Please cite this article in press as: Hultqvist et al., Energetic Pathway Sampling in a Protein Interaction Domain, Structure (2013), http://dx.doi.org/ 10.1016/j.str.2013.05.010
Structure
Article Energetic Pathway Sampling in a Protein Interaction Domain Greta Hultqvist,1 S. Raza Haq,1 Avinash S. Punekar,2 Celestine N. Chi,1 A˚ke Engstro¨m,1 Anders Bach,3 Kristian Strømgaard,3 Maria Selmer,2 Stefano Gianni,4,* and Per Jemth1,* 1Department
of Medical Biochemistry and Microbiology, Uppsala University, BMC Box 582, SE-75123 Uppsala, Sweden of Cell and Molecular Biology, Uppsala University, BMC, Box 596, SE-75124 Uppsala, Sweden 3Department of Drug Design and Pharmacology, Faculty of Health and Medical Sciences, University of Copenhagen, Universitetsparken 2, DK-2100 Copenhagen, Denmark 4Istituto di Biologia e Patologia Molecolari del CNR, Dipartimento di Scienze Biochimiche ‘‘A. Rossi Fanelli’’, Sapienza Universita ` di Roma, Piazzale A. Moro 5, 00185 Rome, Italy *Correspondence:
[email protected] (S.G.),
[email protected] (P.J.) http://dx.doi.org/10.1016/j.str.2013.05.010 2Department
SUMMARY
The affinity and specificity of protein-ligand interactions are influenced by energetic crosstalk within the protein domain. However, the molecular details of such intradomain allostery are still unclear. Here, we have experimentally detected and computationally predicted interaction pathways in the postsynaptic density 95/discs large/zonula occludens 1 (PDZ)-peptide ligand model system using wild-type and circularly permuted PDZ proteins. The circular permutant introduced small perturbations in the tertiary structure and a concomitant rewiring of allosteric pathways, allowing us to describe how subtle changes may reshape energetic signaling. The results were analyzed in the context of other members of the PDZ family, which were found to contain distinct interaction pathways for different peptide ligands. The data reveal a fascinating scenario whereby several energetic pathways are sampled within one single domain and distinct pathways are activated by specific protein ligands.
INTRODUCTION Several computational and experimental studies have recently addressed the issue of intradomain signaling in protein domains. It is believed that this type of allostery has an effect on ligand affinity and specificity. However, it is unclear how such signals are transferred from distal residues to the ligand binding site. In the absence of clear conformational transitions, changes in backbone or side chain dynamics upon binding are possible mechanisms for intradomain signal transfer (Akke et al., 1993; Frederick et al., 2007; Fuentes et al., 2006; Popovych et al., 2006; Tzeng and Kalodimos, 2009). A prime model system for studies of intradomain interaction networks is the postsynaptic density 95/discs large/zonula occludens 1 (PDZ) protein domain family, ubiquitous proteinprotein interaction modules, which are important in signaling
and scaffolding (Wawrzyniak et al., 2012; (Figures 1A and 1B). We previously set up an experimental system in which we could assess energetic crosstalk between the side chains in the PDZ domain and residues in the peptide ligand by performing double mutant cycles and measuring coupling free energies (DDDGc) with high accuracy and precision (Gianni et al., 2011; Figures 1D and 1E). DDDGc values report any energetic interaction between two residues (Carter et al., 1984) and this technique is a straightforward experimental approach to assess allostery within a protein domain (Horovitz, 1996). Here, we take this analysis one step further by combining it with circular permutation and structure-based computational predictions of interaction pathways (Amitai et al., 2004; Hu et al., 2007; Kannan and Vishveshwara, 1999; Martin et al., 2011). In a circularly permuted protein, the original N and C termini are sealed while new termini are created at another site in the polypeptide chain. A circular permutant has the same amino acid residues as the wild-type and folds to the same structure, but the orientation of some residues may be slightly different. Thus, the circular permutant offers an experimental system with locally perturbed side chain interactions. We previously designed a circular permutant (cpSAP97 PDZ2; Hultqvist et al., 2012) of the second PDZ domain from SAP97 (denoted pseudo wild-type [pwt] SAP97 PDZ2; Chi et al., 2009) by connecting the strand b6 with b1 and disconnecting b1 and b2 (Figure 1A). The crystal structures of pwtSAP97 PDZ2 (Haq et al., 2010) and cpSAP97 PDZ2 (Hultqvist et al., 2012) show that there is little change in the backbone structure upon circular permutation except around the termini (Figure 1C). However, there are a few rearrangements of side chains leading to altered side chain packing in cpSAP97 PDZ2. These structural changes allowed us to investigate how energetic interaction pathways depend on the fine details of the tertiary structure. In this work, we experimentally determined the coupling free energies between residues in these domains and their ligands (Figure 1) and found distinct changes in the pattern of DDDGc upon circular permutation. Further, we identified networks of side chain interactions for pwtSAP97 PDZ2 and cpSAP97 PDZ2 using computational analyses. The subtle spatial alterations in side chain interaction pathways for cpSAP97 PDZ2 as compared to pwtSAP97 PDZ2 rationalize the changes in DDDGc values. A re-analysis of previously published DDDGc
Structure 21, 1–10, July 2, 2013 ª2013 Elsevier Ltd All rights reserved 1
Please cite this article in press as: Hultqvist et al., Energetic Pathway Sampling in a Protein Interaction Domain, Structure (2013), http://dx.doi.org/ 10.1016/j.str.2013.05.010
Structure Energetic Pathway Sampling
A
B
C
D
E
Figure 1. The Experimental Setup (A) Sequence alignment of the four PDZ variants analyzed in the study. Data for PSD-95 PDZ3 and PTP-BL PDZ2 were previously published (Gianni et al., 2011). Data for pwtSAP97 PDZ2 were previously published as linear free-energy relationships (Haq et al., 2012) but are presented here in detail and as coupling free energies. (B) Structure of SAP97 PDZ2 (Haq et al., 2010) rainbow colored from blue (N terminus) to red and in complex with the peptide ligand (dark green, indicated by arrow). The sequence of the wild-type and three mutant peptides is numbered according to convention. The side chains of the two unnatural amino acids Abu (2-aminobutyric acid) and Ape (2-aminopentanoic acid) are shown to the right. (C) Superposition of the crystal structures for pwtSAP97 PDZ2 and its circular permutant cpSAP97 PDZ2 (PDB codes 2X7Z and 4AMH, respectively). Twenty point mutants of each were analyzed with regard to coupling free energies (DDDGcs) to three side chains of the peptide (Val0, Thr 2, and Arg 4). (D) Example of kinetic traces from binding and displacement experiments and calculation of Kd values. (E) The calculated DDDGc values reflect energetic crosstalk between side chains in the peptide ligand (e.g., Thr 2, green) and a side chain in the PDZ (e.g., Ala358, dark gray). See Figure S1 for representative examples of kinetic data.
data for two other PDZ domains, PTP-BL PDZ2 and PSD-95 PDZ3 (Gianni et al., 2011), further demonstrate that interaction pathway prediction is an excellent method to explain allosteric signal transfer in protein domains and that allosteric pathways are highly sequence-specific (Gianni et al., 2011; Livesay et al., 2012). Finally, our results demonstrate that, analogous to conformational sampling, several pathways co-exist in any given PDZ domain and could be specifically selected by distinct peptide ligands. We call this phenomenon energetic pathway sampling. RESULTS A circular permutant and the corresponding wild-type protein comprise an ideal system for comparing the fine details of intradomain energetic pathways due to their very similar three-
dimensional (3D) structures and only occasional differences in side chain conformations. To understand the energetic pathways of the PDZ domain family, we used double mutant cycles to study coupling free energies (DDDGc) between specific residues involved in peptide binding and compared the experimental results to predicted network interaction pathways for pwtSAP97 PDZ2 (Chi et al., 2009) and an engineered circular permutant, cpSAP97 PDZ2 (Hultqvist et al., 2012; Figure 1). To calculate the DDDGc between two mutated side chains with the double mutant cycle approach, the Kd of a binding reaction needs to be measured for four combinations of PDZ domains and ligands: the wild-type PDZ and ligand, the double mutant (with mutation in PDZ and in the ligand), and the two single mutants with mutation in either the PDZ domain or the ligand (Figure 1). In the ideal case, the interaction energy between the
2 Structure 21, 1–10, July 2, 2013 ª2013 Elsevier Ltd All rights reserved
Please cite this article in press as: Hultqvist et al., Energetic Pathway Sampling in a Protein Interaction Domain, Structure (2013), http://dx.doi.org/ 10.1016/j.str.2013.05.010
Structure Energetic Pathway Sampling
Table 1. Coupling Free Energies DDDGc for the Interaction between Side Chain Residues in the PDZ Domain and in the Ligand, Val0, Thr–2, and Arg–4, Respectively DDDGc to Val0 (kcal/mol)
DDDGc to Thr-2 (kcal/mol)
Mutant
pwtSAP97 PDZ2
pwtSAP97 PDZ2
L322A
0.09 ± 0.1
0.0077 ± 0.05
DDDGc to Arg-4 (kcal/mol) cpSAP97 PDZ2 –
pwtSAP97 PDZ2 0.032 ± 0.08
cpSAP97 PDZ2 –
F331A
0.4 ± 0.09
0.23 ± 0.07
0.26 ± 0.1
0.039 ± 0.1
0.034 ± 0.07
A334G
0.14 ± 0.08
0.19 ± 0.1
0.0092 ± 0.08
0.039 ± 0.08
0.021 ± 0.06 0.0097 ± 0.05
V350A
0.08 ± 0.09
0.011 ± 0.08
0.017 ± 0.08
0.095 ± 0.1
K352A
0.11 ± 0.1
0.24 ± 0.1
0.12 ± 0.1
0.0089 ± 0.1
0.011 ± 0.08
A358G
0.13 ± 0.1
0.013 ± 0.09
0.14 ± 0.1
0.1 ± 0.1
0.017 ± 0.05
L365A
0.08 ± 0.1
0.2 ± 0.1
0.071 ± 0.08
0.011 ± 0.1
0.066 ± 0.07 0.032 ± 0.06
L372A
0.19 ± 0.08
0.051 ± 0.08
0.017 ± 0.09
0.0095 ± 0.09
V374A
0.34 ± 0.07
0.074 ± 0.08
0.1 ± 0.09
0.045 ± 0.07
0.13 ± 0.06
E380A
0.16 ± 0.09
0.11 ± 0.08
0.0074 ± 0.08
0.029 ± 0.08
0.022 ± 0.05
A387G
0.24 ± 0.1
0.1 ± 0.1
0.076 ± 0.08
0.1 ± 0.08
0.0018 ± 0.05
V388A
0.32 ± 0.08
0.088 ± 0.08
0.0038 ± 0.08
0.21 ± 0.08
0.078 ± 0.05
A390G
0.13 ± 0.08
0.028 ± 0.06
0.08 ± 0.08
0.0096 ± 0.09
0.03 ± 0.05
L391A
0.12 ± 0.08
0.068 ± 0.06
0.098 ± 0.09
0.077 ± 0.08
0.045 ± 0.05
K392A
0.68 ± 0.1
0.16 ± 0.1
0.028 ± 0.2
0.68 ± 0.1
0.003 ± 0.07
T394A
0.27 ± 0.09
0.069 ± 0.06
0.012 ± 0.09
0.091 ± 0.07
0.019 ± 0.05
F397A
0.083 ± 0.07
0.18 ± 0.07
0.042 ± 0.08
0.13 ± 0.08
0.022 ± 0.05
V398A
0.025 ± 0.09
0.18 ± 0.09
0.027 ± 0.08
0.074 ± 0.09
0.022 ± 0.05
L400A
0.18 ± 0.08
0.15 ± 0.06
0.35 ± 0.1
0.0023 ± 0.08
0.07 ± 0.08
K404A
0.24 ± 0.1
0.14 ± 0.1
0.032 ± 0.08
0.064 ± 0.05
0.054 ± 0.1
DDDGc values could not be determined to Val0 for cpSAP97 PDZ2 mutants due to low affinity resulting from high dissociation rate constants. See also Tables S1 and S2 for kinetic rate constants.
two truncated side chains is obtained because any nonintended structural perturbations will be cancelled out in the calculation if the unwanted perturbation is present in both the single and double mutant. It is important to emphasize that this interaction energy (coupling free energy) is the one between the two probed residues in the context of the wild-type complex (Carter et al., 1984; Fersht et al., 1992; Horovitz, 1996). Twenty point mutants were made for cpSAP97 PDZ2, identical to a subset of those made previously for pwtSAP97 PDZ2. Each SAP97 PDZ2 mutant was analyzed in presteadystate binding experiments using the stopped flow technique with a wild-type ligand (LQRRRETQV) and the same ligand mutated at position 0 (Val/Abu, aminobutyric acid), 2 (Thr/Ser), and 4 (Arg/Ape, aminopentanoic acid; Figure 1B). The cpSAP97 PDZ2 variants displayed poor binding to the peptide with the Val0/Abu mutation, resulting in observed rate constants that were too large for the stopped flow technique. Stopped flow experiments (see Figure S1, available online, for examples of binding experiments) yielded values for kon, koff, and the resulting Kd (= koff/kon) for all PDZ/peptide combinations as described in the Experimental Procedures section and in Figure 1. General Effects of the Circular Permutation The difference in affinity between wild-type and mutated peptides was larger for pwtSAP97 PDZ2 than for cpSAP97 PDZ2 (Tables S1 and S2), suggesting a general loss of selectivity upon circular permutation perhaps due to changes in the
binding site or to changes in pathways for allosteric signaling. Most of the coupling free energies in both pwt- and cpSAP97 PDZ2 are low (Table 1; Figure 2) and few differ by more than the error in the measurement. pwtSAP97 PDZ2 has two residues (Lys392 and Val388) with relatively high coupling free energies to the Arg 4 position of the ligand (0.68 and 0.21 kcal/mol, respectively). In the circular permutant, these coupling free energies are much weaker. In fact, none of the analyzed residues in the circular permutant show any strong coupling free energy to Arg 4 (all are below 0.08 kcal/mol; Figure 2). This is consistent with the observation that the Kd for the wild-type and Arg 4/ Ape ligands is the same for cpSAP97 PDZ2 (2.7 and 2.9 mM, respectively), but different for pwtSAP97 PDZ2 (0.25 and 1.1 mM, respectively). It is likely that loss of allosteric signaling to the Arg-4 residue upon circular permutation of SAP97 PDZ2 is the reason for the relatively high loss of affinity for the wildtype peptide (from 0.25 to 2.7 mM). This is further supported by the fact that the coupling free energy between Arg 4 and Lys392 (0.68 kcal mol 1) in pwtSAP97 PDZ2 accounts for much of the loss in affinity for the Arg 4/Ape mutation in the peptide (0.83 kcal/mol). On the other hand, the DDDGc value for Phe331 to peptide position Thr-2 is similar for pwt- and cpSAP97 PDZ2 (Table 1). This shows that the circular permutation does not lead to loss of allosteric strength at all positions and that fine details of the structure govern the energetic crosstalk. It is also obvious that the coupling free energies in SAP97 PDZ2 in general are lower than those in PTP-BL PDZ2 and PSD-95 PDZ3 (Figure 2).
Structure 21, 1–10, July 2, 2013 ª2013 Elsevier Ltd All rights reserved 3
Please cite this article in press as: Hultqvist et al., Energetic Pathway Sampling in a Protein Interaction Domain, Structure (2013), http://dx.doi.org/ 10.1016/j.str.2013.05.010
Structure Energetic Pathway Sampling
pwtSAP97 PDZ2 ∆∆∆GC to Val0 in ligand
>0.8 kcal mol-1
pwtSAP97 PDZ2 ∆∆∆GC to Thr-2 in ligand
cpSAP97 PDZ2 ∆∆∆GC to Thr-2 in ligand
Figure 2. Spatial Distribution of Interaction Pathways
pwtSAP97 PDZ2 ∆∆∆GC to Arg-4 in ligand
Coupling free energies between PDZ side chains and peptide ligand side chains were grafted onto the structures according to a color code. Three PDZ domains and three different positions in the peptide ligand were included in our analysis. Rows, from top to bottom: pwtSAP97 PDZ2, cpSAP97 PDZ2, PSD-95 PDZ3, and PTP-BL PDZ2. Columns, from left to right: coupling to Val0, Thr/Ser 2, and Arg 4, respectively. The ligand residue for which the coupling free energy DDDGc is calculated is shown in green spheres. The figure shows that PSD-95 PDZ3 and PTP-BL PDZ2 (data from Gianni et al., 2011) display both higher DDDGc values and a higher frequency of coupled residues than SAP97 PDZ2.
cpSAP97 PDZ2 ∆∆∆GC to Arg-4 in ligand
0.4 kcal mol-1 0.2 kcal mol-1 0 kcal mol-1 -0.2 kcal mol-1 -0.4 kcal mol-1
PSD95 PDZ3 ∆∆∆GC to Val0 in ligand
PTP-BL PDZ2 ∆∆∆GC to Val0 in ligand
PSD95 PDZ3 ∆∆∆GC to Thr-2 in ligand
PTP-BL PDZ2 ∆∆∆GC to Thr-2 in ligand
Transfer of Allosteric Communication via ‘‘Side Chain Interaction Pathways’’ Distinct energetic communication pathways are thus likely present in the PDZ domain family, as suggested by the present data as well as previous experiments. But how is the coupling free energy transferred through the tertiary structure from distal residues in the protein to the peptide? Several computational methods have been developed to guide the analysis of possible communication pathways in protein domains based on the 3D structure. The methods include residue interaction networks (RINs), residue interaction graphs (RIGs), protein structure networks (PSNs), and protein energy networks (PENs) in which all the interactions between amino acids in the tertiary structure is plotted in a two-dimensional network (Greene, 2012; Vijayabaskar and Vishveshwara, 2010). With the help of such analyses and by determining each residue’s shortest path to all other residues in the domain it is possible to identify residues likely to be important for structure and function (Amitai et al., 2004; Cusack et al., 2007; Hu et al., 2007; Kannan and Vishveshwara, 1999; Vijayabaskar and Vishveshwara, 2010) suggesting that these kind of pathways are important. In a few cases, individual
pathways between the binding of an activator molecule and the active site have been studied (Ghosh et al., 2011; Sethi et al., 2009; Tang et al., 2007) and found to involve conformational changes upon binding of the activator molecule. We used the structures of pwtSAP97 PDZ2 (Haq et al., 2010) and cpSAP97 PDZ2 (Hultqvist et al., 2012) to analyze all interactions in the tertiary structure using online tools such as RING (http:// protein.bio.unipd.it/ring/; Martin et al., 2011), which generates a network of nodes and edges, where the nodes are the residues in the protein and the edges are the interactions between the residues. This type of network greatly simplifies the analysis of possible interaction pathways within the tertiary structure and can be visualized in the open source program cytoscape (Smoot et al., 2011; Figure 3; Figures S2–S6). Using RING, we calculated interaction networks based on closeness. While interactions involving the backbone may be involved in the transmission of energetic signal, we found that side chain to side chain interactions are sufficient to describe the experimentally observed networks. Therefore, the networks presented here only include the noncovalent bonds between side chains, and not covalent bonds and bonds to the peptide backbone. However, we do not rule out the involvement of backbone interactions in the transfer of coupling free energy. The weakest van der Waals interactions were excluded to simplify the figures. Side Chain Interaction Pathways Explain the Transfer of Coupling Free Energies in pwtSAP97 PDZ2 Of the eleven residues with coupling free energies >0.2 kcal/mol in pwtSAP97 PDZ2, ten are either directly interacting with the specific ligand position (first sphere) or located closely in a side chain pathway to such a first sphere residue and not separated by any residues with low coupling energy. The exception
4 Structure 21, 1–10, July 2, 2013 ª2013 Elsevier Ltd All rights reserved
Please cite this article in press as: Hultqvist et al., Energetic Pathway Sampling in a Protein Interaction Domain, Structure (2013), http://dx.doi.org/ 10.1016/j.str.2013.05.010
Structure Energetic Pathway Sampling
A
A 334
T 351
SAP97 PDZ2 ∆∆∆GC to Thr-2 in ligand ∆∆∆GC to Arg-4 in ligand
K 392
G 357
T 406 N 346
G 335
S 347
Q 340
I 354 Y 349
H 360
I 333
K 404
I 320
K 370
The depicted interactions (lines) are predicted using the RING software (Martin et al., 2011). (A) Analysis of side chain interaction pathways to the ligand positions Thr 2 and Arg 4 for pwtSAP97 PDZ2. van der Waals interactions with a contact score lower than 0.03 according to RING (Martin et al., 2011) were excluded to make the illustration more clear. The bonds in green and red are important for the transfer of coupling free energy. (B) Analysis of side chain interaction pathways to the ligand position Val0 for pwtSAP97 PDZ2 (van der Waals contact score above 0.008). Lys392 and Val388 are part of two distinct pathways, one coupled to Val0 and one to Arg 4 (A and B). The thick lines are pathways conserved between the unbound structure and a ligand-bound SAP97 PDZ2 structure (Figure S2). See also Figures S3–S6 for network analyses of other PDZ domains.
K 352
H 341
I 367 I 353
P 312
N 339
H 384 V 388
D 345
Figure 3. Networks of Side Chain to Side Chain Interactions in the Unbound pwtSAP97 PDZ2
S 332
G 330
V 402
F 331 L 400
K 311
D 369
I 348
E 315
L 391
L 365
A 387
S 314
A 358
V 313 P 326
L 371 A 403 L 379
∆∆∆GC: -0.2 - 0.2 kcal mol-1
N 376
V 398 Y 399
∆∆∆GC: 0.2 - 0.3 kcal mol-1
L 322
A 373
K 401
I 317
G 325
Not analyzed
Q 366
L 329
D 362
A 390
F 397
K 321
Residue with direct interaction with Thr-2 in ligand
N 375
V 374 E 319
S 395
T 394
V 377
∆∆∆GC> 0.4 kcal mol-1 ∆∆∆GC: 0.2 - 0.3 kcal mol-1
K 324 D 396 G 328
Residue with direct interaction with Arg-4 in ligand
B
P 405
K 311 P 312
K 404
SAP97 PDZ2 ∆∆∆GC to Val0 in ligand
L 365
Q 366
A 358
L372
E 315
A 403 I317
V 402
K 370
K 321
I 354 D 362
T 406
I 353
V350
D 369
I 320
S 314
V 313
H 360
L 329
S 347
E 319
I 367
K 401
K 352
S 332
F 331 L 371
L 322
V 398
Y 349
T 351
V 382
Y 399 N 376
E 381
A 373
L 400 A 387
L379
I 333
L 391
D 345
structures, but there are some differences to ligand position Val0, probably related to a few differences in primary structure between the two constructs (see legend to Figure S2).
H 341 I 348
A 334
N 375 V 374
G 328
K 392
A390
K 324
∆∆∆GC: 0.3 - 0.4 kcal mol-1
V 377
F 397
D 396
∆∆∆GC> 0.4 kcal mol-1
S 395
V 388
N 346 E 386
∆∆∆GC: 0.2 - 0.3 kcal mol-1
Q 340
∆∆∆GC: -0.2 - 0.2 kcal mol-1 ∆∆∆GC< -0.2 kcal mol-1
W 342
A Strong Correlation between the Loss of Coupling Free Energy and Side Chain Distortion upon Circular Permutation If we compare the side chain position in structures of the pwt- and cpSAP97 PDZ2 for the three side chain interaction pathways highlighted in Figure 3A, a clear pattern emerges. Phe331 has conserved coupling free energies in pwt- and cpSAP97 PDZ2, and the X-ray crystal structures show that the side chains of the two residues predicted to be part of the pathway (Phe331 and Ile333) are superimposable (Figure 4A). On the other hand, Lys392-Arg 4 has a high DDDGc value in pwtSAP97 PDZ2 (0.68 kcal/mol) but low in cpSAP97 PDZ2 (0.0 kcal/mol). The side chain for this residue is heavily distorted in cpSAP97 PDZ2 (Figure 4B) due to the new N-terminal serine, which forms a hydrogen bond with the side chain of Lys392. This interaction shifts the side chain of Lys392 away from the neighboring residue Val388. In the pwtSAP97 PDZ2 these two side chains are connected through van der Waals forces, but in cpSAP97 PDZ2 the distance between the side chains precludes any interaction. Val388 has direct interactions with His384, which in turn has a direct interaction with Arg 4 in the ligand. A possible and likely pathway for this signal to travel from Lys392 to His384 would be via Val388, all located on the same a helix. Val388 in pwtSAP97 PDZ2 also has a moderately strong coupling to the 4 position, while in the circular permutant, this coupling free energy is reduced. The third pathway (Lys352-Ser332) has a slight structural distortion of the side chain in residue Ser332 due to the circular permutation, causing G 335
T 394
V 337
Not analyzed
T 383
N 339 H 384
G 357 G 344 G 330
G 325 P 326
E385
Residue with direct interaction with Val0 in ligand
is Lys404, which is distal to Val0 but not separated by any residue with low coupling free energy (Figures 3A and 3B). Only two of the analyzed 20 residues in pwtSAP97 PDZ2 display high coupling energies (>0.2 kcal/mol) to the methyl group of Thr 2 in the ligand (Phe331 and Lys352) and two more display high coupling energies to the guanidine moiety of Arg 4 (Lys392 and Val388). These residues are located on three different side chain interaction pathways, each leading to residues directly interacting with the specific ligand position (Figure 3A). Residues close to those in the pathway do not have any significant coupling free energy with the peptide ligand side chains. Along a specific pathway and within error due to the different types of mutations, it does not seem to be a relation between distance from the ligand and coupling free energy. For example, the more distal Lys392 has a higher DDDGc value than the second sphere Val388. The above network analysis was made on the unbound pwtSAP97 PDZ2 crystal structure 2X7Z, but the same analysis largely holds also for the ligand-bound nuclear magnetic resonance (NMR) structure (2OQS; Liu et al., 2007; Figure S2). All the predicted networks to ligand positions Thr 2 and Arg 4 are conserved in these two
P 343
Structure 21, 1–10, July 2, 2013 ª2013 Elsevier Ltd All rights reserved 5
Please cite this article in press as: Hultqvist et al., Energetic Pathway Sampling in a Protein Interaction Domain, Structure (2013), http://dx.doi.org/ 10.1016/j.str.2013.05.010
Structure Energetic Pathway Sampling
Figure 4. Differences in Side Chain Positions for pwt- and cpSAP97 PDZ2 Are Reflected in Changes in Coupling Free Energies The pathways highlighted in Figure 3 are illustrated here in the structure for both the pwtSAP97 PDZ2 (green) and the cpSAP97 PDZ2 (purple). (A) Ile333 has a direct interaction and Phe331 has a high coupling free energy to Thr 2 in the ligand. The positions of the side chains of Phe331 and Ile333 are identical for pwt- and cpSAP97 PDZ2, which is consistent with their similar coupling free energies. (B) His384 has a direct interaction with Arg 4 in the ligand and Val388 and Lys392 in pwtSAP97 PDZ2 have high coupling free energies to Arg 4. The side chain in Lys392 has been dislocated in the circular permutant so that the interaction with the side chain of Val388 is broken. Accordingly, the coupling free energy between Lys392 and Arg 4 is close to zero for the circular permutant. (C) Ser332 has a direct interaction with the Thr 2 and Lys352 has a high coupling free energy to this position in the ligand. The position of the side chain of Ser332 is slightly altered relative to Lys352 in the circular permutant. This is reflected in a 50% reduction of the coupling free energy between Lys352 and Thr 2 for the circular permutant. See also Figure S7.
a reduction of the coupling free energy at position Lys352 (Figure 4C). These clear relationships between side chain conformations and coupling free energy show that side chain interaction pathways can explain the transfer of coupling free energies within a protein domain. However, the examples of how energetic pathways are modulated by small changes in structure rely on the quality of the crystal structures. Superimposing the whole cpSAP97 PDZ2 structure onto that of the pwtSAP97 PDZ2 shows that the structures are very similar (Figure 1C) except for the differences related to the circular permutation. The overall root-mean-square deviation (rmsd) for 91 aligned Ca atoms is 0.88 A˚ and the backbone conformations in the four interaction pathways are close to identical. A detailed analysis of side chain conformations is presented in the Supplemental Results. Transfer of Coupling Free Energy via Side Chain Interaction Pathways Reconciles Previous Data on PDZ Domains Encouraged by the identification of side chain interaction pathways in SAP97 PDZ2, we analyzed the interaction pathways for PSD-95 PDZ3 and PTP-BL PDZ2 (Figure 1A), for which we have previously measured coupling free energies (Gianni et al., 2011). These two PDZ domains have several residues with high or very high coupling free energies, compared to those of SAP97 PDZ2. Several of the residues are similar in PSD-95 PDZ3 and SAP97 PDZ2 (Figure 1A), facilitating a direct comparison of side chain positions and coupling free energies. However, side chain pathways differ between the two proteins as we move away from the ligand interacting residues. Nevertheless, two examples (Phe325/Phe331 and Ala375/Ala387) show that residues adjacent to a first sphere residue with conserved side chain position have a similar coupling free energy (Figure S7). All side chain interactions present in the structure of PSD-95 PDZ3 (Protein Data Bank [PDB] codes 1BFE and 1BE9) were plotted together with coupling free energies (Gianni et al., 2011) to the ligand position Val0 (Figures S3A and S3B) and to the
ligand position Thr 2 (Figures S4A and S4B). All residues with high DDDGc values to Thr 2 are located close to a residue that has a direct interaction with the ligand residue, as measured in side chain interaction distance. In fact none of these high coupling free energy pathways contain any analyzed residue with a low DDDGc value (<0.2 kcal/mol). The residue Ile327 seems to be the most important residue for the transfer of coupling free energies to Thr 2 in the ligand for PSD-95 PDZ3, based on several side chain pathways leading to this residue. Some of the other residues that directly interact with Thr 2 also display significant DDDGc values, but unlike Ile327, they do not connect to a network of side chain pathways stretching out into the protein. A similar pattern is seen for coupling free energies to the ligand position Val0. Leu323 transfers coupling free energy to the ligand Val0 in PSD-95 PDZ3 (Figure S3). This residue is coupled to the ligand with a backbone hydrogen bond in PSD-95 PDZ3, which is not present in the NMR structure with bound ligand for SAP97 PDZ2 (Liu et al., 2007). Indeed, the corresponding residue in SAP97 PDZ2 (Leu329) has a low DDDGc value. This example shows how the fine details of the structure govern the observed allosteric signal transfer. Phe325 in PSD-95 PDZ3 interacts directly with the Val0 and is part of a side chain interaction network to Val0. His372 has been previously reported to be important for binding affinity (Chi et al., 2006) and to be the key residue responsible for ligand specificity in the class I PDZ domains (van Ham and Hendriks, 2003), by making a hydrogen bond to the hydroxyl group of Thr 2. But His372 is also connected to Phe325 through a side chain interaction pathway via Ile327. In fact the coupling free energy for His372 to Val0 (0.59 kcal/mol) is much stronger than that to the methyl group of Thr 2 (0.24 kcal/mol) despite the hydrogen bond to the hydroxyl group of Thr 2. Ile 327 has a coupling free energy to Thr 2 of 0.19 kcal/mol. This is exactly what was seen when this particular pathway was studied with a computational method, anisotropic thermal diffusion (ATD) modeling (Ota and Agard, 2005). Thermal energy given to His372 will diffuse through Ile327 to Phe325 but much more thermal energy will end up in
6 Structure 21, 1–10, July 2, 2013 ª2013 Elsevier Ltd All rights reserved
Please cite this article in press as: Hultqvist et al., Energetic Pathway Sampling in a Protein Interaction Domain, Structure (2013), http://dx.doi.org/ 10.1016/j.str.2013.05.010
Structure Energetic Pathway Sampling
Figure 5. Residues with High Coupling Free Energies to Ligand Position –2 in PTP-BL PDZ2 Positive coupling free energies are depicted as green while negative ones are pink. Grey residues are directly interacting with the ligand and have coupling free energies close to zero. All the residues with positive DDDGc values are located on one side of the domain while the ones with negative DDDGc values are located on the other side of the domain. The ligand-interacting residue Val29 is trapped in between the positive and negative coupling free energies, which might explain its low DDDGc value. Leu25 is the residue with the largest negative coupling free energy to the ligand position 2, but it is also the one with the highest coupling energy to ligand position 0, suggesting that the coupling to position 2 cannot be optimized without losing coupling to ligand position 0.
the end of the chain (Phe325) than in the middle (Ile327). Later analyses using rotamerically induced perturbation (Ho and Agard, 2010) and pump-probe molecular dynamics (Sharp and Skinner, 2006) yielded a different result, possibly due to the specific frequencies used in the simulations. In ATD transmission of kinetic or heat energy, all types of motion are considered. All pathways that we have identified for PSD-95 PDZ3 are conserved between the unbound and bound structure, except the one involving Gly331 and Thr 2 (Figures S4A and S4B). Analysis of coupling free energy and side chain position in the tertiary structure of PTP-BL PDZ2 clearly shows that residues with negative and positive coupling free energies form distinct interaction pathways. Interestingly, there is an overlap between residues with negative DDDGc values to ligand position 2 and positive DDDGc values to ligand position 0. In fact, the residue with the highest measured coupling to ligand position 0 (Figure S5A), Leu25, also has the highest negative coupling free energy to ligand position 2 (Figure S6A). This suggests that a simultaneous optimization of energetic coupling to ligand position 0 and 2 is difficult to obtain. PTP-BL PDZ2 has some distinct differences in sequence as compared to PSD-95 PDZ3 and SAP97 PDZ2. The conserved GLGF motif, which binds the
carboxylate of the peptide ligand, is for example modified to SLGI in PTP-BL PDZ2. Replacing the phenylalanine with an isoleucine may have large functional effects on the energetic coupling. In PSD-95 PDZ3 and in SAP97 PDZ2, the phenylalanine (Phe325 and Phe331, respectively) displays positive DDDGc values to Thr 2 in the ligand, while the isoleucine (Ile27) in PTP-BL PDZ2 has a slightly negative DDDGc value (Figure 5; Figure S6). In PSD-95 PDZ3, Phe325 is a hub in a network of positive coupling free energy (Figure S4), whereas in PTP-BL PDZ2, the corresponding residue (Ile27) is only connecting to one network. The other residues with positive coupling free energies are partly conserved between PSD-95 PDZ3 and PTP-BL PDZ2; Leu85 corresponds to Leu379 in PSD95 PDZ3, which also lies on a positive path, while the other residues do not show any strong correlation. For PTP-BL PDZ2, we looked at two structures without ligand, one from mouse and one from human PTP-BL PDZ2 (1GM1 and 3LNX, respectively; Walma et al., 2002; Zhang et al., 2010). The networks generated from these two structures largely resemble each other but with slight differences. We believe that pathways that are intact in both these structures are more likely to be transferring the coupling free energy than the ones only present in one structure. From the joint analysis we conclude that the negative network present in PTP-BL PDZ2 to residue Ser 2 most likely affects residues Leu85 and Leu73, which are residues in the positive network, thus weakening the positive network. If we only consider the 1GM1 structure, then a direct pathway to the ligand interacting with Val29 from Val65 would also be possible. In addition, the pathway of Val33 is different between the mouse and human structures. There are however slight differences in the amino acid sequences of the two orthologs, which could account for the differences in pathway prediction (Figures S5 and S6). DISCUSSION General Aspects of Allosteric Pathways With detailed experimental analyses of coupling free energies and computationally predicted interaction pathways, we provide several examples of how fine details in the structure govern intradomain energetic communication in PDZ domains. Whereas some residues are involved in the energetic wiring of all three investigated PDZ domains, it is clear that distinct energetic interaction pathways are operational in different members of this protein family. In fact, even for the same PDZ domain, different ligand positions are connected to different pathways. We therefore coin the term ‘‘energetic pathway sampling,’’ to capture the observation that different pre-existing pathways are activated by different ligands. For example, the CRIPT peptide, associated with PSD-95 PDZ3 (Niethammer et al., 1998), is likely to activate other pathways in SAP97 PDZ2 than the E6 peptide from human papillomavirus (Figure 1). Our observed interaction pathways, as probed by direct measurements of DDDGc, may arise from multiple parallel allosteric pathways. While such a scenario is difficult to experimentally dissect, we note that the shortest pathway between amino acid side chains is sufficient to explain the data. We can only speculate about how the interaction energy is transferred within the domain. Structural changes are usually very small or absent in PDZ-peptide interactions, including
Structure 21, 1–10, July 2, 2013 ª2013 Elsevier Ltd All rights reserved 7
Please cite this article in press as: Hultqvist et al., Energetic Pathway Sampling in a Protein Interaction Domain, Structure (2013), http://dx.doi.org/ 10.1016/j.str.2013.05.010
Structure Energetic Pathway Sampling
those in the present study (Doyle et al., 1996; Zhang et al., 2010). The energy may therefore be transmitted via changes in side chain and/or backbone dynamics upon binding, in the absence of conformational change (Cooper and Dryden, 1984), as demonstrated in a few cases (Akke et al., 1993; Frederick et al., 2007; Popovych et al., 2006; Tzeng and Kalodimos, 2009). Backbone dynamics usually show that the main chain in folded proteins is quite rigid in regions with secondary structures, and that loops and termini have motions with larger amplitudes. In general, side chain dynamics, at least on the picosecond-to-nanosecond time scale, are more heterogeneous and responsive to changes in functional state such as in binding, when compared to backbone dynamics (Frederick et al., 2007; Marlow et al., 2010; McDonald et al., 2013). We note that side chain dynamics in PDZ domains show some conservation (Law et al., 2009), but the differences in dynamics between PDZ domains appear sufficiently large to accommodate the multiple distinct pathways found here. While side chain to side chain interactions can explain the spatial distribution of coupling free energy in all our analyzed cases (PTP-BL PDZ2, SAP97 PDZ2, and PSD-95 PDZ3), we do not rule out the involvement of backbone interactions. Nevertheless, our hypothesis is that energy transfer through the protein to the ligand is mediated mainly via side chain to side chain interactions to a residue that is directly interacting with a specific ligand position. Importantly, the striking changes in coupling free energy from small structural re-arrangements of side chains upon circular permutation lend support to this model. Because the residues with high coupling energies are located relatively close to the binding site, pathways involving backbone interactions (i.e., side chain backbone, backbone-backbone) can also explain the transfer of coupling free energy rather well. However, there are residues located far away from the binding site, for which only side chain to side chain interactions can explain the observed coupling free energies; for example Val362 in PSD-95 PDZ3, which couples to Val0 in the peptide. Allosteric Pathways in PTP-BL PDZ2 A large number of computational and experimental studies have predicted energetic pathways in PDZ domains using different methods (Cilia et al., 2012; Dhulesia et al., 2008; Fuentes et al., 2006; Ho and Agard, 2010; Kong and Karplus, 2009; Liu et al., 2009; Lockless and Ranganathan, 1999). Overall, there is no clear consensus among the studies although some agreement regarding key positions can be made, for example, residues close to the binding site. Several papers have looked at intradomain allostery in mouse and human PTP-BL PDZ2. Experimental differences can be understood in terms of the different approaches. For example, we measure interaction energies between side chains in the PDZ and peptide ligand. Fuentes and coworkers measure changes in side chain dynamics upon binding (Fuentes et al., 2004, 2006) and assess binding of the whole peptide rather than specific residues in the peptide. The timescale of the dynamics may also explain discrepancies between our data and NMR relaxation experiments. Dhulesia and coworkers (Dhulesia et al., 2008) analyzed human PTP-BL PDZ2 with NMR in combination with molecular dynamics simulation and found two distal regions that display long-range interactions with the binding
site. Distal region 1 becomes more flexible and shows both changes in fluctuations at the nanosecond timescale and structural differences upon binding. Distal region 2 does not become more flexible upon binding and only have changes in fluctuations. Distal region 1 contains amino acids that have high coupling free energies in our measurements, but we did not observe large coupling free energies for residues in distal region 2. In a recent publication Cilia and colleagues used a method called mutual information (MI) to assess how the locations of two neighboring side chains affect each other’s conformation (Cilia et al., 2012). By analyzing the difference in mutual information between the unbound and the bound state (i.e., DMI) in human and mouse PTP-BL PDZ2, they found pathways where allosteric signals can be transferred (Cilia et al., 2012). Some of these pathways agree in part with the ones we experimentally identify, but others do not. This underscores that different approaches may identify different pathways and might be a reflection of the multiple potential pathways sampled by each PDZ domain. Allosteric Pathways in PSD-95 PDZ3 A recent study by McLaughlin et al. (2012), which builds on previous work (Lockless and Ranganathan, 1999), suggests that amino acids that have a high frequency of coexistence in proteins from different organisms are energetically coupled to each other. The most frequently coupled residues were defined as sector residues, and it was suggested that these sector residues are the ones most important for modulating binding affinity. As expected, the interaction pathways that we have identified show considerable overlap around the binding site, with each other and with sector residues. It is however clear that the transmission of energetic signals within PDZ domains depends on fine details in the structure. For example, nonsector residues such as Lys352 (SAP97 PDZ2), Val33 (PTP-BL PDZ2), and Phe340 (PSD-95 PDZ3) may influence ligand binding. A recent appraisal of analysis of correlated mutations (coevolution) concluded that the approach will have little success in predicting allosteric pathways, except for homologs with very high sequence identity, due to the high sequence dependence of protein dynamics (Livesay et al., 2012). This conclusion is in perfect agreement with our energetic pathway predictions and experimental results. Along these lines, we suggest that coevolved residues may indicate regions of a protein where allosteric networks occur, but for a particular case, the exact sequence and side chain packing will determine any intradomain allostery. For example, a comparison between our data set on PSD-95 PDZ3 and that of McLaughlin and colleagues (McLaughlin et al., 2012), in which they used a larger perturbation in the peptide (Thr 2/Phe), and only considering mutations to Ala or Gly in the PDZ domain reveals a limited overlap: Ile327, Ile336, and Ala375 all have high coupling free energies in both data sets, while DDDGc values do not correlate at other positions. But, interestingly, these three overlapping residues are located on a single side chain interaction pathway. For conserved allosteric mechanisms involving large conformational changes such as that in myosin (Tang et al., 2007), it is more likely that co-evolved residues will pinpoint functional allosteric pathways. Conclusions Our studies show that the PDZ domain family sample energetic pathways, i.e., several distinct energetic pathways may be
8 Structure 21, 1–10, July 2, 2013 ª2013 Elsevier Ltd All rights reserved
Please cite this article in press as: Hultqvist et al., Energetic Pathway Sampling in a Protein Interaction Domain, Structure (2013), http://dx.doi.org/ 10.1016/j.str.2013.05.010
Structure Energetic Pathway Sampling
used to connect distal residues with peptide ligand side chains. Fine structural details along the pathways determine the energetic crosstalk, leaving plenty of room for evolutionary adaption of ligand specificity. EXPERIMENTAL PROCEDURES Double Mutant Cycle Constructs, Expression, and Purification Expression and purification of pwtSAP97 PDZ2 (Haq et al., 2010) and cpSAP97 PDZ2 (Hultqvist et al., 2012) were performed as described earlier. The peptide binding of pwtSAP97 PDZ2 mutants were described in a previous publication (Haq et al., 2012) in the context of linear free-energy relationships, but their coupling free energies have not been published before. Single amino acid mutations were originally done at 27 evenly distributed sites in the pwtSAP97 PDZ2 (Haq et al., 2012). Some mutants of both pwtSAP97 PDZ2 and cpSAP97 PDZ2 expressed poorly and were therefore not included here. In this study, 20 successfully expressed point mutants were included for cpSAP97 PDZ2 and compared to the corresponding 20 mutants of pwtSAP97 PDZ2. Mutants of cpSAP97 PDZ2 were expressed and purified as described (Hultqvist et al., 2012) and analyzed with SDS-PAGE and mass spectrometry to ensure purity and identity, respectively. Numbering of amino acid residues is according to the pwtSAP97 PDZ2 sequence, also for the circular permutant (Figure 1A). Stopped Flow Experiments Pre-equilibrium binding experiments between PDZ domain variants and four peptide ligands (Figure 1B) were carried out on an upgraded SX-17 MV stopped flow spectrophotometer (Applied Photophysics, Leatherhead, UK). The Trp342 present in all constructs was exited at 280 nm and emission was recorded using a 330 ± 25 nm interference filter. For wild-type and mutants of pwtSAP97 PDZ2, the experimental buffer was 50 mM potassium phosphate, pH 7.5, while for the cpSAP97 PDZ2 constructs, 0.4 M sodium sulfate was included in the buffer to increase the thermodynamic stability of the mutants and ensuring that they were all folded under the experimental conditions. All experiments were conducted at 10 C. The ligands used in the binding experiments were either bought (GL Biochem, Shanghai, China) or synthesized in house. Two nonnatural amino acids were used in the mutated peptide ligands, to obtain a perturbation of the wild-type affinity that was well suited for the analysis: large enough to allow accurate estimates of the coupling free energy, but not too large because this could lead to ground state effects, such as changes in overall structure. In addition, observed rate constants should be in the window accessible by the stopped flow technique (<500 s 1). These nonnatural amino acids were 2-aminobutyric acid, Abu (Val with a deleted methyl group) and 2-aminopentanoic acid, Ape (Arg with a deleted guanidinium group; Figure 1B). In addition, a Thr/Ser mutation was used in the peptide ligand (deletion of a methyl group). Single exponential binding traces were obtained in the stopped flow by mixing PDZ and peptide (Figure 1D). The monophasic binding showed that all mutants were properly folded under the experimental conditions; otherwise, a phase related to the refolding of the PDZ domain would have been observed (Haq et al., 2010; Hultqvist et al., 2012). To determine the association rate constant kon the PDZ concentration was kept constant at 1 mM of PDZ variant, while varying the peptide ligand concentration between 1 and 20 mM. The observed rate constants were plotted against the peptide concentration and fitted to an equation described elsewhere (Malatesta, 2005) to obtain kon as the slope of the curve (Figure S1). Displacement experiments were performed to get a precise and accurate value of koff (Chi et al., 2009). The peptide ligand, bound to the protein, was displaced by a high concentration (up to 60 mM) of dansylated peptide (Dansyl-RRETQV; Figure 1D). The concentration of pwtSAP97 PDZ2 variant and peptide in the displacement experiments was usually 1 and 0.25 mM, respectively, and that of cpSAP97 PDZ2/peptide was 0.5 and 0.75 mM, respectively. Ratios were adjusted according to the affinity of the complex to optimize the kinetic amplitude. By using a double mutant cycle it is possible to exclude any structural effects and only study the communication between two specific residues (Carter et al., 1984; Gianni et al., 2011; Horovitz, 1996). From a double mutant cycle analysis the coupling free energy DDDGc between the two analyzed residues is obtained, which reflects their interdependence, or ‘‘communication.’’
Because we perturb mainly side chain interactions by mutation in the double mutant cycles, the DDDGc values describe the communication between the two mutated side chains. DDDGc was calculated as described in Figure 1 and elsewhere (Gianni et al., 2011). Allosteric Pathway Prediction Residue interaction networks (RINs) for the structure of unbound pwtSAP97 PDZ2 (PDB code 2X7Z), bound SAP97 PDZ2 (PDB code 2OQS), unbound cpSAP97 PDZ2 (PDB code 4AMH), unbound PSD-95 PDZ3 (PDB code 1BFE), bound PSD-95 PDZ3 (PDB code 1BE9), unbound PTP-BL PDZ2 (PDB code 1GM1), and bound PTP-BL PDZ2 (PDB code 3LNY) were generated using RING (http://protein.bio.unipd.it/ring/; Martin et al., 2011), with the standard settings and network type set to van der Waals contacts. The networks were then visualized in the open source program Cytoscape (www.cytoscape.org; Smoot et al., 2011). Residues interacting with specific ligand positions were detected from the ligand bound RINs, while the subsequent analysis was performed on the unbound structures because no bound structure is available for the circular permutant. Structures with a PDZ sequence and bound peptide that were identical to those used in the experiments were not available for pwtSAP97 PDZ2 and PTP-BL PDZ2. The predicted pathways were therefore analyzed using the most similar available bound structures. All peptide bonds and noncovalent bonds involving the peptide backbone were excluded from the network for the subsequent analysis of data. In the generated network figures presented in this paper, all van der Waals interactions weaker than the ones predicted to be part of the specific side chain interaction were removed, to simplify the figure, by setting a cutoff to the van der Waals contact score calculated by RING. Analyses of Side Chain Orientations of Residues Involved in Putative Pathways The structures of molecules A and B of the circularly permutated PDZ domain (cpSAP97 PDZ2, PDB code 4AMH) and the pseudo wild-type PDZ domain (pwtSAP97 PDZ2, PDB code 2X7Z) were superimposed along with their respective electron density maps using the program Phenix (Adams et al., 2010). The superposed electron density maps were used to validate the significance of observed differences in orientations of side chains implicated in putative pathways. See the Supplemental Results for a detailed discussion of side chain conformations. SUPPLEMENTAL INFORMATION Supplemental Information includes Supplemental Results, seven figures, and two tables and can be found with this article online at http://dx.doi.org/ 10.1016/j.str.2013.05.010. ACKNOWLEDGMENTS This work was supported by the Swedish Research Council (NT; to P.J. and M.S.) and the Italian Ministry of University and Research (PNR-CNR Aging Program 2012-2014; to S.G.). Received: March 28, 2013 Revised: May 24, 2013 Accepted: May 29, 2013 Published: June 27, 2013 REFERENCES Adams, P.D., Afonine, P.V., Bunko´czi, G., Chen, V.B., Davis, I.W., Echols, N., Headd, J.J., Hung, L.-W., Kapral, G.J., Grosse-Kunstleve, R.W., et al. (2010). PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr. D Biol. Crystallogr. 66, 213–221. Akke, M., Skelton, N.J., Ko¨rdel, J., Palmer, A.G., 3rd, and Chazin, W.J. (1993). Effects of ion binding on the backbone dynamics of calbindin D9k determined by 15N NMR relaxation. Biochemistry 32, 9832–9844. Amitai, G., Shemesh, A., Sitbon, E., Shklar, M., Netanely, D., Venger, I., and Pietrokovski, S. (2004). Network analysis of protein structures identifies functional residues. J. Mol. Biol. 344, 1135–1146.
Structure 21, 1–10, July 2, 2013 ª2013 Elsevier Ltd All rights reserved 9
Please cite this article in press as: Hultqvist et al., Energetic Pathway Sampling in a Protein Interaction Domain, Structure (2013), http://dx.doi.org/ 10.1016/j.str.2013.05.010
Structure Energetic Pathway Sampling
Carter, P.J., Winter, G., Wilkinson, A.J., and Fersht, A.R. (1984). The use of double mutants to detect structural changes in the active site of the tyrosyltRNA synthetase (Bacillus stearothermophilus). Cell 38, 835–840. Chi, C.N., Engstro¨m, A., Gianni, S., Larsson, M., and Jemth, P. (2006). Two conserved residues govern the salt and pH dependencies of the binding reaction of a PDZ domain. J. Biol. Chem. 281, 36811–36818. Chi, C.N., Bach, A., Engstro¨m, A., Wang, H.Q., Strømgaard, K., Gianni, S., and Jemth, P. (2009). A sequential binding mechanism in a PDZ domain. Biochemistry 48, 7089–7097. Cilia, E., Vuister, G.W., and Lenaerts, T. (2012). Accurate prediction of the dynamical changes within the second PDZ domain of PTP1e. PLoS Comput. Biol. 8, e1002794. Cooper, A., and Dryden, D.T. (1984). Allostery without conformational change. A plausible model. Eur. Biophys. J. 11, 103–109. Cusack, M.P., Thibert, B., Bredesen, D.E., and Del Rio, G. (2007). Efficient identification of critical residues based only on protein structure by network analysis. PLoS ONE 2, e421.
Kong, Y., and Karplus, M. (2009). Signaling pathways of PDZ2 domain: a molecular dynamics interaction correlation analysis. Proteins 74, 145–154. Law, A.B., Fuentes, E.J., and Lee, A.L. (2009). Conservation of side-chain dynamics within a protein family. J. Am. Chem. Soc. 131, 6322–6323. Liu, Y., Henry, G.D., Hegde, R.S., and Baleja, J.D. (2007). Solution structure of the hDlg/SAP97 PDZ2 domain and its mechanism of interaction with HPV-18 papillomavirus E6 protein. Biochemistry 46, 10864–10874. Liu, Z., Chen, J., and Thirumalai, D. (2009). On the accuracy of inferring energetic coupling between distant sites in protein families from evolutionary imprints: illustrations using lattice model. Proteins 77, 823–831. Livesay, D.R., Kreth, K.E., and Fodor, A.A. (2012). A Critical Evaluation of Correlated Mutation Algorithms and Coevolution Within Allosteric Mechanisms. In Allostery, A.W. Fenton, ed. (New York: Springer), pp. 385–398. Lockless, S.W., and Ranganathan, R. (1999). Evolutionarily conserved pathways of energetic connectivity in protein families. Science 286, 295–299. Malatesta, F. (2005). The study of bimolecular reactions under non-pseudofirst order conditions. Biophys. Chem. 116, 251–256.
Dhulesia, A., Gsponer, J., and Vendruscolo, M. (2008). Mapping of two networks of residues that exhibit structural and dynamical changes upon binding in a PDZ domain protein. J. Am. Chem. Soc. 130, 8931–8939.
Marlow, M.S., Dogan, J., Frederick, K.K., Valentine, K.G., and Wand, A.J. (2010). The role of conformational entropy in molecular recognition by calmodulin. Nat. Chem. Biol. 6, 352–358.
Doyle, D.A., Lee, A., Lewis, J., Kim, E., Sheng, M., and MacKinnon, R. (1996). Crystal structures of a complexed and peptide-free membrane protein-binding domain: molecular basis of peptide recognition by PDZ. Cell 85, 1067–1076.
Martin, A.J.M., Vidotto, M., Boscariol, F., Di Domenico, T., Walsh, I., and Tosatto, S.C.E. (2011). RING: networking interacting residues, evolutionary information and energetics in protein structures. Bioinformatics 27, 2003–2005.
Fersht, A.R., Matouschek, A., and Serrano, L. (1992). The folding of an enzyme. I. Theory of protein engineering analysis of stability and pathway of protein folding. J. Mol. Biol. 224, 771–782.
McDonald, L.R., Whitley, M.J., Boyer, J.A., and Lee, A.L. (2013). Colocalization of fast and slow timescale dynamics in the allosteric signalling protein CheY. J. Mol. Biol. 425, 2372–2381.
Frederick, K.K., Marlow, M.S., Valentine, K.G., and Wand, A.J. (2007). Conformational entropy in molecular recognition by proteins. Nature 448, 325–329.
McLaughlin, R.N., Jr., Poelwijk, F.J., Raman, A., Gosal, W.S., and Ranganathan, R. (2012). The spatial architecture of protein function and adaptation. Nature 491, 138–142.
Fuentes, E.J., Der, C.J., and Lee, A.L. (2004). Ligand-dependent dynamics and intramolecular signaling in a PDZ domain. J. Mol. Biol. 335, 1105–1115.
Niethammer, M., Valtschanoff, J.G., Kapoor, T.M., Allison, D.W., Weinberg, R.J., Craig, A.M., and Sheng, M. (1998). CRIPT, a novel postsynaptic protein that binds to the third PDZ domain of PSD-95/SAP90. Neuron 20, 693–707.
Fuentes, E.J., Gilmore, S.A., Mauldin, R.V., and Lee, A.L. (2006). Evaluation of energetic and dynamic coupling networks in a PDZ domain protein. J. Mol. Biol. 364, 337–351. Ghosh, A., Sakaguchi, R., Liu, C., Vishveshwara, S., and Hou, Y.-M. (2011). Allosteric communication in cysteinyl tRNA synthetase: a network of direct and indirect readout. J. Biol. Chem. 286, 37721–37731. Gianni, S., Haq, S.R., Montemiglio, L., Ju¨rgens, M.C., Engstro¨m, A˚., Chi, C.N., Brunori, M., and Jemth, P. (2011). Sequence specific long-range networks in PDZ domains tune their binding selectivity. J. Biol. Chem. 286, 27167–27175. Greene, L.H. (2012). Protein structure networks. Brief Funct. Genomics 11, 469–478. Haq, S.R., Ju¨rgens, M.C., Chi, C.N., Koh, C.-S., Elfstro¨m, L., Selmer, M., Gianni, S., and Jemth, P. (2010). The plastic energy landscape of protein folding: a triangular folding mechanism with an equilibrium intermediate for a small protein domain. J. Biol. Chem. 285, 18051–18059. Haq, S.R., Chi, C.N., Bach, A., Dogan, J., Engstro¨m, A˚., Hultqvist, G., Karlsson, O.A., Lundstro¨m, P., Montemiglio, L.C., Strømgaard, K., et al. (2012). Sidechain interactions form late and cooperatively in the binding reaction between disordered peptides and PDZ domains. J. Am. Chem. Soc. 134, 599–605. Ho, B.K., and Agard, D.A. (2010). Conserved tertiary couplings stabilize elements in the PDZ fold, leading to characteristic patterns of domain conformational flexibility. Protein Sci. 19, 398–411. Horovitz, A. (1996). Double-mutant cycles: a powerful tool for analyzing protein structure and function. Fold. Des. 1, R121–R126. Hu, Z., Bowen, D., Southerland, W.M., del Sol, A., Pan, Y., Nussinov, R., and Ma, B. (2007). Ligand binding and circular permutation modify residue interaction network in DHFR. PLoS Comput. Biol. 3, e117. Hultqvist, G., Punekar, A.S., Morrone, A., Chi, C.N., Engstro¨m, A˚., Selmer, M., Gianni, S., and Jemth, P. (2012). Tolerance of protein folding to a circular permutation in a PDZ domain. PLoS ONE 7, e50055. Kannan, N., and Vishveshwara, S. (1999). Identification of side-chain clusters in protein structures by a graph spectral method. J. Mol. Biol. 292, 441–464.
Ota, N., and Agard, D.A. (2005). Intramolecular signaling pathways revealed by modeling anisotropic thermal diffusion. J. Mol. Biol. 351, 345–354. Popovych, N., Sun, S., Ebright, R.H., and Kalodimos, C.G. (2006). Dynamically driven protein allostery. Nat. Struct. Mol. Biol. 13, 831–838. Sethi, A., Eargle, J., Black, A.A., and Luthey-Schulten, Z. (2009). Dynamical networks in tRNA:protein complexes. Proc. Natl. Acad. Sci. USA 106, 6620– 6625. Sharp, K., and Skinner, J.J. (2006). Pump-probe molecular dynamics as a tool for studying protein motion and long range coupling. Proteins 65, 347–361. Smoot, M.E., Ono, K., Ruscheinski, J., Wang, P.-L., and Ideker, T. (2011). Cytoscape 2.8: new features for data integration and network visualization. Bioinformatics 27, 431–432. Tang, S., Liao, J.-C., Dunn, A.R., Altman, R.B., Spudich, J.A., and Schmidt, J.P. (2007). Predicting allosteric communication in myosin via a pathway of conserved residues. J. Mol. Biol. 373, 1361–1373. Tzeng, S.-R., and Kalodimos, C.G. (2009). Dynamic activation of an allosteric regulatory protein. Nature 462, 368–372. van Ham, M., and Hendriks, W. (2003). PDZ domains-glue and guide. Mol. Biol. Rep. 30, 69–82. Vijayabaskar, M.S., and Vishveshwara, S. (2010). Interaction energy based protein structure networks. Biophys. J. 99, 3704–3715. Walma, T., Spronk, C.A.E.M., Tessari, M., Aelen, J., Schepens, J., Hendriks, W., and Vuister, G.W. (2002). Structure, dynamics and binding characteristics of the second PDZ domain of PTP-BL. J. Mol. Biol. 316, 1101–1110. Wawrzyniak, A.M., Vermeiren, E., Zimmermann, P., and Ivarsson, Y. (2012). Extensions of PSD-95/discs large/ZO-1 (PDZ) domains influence lipid binding and membrane targeting of syntenin-1. FEBS Lett. 586, 1445–1451. Zhang, J., Sapienza, P.J., Ke, H., Chang, A., Hengel, S.R., Wang, H., Phillips, G.N., and Lee, A.L. (2010). Crystallographic and nuclear magnetic resonance evaluation of the impact of peptide binding to the second PDZ domain of protein tyrosine phosphatase 1E. Biochemistry 49, 9280–9291.
10 Structure 21, 1–10, July 2, 2013 ª2013 Elsevier Ltd All rights reserved