Allostery in the Ferredoxin Protein Motif Does Not Involve a Conformational Switch

Allostery in the ferredoxin protein motif does not involve a conformational switch

Rachel Nechushtaia, Heiko Lammertb, Dorit Michaelia, Yael Eisenberg-Domovicha, John A. Zurisc, Maria A. Lucac, Dominique T. Capraroc, Alex Fisha, Odelia Shimshona, Melinda Royc, Alexander Schugb,d, Paul C. Whitforde, Oded Livnaha, José N. Onuchicb,1, and Patricia A. Jenningsc,1

aLife Science Institute and The Wolfson Centre for Applied Structural Biology, The Hebrew University of Jerusalem, Edmond J. Safra Campus, Givat Ram, Jerusalem 91904, Israel; bCenter for Theoretical Biological Physics and the Department of Physics, University of California at San Diego, 9500 Gilman Drive, La Jolla, CA 92093-0375; cDepartment of Chemistry and Biochemistry, University of California at San Diego, 9500 Gilman Drive, La Jolla, CA 92093- 0375; dDepartment of Chemistry, Umeå University, Umeå, Sweden; and eLos Alamos National Laboratory, Theoretical Biology and Biophysics, MS K710, Los Alamos, NM 87545

Contributed by José N. Onuchic, December 28, 2010 (sent for review December 11, 2010)

Regulation of protein function via cracking, or local unfolding substantial theoretical (9–13) and experimental (14, 15) work and refolding of substructures, is becoming a widely recognized has aimed at revealing the details of the functionally relevant mechanism of functional control. Oftentimes, cracking events are transition-state ensembles. These investigations have shown localized to secondary and tertiary structure interactions between that rearrangements may involve a degree of cracking, or loca- domains that control the optimal position for catalysis and/or the lized unfolding and refolding, which serves to reduce the asso- formation of protein complexes. Small changes in free energy ciated free-energy barriers (16, 17). By modulating the balance associated with ligand binding, phosphorylation, etc., can tip the of these order–disorder transitions that control domain–domain balance and provide a regulatory functional switch. However, interactions/orientations, cells may tune the functional dynamics. understanding the factors controlling function in single-domain Despite these conceptual advances in understanding enzyme reg- proteins is still a significant challenge to structural biologists. We ulation, these principles may not be applicable to single-domain investigated the functional landscape of a single-domain plant- proteins if they do not exhibit large-scale structural rearrange- type ferredoxin protein and the effect of a distal loop on the elec- ments. In those cases, alternative modes of functional control BIOPHYSICS AND

tron-transfer center. We find the global stability and structure are are likely employed by the cell. COMPUTATIONAL BIOLOGY minimally perturbed with mutation, whereas the functional prop- In the present study, we investigate the relationship between erties are altered. Specifically, truncating the L1,2 loop does not structure, dynamics, and function for a single-domain plant-type lead to large-scale changes in the structure, determined via X-ray ferredoxin (Fdx) protein. This Fdx folding motif is among the crystallography. Further, the overall thermal stability of the protein most abundant in nature, and it has been well studied previously is only marginally perturbed by the mutation. However, even biochemically (18–23), making Fdx an excellent candidate for though the mutation is distal to the iron–sulfur cluster (∼20 Å), it quantitative, biophysical characterization. Moreover, the plant- leads to a significant change in the redox potential of the iron– type Fdx folding motif has been found in numerous redox pro- sulfur cluster (57 mV). Structure-based all-atom simulations indi- teins and enzymes, as well as in functionally unrelated proteins cate correlated dynamical changes between the surface-exposed that do not contain the [2Fe-2S] cluster, such as ubiquitin and loop and the iron–sulfur cluster-binding region. Our results suggest the immunoglobulin-binding domain of protein G (24, 25). Cir- intrinsic communication channels within the ferredoxin fold, com- cular permutation of the secondary structure units helped identify posed of many short-range interactions, lead to the propagation of previously unrecognized homologs, such as the mercury transpor- long-range signals. Accordingly, protein interface interactions that ter protein MerP (26). This versatility suggests that a fundamen- ’ involve L1,2 could potentially signal functional changes in distal tal understanding of the fold s dynamical properties may facilitate regions, similar to what is observed in other allosteric systems. the design of proteins with tightly regulated, and unique, functional roles (27). The development of “modular” proteins that electron transfer ∣ functional energy landscape ∣ iron–sulfur proteins ∣ can be artificially assembled into large functional complexes is protein folding also at the heart of modern synthetic biology (28). When com- plexed with iron–sulfur clusters, Fdx proteins possess electron ver the last several decades, our understanding of protein transport properties, which suggests they could be a component function has evolved from a rather static perspective, where in artificial photosynthetic devices and other biologically moti- O vated molecular “circuitry.” signaling and function have been understood through surface The potential role of iron–sulfur (FeS) proteins in synthetic complementarity arguments, such as the “lock and key” paradigm biology is inspired by their role inside the cell. These proteins (1, 2), to a more dynamic view where protein conformational are key players in the three life-essential processes: photosynth- fluctuations are inextricably linked to function (3). As our under- esis, nitrogen fixation, and respiration (29, 30). The plant-type standing of protein dynamics expands, we are revealing many Fdx proteins carry a single surface-exposed [2Fe-2S] iron–sulfur mechanisms by which proteins exploit conformational fluctuations to perform cellular function. In multidomain proteins, relative repositioning of domains is often linked to their levels Author contributions: R.N., H.L., D.M., Y.E.-D., J.Z., M.A.L., D.T.C., A.F., O.S., M.R., A.S., of activity. For example, large-scale domain rearrangements in P.C.W., O.L., J.N.O., and P.A.J. designed research; R.N., H.L., D.M., Y.E.-D., J.Z., M.A.L., D.T.C., A.F., O.S., M.R., and A.S. performed research; R.N., H.L., P.C.W., O.L., and P.A.J. the four-domain Src kinase (4) and C-terminal Src kinase (5, 6) analyzed data; and R.N., H.L., P.C.W., O.L., J.N.O., and P.A.J. wrote the paper. lead to these proteins being in so-called “on” or “off” states, and The authors declare no conflict of interest. functional regulation may be obtained by adjusting the balance Data deposition: The crystallography, atomic coordinates, and structure factors have been between these conformations (7). In proteins such as adenylate deposited in the Protein Data Bank, www.pdb.org (PDB ID code 3P63). kinase, domain rearrangements can be rate limiting during each 1To whom correspondence may be addressed. E-mail: [email protected] or round of catalysis (8), where the enzyme cycles between ligand- [email protected]. competent and ligand-release conformations. Because the kinet- This article contains supporting information online at www.pnas.org/lookup/suppl/ ics of these transitions control the activity level of the protein, doi:10.1073/pnas.1019502108/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1019502108 PNAS Early Edition ∣ 1of6 Downloaded by guest on September 30, 2021 cluster (ISC) (19), and homologous proteins are found in all organisms, e.g., adrenodoxin in human mitochondria (20). The E oxidation-reduction potentials ( m) of these Fdx proteins range C’ from −325 mV to around −450 mV (21), making Fdx one of the strongest soluble reducing agents in nature (22). The plant-type Fdx functions primarily as the electron acceptor of photosystem I, and upon reduction Fdx functions as an electron donor in a myr- iad of key metabolic pathways (e.g., NADPþ reduction, carbon assimilation, nitrite and sulfite reduction, glutamate synthesis, and thioredoxin reduction, etc.) (22, 23, 31–35). The Fdx fold (22, 23, 31, 36), also termed the β-grasp, or UB roll, is found throughout the cell. The minimum core elements N’ consist of a four-stranded mixed β-sheet with an α-helix packed across one face. Loop insertions and additional strands/helices may be incorporated into this fold for functional purposes. A ribbon diagram of the tertiary fold of Mastigocladus laminosus Fdx (mFdx), is shown in Fig. 1 (19, 37). Substitution of the flexible L1,2 loop (residues 10–14) of wild-type Fdx (mFdxWT) with the rigid β-turn of a mesophilic Fdx (residues 10–12) (here we refer to this substituted mutant as mFdxΔL1;2),* yields a protein that is functional at 23 °C but loses electron-transfer (ET) activity at 55 °C (19), whereas mFdx is fully active. This loss of function is not due to dissociation of the [2Fe-2S] cluster, as this moiety is stable to 75 °C. The highly spe- L1,2 cific effects of this isolated mutation is a well-defined feature, β1 making mFdx ideal for exploring the relationship between a protein’s functional and dynamical properties. Through a combined experimental-theoretical investigation, we find that the network of short-range residue–residue interactions that define the Fdx fold directly facilitates long-range communication between L1,2 and the ISC pocket. Protein–film voltammetry (PFV) showed that the deletion of loop L1,2 alters the redox potential of the functional [2Fe-2S] center, ca. 20 Å removed from the site of mutation. X-ray crystallography and temperature-dependent CD studies revealed that, despite changes in β2 the ET rates at high temperatures, the global structure and stability of the protein are robust. The only noticeable structural Fig. 1. Loop deletion does not have global effects on the structure of mFdx. changes (i.e., larger than the uncertainties in the crystallographic (Top) Ribbon representation of M. laminosus Fdxwt (orange) and mFdxΔL1;2 coordinates) are in residues that directly interact with the mutated deletion mutant (blue). The overall fold of the two molecules, as well as loop. Because these functional changes could not be attributed to the position of the 2Fe-2S cluster (shown as spheres), is similar. (Bottom) changes in structure, or stability, we employed all-atom structure- The L1,2 region connecting β1 and β2 in mFdxΔL;1;2 is two residues shorter based simulations (38) to characterize the native-basin fluctua- than the WT, resulting in a type-I β-turn conformation. tions. These simulations revealed transdomain dynamic coupling (i.e., between two regions on opposing sides of a single domain) of restrained refinement using REFMAC5 (40), clearly indicated between the L1,2 loop and the [2Fe-2S] cluster. Taken together, the presence of a β-turn between β1 and β2. The structure these results demonstrate that the network of short-range native was further refined in the resolution range of 4.0–2.3 Å using interactions in the Fdx fold possesses intrinsic communication REFMAC5. The structure was fitted into electron density maps channels that span the domain. These modes of communication using the graphics program O (41). The final model of mFdxΔL1;2 connect functional centers to distal sites, which suggest previously R ¼ 20 9 R ¼ 28 0 – ( value . ; free . ) consists of residues 1 97 for both undescribed modes of functional control in single-domain pro- molecules and two [2Fe-2S] clusters (Table S1). teins within the cell. The structures of mFdxWT and mFdxΔL1;2 are shown in Fig. 1. The overall backbone fold of mFdxΔL1;2 is superimposable with Results that of mFdxWT, with a rmsd of 0.35 Å (excluding the L1,2 resi- Minimal Structural Perturbation Observed upon Loop Deletion in dues). The only major difference observed between the two struc- mFdx. Because conformational rearrangements are often coupled tures is localized to the mutated L1,2 loop (Fig. 1), which is over to changes in the activity of proteins, we first determined the crys- β ΔL1;2 20 Å removed from the ISC and adopts a shorter type-I -turn tal structure of mFdx (2.3-Å resolution) to see if structural conformation in mFdxΔL1;2. The [2Fe-2S] moieties and the changes associated with loop deletion are sufficient to rationalize surrounding side chains (residues within 4 Å of the ISC) in the the large effect it has on the protein’s function. The structure was cluster-binding pocket are also superimposable, with a rmsd of solved via molecular replacement methods using the AMoRe 0.15 Å. Thus, the change in electron-transfer efficiency upon module of CCP4i (39), where mFdxWT (PDB ID code 1RFK) mutation cannot be attributed to any apparent shift in the pro- was used as a search model (excluding solvent). The solution ’ R tein s configuration. resulted in an value of 34.4% and correlation coefficient of 56.5% in the resolution range of 10–4.0 Å. The initial F -F obs calc Loop Deletion Alters the Redox Potential in mFdx. Despite the mini- and 2F -F electron density maps, calculated after five cycles obs calc mal effect that L1,2 deletion has on the tertiary structure, this deletion maintains activity well below the optimal temperature *Starting at E10, the sequence EAEGL of mFdx is replaced with PTG- -. of the thermophilic organism but abolishes the electron transport

2of6 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1019502108 Nechushtai et al. Downloaded by guest on September 30, 2021 properties of this protein in vitro at 55 °C (19). Electron transport of molar quantities of salt and extremes in pH) will be necessary efficiencies are the result of a combination of intra- and interpro- to identify the origins and limits of this kind of robust functional tein interactions and properties including distance, orientation, dynamics to changes in solution conditions, though this is beyond hydrogen bonding, van der Waals, electrostatic interactions, and the scope of the present investigation. folding characteristics. To determine if loop mutation has a spe- When studying electron transport processes, it is often difficult cific effect on the individual Fdx domain, we measured the redox to partition the contributions of the donor and the acceptor potential and global stabilities (see below) of the WTand mutant because intermolecular interactions will also affect the transport proteins. properties. Here, by measuring the effects of distal loop swap on Using protein–film voltammetry (42). we measured the redox the redox potential of an isolated Fdx protein, we have identified WT potential (Em) of the ISC at pH 7.0 for both mFdx and a direct intramolecular effect that loop characteristics have on the mFdxΔL1;2. Plots of applied potential versus the measured current active site. for mFdxWT and mFdxΔL1;2 are given in Fig. 2. mFdxWT has a maximum cathodic peak at −345 mV and maximum anodic peak Loop Deletion Has Minimal Effects on Global Protein Stability. As at −305 mV, yielding an Em of −325 mV. This value is in excel- deletion of L1,2 does not lead to substantial structural changes, lent agreement with previous reports achieved with solution titra- an alternative explanation for the observed functional coupling tions coupled with electron paramagnetic detection methods between these distal sites is that mutation reduces the stability (21). It is important to note that while the addition of neomycin of the protein, which results in frequent excursions to denatured, is extremely useful in the collection of protein–film voltammetry or partially denatured, conformations. To address this potential measurements, it has the potential to alter the observed redox effect, we measured the thermal stability of the WTand mutated F mFdx proteins. The fraction of unfolded protein, app, as a func- potential of specific systems (43). However, in the case of mFdx Δ 1;2 the observed redox potential for mFdx is unchanged in the tion of temperature for mFdxWT and mFdx L is shown in presence/absence of neomycin (21). mFdxΔL1;2 exhibits a cathodic Fig. 3. Both proteins are thermally stable and fully folded at peak at −400 mV and maximum anodic peak at −364 mV yield- temperatures below 70 °C. They exhibit cooperative transitions ing an Em of −382 mV and a change in Em values of ∼57 mV. between 70 °C and 80 °C and are fully unfolded at temperatures ΔL1;2 Thus mFdx has a significantly altered Em value, though it above 80 °C. The midpoints of the transitions are 76.1±0.2 °C and WT ΔL1;2 is a suitable Em for accepting electrons from photosystem I 74.0±0.3 °C for mFdx and mFdx , indicating minimal (Em of −550 mV). We next asked whether alterations in the solu- perturbation to the stability of the protein upon loop deletion. tion ionic strength, compatible with the typical ranges used in Because the physiological temperature (55 °C) is far below the BIOPHYSICS AND

enzymatic assays (19, 44), would modify the observed potentials. folding temperatures, this change in stability should lead to a COMPUTATIONAL BIOLOGY We found that addition of 100 mM NaCl does not have a mea- negligible increase in the probability of being unfolded in vivo. surable effect on the respective redox potentials of the WT or Thus, it is unlikely that the large changes in functional dynamics mutant proteins (Fig. 2). This robustness of the redox potential, are the result of this modest change in stability. under physiological changes in ionic strength, may have evolved The lack of coupling between the thermal stability and functional to allow Fdx to function under a wider range of cellular condi- dynamics in the single-domain mFdx protein contrasts with our tions. Much more dramatic changes in conditions (i.e., addition understanding of enzymatic catalysis in many multidomain proteins, such as adenylate kinase. In adenylate kinase, the dynamics of the large-scale rearrangements that govern catalysis are often correlated with the thermal stability of the protein (8). It has been argued thatthecouplingofstabilityandfunctionaldynamicsallowsthesame protein fold to carry out similar functional roles in organisms with different physiological temperatures. Additionally, as a protein’s stability evolves, so does the propensity to crack (9, 10, 17). This phy-

Fig. 2. Redox changes are observed for the ΔL1,2 mutant via protein–film voltammograms of mFdxWT (shown in black) and mFdxΔL1;2 (shown in red). Baseline-subtracted voltammograms of mFdxWT (shown in gray) and mFdxΔL1;2 mutant (shown in red) in 100 mM Bis-Tris 100 mM NaCl pH 7.0 (not shown to scale). Baseline-subtracted voltammograms in low salt (100 mM Bis-Tris pH 7.0) for mFdxWT (black squares) and mFdxΔL1;2 (blue squares) show no major effect of salt concentration on the redox potential. Midpoint potentials (Em) were found by determining the potential equidi- stant from the cathodic and anodic peak potentials of the oxidative and re- WT ΔL1;2 ductive curves. Em values for mFdx and mFdx were determined to be −325 10 mV and −382 10 mV, respectively. Voltammograms were gener- Fig. 3. The thermostabilities of mFdxWT and mFdxΔL1;2 are comparable. Frac- ated using highly oriented edge-plane graphite electrode containing neomy- tion of unfolded protein as a function of temperature for mFdxWT (black cin as a coadsorbate. The reference electrode was 3 MAg∕AgCl. All values triangles) and mFdxΔL1;2 (blue triangles) as measured from the mean-residue shown above are corrected to standard hydrogen electrode (SHE) values. ellipticity at 222 nm.

Nechushtai et al. PNAS Early Edition ∣ 3of6 Downloaded by guest on September 30, 2021 sical-chemical coupling between stability and function enables a A range of sequences to have similar functional properties, so long as sufficient stability is selected. In contrast to this picture, for mFdx we find that substitution of the distal L1,2 loop from a mesophilic sequence into a thermophilic sequence has a negligible effect on its stability and conformation, whereas it has a pronounced effect on the functional properties of the thermophilic protein. These results suggest an alternative mechanism of regulation in mFdx. Specifically, it suggests that the Fdx fold has internal communication channels that allow it to “sense” signals on one side of the domain and translate B them into functional changes in distal residues. By mutating the L1,2 loop, we have artificially “flipped the switch” in mFdx. Because each organism has a unique set of potential interaction partners for Fdx-like proteins, the precise chemical details of each may lead to highly specific effects on the functional residues.

Simulations Suggest a Mechanism of Communication in mFdx. Because we cannot rationalize the functional impact of L1,2 deletion in terms of a conformational change or a change in thermal stability, we employed simulations to explore the possibility that the connectivity of short-ranged native interactions leads to the observed long- range functionally relevant transdomain communication. Specifi- cally, we used simulations that employed a single energetic basin, where only interactions that are formed in the crystal structure are stabilizing, for the apo forms of both mFdxWT and mFdxΔL1;2†. These simulations enabled us to determine whether L1,2 deletion affects the ISC-binding pocket as part of a global change in the dynamics, or through a specific communication route. Because structure-based simulations (38, 45), which are grounded in energy C landscape theory (46), accurately describe the native-ensemble structural fluctuations of biomolecular systems (10, 38, 47–51), we utilized an all-atom structure-based model (38) to explore the mechanism of L1,2-ISC-pocket communication. This model does not explicitly include long- or short-range electrostatic interactions, but it allows for the dissection of the geometric features that may impact function and folding. Therefore, the observed coupling between L1,2 and the binding site would be a consequence of the local connectivity of interactions, which is a property of the Fdx fold. Our simulations reveal that deletion of L1,2 leads to temperature-dependent differences in the structural fluctuations of the WT and mutant proteins as shown in Fig. 4. Specifically, the spatial root-mean-squared fluctuations (rmsf) for mFdxWT and ΔL1;2 Fig. 4. Dynamical effects of the ΔL1,2 mutation are observed in simulations. mFdx are similar across the protein (with the exception of ΔL1;2 WT the site of mutation) in the folded state at lower temperature, (A) Differences in fluctuations between mFdx and the mFdx ,atlow (blue squares) and higher (open triangles) temperature. Red triangles indi- though appreciable differences surface as the temperature is cate the changes at higher T when all non-L1,2 native contacts are exchanged increased. Near the folding temperature, three regions distal between mFdxΔL1;2 and mFdxWT. The plot shows the average taken from the WT to the L1,2 region exhibit large fluctuations in mFdx , which two sets of hybrid simulations. Yellow lines mark the cysteines that coordi- Δ 1;2 are attenuated in mFdx L (Fig. 4A). Changes are observed nate the [2Fe-2S] cluster. (B) Cartoon representation of mFdxWT, with the in the regions proximal to residue A43 and D68 (denoted here Fe-coordinating Cys shown in yellow. Translucent clouds indicate which residues have an rms fluctuation greater than 1.5 Å. The red shading is propor- as r43 and r68) and the C-terminal residues. r43 and r68 are located ’ near the ISC-binding cysteine residues (positions 41, 46, 49, tional to the residue s reduction in rms upon mutation (for the hybrid force fields). The most strongly affected residues (positions 41–48, 63–71, 93–97, and 79). which decrease by more than 5%) are colored solid red. All contacts between To probe whether the dynamic changes in the ISC-binding the mutated residues 10 and 14 (colored sticks) with residues outside of the pocket are the result of communication between it and the L1,2 L1,2 (in blue) are shown as green lines. (C) Spatial cross-correlation matrix loop, as opposed to being the result of many small perturba- showing correlated motion between Cα atoms. tions to the structure, additional “control” simulations were performed. That is, due to minor differences in the atomic coordi- the differences in fluctuations, we also calculated the fluctuations nates of mFdxWT and mFdxΔL1;2, the native atomic contacts for models that employed “hybrid” contact maps. The two hybrid between non-L1,2 residues are 81% identical. To test whether these disparities in the contact maps contribute significantly to force fields were constructed by defining the native contacts as (i) all the non-L1,2 contacts of mFdxWT and the L1,2 contacts ΔL1;2 WT † of mFdx or (ii) including the L1,2 contacts of mFdx and Although the crystal structures have bound ISCs, the simulations were performed for Fdx ΔL1;2 without the ISC. Because communication of dynamics between L1,2 and the ISC-binding the non-L1,2 contacts of mFdx . Regardless of which set of site is the result of the network of interactions within the Fdx fold, and these interactions non-L1,2 contacts was employed, simulations that employed the are present in the simulation (because the ISC-bound configuration was used), it is ΔL1;2 appropriate to probe the degree of coupling from simulations in which the ISC is not L1,2 contacts from mFdx displayed reduced fluctuations in explicitly represented. regions r43 and r68.

4of6 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1019502108 Nechushtai et al. Downloaded by guest on September 30, 2021 Careful examination of the native contacts of the L1,2 regions communication channel is for it to obtain a Fdx-like fold. Here, suggests two mechanisms by which loop deletion may alter the a common feature of multidomain functional dynamics and structural fluctuation in the ISC-binding site. The most mobile mFdx may be seen. That is, similar to adenylate kinase, the glo- regions, which are also the most affected by ΔL1,2 mutation, bal dynamics are encoded in the tertiary fold, and the scale of WT are r43, r68, and the C terminus (Fig. 4 A and B). In mFdx , the motions is coupled to the global stability. For proteins that the L1,2 residues (excluding those in strands β1 and β2) form take on a Fdx-like fold, the network of residue–residue interac- Δ 1;2 14 contacts with residues 35, 37, 54, and 91–94. In mFdx L , tions will necessarily be very similar, because the finite size of L1,2 has only a total of five contacts, all of which are with residues protein side chains will limit the possible residue–residue con- 91 and 92. Because loop deletion results in the loss of all contacts nectivity. If the native interactions are similar between Fdx-like r between L1,2 and residues near 43 (positions 35, 37, and 54), the folded proteins, these communication avenues should also be r observed changes in 43 are reasonable to expect. Similarly, L1,2 preserved. deletion changes the contacts between L1,2 and residues near the From a theoretical perspective, the role of local connectivity C terminus (positions 91 and 92), which explains the changes in demonstrates how high-level functional dynamics may be under- the dynamics of the C-terminal residues. Although there is no stood in terms of simple considerations. If local connectivity gov- obvious, direct, structural connection between changes in L1,2 r r erns the functional modes of communication, one must account and 68, both the C terminus and 43 are located proximal to only for the native interactions in order to reveal these dynamical it. Thus, the changes in dynamics of the C terminus and r43 likely properties. For single-domain proteins that do not undergo large contribute to the changes in r68 dynamics. Inspection of the posi- conformational rearrangements, then one must employ only a tional covariance matrix for the Cα atoms (Fig. 4C) reveals cor- model that accurately describes the fluctuations about a particu- related motion of these three regions. Because these regions contribute to the electronic environment surrounding the iron– lar energetic minimum, such as the structure-based model used sulfur cluster, these results provide a clear mode of communica- here. Although this approach captures the dynamical coupling tion between L1,2 and the system’s electrochemical activity. This between distal regions, future methodological advances may al- type of communication may signal changes across a domain, low one to predict the degree of functional (in this case, the redox which might then affect the electrostatic environment at distal potential) changes from this type of dynamic information. sites and influence protein–protein and protein–ligand interac- This study has revealed a mode of transdomain communica- tions and is the subject of ongoing investigations (52). tion between the distal ISC-binding pocket and the L1,2 loop. Further, this communication does not appear to require long- Discussion range nonspecific interactions, but rather a network of short- BIOPHYSICS AND This study demonstrates that within a single protein domain fold range native interactions. Although these findings shed light COMPUTATIONAL BIOLOGY the network of short-range native interactions can give rise to on the relationship among protein structure, dynamics, and func- long-range modes of communication. Transdomain “cross-talk” tion, the standing challenge will be to identify these functional between a catalytic center and a distal loop is an emerging theme channels a priori. Here, we have explained the basis for an experi- by which single domains can transmit cellular signals (7, 53–55). mentally suggested communication channel between distal sites, In the single-domain protein interleukin-1β, point mutations that but the communication was not predicted without explicitly in- are distal to the receptor binding interface have no effect on sta- cluding a particular perturbation (mutation of the L1,2 loop). bility, yet they have significant effects on the binding properties These results suggest that analyzing interaction networks in pro- (55). In multidomain proteins, there are also examples of trans- teins is a viable approach to identify “hidden” modes of dynamic domain communication across individual domains, including in coupling in proteins. Finally, a complete understanding of how cAMP-dependent protein kinase (56) and in COOH-terminal short-ranged interactions give rise to long-range communication Src kinase (Csk) (7). It was demonstrated (57) that chemical per- will provide guiding principles for designing molecular devices turbations are linked to activity regulation through a “communi- ” that are capable of transmitting and converting chemical signals cation pathway between a distal SH2 domain and the active site. into mechanical activity. In contrast to the present study, NMR experiments revealed that the SH2 domain of Csk populates two structurally similar confor- Methods mations and selection of one also contributes to the functional Crystals of mFdxΔL1;2 were obtained using the hanging-drop vapor diffusion changes. In mFdx, we found no evidence of a conformational method. Crystallographic data were collected at the European Synchrotron change within the domain, yet changes in the fluctuations of the Radiation Facility beamline ID29 (λ ¼ 0.976 Å) on an Area Detector Systems distal L1,2 loop are coupled to the functionally relevant dynamics Corporation Quantum 210 CCD detector and were integrated and scaled of the ISC pocket. using the HKL2000 suite (59). Protein–film voltammograms were generated In light of these results, it is worthwhile to consider the notion using a CHI730C Electrochemical Workstation (CH Instruments Inc.). The elec- of allosteric communication in proteins. The term allostery trode was prepared as reported earlier (60). All PFV measurements were ta- describes the observation of regulation of activity within a protein ken at 4 °C to enhance signal to noise (42). The SOAS software package (61) upon ligand binding at one site, and transmission of information (courtesy of Christophe Léger, Centre National de la Recherche Scientifique, to a second, distal site. The conventional view holds that confor- Paris, France) was used to perform the baseline subtraction of voltammetry data. All-atom structure-based force fields (38) for the Apo forms of mFdxWT mational perturbations or rearrangements convey the regulatory and mFdxΔL1;2 were prepared from the X-ray structures using the effect. However, it now appears that the mechanism of allosteric SMOG@ctbp Web tool (http://smog.ucsd.edu) (62), with tertiary contacts re- regulation may be more universal than the classical interpretation presented by Gaussian-based potentials (63). Simulations were performed (58). In this updated view, allosteric signals can be transmitted with the Gromacs software package (64). Structural figures were prepared through multiple, preexisting pathways and which pathway is cho- with VMD (65). Further details are provided in SI Text. sen depends on topologies, binding events, and environmental conditions. Hence, long-range coupling via local interactions, ACKNOWLEDGMENTS. We are thankful to Sichun Yang for many scientific as we observe here, can relay effects without protein structure discussions in the early stages of this research. This work was supported by being affected. the Center for Theoretical Biological Physics sponsored by the National Science Because the physical origin of transdomain communication in Foundation (NSF) (Grant PHY-0822283) and NSF Grant NSF-MCB-1051438, National Institutes of Health Grant GM-54038, and the Los Alamos National mFdx can be traced back to the network of short-range, native, Laboratory Laboratory Directed Research and Development program. R.N. residue–residue interactions, it is reasonable to infer that the acknowledges Israel Science Foundation ISF-863/09. P.C.W. is funded by a primary requirement for a protein to possess this dynamical Los Alamos National Laboratory Director’s Postdoctoral Fellowship.

Nechushtai et al. PNAS Early Edition ∣ 5of6 Downloaded by guest on September 30, 2021 1. Haldane JBS (1930) Enzymes (Longmans Green, New York). 34. Xu X, et al. (2009) Ternary protein complex of ferredoxin, ferredoxin:thioredoxin 2. Pauling L (1948) Nature of forces between large molecules of biological interest. reductase, and thioredoxin studied by paramagnetic NMR spectroscopy. J Am Chem Nature 161:707–709. Soc 131:17576–17582. 3. Frauenfelder H, Sligar SG, Wolynes PG (1991) The energy landscapes of motions of 35. Muraki N, Seo D, Shiba T, Sakurai T, Kurisu G (2010) Asymmetric dimeric structure of proteins. Science 254:1598–1603. ferredoxin-NAD(P)+ oxidoreductase from the green sulfur bacterium Chlorobaculum þ 4. Huse M, Kuriyan J (2002) The conformational plasticity of protein kinases. Cell tepidum: Implications for binding ferredoxin and NADP . J Mol Biol 401:403–414. 109:275–282. 36. Overington JP (1992) Comparison of three-dimensional structures of homologous 5. Ogawa A, et al. (2002) Structure of the carboxyl-terminal Src Kinase, Csk. J Biol Chem proteins. Curr Opin Struct Biol 2:394–401. 277:14351–14354. 37. Fish A, Lebendiker M, Nechushtai R, Livnah O (2003) Purification, crystallization and 6. Jamros MA, et al. (2010) Proteins at work: A combined small angle x-ray scattering and preliminary X-ray analysis of ferredoxin isolated from thermophilic cyanobacterium theoretical determination of the multiple structures involved on the protein kinase Mastigocladus laminosus. Acta Crystallogr D 59:734–736. functional landscape. J Biol Chem 285:36121–36128. 38. Whitford PC, et al. (2009) An all-atom structure-based potential for proteins: Bridging 7. Mill JE, et al. (2007) A novel disulfide bond in the SH2 domain of the C-terminal Src minimal models with empirical forcefields. Proteins 75:430–441. kinase controls catalytic activity. J Mol Biol 365:1460–1468. 39. Anonymous (1994) The CCP4 suite: Programs for protein crystallography. Acta 8. Wolf-Watz M, et al. (2004) Linkage between dynamics and catalysis in a thermophilic- Crystallogr D 50:760–763. mesophilic enzyme pair. Nat Struct Biol 11:945–949. 40. Murshudov GN, Vagin AA, Dodson EJ (1997) Refinement of macromolecular structures – 9. Miyashita O, Onuchic JN, Wolynes PG (2003) Nonlinear elasticity, proteinquakes, and by the maximum-likelihood method. Acta Crystallogr D 53:240 255. the energy landscapes of functional transitions in proteins. Proc Natl Acad Sci USA 41. Jones TA, Zou JY, Cowan SW, Kjeldgaard M (1991) Improved methods for building 100:12570–12575. protein models in electron density maps and the location of errors in these models. – 10. Whitford PC, Miyashita O, Levy Y, Onuchic JN (2007) Conformational transitions of Acta Crystallogr A 47:110 119. adenylate kinase: Switching by cracking. J Mol Biol 366:1661–1667. 42. Armstrong FA, Butt JN, Sucheta A (1993) Voltammetric studies of redox-active centers – 11. Lu Q, Wang J (2009) Kinetics and statistical distributions of single-molecule conforma- in metalloproteins adsorbed on electrodes. Methods Enzymol 227:479 500. tional dynamics. J Phys Chem B 113:1517–1521. 43. Zu Y, et al. (2003) Reduction potentials of rieske clusters: Importance of the 12. Brokaw JB, Chu JW (2010) On the role of substrate binding and hinge unfolding in coupling between oxidation state and histidine protonation state. Biochemistry – conformational changes of adenylate kinase. Biophys J 99:3420–3429. 42:12400 12408. 13. Arora K, Brooks CL (2007) Large-scale allosteric conformational transitions of adeny- 44. Fish A, Lebendiker M, Nechushtai R, Livnah O (2003) Purification, crystallization and late kinase appear to involve a population shift mechanism. Proc Natl Acad Sci USA preliminary X-ray analysis of ferredoxin isolated from thermophilic cyanobacterium – 104:18496–18501. Mastigocladus laminosus. Acta Crystallogr D 59:734 736. 14. Bae E, Phillips GN (2006) Roles of static and dynamic domains in stability and catalysis 45. Clementi C, Nymeyer H, Onuchic JN (2000) Topological and energetic factors: What determines the structural details of the transition state ensemble and “en-route” of adenylate kinase. Proc Natl Acad Sci USA 103:2132–2137. intermediates for protein folding? An investigation for small globular proteins. 15. Hanson JA, et al. (2007) Illuminating the mechanistic roles of enzyme conformational J Mol Biol 298:937–953. dynamics. Proc Natl Acad Sci USA 104:18055–18060. 46. Bryngelson JD, Onuchic JN, Socci ND, Wolynes PG (1995) Funnels, pathways, and the 16. Whitford PC, Onuchic JN, Wolynes PG (2008) Energy landscape along an enzymatic energy landscape of protein-folding—A synthesis. Proteins 21:167–195. reaction trajectory: Hinges or cracks? HFSP J 2:61–64. 47. Roy M, et al. (2005) The native energy landscape for interleukin-1b. Modulation of the 17. Olsson U, Wolf-Watz M (2010) Overlap between folding and functional energy population ensemble through native-state topology. J Mol Biol 348:335–347. landscapes for adenylate kinase conformational change. Nat Commun 10.1038/ 48. Hyeon C, Jennings PA, Adams JA, Onuchic JN (2009) Ligand-induced global transitions ncomms1106. in the catalytic domain of protein kinase A. Proc Natl Acad Sci USA 106:3023–3028. 18. Beinert H, Holm RH, Münck E (1997) Iron-sulfur clusters: Nature’s modular, multipur- 49. Whitford PC, et al. (2010) Accommodation of aminoacyl-tRNA into the ribosome pose structures. Science 277:653–659. involves reversible excursions along multiple pathways. RNA 16:1196–1204. 19. Fish A, Danielli T, Ohad I, Nechushtai R, Livnah O (2005) Structural basis for the 50. Ratje AH, et al. (2010) Head swivel on the ribosome facilitates translocation by means thermostability of ferredoxin from cyanobacterium Mastigocladus laminosus. J Mol of intra-subunit tRNA hybrid sites. Nature 468:713–716. Biol 350:599–608. 51. Haliloglu T, Bahar I (1998) Coarse-grained simulations of conformational dynamics of 20. Grinberg AV, et al. (2000) Adrenodoxin: Structure, stability and electron transfer proteins: Application to apomyoglobin. Proteins 31:271–281. properties. Proteins 40:590–612. 52. Pyaram K, Kieslich CA, Yadav VN, Morikis D, Sahu A (2010) Influence of electrostatics 21. Cammack R, et al. (1977) Midpoint redox potentials of plant and algal ferredoxin. on the complement regulatory functions of kaposica, the complement inhibitor of Biochem J 168:205–209. Kaposi’s sarcoma-associated herpesvirus. J Immunol 184:1956–1967. 22. Fukuyama K (2004) Structure and function of plant-type ferredoxin. Photosynth Res 53. Tsai C-J, del Sol A, Nussinov R (2008) Allostery: Absence of a change in shape does not 81:289–301. imply that allostery is not at play. J Mol Biol 378:1–11. 23. Rypniewski WR, et al. (1991) Crystallization and structure determination of 2.5-.ANG. 54. Bahar I, Chennubhotla C, Tobi D (2007) Intrinsic dynamics of enzymes in the unbound resolution of the oxidized iron-sulfur [2Fe-2S] ferredoxin isolated from Anabaena. state and relation to allosteric regulation. Curr Opin Struct Biol 17:633–640. – Biochemistry 30:4126 4131. 55. Heidary DK, Roy M, Daumy GO, Cong Y, Jennings PA (2005) Long-range coupling 24. Vriend G (1991) Detection of common three-dimensional substructures in proteins. between separate docking sites in interleukin-1β. J Mol Biol 353:1187–1198. – Proteins 11:52 58. 56. Andersen MD, Shaffer J, Jennings PA, Adams JA (2001) Structural characterization of 25. Overington J (1992) Structural constraints on residue substitution. Genet Eng protein kinase A as a function of nucleotide binding. J Biol Chem 276:14204–14211. – 14:231 249. 57. Hamuro Y, et al. (2002) Phosphorylation driven motions in the COOH-terminal Src 26. Steele RA, Opella SJ (1997) Structures of the reduced and mercury-bound forms of kinase, Csk, revealed through enhanced hydrogen-deuterium exchange and mass – MerP, the periplasmic. Biochemistry 36:6885 6895. spectrometry (DXMS). J Mol Biol 323:871–881. 27. Frolow F, Harel M, Sussman JL, Mevarech M, Shoham M (1996) Insights into protein 58. del Sol A, Tsai C-J, Ma B, Nussinov R (2009) The origin of allosteric functional adaptation to a saturated salt environment from the crystal structure of a halophilic modulation: Multiple pre-existing pathways. Structure 17:1042–1050. – 2Fe-2S ferredoxin. Nat Struct Biol 3:452 458. 59. Otwinowski Z, Minor W (1997) Processing of X-ray diffraction data collected in 28. Purnick PEM, Weiss R (2009) The second wave of synthetic biology: From modules to oscillation mode. Methods Enzymol 276:307–326. systems. Nat Rev Mol Cell Bio 10:410–422. 60. Bak DW, Zuris JA, Paddock ML, Jennings PA, Elliott SJ (2009) Redox characterization of 29. Sheftel AD, et al. (2010) Humans possess two mitochondrial ferredoxin, Fdx1 and Fdx2, the FeS protein MitoNEET and impact of thiazolidinedione drug binding. Biochemistry with distinct roles in steroidogenesis, heme and Fe/D cluster biosynthesis. Proc Natl 48:10193–10195. Acad Sci USA 107:11775–11780. 61. Fourmond V, et al. (2009) SOAS: A free software to analyse electrochemical data and 30. Rees DC, Howard JB (2003) The interface between the biological and inorganic worlds: other one-dimensional signals. Bioelectrochemistry 76:141–147. Iron-sulfur metalloclusters. Science 300:929–931. 62. Noel JK, Whitford PC, Sanbonmatsu KY, Onuchic JN (2010) SMOG@ctbp: 31. Fukuyama K, Ueki N, Nakamura H, Tsukihara T, Matsubara H (1995) Tertiary structure Simplified deployment of structure-based models in GROMACS. Nucleic Acids Res of [2Fe-2S] ferredoxin from Spirulina platensis refined at 2.5 A resolution: Structural 38:W657–W661. comparisons of plant-type ferredoxin and an electrostatic potential analysis. 63. Lammert H, Schug A, Onuchic JN (2009) Robustness and generalization of structure- J Biochem-Tokyo 117:1017–1023. based models for protein folding and function. Proteins 77:881–891. 32. Crossnoe CR, Germanas JP, LeMagueres P, Mustata G, Krause KL (2002) The crystal 64. Hess B, Kutzner C, van der Spoel D, Lindahl E (2008) GROMACS 4: Algorithms for highly structure of trichomonas vaginalis ferredoxin provides insight into metronidazole efficient, load-balanced, and scalable molecular simulation. J Chem Theory Comput activation. J Mol Biol 318:503–518. 4:435–447. 33. Bes MT, et al. (1999) Crystal structure determination at 1.4 Å resolution of ferredoxin 65. Humphrey W, Dalke A, Schulten K (1996) VMD—Visual Molecular Dynamics. J Mol from the green alga Chlorella fusca. Structure 7:1201–1211. Graphics 14:33–38.

6of6 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1019502108 Nechushtai et al. Downloaded by guest on September 30, 2021