Förster resonance energy transfer studies of calmodulin produced by native protein ligation reveal inter-domain electrostatic repulsion

Authors


Correspondence

S. Linse, Biochemistry and Structural Biology, Lund University, Chemical Centre, PO Box 124, SE-221 00 Lund, Sweden

Fax: +46-46-2224116

Tel: +46-46-2228246

E-mail: sara.linse@biochemistry.lu.se

K. S. Åkerfeldt, Department of Chemistry, Haverford College, Haverford, PA 19041, USA

Fax: +1-610-896-4904

Tel: +1-610-896-1213

E-mail: kakerfel@haverford.edu

Abstract

This study explores the influence of long-range intra-protein electrostatic interactions on the conformation of calmodulin in solution. Ensemble Förster resonance energy transfer (FRET) is measured for calmodulin with a fluorophore pair incorporated specifically with a donor at residue 17 and an acceptor at position 117. This construct was generated by a combination of solid phase peptide synthesis, cloning, expression and native chemical ligation. This labelling method has not previously been used with calmodulin and represents a convenient method for ensuring the explicit positioning of the fluorophores. The ensemble FRET experiments reveal significant electrostatic repulsion between the globular domains in the calcium-free protein. At low salt, calmodulin has a relatively extended conformation and the distance between the domains is further increased by denaturation, by heat or by non-ionic denaturants. The repulsion between domains is screened by salt and is also diminished by calcium binding, which changes the protein net charge from −23 to −15. Compared with the calcium-free form at low salt, the FRET efficiency for the calcium-bound form has, on average, increased 10-fold. The conformation of the calcium form is insensitive to salt screening. These results imply that when the two globular domains of calmodulin interact with target, there is no significant free energy penalty due to electrostatic interactions.

Abbreviations
CaM-DA

calmodulin donor–acceptor construct

CaM-D

donor-labelled calmodulin

FRET

Förster resonance energy transfer

MESNA

sodium 2-sulfanylethanesulfonate

Introduction

Electrostatic interactions play important roles in the Ca2+-signalling protein calmodulin [1-4]. This protein is composed of two globular domains, each containing two Ca2+ binding sites. The domains are connected by a flexible linker [5, 6]. Ca2+-induced structural changes within each domain lead to increased exposure of hydrophobic groups and favour target protein binding. In addition, the domains may change their relative orientation and separation in response to Ca2+ binding, target binding, pH modulation or salt screening. Because of the highly negative net charge (−23 for the apo form, i.e. in the absence of bound ligand), the populations of different conformational substates are regulated by modulation of electrostatic interactions. Depending on metal ion concentration, pH and ionic strength, calmodulin is bound to zero, two or four Ca2+ ions [7, 8]. Each Ca2+ form may further exist in multiple conformational substates, ranging from extended to compact [9-19]. Target binding or addition of denaturant are other factors that may shift the population distribution.

Calmodulin is an essential protein present in all eukaryotic cells. It transmits the signal of increased cytoplasmic Ca2+ concentration to the activation or attenuation of several hundred enzymes, transporters and other target proteins [20-28]. Calmodulin is thereby involved in the regulation of higher order functions such as memory and learning, cell motility and growth, and immune response [25, 26, 28]. One of the most striking aspects of calmodulin is its ability to blend promiscuity with specificity in target binding. A high abundance of flexible methionine side chains at the hydrophobic binding surfaces allows calmodulin to optimize its van der Waals and steric interactions with a variety of targets and diverse binding sequences [29, 30]. Only a small portion of the target is in direct contact with calmodulin in the complex. While there is no strict consensus binding sequence, dominant features include a number of hydrophobic residues along with a net positive charge, typically within the structural context of an amphiphilic α-helix [31-33].

Interactions involving hydrophobic surfaces of calmodulin and target protein provide the dominant driving force for complex formation. The electrostatic component appears to be less important than one might expect from the opposite net charges of target and calmodulin. At neutral pH, Ca4-calmodulin has a net charge of −15, while many binding sequences carry a net charge +4 to +8 over ~ 20 residues. Still, a remarkable insensitivity was observed in the binding of calmodulin to charge modulated variants of a target peptide from smooth muscle myosin light chain kinase [1]. Interestingly, no measurable effect on binding was discerned when the formal charge of this peptide was reduced from +7 to +4, nor when the net charge of calmodulin was partially neutralized through pH variation [1]. These findings were reconciled with a charge regulation mechanism by which target and calmodulin modulate each other's net charge upon complex formation. pKa values of ionizable amino residues are very sensitive to the electrostatic environment. They are upshifted in a negatively charged environment (calmodulin) and downshifted in a positively charged environment (target peptide), but when the two oppositely charged species come together the pKa values move closer to the non-shifted values. Therefore, in a complex formed from two highly and oppositely charged species, the effect of charge modulation mutations may be counteracted by this charge regulation mechanism [34].

In a subsequent study, and perhaps even more unexpectedly, complex formation was found to occur with higher affinity at physiological salt concentration (150 mm KCl) compared with at low salt, indicating that the net electrostatic component actually counteracts complex formation between the negatively charged Ca2+-calmodulin and the positively charged target peptide [2]. The latter finding was proposed to originate from inter-domain electrostatic repulsion between the two lobes of calmodulin (carrying a −7 and −8 net charge, respectively, in the Ca2+-bound state), which becomes more pronounced as the domains are brought in closer proximity to wrap around the target peptide.

These findings prompted the current study in which we have explored conditions that affect the distribution of conformational substates of calmodulin, as a result of inter-domain repulsion. Using native chemical ligation [35-37], we have produced a doubly labelled variant of calmodulin with Alexa488 at position 17 and Alexa546 at position 117. We used this construct to measure ensemble Förster resonance energy transfer (FRET) [38, 39]. These studies reveal the average inter-domain distance of calmodulin under a range of conditions, including the absence and presence of Ca2+, high and low salt, and native versus denaturing conditions.

Results

Site-specific incorporation of fluorophores

The calmodulin donor–acceptor construct (CaM-DA) was obtained through native chemical ligation and was labelled with a donor (AlexaFluor488) in position 17 and an acceptor (AlexaFluor546) in position 117. The larger fragment, EF123, encompassing the three N-terminal EF-hands, residues 1–116, with a serine to cysteine mutation in position 17, was produced through cloning (see 'Materials and methods'). The smaller fragment, CysEF4, residues 117–148, comprising the C-terminal EF-hand, with threonine 117 replaced by cysteine, was produced by solid phase methodology. The steps involved in the intein-mediated chemical ligation of EF123 and CysEF4 are outlined in Fig. 1. Cys17 was labelled with Alexa488 before ligation, and Cys117 was labelled with Alexa546 after ligation. In this manner, one defined site-specifically labelled product was obtained.

Figure 1.

Reaction scheme for the production of calmodulin with Alexa488 at position 17 and Alexa546 at position 117 using native chemical ligation.

Intein fusion offers a means for producing Alexa488-EF123 with a C-terminal thioester. In the presence of thiols, such as dithiothreitol or sodium 2-sulfanylethanesulfonate (MESNA), the intein undergoes a specific self-cleavage which releases the fused protein [35-37]. The identity of the thioester depends on the thiol compound used in the cleavage reaction and, in our hands, MESNA proved to furnish the most effective leaving group in the subsequent nucleophilic attack by the N-terminal cysteine of the second fragment, CysEF4. There was close to 100% conversion of starting materials to product, as judged by analytical HPLC. Minor unreacted components were removed using gel filtration prior to the FRET studies.

FRET studies

The influence of temperature, chemical denaturants (GuHCl and urea) and ionic strength of the surrounding medium on the distribution of conformational substates of calmodulin, in the presence and absence of Ca2+, was studied employing ensemble averaged FRET. The FRET efficiency E was calculated using

display math(1)

where IDA and ID represent the peak intensity for donor- and acceptor-labelled calmodulin, CaM-DA, and donor-labelled calmodulin, CaM-D, respectively. In the pairwise measurements of IDA and ID, the donor concentration was kept constant.

The FRET efficiency E is related to the distance between the fluorophores, r, according to

display math(2)

R0 is defined as the characteristic Förster distance between the donor and acceptor at which the transfer efficiency is 50%. This distance is dependent on the identity of the fluorophore pair used, their orientation and their surrounding environment. The given R0, 64 Å, as calculated by the manufacturer of the AlexaFluor488–AlexaFluor546 pair, assumes an aqueous environment in which the chromophores experience full rotational freedom [40]. Figure 2 shows the fluorescence emission spectra obtained for apo and holo CaM-D and CaM-DA, at low (no added) and physiological (150 mm) KCl. At low salt, the signal intensity of the donor peak at 515 nm decreases by 64% upon Ca2+ binding for CaM-DA, reflecting a decrease in the distance between the chromophores. The r values obtained using R0 contain a large systematic error due to the large size of the chromophores, their orientation relative to Cα etc. [41]. Therefore all distances are reported relative to R0, and the discussion is focused on comparative values between different solution conditions. Hence at low salt, Ca2+-loaded calmodulin displays a compact conformation, = 0.6 (r/R0 = 0.9), in contrast to a less compact apo state, = 0.06 (r/R0 = 1.6). Measurement at a physiological salt concentration reveals a similar conformation for the compact holo state, = 0.6 (r/R0 = 0.9), with a less extended apo state, = 0.4 (r/R0 = 1.1). Mg2+-bound calmodulin (in the presence of 2 mm Mg2+ [42]) displays a compact conformation with = 0.4 (r/R0 = 1.1) at both low and physiological salt concentrations (Fig. S1), although more extended than the Ca2+-loaded form.

Figure 2.

Fluorescence emission spectra (500–660 nm, with excitation at 480 nm) at 25 °C for CaM-D (A, C) and CaM-DA (B, D) in 2 mm Tris/HCl pH 7.5 without (A, B) and with (C, D) 150 mm KCl: blue, apo protein, spectra recorded in the presence of 0.2 mm EDTA; red, Ca2+-loaded protein, spectra recorded after addition of 2 mm CaCl2.

To further investigate the electrostatic component of the compaction of calmodulin upon Ca2+ binding, the effect of salt concentration was studied at 0, 10, 20, 40, 80, 160 and 320 mm KCl (Fig. 3). For apo calmodulin, the addition of salt is found to lead to a significant decrease in the inter-chromophore distance. The largest effect per titration step, with a 12% reduction in r, is observed between 0 and 10 mm KCl. In contrast, the inter-chromophore distance in Ca2+-loaded calmodulin appears insensitive to ionic strength and is shorter than in apo calmodulin at all salt concentrations investigated.

Figure 3.

Salt dependence of inter-chromophore distance, plotted as r/R0 versus KCl concentration for apo (blue circles) and Ca2+-loaded (red squares) calmodulin. r/R0 was calculated from spectra obtained for CaM-DA and CaM-D in 2 mm Tris/HCl pH 7.5 at 25 °C, with 0.2 mm EDTA (apo) after addition of 2 mm CaCl2 (Ca2+-loaded) and 0–320 mm KCl using Eqns (1) and (2).

Calmodulin denaturation studies

In an effort to investigate the change in inter-domain distance upon unfolding, denaturation was attempted with GuHCl, urea or heat (Fig. 4). In the presence of 6 m GuHCl, no difference in FRET was observed between the calmodulin apo and holo states, with = 0.03 (r/R0 = 1.8) (Fig. 4, right panel). Under these conditions, the Ca2+-bound form is denatured (Fig. S2). The effect of the non-ionic denaturant urea is quite different in which case the apo protein is clearly more extended than the Ca2+-loaded form, with = 0.01 (r/R0 = 2.0) and = 0.26 (r/R0 = 1.2), respectively. Ca2+ remains bound to the protein in the presence of 8 m urea (see 'Discussion'). Upon an increase in ionic strength, the spectra in urea gradually become closer in appearance to those recorded in 6 m GuHCl. At 90 °C and low ionic strength, there is again a clear difference between the spectra recorded in the absence and presence of Ca2+, with < 0.01 (r/R0 > 2.2) and = 0.55 (r/R0 = 1.0), respectively. To summarize, screening by salt affects Ca2+-free calmodulin significantly whereas the Ca2+-loaded form is only affected to a minor extent, if at all.

Figure 4.

Apparent r/R0 calculated from ensemble FRET mea-surements of apo and holo CaM under different denaturing conditions.

Target peptide binding

The side chains at positions 17 and 117 are fully exposed in all available calmodulin–target structures, and were specifically chosen for chromophore labelling of calmodulin to avoid obstruction in target binding. To verify the function of the construct, we monitored binding of a peptide from smooth muscle myosin light chain kinase. A peptide titration experiment shows that the FRET-CaM construct binds this peptide with high affinity in the presence of Ca2+ and with lower affinity in the absence of Ca2+ (Fig S3).

Discussion

In the present study, each calmodulin carries the donor Alexa488 at position 17 and the acceptor Alexa546 at position 117 (Fig. 1). To achieve the homogeneous incorporation of chromophores site specifically, calmodulin was built from two pieces containing EF123 and EF4, respectively, employing native chemical ligation. This method offers a practical alternative to the incorporation of chromophore-labelled amino acids using in vitro technology and two-four-base codons [43]. Native chemical ligation provides the opportunity to build proteins by linking polypeptide fragments through a standard peptide bond. The native peptide bond is formed by the nucleophilic displacement of a C-terminal thioester of one fragment with an N-terminal cysteine side chain thiol of a second fragment [35] to first produce a thioester linking the two fragments. This thioester then undergoes an S → N acyl transfer reaction in a manner that was first observed to cause peptide bond formation through an intra-molecular rearrangement in the early 1950s [44]. The C-terminal thioester is obtained by chemical synthesis [35] or by the expression of the fragment fused to an intein, followed by a thiol-induced cleavage [36]. When inteins are used, the technique of native chemical ligation is sometimes referred to as intein-mediated ligation or expressed protein ligation. The approach has been used in the chemical synthesis of full-length proteins from fragments, for segmental isotope labelling and for introduction of fluorophores [37, 45-49]. The number of steps involved potentially lowers the overall yield; however, the ease of cloning and expressing large amounts of intein-tagged protein, the ability to ensure one well-defined product and the fact that only a low concentration of the final product is needed in fluorescence studies makes expressed protein ligation an attractive approach.

Since the 1970s, fluorophore-labelled calmodulin has been used to probe conformational changes upon Ca2+ or target binding, and as a vehicle to sense free Ca2+ in vivo. Many different types of dyes have been employed (for representative examples see Table 1). A single solvatochromic fluorophore can be used to monitor conformational changes within the protein due to a change in polarity of the environment surrounding the dye (for a recent review see [46]). Methods for labelling calmodulin with two fluorophores have been developed more recently, particularly for the purpose of FRET studies. Conventionally, two Cys residues are incorporated through site-directed mutagenesis and modified by the simultaneous addition of equimolar amounts of donor (D) and acceptor (A) chromophores. This approach results in a mixture of six components: A, D, AD, DA, DD and AA. The method is straightforward, in principle, but has the limitation that it is difficult, if not impossible, to separate AD, DA, DD and AA. This problem is of less importance in single molecule FRET measurements but can be particularly troublesome in bulk measurement studies as it may lead to incorrect A–D distance measurements. In contrast, by expressed protein ligation we obtain a single homogeneous product, DA.

Table 1. Examples of fluorophore-labelled calmodulin
Scientific questionMethod (position)ChromophoreReference
  1. Methods of chromophore incorporation and applications: A, Cys-specific labelling; B, Lys-specific labelling; C, fusion to other chromophoric protein; D, recombinant chromophore labelling; E, amber codon labelling; F, chromophore incorporation through native chemical ligation. AEDANS, attached to protein with 1,5-IAEDANS: N-iodoacetyl-N'-5(sulfo-1-naphthyl)ethylenediamine; BODIPY, 4,4-difluoro-4-bora-3a,4a-diaza-s-indacene; BODIPY-FL, 4,4-difluoro-5,7-dimethyl-4-bora-3a,4a-diaza-s-indacene-3-propionic acid; CFP, cyan fluorescent protein; DAB, 4-dimethylaminophenylazophenyl-4′-maleimide; dabsyl, 4-((4-(dimethylamino)phenyl)azo)-benzenesulfonylchlorid;4-[[4-(dimethylamino)phenyl]azo]-benzenesulfonylchlorid; N,N''-dimethyl-N-(iodoacetyl)-N''-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)ethylenediamine; dansyl, 5-(dimethylamino)naphthalene-1-sulfonyl; GFP, green fluorescent protein; MDCC, N-[2-(1-maleimidyl)ethyl]-7-(diethylamino)coumarin-3-carboxamide; TBA, tetramethinemerocyanine; YFP, yellow fluorescent protein; BFP, blue fluorescent protein.

One chromophore
Peptide/protein binding to CaMB (K-residues)Dansyl [58-60]
Ca2+ induced conformational changes to CaM(K-residues)TBA analogues [61, 62]
Ca2+ sensing in vivo?A (M109C)MDCC [63]
Antidepressant binding to CaMA (M109C)MDCC, IANBD Acrylodan [64]
Protein array screening of CaM targetsA (S17C)Alexa488 [27]
Labelling of plant calmodulinA (C26)Fluorescein [65]
Two chromophores
Structural coupling between calmodulin domainsA (Y99W)AEDANS [66]
Edema factor binding to CaMA (T34C, T110C)AEDANS (D) DAB (A) [67]
Ca2+ sensing in vivoC (N-terminus) C (C-terminus)BFP or CFP GFP or YFP [68, 69]
Affinity chromatography, fluorescence anisotropy measurements, localization in vivoC, D (4-Cys tag)FlAsH [70]
Ca2+ induced conformational changes and target binding: single molecule FRETC (T34C, T110C)Alexa [71-74]
Conformational changes of CaM monitored by FRETE (N-terminus, Gly40, Phe99 Tyr?, Leu112)BODIPY derivatives BODIPY-FL derivatives [43]
Nitric oxide synthase isozyme binding to CaM: orientation of binding A (T34C, T110C)Dabsyl Alexa [75]
CaM binding to ryanodine receptorN-domain, C-domain central portion of CaMFluorescein FK506 labelled [76]
Conformational changes of CaMA, E (T34Pyl, T110C)Azidocoumarine (D) Alexa488 (A) [77]
CaM domain electrostatic repulsionF (S17C, T117C)Alexa488 and Alexa546Hellstrand (this work)

The present results show that the inter-domain distance in calmodulin is strongly affected by electrostatic interactions. Figure 5 summarizes the inter-chromophore distance as a function of KCl concentration for apo and holo calmodulin under native and denaturing conditions. As a relative comparison, the distance between the α carbons in available structures (Table S1) are shown on a separate axis. Although only relative distances are measured, it is clear that the span of distances is within the range of distances calculated from available NMR and X-ray crystal structures. It is clear from Fig. 5 that the measured distances for apo calmodulin decrease with increased ionic strength as well as upon Ca2+ binding. Interestingly, of the available crystal structures, the Ca2+-bound form is represented by the most compact conformation as well as by the most extended conformation. This discrepancy is presumably caused by constraints imposed by crystal packing, and by the very high concentration of substances in the crystallization solutions which may shift the distribution over conformational substates. This may lead to inter-domain orientations and distances that are less relevant for more dilute solutions.

Figure 5.

Comparison of the measured Cα(17)–Cα(117) distance for apo (blue circles) and holo (red squares) calmodulin in native and denaturing conditions of different salt strengths (left axis). As a relative comparison of length scales, the calculated corresponding distance in reported structures [9-16, 55-57] is plotted on the right axis.

Electrostatic repulsion under native conditions

Ca2+-loaded calmodulin displays an inter-chromophore distance that is largely insensitive to salt over the range 0–500 mm KCl (Fig. 3). In contrast, the inter-chromophore distance for apo calmodulin is strongly affected by the presence of salt. Most of the screening effect is observed already at the rather modest salt concentration of 10–20 mm KCl (Fig. 3). This observation indicates that there is an electrostatic repulsion between the globular domains in apo calmodulin, which is screened by salt. One reason for the more pronounced repulsion within the apo form is the higher net negative charge (−23) compared with −15 when calmodulin is loaded with Ca2+. At all salt concentrations, the Ca2+ form is more compact than the apo form, but the difference levels off with increasing salt. Although the FRET efficiency does not change, an overall decrease in the fluorescence intensity for both the donor and acceptor can be seen for the compact holo form at increasing salt concentration. This is most probably due to the fluorophores being in close proximity, resulting in quenching of both donor and acceptor.

Electrostatic interactions between two charged groups are proportional to Z1Z2/r, where Z1 and Z2 are the two charges and r the distance between them. If we approximate the domains as macro-ions, this would give a similar level of repulsion between domains of charge −12 and −11 (apo calmodulin) at = 2.36R, if the distance is R between domains of charge −8 and −7 (Ca2+-loaded calmodulin; 11 × 12/(7 × 8) = 2.36). The inter-domain distance is found experimentally to increase from 0.9R0 to 1.6R0 going from Ca2+-loaded to apo calmodulin, i.e. by a factor of 1.8, of similar magnitude to that calculated above. Presumably, 1.6R0 is as far as the distance can be without partial unfolding.

Electrostatic repulsion in the presence of denaturant

It is intriguing to find that for the apo form in the presence of the non-ionic denaturant urea the inter-chromophore distance is 2R0, i.e. it has increased by a factor of 2.2 relative to the folded Ca2+-loaded calmodulin and is close to the calculated value using the macro-ion approximation. The most extended conformations are found for the apo protein denatured by 8 m urea or by high temperature with no salt added (low salt conditions). These conformations display a 25% longer inter-chromophore distance compared with the folded apo protein, which implies that the urea denatured protein exists in an extended conformation. The apo state denatured in 6 m GuHCl is somewhat less extended, most probably due to considerable screening of the repulsive electrostatic interactions at this extreme salt concentration. Under most conditions, including low, physiological and high salt, as well as in the presence of urea, the Ca2+ form (at 2 mm Ca2+) is significantly more compact than the apo form. This indicates that the Ca2+ form is partially folded under these conditions. The one exception is calmodulin in the presence of 6 m GuHCl and 2 mm Ca2+, where virtually no difference is seen relative to the apo protein. Denaturation studies, however, show that calmodulin is denatured by 6 m GuHCl also in the presence of 2 mm CaCl2 (Fig. S2), suggesting that Ca2+ is not bound under these conditions.

Electrostatic repulsion at 90 °C

In agreement with its high thermal stability, Ca2+-loaded calmodulin is only slightly more extended at 90 °C (r/R0 = 0.96) compared with at room temperature (22 °C). In addition, at 90 °C, the inter-chromophore distance is largely insensitive to the addition of salt. The slight increase in distance upon salt addition might, again, be related to a loss in Ca2+ affinity. Apo calmodulin, however, is more sensitive to heat and denatures between 50 and 60 °C [50, 51]. Due to the high net negative charge and the resulting intra protein domain repulsion in calmodulin, the acceptor–donor chromophore distance in the denatured state is highly salt dependent with a shorter distance observed under increased ionic strength.

Conformational substates

There are currently two crystal structures available for Ca2+-saturated calmodulin, the classical extended structure [10, 12] and a recently determined structure which is compact [9]. In single molecule FRET experiments [18], using calmodulin labelled with fluorophores in positions 34 and 110, a cluster of three major conformational substates was found for the Ca2+-saturated form: a compact state (with a donor–acceptor distance of 20–30 Å), an intermediate state (30–40 Å) and an extended state (50–60 Å). Bulk FRET may not resolve these substates. FRET effects depend on 1/R6 and conformations with short inter-chromophore distance give a much larger contribution to the observed decrease of donor fluorescence for DA relative to D. The observed average distance is 1/R6 averaged and skewed towards shorter distances.

Concluding remarks

Ensemble FRET studies with site-specifically incorporated fluorophores reveal the role of electrostatic repulsion and salt screening on the inter-domain separation in calmodulin. At conditions of strong electrostatic repulsion, the domains are found to have a large average separation, as demonstrated for the apo protein at low salt. This effect is even more pronounced in the presence of a non-ionic denaturant. In contrast, when the electrostatic repulsion is diminished by Ca2+ binding, which makes each domain four units less negative, or by salt screening, the domains move closer together. In fact the separation appears insensitive to electrostatics in a range that includes the Ca2+-bound form at physiological salt concentration. This result implies that the domains of calmodulin can come together and interact with multiple targets with no significant free energy penalty due to electrostatics.

Materials and methods

Synthesis and purification of CysEF4

CysEF4, comprising residues 117–148 of human calmodulin and containing an N-terminal Cys residue in position 117 in place of the native Thr, was synthesized using solid phase methodology, employing Fmoc chemistry and the PAL resin, as described elsewhere [52]. The cleavage, using trifluoroacetic acid containing anisole, thioanisole and 1,2-ethanedithiol (ratio v/v 4.5 : 0.10 : 0.25 : 0.15) furnished the peptide with a C-terminal carboxamide and a free N-terminal amine. The peptide was purified to homogeneity by reversed phase HPLC using a C4 Vydac column and its identity was confirmed by mass spectrometry.

Cloning and expression of EF123-intein-CBD

The gene for the calmodulin EF123 fragment (residues 1–116) with the S17C mutation was amplified from the gene for intact human calmodulin-S17C in the PetSac vector [25] using PCR and two primers, Pet1 (5′GAGACCACAACGGTTTCCCTCTAGA) and CaMSap116rev (5′CGCGCTTAGGGCTCTTCCGCACAGCTTCTCACCAAGGTTTGTC). The underlined sequence in CaMSap116rev anneals with the DNA segment coding for residues 108–116. The DNA was digested with NdeI and SapI (New England Biolab, Beverly, MA, USA) to yield a fragment coding for residues 1–116 of calmodulin with the capability to ligate directly before the codon for the N-terminal Cys residue of the intein with no extra amino acids in between. The digested fragment was cloned into the pTXB1 plasmid of the Impact system (New England Biolab), which had been digested by the same (NdeI and SapI) enzymes. This construct allows for the expression of EF123 fused to the N-terminus of an intein connected to a chitin binding domain. Plasmids containing the correct sequence were transformed into the Escherichia coli strain ER2566 using electroporation. Single colonies were picked to inoculate 50 mL LB containing 50 μg·mL−1 ampicillin for overnight cultures. Overnight culture (5 mL) was transferred to each 500 mL day culture of LB with 50 μg·mL−1 ampicillin. Expression of the EF123-intein-CBD fusion protein was initiated through induction with 100 μg·mL−1 isopropyl thio-β-d-galactoside when D600 reached 0.6. Four hours after induction, the cells were harvested by centrifugation at 5000 g for 5 min and the pellet was stored at −80 °C until purification.

Purification of EF123-intein-CBD

Pellets were thawed and lysed by suspension in H2O (30 mL per pellet from 1 L culture). The suspension was poured into boiling buffer A (20 mm Tris/HCl, 1 mm EDTA, pH 7.5; 50 mL per 30 mL suspension). The mixture was heated to 80 °C, followed by immediate cooling on ice and centrifugation at 27 000 g for 10 min. The supernatant was supplemented with 0.1 m NaCl from a concentrated stock and applied to a 3.4 × 10 cm DEAE-cellulose column equilibrated in buffer A with 0.1 m NaCl. The pellet was sonicated in 30 mL buffer A and then poured into 50 mL boiling buffer A, with heating of the mixture to 80 °C followed by immediate cooling on ice and centrifugation at 27 000 g for 10 min. The supernatant was supplemented with 0.1 m NaCl and applied to the same DEAE-cellulose column as above. CaM123-intein-CBD was eluted using a linear gradient in buffer A from 0.1 to 0.5 m NaCl (total gradient volume 800 mL). The fractions containing CaM123-intein-CBD from one purification from a 4 L culture were pooled and lyophilized in eight identical aliquots of 25 mg CaM123-intein-CBD each. The yield was ~ 200 mg CaM123-intein-CBD from 4 L culture.

On column labelling and elution of Alexa488-EF123-thioester

One CaM123-intein-CBD aliquot from above (25 mg) was dissolved in 20 mL H2O, supplemented with 2 mL of chitin matrix equilibrated in buffer B (10 mm Tris, 0.5 m NaCl, 1 mm EDTA, pH 7.5) and placed on a rocker for 1 h at 4 °C. The suspension was poured into an empty column and the unbound material was collected. The column was then washed with 10 column volumes of buffer B. Labelling was achieved by the addition of 2 mL of buffer B containing 5 m equivalents of Alexa488-maleimide (added from a 5 mg·mL−1 stock in dimethylsulfoxide). The flow of the column was halted and the column was wrapped in aluminium foil and incubated for 2 h. Cleavage of the intein tag was initiated by flushing the column with one column volume of 10 mm Tris, 0.4 m NaCl, 50 mm MESNA, pH 8.0, to release Alexa488-EF123-thioester. The flow of the column was stopped and the column was left at 4 °C overnight. Alexa488-EF123-thioester was eluted by applying a column volume of a freshly prepared solution of 50 mm MESNA in 10 mm Tris, 0.15 m NaCl, pH 8.0. This elution protocol was repeated three times with 24-h intervals. The collected fractions were analysed by SDS/PAGE and agarose gel electrophoresis and EF123-containing fractions were pooled and lyophilized. From the design of the Impact system, one would expect only the EF123-thioester to be eluted from the column; however, it was found to be contaminated with the cleaved intein tag. In order to obtain pure Alexa488-EF123-thioester, the lyophilized fractions were therefore purified by gel filtration on a 3.4 × 180 cm Sephadex G50 superfine column (GE Healthcare, Uppsala, Sweden) using 50 mm ammonium acetate, pH 6.5, as running buffer, and subsequently desalted on a Sephadex G25 column using water as eluent. Based on analysis by MALDI-TOF mass spectrometry, the thioester remained completely stable under these conditions. Lyophilized fractions were dissolved and purified by gel filtration. The yield obtained was 4 mg Alexa488-EF123-thioester from one CaM123-intein-CBD aliquot (25 mg).

Ligation

Fractions containing Alexa488-EF123-thioester were supplemented with CysEF4, lyophilized and dissolved in H2O to increase the total concentration of both fragments. The ligation was carried out in 10 mm Tris buffer at pH 7.5, with 1 mm CaCl2 or 1 mm EDTA, with and without 8 m urea. The ligation step was monitored as a function of time, over 2–96 h, by reversed phase HPLC (Vydac, C4, analytical column 124TP54, 5 μm particle size, column dimensions 4.6 mm inner diameter × 250 mm), employing a gradient of 20–60% CH3CN : H2O 9 : 1 and H2O, both containing 0.1% CH3COOH, over 40 min. Under these conditions Alexa488-EF123-thioester and CysEF4 eluted together as a broad peak with a retention time of ~ 16 min. The retention time of the ligated product was ~ 27 min, which corresponds to the same retention time as native calmodulin. The HPLC analysis indicated a near complete conversion of starting materials to product over 96 h. A residual broad hump on the baseline remained due to unreacted starting materials, with some overlap with the product peak which made it difficult to assess reaction yield. Although the reconstitution of the C-terminal domain of calmodulin from peptide fragments corresponding to EF3 and EF4 is favoured by the presence of Ca2+ [53, 54], the highest ligation yields in this study, between Alexa488-EF123-thioester and CysEF4, were obtained in the presence of EDTA without urea. Evidence for successful ligation was also shown by SDS/PAGE (Fig. S4) and gel filtration on a Superdex 75 column. A larger scale ligation reaction was carried out and the product was purified on a preparative C4 column, followed by desalting on a Sephadex G25 gel filtration column (GE Healthcare) in water. Prior to fluorescence measurements, the ligated product was further purified from unligated components using size exclusion chromatography on a 1 × 30 cm Superdex 75 column.

Chromophore labelling of ligation product with Alexa546

In the last step, the acceptor chromophore was incorporated. The acceptor concentration was 10 times higher than the donor concentration in the ligation process of the doubly labelled CaM, ensuring that almost all molecules with donor have an acceptor. Alexa546-maleimide was added in a 10 m excess to the ligation product followed by incubation in the dark for 2 h at room temperature and gel filtration to remove excess chromophore.

CaM-D

CaM-D used in the fluorescence studies was produced by Alexa488-maleimide labelling of the calmodulin S17C mutant [25], following a similar protocol to that described above.

Fluorescence spectroscopy

Fluorescence emission spectra were recorded on a PerkinElmer (Foster City, CA, USA) Luminescence Spectrometer LS 50 B connected to a thermostatted circulating water bath set at 25 ˚C. The donor and acceptor concentrations used in the fluorescence experiments were quantified by peak convolution of a measured absorbance spectrum from the individual absorbance spectra of the two fluorophores (Fig. S5), coupled to IgG as given from the fluorophore manufacturer. The concentrations of calmodulin labelled with donor or with donor and acceptor were then adjusted so that the donor concentration was kept constant in all experiments.

Acknowledgements

This work was financed by grants from the Swedish Research Council (SL), the Royal Physiographic Society, Lund (CS) and the Crafoord Foundation (SL) and with funds from the Howard Hughes Medical Institute to Haverford College and the Koshland Integrated Science Center Program at Haverford College (KSÅ).

Ancillary