). Therefore, the epitope of mAb 10B11 could be mapped to residues PAPIRRR, which are common to the two reactive peptides (Table 1 ). Similarly, mAb 11E12 recognized three overlapping peptides (spots 9, 10, 11; Fig. 1B ), the common sequence of which was TEPH (Table 1 ). A faint but reproducible reactivity of mAb 11E12 with spot 38 (Fig. 1C ) was observed; it was attributed to the presence in this peptide (sequence RYDIEAKVTK) of a Y- -E-K sequence motif similar to the motif Y- -E-H present in the most reactive peptides, i.e., peptides 9 and 10 (Table 1 ). All 16 mAbs tested were found to bind to at least one peptide from the membrane, indicating that, despite the fact that hybridomas were obtained through an immunization-selection process by using the hcTnI protein, the epitope recognition of the resulting antibodies was not dependent on the tertiary structure of the protein. No binding was observed in the absence of any mAb (Fig. 1C ).

View larger version (56K): [in this window] [in a new window]   Figure 1. Identification by the Spot method of peptide epitopes recognized by anti-hcTnI mAbs 10B11 and 11E12. The reactive peptide sequences are indicated in Table 1 .

Table 1 lists the epitopes that were thus identified by using the complete panel of anti-hcTnI mAbs. These epitopes were distributed all along the sequence of the protein. At the N terminus, residues 16–22 were recognized by a single mAb (10B11), whereas amino acids NYRAYATEPHAKK (residues 25 to 37) were recognized by six different mAbs (3B9, 8D5, 3B8, 3C6, 11E12, and 10F4) with identical or similar specificities. In the central part of the protein, the sequence LGFAELQ (residues 88–94) was recognized by mAbs 8E10, 2A3, and 5F1, and the sequence ADAMMQALLG (residues 151–160) was recognized by mAbs 2E6, 8G2, 7D1, and 8E1. In the C-terminal part of hcTnI, residues 190–196 were mapped by antibodies 10F2 and 7F4, which exhibited identical specificity. Most of the epitopes recognized by anti-hcTnI included one or several charged residues (e.g., three Arg for 10B11 and one His and two Lys for 10F4); the only exception was mAb 8E1, which recognized the neutral and quite apolar sequence ALLG.

Peptides identified as epitopes by the Spot method might in fact represent only a part of the epitope that the mAb actually recognizes at the surface of the protein. To verify this possibility, the binding characteristics of two model anti-hcTnI mAbs (8E1 and 11E12) to hcTnI and to synthetic peptide epitopes were assessed by using BIAcore methodology (20). Each mAb bound specifically to its cognate synthetic peptide epitope with characteristic association–dissociation curves (Fig. 2 , A and B). The equilibrium affinity constants derived from the binding of these mAbs to the peptides compared very well with those obtained from their interaction with the hcTnI molecule (Fig. 2C ). Therefore, at least in these two cases, the peptide identified by the Spot technique as an epitope is a good mimic of the antigenic epitope recognized by the mAb on the protein.

View larger version (20K): [in this window] [in a new window]   Figure 2. Real-time interaction analysis of mAbs 8E1 and 11E12 with synthetic peptides and hcTnI. RU, resonance unit.

antigenic regions defined by reactivity with polyclonal anti-hcTnI antibodies
It was important to verify whether the epitope localization obtained with the individual mAbs reflects the overall antigenic structure of the protein. Two mouse and three rabbit polyclonal anti-hcTnI were used to identify those parts of the protein that are antigenic. The results are shown in Fig. 3 . Peptides from the N-terminal sequence (peptides 1–10) and from the C-terminal sequence of hcTnI (peptides 62–66) were most frequently recognized by polyclonal antibodies (e.g., peptides 2, 3, 9, 63, and 64). This indicates that the extremities of the protein correspond to the strongest antigenic regions. The central region of the sequence (represented by peptides 28–51) was recognized less frequently. Some peptides of the hcTnI sequence were never bound by polyclonal antibodies, possibly indicating that the corresponding protein regions were weakly or not antigenic. The epitopes identified by the anti-hcTnI mAbs (Fig. 3 ) always corresponded to peptides that are part of the antigenic regions identified on the basis of peptide reactivity with polyclonal antibodies; this shows that the mAbs can be considered as representative elements of the polyclonal response against this molecule.

View larger version (34K): [in this window] [in a new window]   Figure 3. The frequency of the antigenic response given by the number of anti-hcTnI rabbit IgG and mouse sera that reacted with each decapeptide from the sequence of hcTnI. The reactivity of the three polyclonal rabbit IgGs (212, 245, and 232) and the two mouse sera (G and E) with the set of 68 overlapping peptides was assessed as described in Materials and Methods. Horizontal lines, peptides recognized by each mAb.

relationship between predicted secondary structure of hcTnI and epitope positions
The method of Rost and Sander (22) was used to predict the secondary structural features of hcTnI from its amino acid sequence. The results are shown in Fig. 4 . The distinctive features are: (a) the existence of a large unordered N-terminal region (residues 1–41) enclosing the 32 N-terminal residues of the protein, which are specific for the cardiac isoform; (b) the existence of five regions predicted to be in -helical conformation. Although there remain some uncertainties as to the exact residue at which helices start and stop, the helices were tentatively assigned as: helix A, residues 42–77; helix B, residues 90–135; helix C, residues 145–159; helix D, residues 168–176; and helix E, residues 182–197; and (c) the helical regions are separated from each other by short unordered sequences. Altogether, 63.3% of the residues were predicted to be in an -helical conformation, and no ß-strand regions were predicted. On the basis of this information, hcTnI can be said to be an all-alpha type protein (23). The position of antigenic epitopes recognized by mAbs on the molecule is also given in Fig. 4 . The unordered structure of the first 42 residues is recognized by several mAbs, which is consistent with the general observation that the extremities of proteins are particularly antigenic. Furthermore, the fact that several differences between the human sequence and the rabbit and mouse cTnI sequences are clustered in the N-terminal part probably contributes to the immunogenicity of this region. Other unordered regions connecting helices A and B, B and C, C and D, and D and E were not recognized by the mAbs; this is different from what is known for globular proteins, for which antigenicity is often associated with loop regions connecting elements of regular secondary structure. Helices B, C, and E enclosed one or several antigenic epitopes, indicating that they are exposed to the solvent; no epitopes were associated with the large helix A and the short helix D.

View larger version (56K): [in this window] [in a new window]   Figure 4. A scheme of the predicted secondary features of hcTnI and of the position of antigenic epitopes recognized by mAbs on the protein. Predicted helices are represented as grey cylinders, and loop regions are shown as freely drawn lines. The boxed parts of the sequence indicate epitopes recognized by the mAbs.

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

 
Our results provide both a complete description of the antigenic structure and a model of the structural organization of hcTnI, a protein that is considered of great value for diagnosing AMI. The polyclonal antibodies and all of the 16 mAbs raised against the entire molecule were found to react with the nested set of synthetic peptides we prepared. Although all mAbs were reactive with short peptides, all of them also recognized the hcTnI protein in human sera (3)(19), indicating that they do not bind to denatured forms of the protein. The location of the epitopes recognized by mAbs coincides with the antigenic regions defined by the use of polyclonal antibodies, indicating that no bias in the selection procedure of the mAbs was introduced.

The two main antigenic regions were found to be at the N and C termini of the hcTnI sequence. This observation is of interest for the improvement of immunoassays for hcTnI; the antigenicity of the N-terminal, cardiac-specific part of hcTnI could advantageously be used by immunizing mice with N-terminal peptides or by selecting hydridomas on the basis of their reactivity with such peptides. mAbs recognizing the 32 N-terminal residues region should be specific for the cardiac isoform, which only exhibits this sequence. We identified mAbs 10B11, 3B9, 8D5, 3B8, and 3C6 as possessing this kind of epitope specificity. Therefore, these mAbs could possibly constitute useful tools in the immunodetection of the cardiac isoform of TnI for the clinical diagnosis of AMI, as is mAb 11E12 (3).

Our results not only corroborate those published previously concerning the epitope localization of the mAb 11E12 (6) but provide detailed information on the epitopes of many other mAbs. Remarkably, the precise analysis of the kinetics of interaction of mAbs 8E1 and 11E12 with synthetic replicas of the peptides identified by the Spot method indicated that they bound these short peptides with an affinity comparable with that for the entire hcTnI molecule. Thus, the epitope recognized by 8E1 and 11E12 at the surface of the protein (the "structural" epitope) probably does not enclose other hcTnI residues and is almost perfectly mimicked by the "functional" epitope (24) defined by peptide analysis. Consistent with the highly polar nature of the protein, all but one of the antibodies recognized hydrophilic, electrically charged sequences of hcTnI. The exquisite specificity of anti-hcTnI mAbs can be appreciated by the observation that, for example, mAb 10F4, which recognized the peptide TEPHAKKKSK, does not cross-react with peptide MEGRKKKFES, which encloses a similar stretch of three lysines (underlined residues). Furthermore, mAbs 8E10 and 2A3 bound peptide AELQDLCRQL but not peptide QIAKQELERE, which shares a highly similar motif (underlined residues). Only in the case of mAb 11E12 could a faint cross-reactivity be observed, which we attributed to partial sequence homology.

The secondary structure prediction that was performed by using one of the most accurate prediction methods (15) shows that hcTnI is an all-alpha type molecule. Several implications of the epitope analysis we have performed argue against hcTnI being globular, at least in the uncomplexed form that was used for immunization: (a) all 16 anti-hcTnI mAbs recognized a continuous epitope markedly; this contrasts with the observation that most mAbs to globular proteins show a conformation-dependent recognition (12) and, therefore, do not react with short peptides; (b) several helices of hcTnI are antigenic, which means that they are probably exposed to the solvent and not packed together as in globular proteins; and (c) we observed that regions connecting the predicted helices of hcTnI are not antigenic; this is in sharp contrast with the fact that turn regions in globular proteins are generally antigenic. Taken together, our results lead to the view that hcTnI is probably in an extended conformation, allowing most of the amino acid sequence of this protein to be recognized by the immune system. Biophysical measurements have shown that when complexed to human cardiac troponin C, a form that is likely released during AMI (19), hcTnI also appears to be in an extremely extended conformation (25).

In conclusion, the data presented here suggest that the improved knowledge of the antigenic and structural properties of hcTnI resulting from this study might yield methods for developing new antibodies and immunoassays for use in the clinical diagnosis of myocardial infarction.

	Acknowledgments

 
We are greatly indebted to Sharon Lynn Salhi for assistance with the manuscript, and we acknowledge the technical help of Pierrette Monmouton. This work was supported by a special grant from ELF-Aquitaine. G.F. was a recipient of a bursary from the Ministère de la Recherche et de la Technologie.

	Footnotes

 
¹ Nonstandard abbreviations: AMI, acute myocardial injury; hcTnI, human cardiac troponin I; and HBS, HEPES-buffered saline.

	References

Top Abstract Introduction Materials and Methods Results Discussion References

 

1. Chaitman BR, Jaffe AS. What is the true periprocedure myocardial infarction rate? Does anyone know for sure?. The need for clarification. Circulation 1995;91:1609-1610.
2. Adams JE, Bodor GS, Davila-Roman VG, Delmez JA, Apple FS, Ladenson JH, Jaffe AS. Cardiac troponin I. A marker with high specificity for cardiac injury. Circulation 1993;88:101-106. [Abstract/Free Full Text]
3. Larue C, Calzolari C, Bertinchand JP, Leclercq F, Grolleau R, Pau B. Cardiac-specific immunoenzymometric assay of troponin I in the early phase of acute myocardial infarction. Clin Chem 1993;39:972-979. [Abstract/Free Full Text]
4. Bodor GS. Cardiac troponin I: a highly specific biochemical marker for myocardial infarction. J Clin Immunoassay 1994;17:40-44.
5. Jaffe A, Landt Y, Parvin CA, Abendschein DR, Geltman EM, Ladenson JH. Comparative sensitivity of cardiac troponin I and lactate dehydrogenase isoenzymes for diagnosing acute myocardial infarction. Clin Chem 1996;42:1770-1776. [Abstract/Free Full Text]
6. Larue C, Defacque-Lacquement H, Calzolari C, LeNguyen D, Pau B. New monoclonal antibodies as probes for human cardiac troponin I: epitopic analysis with synthetic peptides. Mol Immunol 1992;29:271-278. [ISI][Medline] [Order article via Infotrieve]
7. Bodor GS, Porter S, Landt Y, Ladenson GH. Development of monoclonal antibodies for an assay of cardiac troponin I and preliminary results in suspected cases of myocardial infarction. Clin Chem 1992;38:2203-2214. [Abstract/Free Full Text]
8. Cummins P, Perry V. Troponin I from human skeletal and cardiac muscles. Biochem J 1978;171:251-259. [ISI][Medline] [Order article via Infotrieve]
9. Vallins WJ, Brand NJ, Dabhade N, Butler-Browne G, Yacoub MH, Barton PJR. Molecular cloning of human cardiac troponin I using polymerase chain reaction. FEBS Lett 1990;170:57-61.
10. Geysen MH, Rodda SJ, Mason TJ, Tribbick G, Schoofs PG. Strategies for epitope analysis using peptide synthesis. J Immunol Methods 1997;102:259-274.
11. Frank R. Spot-Synthesis: an easy technique for the positionally addressable, parallel chemical synthesis on a membrane support. Tetrahedron 1992;48:9217-9232. [ISI]
12. Benjamin DC, Berzofsky JA, East IJ, Gurd FRN, Hannum C, Leach SJ, et al. The antigenic structure of proteins: a reappraisal. Annu Rev Immunol 1984;2:67-101. [ISI][Medline] [Order article via Infotrieve]
13. Byers DM, Kay CM. Hydrodynamic properties of bovine cardiac troponin-I and troponin-T. J Biol Chem 1983;258:2951-2954. [Abstract/Free Full Text]
14. Leszyk J, Dumaswala R, Potter R, Collins JH. Amino acid sequence of bovine cardiac troponin I. Biochemistry 1988;27:2821-2827. [Medline] [Order article via Infotrieve]
15. Rost B, Sander C. Prediction of protein structure at better than 70% accuracy. J Mol Biol 1993;232:584-599. [ISI][Medline] [Order article via Infotrieve]
16. Molina F, Laune D, Gougat C, Pau B, Granier C. Improved performances of Spot multiple peptide synthesis. Peptide Res 1996;9:151-155.
17. Gausepohl H, Boulin C, Kraft M, Franck RW. Automated multiple peptide synthesis. Peptide Res 1992;5:315-320.
18. Nirenberg U. Reversed-phase HPLC. In: Dunn BM, Pennington MW, eds. Methods in molecular biology (Peptide analysis protocols, Vol. 36). Totowa, NJ: Humana Press, 1992:23–35..
19. Kathrukha AG, Bereznikova AV, Esakova TV, Petterson K, Lövgren T, Severina ME, et al. Troponin I is released in bloodstream of patients with acute myocardial infarction not in free form but as a complex. Clin Chem 1997;43:1379-1385. [Abstract/Free Full Text]
20. Malmqvist M. Biospecific interaction analysis using biosensor technology. Nature 1993;361:186-187. [Medline] [Order article via Infotrieve]
21. Karlsson R, Roos H, Fägerstam P, Persson B. Kinetic and concentration analysis using BIA technology. Methods (Orlando) 1994;6:99-110.
22. Rost B, Sander C. Combining evolutionary information and neural networks to predict protein secondary structure. Proteins 1994;19:55-72. [ISI][Medline] [Order article via Infotrieve]
23. Rost B. PHD: predicting one-dimensional protein structure by profile based neural networks. Methods Enzymol 1996;266:525-539. [ISI][Medline] [Order article via Infotrieve]
24. Smith-Gill SJ. Protein epitopes: functional vs. structural definitions. Res Immunol 1994;145:67-70. [ISI][Medline] [Order article via Infotrieve]
25. Olah GA, Trewhella JA. A model structure of the muscle protein complex 4 Ca²⁺ troponin C-troponin I derived from small angle scattering data: implications for regulation. Biochemistry 1994;33:12800-12806. [Medline] [Order article via Infotrieve]