doi:10.1016/j.jmb.2011.09.026 J. Mol. Biol. (2011) 413, 908–913

Contents lists available at www.sciencedirect.com

Journal of Molecular Biologyj ourna l homepage: ht tp : / /ees .e lsev ie r.com. jmb

COMMUNICATION

The C0C1 Fragment of Human Cardiac Myosin BindingProtein C Has Common Binding Determinants for BothActin and Myosin

Yanling Lu, Ann H. Kwan, Jill Trewhella and Cy M. Jeffries⁎School of Molecular Bioscience, Building G08, University of Sydney, NSW 2006, Australia

Received 28 July 2011;received in revised form9 September 2011;accepted 14 September 2011Available online28 September 2011

Edited by M. Moody

Keywords:MyBP-C;C-protein;small-angle neutronscattering;heart muscle;NMR titration

*Corresponding author. E-mail addcy.jeffries@sydney.edu.au.Abbreviations used: SANS, small

scattering; HSQC, heteronuclear sincoherence; RLC, regulatory light chbinding protein C.

0022-2836/$ - see front matter © 2011 E

TheN-terminal domains of cardiacmyosin binding protein C (MyBP-C) playa regulatory role in modulating interactions between myosin and actinduring heart muscle contraction. Using NMR spectroscopy and small-angleneutron scattering, we have determined specific details of the interactionbetween the two-module human C0C1 cMyBP-C fragment and F-actin. Thesmall-angle neutron scattering data show that C0C1 spontaneouslypolymerizes monomeric actin (G-actin) to form regular assemblies com-posed of filamentous actin (F-actin) cores decorated byC0C1, similar towhatwas reported in our earlier four-module mouse cMyBP-C actin study. Inaddition, NMR titration analyses show large intensity changes for a subset ofC0C1 peaks upon addition of G-actin, indicating that human C0C1 interactsspecifically with actin and promotes its assembly into filaments. During theNMR titration, peaks corresponding to cardiac-specific C0 domain are thefirst to be affected, followed by those from the C1 domain. No peak intensityor position changes were detected for peaks arising from the disorderedproline/alanine-rich (P/A) linker connecting C0 with C1, despite previoussuggestions of its involvement in binding actin. Of considerable interest isthe observation that the actin-interaction “hot-spots” within the C0 and C1domains, revealed in our NMR study, overlap with regions previouslyidentified as binding to the regulatory light chain of myosin and to myosinΔS2. Our results suggest that C0 andC1 interact withmyosin and actin usinga common set of binding determinants and therefore support a cMyBP-Cswitching mechanism between myosin and actin.

© 2011 Elsevier Ltd. All rights reserved.

Introduction

The thick filament accessory protein myosinbinding protein C (MyBP-C) contributes to bothstructural and regulatory roles during the cycle of

ress:

-angle neutrongle quantumain; MyBP-C, myosin

lsevier Ltd. All rights reserve

muscle contraction and relaxation.1,2 The cardiacisoform of the protein (cMyBP-C) consists of eightimmunoglobulin (Ig)-type and three fibronectin(Fn)-like domains, designated C0 through C10from the N-terminus. The so-called regulatorydomains (C0C2) include a cardiac-specific Ig domain(C0) and a “motif” (or m-domain) between C1 andC2 that undergoes phosphorylation in response toinotropic stimuli3 to affect cardiac output and aproline/alanine-rich linker connecting C0 and C1that has been proposed to contain a possible actinbinding region.4 The precise details as to how this N-terminal region of cMyBP-C performs its regulatory

d.

909The C0C1 Fragment of Human MyBP-C

role in muscle action remain a topic of debate dueto the different isoforms of cMyBP-C being used indifferent studies and the capacity of differentregions to bind to elements of myosin or actin withdiverse binding affinities (millimolar to micromolarranges5–10). For example, with respect to bindingactin, small-angle neutron scattering (SANS)11 andelectron microscopy and tomography inves-tigations12–15 indicate that the N-terminal regionscan uniformly decorate F-actin “in vitro” or canextend across the interfilament space betweenmyosin thick and actin thin filaments in intactmuscle sarcomeres,13 suggesting that specific in-teractions do exist between cMyBP-C and F-actin.Conversely, co-sedimentation analyses have beeninterpreted as indicating that the interaction of thefirst two N-terminal domains (C0C1) with actin isvery weak (∼40 μM)10 and nonspecific9,10 and thatan interaction of the N-terminal domains with actinrequires the inclusion of the m-domain.10 It also hasbeen suggested that the C-terminal end of theprotein (C5C10) is primarily responsible for actininteractions.9

The canonical myosin-tether mechanism6,7,16,17 iscurrently hypothesized to be the key regulatorymechanism of cMyBP-C in modulating contraction.In this mechanism, phosphorylation of the m-domain detaches the N-terminal domains ofcMyBP-C from myosin ΔS2,7 a region of myosinproximal to the myosin head, enabling myosinheads to reposition and attach to thin filamentactin. Kulikovskaya et al. have shown that thephosphorylation state of the N-terminal domainsmodulates the distribution of states bound to eithermyosin or actin, suggesting that the N-terminaldomains could “switch” between myosin and actinto regulate cardiac contractility. 18 Furthermore,there is increasing evidence that the N-terminaldomains of cMyBP-C affect thin filamentactivation,19–23 suggesting that the modulation ofcontraction by the N-terminal domains of cMyBP-Cextends beyond a simple tether mechanism tomyosin ΔS2 and involves actin thin filaments.Here, we use SANS and NMR to determine detailsof a specific interaction between human C0C1 andactin that includes the identification of regionswithin C0 and C1 likely to be responsible formediating C0C1/F-actin interactions.

The addition of C0C1 to G-actin produces aregular C0C1/F-actin assembly

SANS with deuterium (2H) labeling and contrastvariation was used to characterize the organizationof a higher-order human C0C1/actin assembly.Similar to previous SANS investigations of mouse[2H]C0C2 and actin,11 the addition of the human[2H]C0C1 fragment to a solution of monomericG-actin resulted in the formation of soluble filamen-tous macromolecular assemblies comprising of F-actin cores decorated in a symmetric fashion by [2H]C0C1. Figure 1 displays the two-dimensional crosssection of human C0C1/F-actin derived from theSANS data (Supplementary Fig. 1a) showing thatC0C1 binds on either side of F-actin and portions ofits structure can extend into solution away from thelong axis of the assembly; cross-sectional P(r)analysis indicates the C0C1 extends 40–50 Å in thisdirection (Supplementary Fig. 1c). Based on themass distribution across the assembly, our scatter-ing results indicate that the F-actin is almostsaturated by C0C1 binding in an ∼1:1 stoichiometry[Mr

c, Supplementary Table 1; the average Mrc for the

assembly is ∼20% lower that what would beexpected for a completely saturated C0C1/F-actinfilament (20.5 × 10 10 g mol − 1 cm − 1 versus25.9×1010 g mol−1 cm−1)].

NMR titration study shows C0C1 from humancMyBP-C binds actin specifically

In order to further characterize the interactionbetween C0C1 and actin, we carried out NMRtitration studies using 15N-labeled C0C1 and unla-beled G-actin. The 15N–1H heteronuclear singlequantum coherence (HSQC) NMR spectrum of[15N]C0C1 is of high quality (Fig. 2a and Supple-mentary Fig. 3) and shows mainly sharp and well-dispersed peaks. These peaks overlay well withassignments reported for the isolated C0 and C1domains8,24 and indicate that the C0 and C1 Igmodules in C0C1 retain the same fold as theindividual domains. Chemical shift assignments ofhuman C0C1 (Supplementary Table 3) were madeusing standard backbone triple-resonance experi-ments in combination with previously reportedassignments.8,24 Overall, 83%, 15% and 96% of

Fig. 1. Two-dimensional cross-sectional model of the assemblythat forms between human C0C1(red) and F-actin (blue) derivedfrom SANS with contrast variationdata. For details, see SupplementaryInformation.

Fig. 2. C0C1 NMR actin titration spectra and relative chemical shift peak height changes. (a) Overlay of 15N-1H-HSQCspectra of human C0C1 before (red) and after (cyan) addition of G-actin at a [15N]C0C1:G-actin molar ratio of 1:40. (b)Chemical shift changes in representative residues as observed during titration with G-actin at 1:40, 1:20, 1:10 and 1:5(molar ratios) are shown (apo, unbound C0C1). A116 and A149 are residues within the P/A linker. V68 and K185 areexamples of the “least perturbed” residues in C0 and C1, respectively. V81 and Y237 are examples of the “mostperturbed” residues in C0 and C1, respectively. (c) Graph showing relative changes in chemical shift peak height forassigned [15N]C0C1 amino acids plotted against the amino acid sequence of C0C1 on the addition of unlabeled G-actin to1:40 (cyan) or 1:5 (green) G-actin:C0C1 molar ratios relative to unbound [15N]C0C1 (100% line) in G-actin buffer. Fordetails, see Supplementary Information.

910 The C0C1 Fragment of Human MyBP-C

amide chemical shifts of residues in the C0 domain,the P/A linker connecting C0 and C1 and the C1domain could be identified, respectively.While a more complete backbone assignment of

the P/A linker was hampered by severe spectraloverlap and the repetitive amino acid sequence inthe linker region, we were able to identify shortsequence fragments that allowed the unambiguousidentification of a subset of peaks in the 15N-1H-

HSQC as arising from P/A linker residues (Fig. 2and Supplementary Table 3). Peaks correspondingto the P/A linker are exclusively located within theregion typically associated with “disordered” resi-dues (between 8 and 9 ppm for 1H). Indeed, 15N-relaxation measurements performed on [15N]C0C1demonstrated that the residues corresponding tothe P/A linker are highly flexible in solution (datanot shown).

911The C0C1 Fragment of Human MyBP-C

Upon titration of unlabeled G-actin into [15N]C0C1, many peaks in the 15N–1H-HSQC NMRspectrum broaden significantly and experience asevere drop in signal intensity, with some peaksdisappearing all together (Fig. 2). We postulate thatthe signal broadening is ultimately a result of theformation of large C0C1/F-actin assemblies (asobserved during SANS with contrast variation)that experience much slower molecular tumblingthan C0C1 alone, although a contribution fromexchange broadening may also be present. Consis-tent with filament formation, the solution becameincreasingly viscous as G-actin was added to [15N]C0C1, indicative of filament formation.

IdentificationofC0andC1domain actin interactionhot-spots

Significantly, the observed signal broadening ishighly selective. While peaks corresponding to boththe C0 and C1 domains experience significant dropsin intensities, peaks in the disordered region of the

spectrum, of which a significant proportion isassigned to the P/A linker, are unaffected even atthe highest concentration of actin used (Fig. 2). Thelack of signal broadening experienced by these P/Alinker residues suggests that the P/A linker un-dergoes much faster motions than the overalltumbling of the C0C1/actin complex as a whole.Taken together, these results suggest that the P/Alinker remains largely unstructured upon thebinding of C0C1 to actin and is not directly involvedin the actin/C0C1 interaction.A close inspection of the 15N-1H-HSQC spectra at

the earlier titration points reveals that there is amarked difference in how quickly signals disappearin C0 compared to C1 (Fig. 2c). In particular,residues in the C0 domain are more affected thanthose in the C1 domain; for example, with the firstactin addition (1:40 molar ratio of actin:C0C1), weobserved an average decrease of 75±6% in peakheight for C0 residues but only a decrease of 45±9%for C1 residues. The C0 domain is likely to havehigher affinity toward actin compared to the C1

Fig. 3. Key actin and myosininteraction regions within C0 andC1. (a) Themost and least perturbedamino acid chemical shifts (redspheres and yellow spheres, respec-tively) observed on titrating actininto [15N]C0C1 (1:40 molar ratio)form regions that cluster on oppo-site sides of the Ig folds of either theC0 or C1 domain (left and right.)The most and least perturbed re-gions are defined by those aminoacids in each individual domaindisplaying peak height changes ofmore than 1 SD “above” or “below”the average domain decrease inpeak height intensity, respectively(derived from Fig. 2c.). (b) Overlapof the most perturbed amino acidswithin C0 or C1 during the course ofthe actin/C0C1 titration (redspheres) with regions of C0 and C1that have been previously identifiedas interactingwith themyosin RLC8

(C0, cyan mesh) or myosin ΔS224

(C1, brown mesh). The SSKVKsequence (red mesh) in C0 formspart of the interface with themyosinRLC (cyan mesh) and has beenpreviously hypothesized as a po-tential actin binding site.4,25 Thestructures of C0 and C1 are derivedfrom Protein Data Bank codes2K1M and 2AVG, and molecularvisualization was performed inPyMOL.26

912 The C0C1 Fragment of Human MyBP-C

domain, as lower actin concentrations are able toinduce a similar level of peak height reduction.While it was not possible to observe peaks in the

15N-1H-HSQC corresponding to the bound form ofC0C1,we are nonetheless confident that the Ig fold inboth C0 and C1 was essentially maintained uponbinding to actin. The Ig fold is generally robust,consisting of a β-sandwich composed of two β-sheets,8,24 with its hydrophobic core stabilized by atryptophan residue. Recent electron microscopyreconstructions have shown that when bound to F-actin, the dimensions of both C0 and C1 areconsistent with the dimensions of Ig-like domains.15

It is unlikely that large conformational changes (e.g.,unfolding) have occurredwithin C0 and C1 domainswhen they are bound to F-actin, and therefore, thepeak intensity changes observed in NMR experi-ments for C0C1 upon actin addition are likely toresult from direct interactions of the correspondingresidues with actin. Consequently, when we exam-ine the C0 and C1 data independently and map the“most” and the “least” perturbed regions of thedomains onto their previously determinedstructures,8,24 the most and least perturbed regionswithin either domain during the actin titrationlocalize on opposite sides of each of the domains'Ig fold (Fig. 3a). Of the unambiguously assignedresidues that havewell-isolated peaks in the 15N-1H-HSQC spectrum, the following C0 residues areperturbed the most: A31, W42, I49, G77, A80, V81,A83 and F90; while in the C1 domain, these areW191, V197, D198, K202, Q230, Y237 and S242.

Summary

The SANS data show that the addition of humanC0C1 promotes the formation of soluble filamentousmacromolecular assemblies, which are composed ofF-actin cores that are regularly decorated by C0C1,with a similar symmetric arrangement as previouslyobserved for mouse-C0C2 and actin.11 The NMRdata demonstrate that specific regions on C0C1 areimplicated in the actin interaction. Based onsequence homology with the actin-binding N-terminal proline-rich extension of the myosinessential light chain, Squire et al. have proposedthat a basic amino acid sequence SSKVK proximal tothe proline/alanine-rich segment of C0C1 could be aputative actin binding site.4,25 Recently, the solutionstructure of human C0 domain has been solved byNMR,8 and it is apparent that the SSKVK sequencecomprises the C-terminal end of the last β-strand ofthe C0 domain and forms part of the Ig fold.Interestingly, the SSKVK sequence overlaps withpart of the region within C0 that we have identifiedas undergoing significant peak perturbations on theaddition of actin, confirming that portions of thestructure around these C0 domain residues areinvolved in the C0C1/actin interaction (Fig. 3b).

Ratti et al. have also recently reported that thehuman C0 domain interacts with the regulatorylight chain (RLC) of myosin and that amino acidsincluding the SSKVK region are responsible for theC0/RLC interaction.8 Our results, combined withthese observations, suggest that there is a commonset of amino acid binding determinants that com-prise an interaction interface in C0 responsible formediating contacts with either actin or the myosinRLC (Fig. 3b). Quite interestingly, a similar patternis also seen on comparing amino acid residuespreviously identified in C1 that interact with myosinΔS224 and the actin interaction region we haveidentified in C1: some of the most perturbed actinbinding regions in C1 also form part of the myosinΔS2 binding interface (Fig. 3b).Our results support the existence of a specific

switching mechanism between actin and myosin bythe N-terminal domains of cMyBP-C. It seemsreasonable to assume that when C0C2 is tetheredto myosin ΔS2 in an unphosphorylated state, the C0domain is brought into close proximity and binds tothe RLC so that, in effect, the C0, C1, m- and C2domains are “bound up” with myosin. Subsequentphosphorylation of the motif between C1 and C2,which releases the myosin tether,6,7,16,18 may allowthe higher-affinity C0 domain to extend at the end ofthe P/A linker to establish contacts with actinfollowed by the lower-affinity C1 domain. Havingcommon myosin and actin interaction interfaces onC0C1 could ensure that the timing of the attachmentand detachment of the N-terminal domains ofcMyBP-C between myosin and actin are coordinatedwith phosphorylation. Indeed, Kulikovskaya et al.have observed that the phosphorylation state of theN-terminal domains of cMyBP-C alters the distribu-tion of binding of the domains between myosin andactin in skinned muscle fibres.18 The realization thatthe actin and myosin binding determinants on C0C1overlap contributes to the developing molecularpicture of specific interactions underlying a cMyBP-C mediated switch between the thick and thinfilaments of heart muscle sarcomeres.

Acknowledgements

We thank Ben Crossett for performing massspectrometry using the Australian Proteome Analy-sis Facility established under the Australian Govern-ment's Major National Facilities program. This workwas supported by the U.S. Department of EnergyGrant DE-FG02-05ER64026 (to J.T.). The use ofNeutron facilities were supported by NationalScience Foundation Grant DMR-0454672. Weacknowledge the support of the National Instituteof Standards and Technology, U.S. Department of

913The C0C1 Fragment of Human MyBP-C

Commerce and Boualem Hammouda for assistancein using these facilities.

Supplementary Data

Supplementary data associated with this articlecan be found, in the online version, at doi:10.1016/j.jmb.2011.09.026

References

1. Oakley, C. E., Chamoun, J., Brown, L. J. &Hambly, B.D.(2007). Myosin binding protein-C: enigmatic regulatorof cardiac contraction. Int. J. Biochem. Cell Biol. 39,2161–2166.

2. Oakley, C. E., Hambly, B. D., Curmi, P. M. & Brown,L. J. (2004). Myosin binding protein C: structuralabnormalities in familial hypertrophic cardiomyopa-thy. Cell Res. 14, 95–110.

3. Hartzell, H. C. & Titus, L. (1982). Effects of cholinergicand adrenergic agonists on phosphorylation of a165,000-dalton myofibrillar protein in intact cardiacmuscle. J. Biol. Chem. 257, 2111–2120.

4. Squire, J. M., Luther, P. K. & Knupp, C. (2003).Structural evidence for the interaction of C-protein(MyBP-C) with actin and sequence identification of apossible actin-binding domain. J. Mol. Biol. 331,713–724.

5. Ababou, A., Gautel, M. & Pfuhl, M. (2007). Dissect-ing the N-terminal myosin binding site of humancardiac myosin-binding protein C. Structure andmyosin binding of domain C2. J. Biol. Chem. 282,9204–9215.

6. Gruen, M. & Gautel, M. (1999). Mutations in beta-myosin S2 that cause familial hypertrophic cardiomy-opathy (FHC) abolish the interaction with the regula-tory domain of myosin-binding protein-C. J. Mol. Biol.286, 933–949.

7. Gruen, M., Prinz, H. & Gautel, M. (1999). cAPK-phosphorylation controls the interaction of the regu-latory domain of cardiac myosin binding protein Cwith myosin-S2 in an on–off fashion. FEBS Lett. 453,254–259.

8. Ratti, J., Rostkova, E., Gautel, M. & Pfuhl, M. (2011).Structure and interactions of myosin-binding proteinC domain C0: cardiac-specific regulation of myosin atits neck? J. Biol. Chem. 286, 12650–12658.

9. Rybakova, I. N., Greaser, M. L. & Moss, R. L. (2011).Myosin binding protein C interaction with actin:characterization and mapping of the binding site. J.Biol. Chem. 286, 2008–2016.

10. Shaffer, J. F., Kensler, R. W. & Harris, S. P. (2009). Themyosin-binding protein C motif binds to F-actin in aphosphorylation-sensitive manner. J. Biol. Chem. 284,12318–12327.

11. Whitten, A. E., Jeffries, C. M., Harris, S. P. &Trewhella, J. (2008). Cardiac myosin-binding proteinC decorates F-actin: implications for cardiac function.Proc. Natl Acad. Sci. USA, 105, 18360–18365.

12. Mun, J. Y., Gulick, J., Robbins, J., Woodhead, J.,Lehman, W. & Craig, R. (2011). Electron microscopyand 3D reconstruction of F-actin decorated with

cardiac myosin-binding protein C (cMyBP-C). J. Mol.Biol. 410, 214–225.

13. Luther, P. K., Winkler, H., Taylor, K., Zoghbi, M. E.,Craig, R., Padron, R. et al. (2011). Direct visualizationof myosin-binding protein C bridging myosin andactin filaments in intact muscle. Proc. Natl Acad. Sci.USA, 108, 11423–11428.

14. Kensler, R. W., Shaffer, J. F. & Harris, S. P. (2011).Binding of the N-terminal fragment C0-C2 ofcardiac MyBP-C to cardiac F-actin. J. Struct. Biol.174, 44–51.

15. Orlova, A., Galkin, V. E., Jeffries, C. M., Egelman, E. H.& Trewhella, J. (2011). The N-terminal domains ofmyosin binding protein C can bind polymorphically toF-actin. J. Mol. Biol. 412, 379–386.

16. Gautel, M., Zuffardi, O., Freiburg, A. & Labeit, S.(1995). Phosphorylation switches specific for thecardiac isoform of myosin binding protein-C: amodulator of cardiac contraction? EMBO J. 14,1952–1960.

17. Kunst, G., Kress, K. R., Gruen, M., Uttenweiler, D.,Gautel,M.& Fink, R. H. (2000). Myosin binding proteinC, a phosphorylation-dependent force regulator inmuscle that controls the attachment of myosin headsby its interaction with myosin S2. Circ. Res. 86, 51–58.

18. Kulikovskaya, I., McClellan, G., Flavigny, J., Carrier,L. & Winegrad, S. (2003). Effect of MyBP-C binding toactin on contractility in heart muscle. J. Gen. Physiol.122, 761–774.

19. Razumova, M. V., Shaffer, J. F., Tu, A. Y., Flint, G. V.,Regnier, M. & Harris, S. P. (2006). Effects of the N-terminal domains of myosin binding protein-C in anin vitro motility assay: Evidence for long-lived cross-bridges. J. Biol. Chem. 281, 35846–35854.

20. Saber, W., Begin, K. J., Warshaw, D. M. & VanBuren,P. (2008). Cardiac myosin binding protein-C modu-lates actomyosin binding and kinetics in the in vitromotility assay. J. Mol. Cell. Cardiol. 44, 1053–1061.

21. Shaffer, J. F., Razumova, M. V., Tu, A., Regnier, M. &Harris, S. P. (2008). The cardiac myosin bindingprotein-C motif and C1 domain activate actomyosinmotility independent of Ca2+. Biophys. J. 94, 293b.

22. Shaffer, J. F., Razumova, M. V., Tu, A. Y., Regnier, M.& Harris, S. P. (2007). Myosin S2 is not required foreffects of myosin binding protein-C on motility. FEBSLett. 581, 1501–1504.

23. Shchepkin, D. V., Kopylova, G. V., Nikitina, L. V.,Katsnelson, L. B. & Bershitsky, S. Y. (2010). Effects ofcardiac myosin binding protein-C on the regulation ofinteraction of cardiac myosin with thin filament in anin vitro motility assay. Biochem. Biophys. Res. Commun.401, 159–163.

24. Ababou, A., Rostkova, E., Mistry, S., Le Masurier, C.,Gautel, M. & Pfuhl, M. (2008). Myosin binding proteinC positioned to play a key role in regulation of musclecontraction: structure and interactions of domain C1.J. Mol. Biol. 384, 615–630.

25. Govada, L., Carpenter, L., da Fonseca, P. C., Helliwell,J. R., Rizkallah, P., Flashman, E. et al. (2008). Crystalstructure of the C1 domain of cardiac myosin bindingprotein-C: implications for hypertrophic cardiomyop-athy. J. Mol. Biol. 378, 387–397.

26. Delano, W. L. (2002). The PyMOL Molecular GraphicsSystem. DeLano Scientific, San Carlos, CA, USA.