document

letters

888 nature structural biology • volume 8 number 10 • october 2001

Observation of signaltransduction in three-dimensional domain swappingJoost W.H. Schymkowitz, Frederic Rousseau, Hannah R. Wilkinson, Assaf Friedler and Laura S. Itzhaki

Centre for Protein Engineering, University Chemical Laboratory, LensfieldRoad, Cambridge, CB2 1EW and MRC Centre, Hills Road, Cambridge CB22QH, UK.

p13suc1 (suc1) has two native states, a monomer and a domain-swapped dimer. The structure of each subunit in the dimer isidentical to that of the monomer, except for the hinge loopthat connects the exchanging domains. Here we find that sin-gle point mutations at sites throughout the protein and ligandbinding both shift the position of the equilibrium betweenmonomer and dimer. The hinge loop was shown previously toact as a loaded molecular spring that releases tension presentin the monomer by adopting an alternative conformation inthe dimer. The results here indicate that the release of strainpropagates throughout the entire protein and alters the ener-getics of regions remote from the hinge. Our data illustratehow the signal conferred by the conformational change of aprotein loop, elicited by domain swapping, ligand binding ormutation, can be sensed by a distant active site. This workhighlights the potential role of strained loops in proteins: theenergy they store can be used for both signal transduction andallostery, and they could steer the evolution of protein func-tion. Finally, a structural mechanism for the role of suc1 as anadapter molecule is proposed.

Three-dimensional domain swapping is the process by whichone protein molecule exchanges a domain with an identical part-ner1,2. The swapped ‘domain’ can be a single element of sec-ondary structure or an entire tertiary globular domain. Thesubunits of the resulting oligomer generally have the same struc-ture as the monomer, except in the so-called ‘hinge loop’ con-necting the exchanging subunit with the rest of the protein andany new interfaces that may form between the subunits in theoligomer. Because no new intermolecular interface is introducedin domain-swapped dimeric suppressor of cyclin dependentkinase 1 (suc1)3–5 as compared to monomeric suc1, the presenceof this form must be explained solely by conformational changesoriginating in the hinge loop.

Although there are increasing examples of domain-swappedstructures and evidence is accumulating of the biological signifi-cance of this phenomenon6–12, only limited knowledge exists ofthe thermodynamics and kinetics of domain-swapped systemsin solution13,14. Recently, we measured the effect of mutating theresidues constituting the suc1 hinge loop, which connects thetwo exchanging domains, in order to determine how they con-tribute to its domain swapping ability15. The energetics are dom-inated by two conserved Pro residues, one of which destabilizesthe monomer hinge conformation and the other destabilizes thedimer hinge conformation. Thus, the equilibrium betweenmonomer and dimer is a balance of the destabilizing effects inthe two conformations. Here we extend this analysis using a pro-tein engineering study of 30 sites located throughout the struc-ture. Single point mutations were made that deleted part of a

side chain, and the effect on the monomer-dimer equilibriumwas determined.

Remarkably, in spite of the structural identity outside the hingeloop, mutations have different energetic consequences in the twostates, resulting in a shift in equilibrium. Similarly, specific bind-ing of the small ligand phosphate and of a phosphopeptide at asite distant from the hinge loop, equivalent in the two forms,induces a shift in the monomer-dimer equilibrium of wild typesuc1. Thus, the change in conformation of the loop induced bydomain swapping alters the energetic response of remote sites inthe protein, both in terms of mutation and ligand binding. Ourresults provide a structural explanation for the mechanism ofaction of suc1 and illustrate how signal transduction can beachieved by propagation of an energetic perturbation throughthe cooperative network of interactions in a protein.

Fig. 1 Schematic representation of the suc1 structures. a, Monomer andb, domain-swapped dimer with the hinge loop and the exchanging β-strand 4 indicated in red, generated using the program MOLSCRIPT28.suc1 has a four-stranded antiparallel β-sheet that packs against three α-helices and two long loops3. c, Superposition of monomer and dimer(picture taken from Alonso et al.29) shows the extent of structural identi-ty. No new interfaces are formed in the dimer, and there are no inter-molecular interactions between the two subunits other than in thehinge. Further, the structure of the dimer has been solved in two differ-ent crystal forms4,5, and both show the same relative orientation of thetwo subunits of the dimer, indicating that this is the most energeticallyfavorable orientation rather than a crystal packing artifact. On the basisof dihedral angles, contact maps or r.m.s. deviation of the Cα (or Cβ)atoms, there is no greater difference between the monomer and a sub-unit of the dimer than between the two dimer subunits. The 'phosphatebinding site' is at the far end of the β-sheet from the hinge loop. The siteis occupied by an anion in the structures of the monomer, dimer and thehuman homolog30. Also, a phosphopeptide was shown to bind to suc1 atthis site16.

a

b

c

©20

01 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://s

tru

ctb

io.n

atu

re.c

om

© 2001 Nature Publishing Group http://structbio.nature.com

letters

nature structural biology • volume 8 number 10 • october 2001 889

Monomer and dimer superimpose outside the hingeThe domain-swapped dimer results from exchange of an innerβ-strand, β4, by opening up a turn (the ‘hinge loop’) between β3and β4 (refs 3–5) (Fig. 1a,b). The ‘phosphate binding site’referred to in this paper is located at the far end of the β-sheetfrom the hinge. Monomer and dimer structures are highlysuperimposable, as judged by a number of different parameters,except for the hinge loop (Fig. 1c).

Point mutations alter the monomer-dimer equilibriumThe dissociation constants (Kd) for the majority of the mutantsdiffer from the wild type value, indicating that the equilibriumbetween monomer and dimer is affected by the mutations(Fig. 2; Table 1). The Kd of some of the mutants was measured atdifferent temperatures than others (see Methods), and, there-fore, the parameter ∆∆Gdissociation more accurately reflects theeffects of the different mutations on the equilibrium. It isdefined as follows:

∆∆Gdissociation = –RT (ln KdMutant / Kd

WildType) (1)

where KdMutant and Kd

WildType are measured at the same tempera-ture. The observed shift in the equilibrium upon mutation issurprising because corresponding positions in the monomerand the dimer are identical and occupy the same environmentaccording to the crystal structures. Consequently, mutationswould be expected to affect the stability of the two states in thesame way and, thus, not alter the position of the equilibrium.Because of the equivalence, theoretical models emphasize thatthe differences between monomer and domain-swapped dimerare located either in the hinge loop or in interfaces that are creat-ed in the dimer form that are absent in the monomer2. Our pre-vious work showed that the two conserved Pro residues in thehinge loop indeed play a determining role in the equilibriumbetween monomer and dimer15. However, here we see that ener-getic differences between monomer and domain-swappeddimer are not restricted to the hinge loop but are distributedthroughout the protein.

Mutation of residues in the central strands β4 (the swappedstrand) and β2 (which interacts with β4), whose side chainsform one face of the β-sheet, shift the equilibrium towards themonomer (Fig. 3). This includes residues Arg 39 and Arg 99,which are part of the phosphate-binding site3. Mutation ofSer 79, another residue involved in the phosphate binding site,and residues in α-helices 1 and 2 also shift the equilibriumtowards the monomer (Fig. 3). This is also the case with Glu 86and Ile 94, residues that are adjacent to the hinge but not, bystrict definition, part of it. In contrast, mutation of residues inβ2 and β4, whose side chains form the other face of the β-sheet, shifts the equilibrium towards the dimer form.

The wild type equilibrium is altered upon ligand bindingThe shift in the monomer-dimer equilibrium observed uponmutation of residues in the phosphate binding site suggests thatthe monomer and dimer may have different binding affinitiesfor phosphate ions. To investigate this, we measured the effect ofphosphate ions on the position of the wild type monomer-dimerequilibrium. The equilibrium is shifted towards the monomerupon addition of an excess of phosphate, indicating that phos-phate has a different effect on the stabilities of the two forms ofsuc1 (Fig. 4a,b). The dissociation constant for the dimer increas-es from 1.8 ± 0.1 mM to 6.1 ± 0.1 mM in the presence of 50 mMphosphate, corresponding to a stabilization of the monomer rel-ative to the dimer of 0.47 ± 0.01 kcal mol–1. Moreover, the free

Fig. 2 Measurement of the monomer-dimer equilibrium of suc1. a, Size-exclusion chromatogram for a sample of a representative mutant, L96A,loaded on an analytical superdex 75 column (Pharmacia) and monitoredby absorbance at 280 nm. Dimer elutes at 10.5 ml, whereas monomerelutes at 12.4 ml. The amounts of monomer and dimer can be deter-mined by integration of the area of the respective elution peaks. b, Plotof the square of the concentration of monomer versus the concentrationof dimer for samples of L96A at a range of protein concentrations. Thesolid line is the best fit of the data to a straight line. The linearity of theplot indicates that the system is at equilibrium. c, Plot of the free energychange of dissociation, ∆Gdissociation, versus temperature for wild typesuc1. The dissociation constant was measured for the wild type at sixtemperatures between 35 °C and 50 °C, and the value of ∆Gdissociation isderived for each temperature using the following equation: ∆Gdissociation =–RT ln(Kd / PT), where PT is the total protein concentration. The data fitvery well to a straight line (R > 0.999), indicating that the ∆Cp of thedimerization process is small. The slope of the fit to the data gives avalue of 10.5 kcal mol–1 for the enthalpy change of dimerization. The sol-vent accessible surface area (ASA), which gives a good indication of theexpected value of ∆Cp

dissociation, is very similar for the monomer and thedimer (calculated using NACCESS (S. Hubbard and J. Thornton)). Apartfrom confirming the similarity of the structures, this result allows linearextrapolation of ∆Gdissociation to other temperatures and, thus, the calcula-tion of Kd of the wild type at that temperature.

a

b

c

©20

01 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://s

tru

ctb

io.n

atu

re.c

om


letters


energy of unfolding of the wild typemonomer was measured in the presence andabsence of 50 mM phosphate using fluores-cence-monitored equilibrium urea denatu-ration (Fig. 4e,f ). The results show that thestability of the monomer increases in thepresence of phosphate from 7.1 ±0.2 kcal mol–1 to 7.8 ± 0.1 kcal mol–1. Theeffect is specific because the addition ofcations does not change the stability. Thus,the results of these two experiments showthat the monomer binds phosphate withgreater affinity than does the dimer.

The affinity of the phosphate binding sitewas investigated further using a modelphosphopeptide, EQPLpTPVTDL, from thecandidate substrate cdc25 (ref. 16). A shiftin the wild type monomer-dimer equilibri-um towards the monomer is again observedupon addition of peptide. The dissociationconstant of wild type suc1 increases to 2.7 ±0.1 mM in the presence of 2 mM peptide,corresponding to a relative stabilization of0.40 ± 0.02 kcal mol–1 of monomer relativeto dimer.

The relative binding affinity of monomerand dimer forms of the double mutantP90A/P92A was also measured. No shift inthe monomer-dimer equilibrium wasobserved upon addition of 50 mM phos-phate (Fig. 4c,d), and the dissociation con-stant for the dimer was calculated to be 11.2 ± 1.0 mM and 11.5 ± 1.0 mM in theabsence and presence, respectively, of phos-phate ions. The mutation does not abolishthe binding of phosphate to the monomer,as shown by an increase in the free energy ofunfolding of the monomer in the presenceof phosphate ions (8.0 ± 0.1 kcal mol–1

compared with 8.5 ± 0.1 kcal mol–1)(Fig. 4e,f ). The affinity of P90A/P92A forthe phosphopeptide could not be testedbecause the peptide aggregated under theconditions required for the mutant proteinto reach equilibrium between monomerand dimer.

Hinge strain propagates through suc1The changes in Kd upon mutation or ligand binding must origi-nate in the hinge loop, because this is the only part of the struc-ture that is different in the two states. Pro 90 is strained in themonomer hinge conformation15; when Pro 90 is mutated to Ala,the free energy of unfolding of the monomer increases by >1 kcalmol–1. In the dimer, Pro 92 is strained, and mutation to Alaincreases the free energy of unfolding of the dimer by ∼ 0.8 kcalmol–1. It is apparent from the mutations studied here that thestrain is propagated throughout the entire protein. The structureoutside the hinge loop is identical in monomer and dimer, butthe frustration imposed by the hinge Pro residues is accomodat-ed by different parts of the protein scaffold in the two states.Thus, the difference in conformation of the hinge loop in thetwo states leads to a different solution for minimizing strain inthe same structure.

Most of the mutations shift the monomer-dimer equilibriumtowards the monomer — that is, they destabilize the monomerform less than the dimer form. This indicates that the monomerhinge conformation imposes more strain on the protein struc-ture than does the dimer hinge conformation. This result is con-sistent with our previous measurements of the energetics of thehinge itself, showing that Pro 90 creates more strain in themonomer hinge conformation than does Pro 92 in the dimerhinge conformation15,17.

Dimerization alters the ligand binding affinityMonomer and domain-swapped dimer forms of suc1 responddifferently to mutation. We attribute this effect to the change inconformation of the hinge loop upon dimerization and, specifi-cally, to the strain present in the hinge loop. Is there also a differ-ent response to ligand binding? The binding affinities of the

Table 1 Thermodynamic data for domain swapping of suc1 wild type and mutants1

Mutant Location Temperature (°C) Kd (µM) ∆∆Gdissociation (kcal mol–1)WT 35 1,260 ± 50

40 1,470 ± 2042 1,550 ± 1045 1,660 ± 3048 1,800 ± 4050 1,850 ± 30

L10A Core 40 13 × 103 ± 40 –1.36 ± 0.01S13A α1 40 4,860 ± 110 –0.74 ± 0.02L18A α1/core 40 4,490 ± 100 –0.69 ± 0.02R30A β1 50 2,600 ± 20 –0.36 ± 0.01D34A Turn 40 2,990 ± 190 –0.44 ± 0.04E37A β2 40 4,520 ± 20 –0.70 ± 0.03Y38A β2 40 260 ± 20 1.08 ± 0.05R39A β2 50 1,980 ± 260 –0.19 ± 0.08H40A β2 50 1,080 ± 110 0.20 ± 0.06V41A β2/core 50 5,180 ± 250 –0.80 ± 0.03L43A β2/core 30 1,120 ± 30 0.17 ± 0.02L48A α2/core 30 640 ± 10 0.52 ± 0.01K49A α2 50 6,880 ± 420 –0.98 ± 0.04L63A Loop1/core 50 2,020 ± 130 –0.19 ± 0.04L74A Loop2/core 50 9,460 ± 190 –1.12 ± 0.02S79A Loop2 50 5,180 ± 180 –0.81 ± 0.02E86A β3 50 25 × 103 ± 310 –1.82 ± 0.01V87A3 β3 50 1,970 ± 80 –0.19 ± 0.03H88A3 Hinge 50 5,140 ± 240 –0.80 ± 0.03V89A3 Hinge 50 4,370 ± 380 –0.70 ± 0.06P90A2,3 Hinge 50 8.9 × 105 –4.11 ± 0.01E91A3 Hinge 50 640 ± 50 0.53 ± 0.05E91G Hinge 50 1,220 ± 110 0.12 ± 0.06P92A3 Hinge 50 180 ± 10 1.34 ± 0.04H93A3 Hinge 50 1,780 ± 90 –0.12 ± 0.03I94V β4 50 4,480 ± 120 –0.71 ± 0.02L95A β4/core 40 1,580 ± 110 –0.04 ± 0.05L96A β4 50 670 ± 20 0.49 ± 0.01F97L β4/core 45 1,170 ± 10 0.15 ± 0.01K98A β4 40 490 ± 20 0.69 ± 0.02R99A β4 40 3,980 ± 60 –0.62 ± 0.01

1Kd is the dissociation constant for the monomer-dimer equilibrium. The temperature is given atwhich the Kd was measured. This was then combined with the Kd of the wild type at the sametemperature to obtain ∆∆Gdissociation, the change in the equilibrium upon mutation (using Eq.1).2The Kd for this mutant is very high and, therefore, because measurements could not be made atconcentrations in the range of the Kd, it is an estimate only.3From ref. 15.

©20

01 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://s

tru

ctb

io.n

atu

re.c

om


letters

nature structural biology • volume 8 number 10 • october 2001 891

monomer and dimer were compared by measuring the effect ofphosphate ions or phosphopeptide on the position of themonomer-dimer equilibrium. Addition of ligand shifts the equi-librium in favor of the monomer, showing that the ligand bindswith higher affinity to the monomer than to the dimer. Thus, thestructurally equivalent binding sites of monomer and dimerhave different affinity for ligand. Again, the effect must be a con-sequence of the different conformation of the hinge loop. This is

a form of signal transduction: the change in the conformation ofthe loop is sensed at the distant phosphate binding site.

To test whether the observed signal transduction arisesbecause of the strain in the hinge loop, we measured the phos-phate-binding affinity of the double mutant P90A/P92A becausethis mutation should significantly reduce the strain in bothmonomer and dimer forms. As predicted, the monomer-dimerequilibrium of the mutant, in contrast to that of the wild type, isnot affected by the addition of phosphate (the monomer form ofthe mutant is still able to bind phosphate). Thus, when the strainis reduced, monomer and dimer forms of suc1 have similarphosphate-binding affinities, suggesting that the different bind-ing affinity of wild type monomer and dimer can be attributedto the strain in the hinge loop.

ConclusionsEnergetic differences between the monomer and the domain-swapped dimer forms of suc1 are distributed throughout the pro-tein. With the exception of the hinge loop, monomer and dimer ofsuc1 are superimposable from the static picture that is presentedby their crystal structures, and there are no new interfaces in thedimer that could account for the observed effects. Therefore, thedifferent energetics must result from a difference in the dynamicproperties of the ensemble of native states that is present in solu-

a

b

c

d

e

f

Fig. 4 Ligand binding experiments of suc1. Size-exclusion chromatogram of wild type suc1 and the mutant P90A/P92A, illustrating the effect ofphosphate binding on the monomer-dimer equilibrium. a, Wild type in Tris buffer (50 mM Tris, pH 7.5, 300 mM NaCl and 1 mM EDTA). b, Wild typein Tris buffer containing 50 mM sodium phosphate. c, P90A/P92A in Tris buffer. d, P90A/P92A in Tris buffer containing 50 mM sodium phosphate. Thesamples were incubated in the respective buffer for an interval to allow for equilibration between monomer and dimer, at a protein concentrationof 1 mM for wild type and 4.8 mM for P90A/P92A, before loading on the gel-filtration column. Equilibrium denaturation curves of e, wild type suc1and f, P90A/P92A in Tris buffer (50 mM Tris, pH 7.5, and 1 mM EDTA) (open circles) and in Tris buffer containing 50 mM sodium phosphate (filled cir-cles). The protein concentration was 1 µM.

Fig. 3 Mapping the difference in the effect of mutations on themonomer-dimer equilibrium onto a schematic representation of themonomer structure of suc1. Blue coloring indicates that the equilibriumis shifted in favor of the dimer. Red coloring indicates a shift in favor ofthe monomer. Residues that do not shift the equilibrium significantly areshown in yellow. The picture was generated using the programMOLSCRIPT28.

©20

01 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://s

tru

ctb

io.n

atu

re.c

om


letters


described26,27. The separation of monomeric and dimeric suc1 byanalytical size-exclusion chromatography and the calculation of thedissociation constants of wild type and mutants have beendescribed in detail15. The phosphopeptide EQPLpTPVTDL was syn-thesized and purified by standard methods. The free energy ofunfolding of monomeric suc1 was measured by fluorescence-moni-tored urea denaturation, as described27.

Calculation of the change in free energy of dissociation ofdimeric suc1 upon mutation. Monomer-dimer equilibration wascarried out at elevated temperatures between 40 °C and 50 °C(depending on the mutant) because the process is prohibitivelyslow at room temperature15. The dissociation constant of a mutantprotein can be combined with that of the wild type to obtain∆∆Gdissociation, the effect of mutation on the equilibrium betweenmonomer and dimer, when Kd

WildType and KdMutant are compared at

the same temperature using Eq. 1. The value of ∆∆Gdissociation is inde-pendent of the total protein concentration (PT) and allows directcomparison of the effect of mutation on monomer and dimer. Notethat the values of ∆∆Gdissociation were measured at different tempera-tures depending on the mutant. However, the change in the freeenergy of dissociation upon mutation (of a conservative nature) canreasonably be assumed to be independent of temperature, becausethe heat capacity of the native state of the protein is unlikely tochange upon mutation. This allows us to compare values of∆∆Gdissociation from one mutant to another in spite of the differenttemperatures at which they were measured.

AcknowledgmentsThis work was supported by the MRC of the UK. L.S.I. was supported by a CareerDevelopment Award from the MRC of the UK. J.W.H.S. and F.R. were supportedby Marie Curie Training and Mobility of Research Fellowships from the E.C. A.F. issupported by a Human Frontier Science Program long term fellowship. We thankA.R. Fersht and J. Clarke for useful discussions and S. Kazmirski for help withanalyzing the crystal structures.

Correspondence should be addressed to L.S.I. email: [email protected]

Received 7 March, 2001; accepted 13 August, 2001.

1. Bennett, M.J., Choe, S. & Eisenberg, D. Protein Sci. 3, 1444–1463 (1994).2. Bennett, M.J., Choe, S. & Eisenberg, D. Proc. Natl. Acad. Sci. USA 91, 3127–3131

(1994).3. Endicott, J.A. et al. EMBO J. 14, 1004–1014 (1995).4. Bourne, Y. et al. Proc. Natl. Acad. Sci. USA 92, 10232–10236 (1995).5. Khazanovich, N., Bateman, K.S., Chernaia, M., Michalak, M. & James, M.N.G.

Structure 4, 299–309 (1996).6. Green, S.M., Gittis, A.G., Meeker, A.K. & Lattman, E.E. Nature Strut. Biol. 2,

746–751 (1995).7. SaintJean, A.P., Phillips, K.R., Creighton, D.J. & Stone, M.J. Biochemistry 37,

10345–10353 (1998).8. Vitagliano, L. et al. J. Mol. Biol. 293, 569–577 (1999).9. Qu, C. et al. Struct. Folding Des. 8, 1095–1103 (2000).

10. Liu, Y., Gotte, G., Libonati, M. & Eisenberg, D. A. Nature Strut. Biol. 8, 211–214.(2001).

11. Ogihara, N.L. et al. Proc. Natl. Acad. Sci. USA 98, 1404-1409. (2001).12. Janowski, R. et al. Nature Strut. Biol. 8, 316-320. (2001).13. Murray, A.J., Head, J.G., Barker, J.J. & Brady, R.L. Nature Struct. Biol. 5, 778–782

(1998).14. Hayes, M.V., Sessions, R.B., Brady, R.L. & Clarke, A.R. J. Mol. Biol. 285, 1857–1867

(1999).15. Rousseau, F., Schymkowitz, J.W.H., Wilkinson, H.R. & Itzhaki, L.S. Proc. Natl.

Acad. Sci. USA 98, 5596–5601 (2001).16. Landrieu, I. et al. J. Biol. Chem. 276, 1434–1438 (2001).17. Schymkowitz, J.W.H., Rousseau, F. & Itzhaki, L.S. J. Mol. Biol. 301, 201–206 (2000).18. Hilser, V.J. & Freire, E. J. Mol. Biol. 262, 756–772 (1996).19. Freire, E. Proc. Natl. Acad. Sci. USA 96, 10118–10122 (1999).20. Freire, E. Proc. Natl. Acad. Sci. USA 97, 11680–11682 (2000).21. Todd, M.J. & Freire, E. Proteins 36, 147–156 (1999).22. Pines, J. Curr. Biol. 11, 1399–1402 (1996).23. Patra, D. & Dunphy, W.G. Genes Dev. 12, 2549–2559 (1998).24. Patra, D., Wang, S.X., Kumagai, A. & Dunphy, W.G. J. Biol. Chem. 274,

36839–36842 (1999).25. Bourne, Y. et al. Cell 84, 863–874 (1996).26. Schymkowitz, J.W.H., Rousseau, F., Irvine, L.R. & Itzhaki, L.S. Struct. Folding Des.

8, 89–100 (2000).27. Rousseau, F., Schymkowitz, J.W.H., delPino, M.S. & Itzhaki, L.S. J. Mol. Biol. 284,

503–519 (1998).28. Kraulis, P.J. J. Appl. Crystallogr. 24, 946–950 (1991).29. Alonso, D.O.V., Alm, E. & Daggett, V. Struct. Folding Des. 8, 101–110.30. Parge, H.E., Arvai, A.S., Murtari, D.J., Reed, S.I. & Tainer, J.A. Science 262,

387–394 (1993).

tion. The key lies in our earlier result, which is further confirmedin this paper, showing that the hinge loop in suc1 is strained —that is, upon mutation of residues in the hinge, the free energy ofunfolding of the protein increases15,17. In the case of suc1, releaseof strain in the hinge loop upon dimerization appears to result in aredistribution of the native state ensemble.

Recent theoretical work by Freire et al.18–21 showed that energyperturbations at one site — for example, induced upon ligandbinding — could be propagated to remote locations in a proteinstructure by altering the dynamic network of interactions in theprotein. The propagation effect required the presence of a signifi-cant fraction of residues with ‘low structural stability’ in the‘unperturbed’ (or unbound) state21 — that is, those residues thatcontribute little to stability. By analogy, the monomeric form ofsuc1 can be viewed as the unperturbed state in which the hingeloop residues are strained (similar to the ‘low structural stability’terminology of Freire et al.21). The loop can be seen as perturbed inthe dimer, and here we see that this perturbation is sensed through-out the structure.

Biological and biophysical implicationsIn suc1, the energy perturbations elicited by domain swappingchange the ligand binding affinity at a site >20 Å away from thehinge loop. Domain swapping thus provides a regulatory mecha-nism of sites that are distant from the swapping part of the pro-tein. The effect is completely abolished when the two Pro residuesin the hinge are mutated to Ala. Further, the two Pro residues arehighly conserved in the cks family; therefore, the strain that theyimpose is likely also conserved. There are two possible ways ofinterpreting the strain and explaining its likely conservation in thecks family. Domain swapping may be an essential property ofthese proteins22, and the strain would be required to drive theprocess: in the dimer the hinge loop has a conformation that car-ries less strain while the overall fold of the protein is retained.

An alternative view is suggested by known function of suc1.suc1 is an adapter molecule with two potential molecular recogni-tion sites located on opposite sides of the structure22–25. At specificstages in the cell cycle, suc1 binds to a cyclin-dependent proteinkinase (CDK); this interaction involves residues in the hingeloop25. suc1 is thought to then target the CDK to specific phos-phoprotein substrates, such as cdc25, by binding them at its othersite, which we have referred to as the phosphate binding site.Recently a model peptide of cdc25 was shown to bind via thephosphate binding site16. We have shown that a change to thehinge loop induced by altering the oligomerization state of suc1 ispropogated to remote locations. In the same way, a perturbationof the hinge loop induced by CDK binding could be sensed by thephosphate binding site. Thus, binding to CDK may increase theaffinity of this site, thereby allowing suc1 to bind to a phosphory-lated target protein such as cdc25. In this case, the strain present inthe hinge loop would be required for the signal transduction.

As a model system, suc1 possesses all the elements thatdescribe signal transduction in proteins as seen, for example, inallosteric enzymes. Moreover, the results presented here couldhave implications for the evolution of proteins. Mutations inloops could create strain that could propagate throughout theprotein and alter affinities or introduce new activities. This pro-vides a starting point upon which natural selection can act byaccumulation of further mutations.

MethodsMutagenesis and protein purification. Site-directed mutagene-sis, protein expression and purification were performed as

©20

01 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://s

tru

ctb

io.n

atu

re.c

om


Documents

document