Search tips
Search criteria 


Logo of plosonePLoS OneView this ArticleSubmit to PLoSGet E-mail AlertsContact UsPublic Library of Science (PLoS)
PLoS One. 2010; 5(8): e12214.
Published online 2010 August 16. doi:  10.1371/journal.pone.0012214
PMCID: PMC2922380

Binding of Protein Kinase Inhibitors to Synapsin I Inferred from Pair-Wise Binding Site Similarity Measurements

Floyd Romesberg, Editor


Predicting off-targets by computational methods is getting increasing importance in early drug discovery stages. We herewith present a computational method based on binding site three-dimensional comparisons, which prompted us to investigate the cross-reaction of protein kinase inhibitors with synapsin I, an ATP-binding protein regulating neurotransmitter release in the synapse. Systematic pair-wise comparison of the staurosporine-binding site of the proto-oncogene Pim-1 kinase with 6,412 druggable protein-ligand binding sites suggested that the ATP-binding site of synapsin I may recognize the pan-kinase inhibitor staurosporine. Biochemical validation of this hypothesis was realized by competition experiments of staurosporine with ATP-γ35S for binding to synapsin I. Staurosporine, as well as three other inhibitors of protein kinases (cdk2, Pim-1 and casein kinase type 2), effectively bound to synapsin I with nanomolar affinities and promoted synapsin-induced F-actin bundling. The selective Pim-1 kinase inhibitor quercetagetin was shown to be the most potent synapsin I binder (IC50  = 0.15 µM), in agreement with the predicted binding site similarities between synapsin I and various protein kinases. Other protein kinase inhibitors (protein kinase A and chk1 inhibitor), kinase inhibitors (diacylglycerolkinase inhibitor) and various other ATP-competitors (DNA topoisomerase II and HSP-90α inhibitors) did not bind to synapsin I, as predicted from a lower similarity of their respective ATP-binding sites to that of synapsin I. The present data suggest that the observed downregulation of neurotransmitter release by some but not all protein kinase inhibitors may also be contributed by a direct binding to synapsin I and phosphorylation-independent perturbation of synapsin I function. More generally, the data also demonstrate that cross-reactivity with various targets may be detected by systematic pair-wise similarity measurement of ligand-annotated binding sites.


For long, drug designers had been focusing on a single macromolecular target and a single or very few chemical series [1]. The selectivity of preclinical candidates for the intended target was only addressed relatively at a late stage by profiling the compound against neighboring targets (e.g. receptor subtypes). Therefore, a significant attrition rate in clinical trials in the last decades [2] was due to the unexpected binding of drug candidates to additional targets (off-targets [3] or anti-targets [4]) resulting in dubious pharmacological activities, side effects and sometimes adverse drug reactions [5]. Remarkable advances in structural genomics [6], [7] and diversity-oriented chemistry [8], [9] have changed these practices. On the biological side, the Protein Data Bank [10] which stores publicly available three-dimensional (3-D) structures of macromolecules currently stores over 65 000 entries. Outstanding efforts of structural genomic consortia to complete the structural proteome let us anticipate an acceptable coverage of the UniProt database [11] in only 15 years [7]. On the chemical side, about 27 million unique structures and 435 000 bioactivity screens are available in the PubChem repository [12]. Mapping pharmacological space in 2006 [13] resulted in more than 1 300 targets with significant affinities (<10 µm) for small molecular-weight ligands. Global chemogenomic approaches [14] targeting arrays of ligands (rows) and proteins (columns) to generate huge two-dimensional binding matrices enlarge our vision of how chemical and biological spaces match [15]. Experimental chemogenomics is however expensive, time-consuming and addresses only a restricted subset of chemical (a few thousand ligands) and biological space (a few hundred targets). Combining bio- and chemoinformatic structural approaches [13], [16], [17] to fill chemogenomic matrices presents the noticeable advantage to considerably extend space coverage and limit the number of supporting experimental validations. Predicting missing data in chemogenomic matrices can be operated on a column-by-column (virtual screening of ligand libraries [18]) or on a row-by-row basis (virtual profiling of a ligand against an array of targets [19]). Two main computational strategies are possible to profile a ligand against a panel of putative targets. On one side, ligand-based methods [9], [20], [21] aim at comparing chemical descriptors of biologically-characterized ligands to transfer the target annotation of similar molecules to the query ligand. To overcome structure-activity cliffs [22] and gain statistical relevance, it is preferable to compare sets of diverse ligands. Diverse descriptors and methods have already been validated on existing data [23], [24], [25]. This approach led to the discovery of several off-targets for known drugs [20], [21]. However, pure ligand-based methods have two main drawbacks : (i) they are restricted by the incomplete coverage of target space by known ligands and thus cannot be applied to orphan proteins, (ii) the dogma stating that chemical similarity implies biological similarity is only true in 30% of test cases [26].

On the other side, target-based approaches can also be used to profile a ligand of interest. The most straightforward method is docking a ligand to a collection of protein cavities [27], [28], [29], [30]. This strategy led to the identification of novel targets for existing ligands [5], [31], [32], [33], [34] or for a novel chemotype [35]. Molecular docking is however notoriously hampered by the lack of reliable binding free energy scoring functions [36] and the extreme difficulty to automate the set-up of heterogeneous binding sites [30]. Acknowledging that similar binding sites should recognize similar ligands, a structure-based alternative to docking, is the 3-D comparison of protein-ligand binding sites [37]. As for ligand-based methods, structural descriptors of ligand-characterized binding sites are used to transfer the ligand annotation of putative targets to the query binding site. The method requires a proper metric to compare binding sites in 3-D space and should be able to detect global as well as local similarities among unrelated 3-D structures. Despite the numerous methods described for measuring 3-D similarities between protein-ligand binding sites [37], there are still very few reports of predictive target identifications by systematic binding site comparisons (for a recent review see [38]). We herewith present a predictive study supported by biochemical and functional studies that successfully assigns an unexpected target (synapsin I) to a series of therapeutically important bioactive ligands (serine/threonine protein kinase inhibitors).


The full computational protocol used to detect binding site similarity between synapsin-I and some protein kinases is displayed in Fig.1. Over 6,000 druggable protein-ligand binding sites from the sc-PDB database [39] were screened (step a, Fig. 1) for their similarity to the ATP-binding site of Pim-1 kinase (PDB entry 1yhs with bound inhibitor staurosporine) using the previously described SiteAlign algorithm [40]. From the list of similar binding sites (step b), ATP-binding sites of protein kinases were removed due to their obvious similarity (step c) and only proteins with at least 2 copies (two different sc-PDB entries) were kept (step d). Synapsin-I is the only hit (PDB entries 1aux and 1px2) and was used in a second similarity screen (step e), yet as a reference, for finding among ATP-binding sites which are similar. Among the list of possible hits (step f), the sc-PDB entries were ranked by decreasing similarity to 1aux and corresponding proteins were ranked (step g) according to a Receiver Operating Characteristic (ROC) classifying scheme [40] from the statistically most similar (Pim-1 kinase) to the least similar (panthothenate synthase).

Figure 1
Computational protocol used to detect local similarities between ATP-binding sites in Pim-1 kinase and Synapsin I.

Details of the multi-step protocol and the subsequent experimental validations will be described from here on.

ATP-binding sites of synapsin I and of Pim-1 kinase share strikingly similar features

In benchmarking our 3-D binding site comparison algorithm (SiteAlign) [41], we have previously compared ATP-binding sites of protein kinases with other druggable protein-ligand cavities from the sc-PDB database [39]. The ATP-binding site of synapsin I was predicted to be similar to that of a pan-kinase inhibitor (staurosporine) [42] with the proto-oncogene Pim-1 serine/threonine protein kinase (Fig. 2A).

Figure 2
Similarity of the ATP-binding sites of Pim-1 kinase and of synapsin I.

Protein kinases catalyze the reversible phosphorylation of proteins and constitute a family of macromoleuclar targets of utmost interest for their central implication in signal transduction pathways [43]. Thanks to existing X-ray structures [44], various inhibitors competing with the ATP substrate and exhibiting different selectivity profiles [42] towards the 518 human protein kinases have been designed, and some of them have reached the market as anti-cancer drugs [45].

Synapsin I belongs to an evolutionary conserved family of neuron-specific, synaptic vesicle-associated phosphoproteins involved in the regulation of neurotransmitter release, synaptic plasticity and synaptogenesis [46], [47], [48]. Synapsin isoforms are composed of a mosaic of shared and individual domains, among which the amino-terminal domain A and the large central domain C are the most conserved across isoforms and species [49], [50]. The crystal structure of the recombinant C domain [51] or ABC domains [52] of synapsin I revealed a high similarity to proteins of the ATP-grasp superfamily, notably glutathione synthase, and the presence of tightly associated dimers that can associate in a tetramer. Indeed, in vitro studies showed that ATP binds to all synapsins and that synapsins form homo- and hetero-oligomers [53], [54], [55]. The binding of ATP affects the oligomerization state of the synapsin ABC domains [52] and the interaction of synapsin I with the immunophilin cyclophillin B [56]. Moreover, synapsin I is a major presynaptic substrate of distinct protein kinases including PKA, CaM kinases I/II/IV, MAPK/Erk, cdk5, PAK and Src [46], [57], [58], [59], [60] that regulates synaptic vesicle trafficking, synaptic plasticity and neuronal development in a phosphorylation-dependent fashion [61], [62], [63], [64], [65], [66], [67], [68]. Based on the structural similarity between the crystal structure of the synapsin C domain with ATPases, a highly evolutionary conserved ATP binding site has been mapped in domain C [50], [51] and found to bind ATP with nanomolar affinity in a Ca2+-dependent manner [54]. Although very few data exist in the literature, it has been reported that a domain C peptide corresponding to a sequence between the ATP binding site and the Ca2+-binding site specifically inhibits the binding of synapsin I to F-actin [69]. ATP binding to synapsin I facilitates the transition from dimer to tetramer [52] and inhibits cyclophilin B binding [56].

We showed in the first series of computations (Fig. 2A) that ATP-binding sites of protein kinases do not resemble neither ATP-binding sites of other kinases nor other ATP-binding cavities [41], [70]. It is therefore not surprising that 123 out of the 134 binding sites (92%) scored above an acceptable similarity threshold (SiteAlign d2 score <0.2)[41] are annotated as ATP-binding sites in protein kinases (Fig. 2A, Supplementary Table S1). Out of the 11 outliers, two entries (PDB entries 1aux and 1px2) drew our attention since they both describe the ATP-binding site of synapsin I. Despite a low homology (21%) between human Pim-1 (1yhs) and bovine synapsin I (1aux) amino acid sequences, the proposed 3-D alignment between both binding sites reveal remarkable shared features. Although both proteins adopt distinct 3-D folds, their bound ligands (staurosporine in Pim-1, ATP-γS in synapsin I) in the cognate X-ray structures occupy a similar orientation in their respective binding sites (Fig. 2B). Out of the 32 and 24 cavity-lining residues in 1yhs and 1aux, respectively, 11 amino acids matched in both their chemical properties and 3-D spatial coordinates (Fig. 2C). 6 pairs of short side-chain aliphatic residues, one pair of lysine residues, two pairs of negatively-charged amino acids, one pair of glycine and one pair of serine residues are absolutely conserved in both binding sites (Fig. 2C). To be sure that non-conserved residues in the synapsin I site would not impair staurosporine recognition, we attempted preliminary docking experiments of the latter ligand to the 1aux structure. Only the floppy Lys67 side chain which points inward the ATP-γS binding site was rendered flexible during docking to putatively enlarge the cavity. Docking staurosporine in synapsin I with the GOLD software (see structure in Fig. 3A) provided a single set of similar binding poses with mostly hydrophobic intermolecular contacts and a bidentate hydrogen bond to main chain atoms of a hinge region (Pro306, Ile308; see top-ranked pose Fig. 2D).

Figure 3
Staurosporine binds to synapsin I and affects synapsin I-dependent F-actin bundling.

Staurosporine binds to synapsin I and enhances synapsin- F-actin interactions

The in silico predicted interaction between staurosporine and synapsin I was tested by in vitro experiments aimed at analyzing the ability of staurosporine to competitively inhibit ATPγ35S binding to synapsin I [53], [54] or to affect the interactions of synapsin I with actin which occur through a major binding site in domain C [71], [72], [73]. When purified bovine synapsin I was incubated with ATPγ35S in the absence or presence of increasing concentrations of either staurosporine or cold ATP (Fig. 3A), both ligands were able to inhibit ATPγ35S in a concentration-dependent fashion (IC50 0.07±0.01 µM and maximal inhibition 95.8±2.2% for ATP; IC50 0.31±0.09 µM and maximal inhibition 84.3±1.2% for staurosporine), indicating that staurosporine bound synapsin I at the ATP binding site.

Since the Synapsin I ATP binding pocket is localized in the synapsin domain primarily involved in actin binding [69], [72], we investigated whether staurosporine binding is able to affect the synapsin I-actin interaction. To this aim, we evaluated the F-actin binding/bundling activity under conditions of ATP binding to synapsin I in the presence of increasing concentrations of staurosporine ranging from to 1 to 20 µM. The amount of F-actin/synapsin I bundles recovered by low speed sedimentation was increased by staurosporine by approximately 30% at 5 µM (Fig. 3B), indicating that binding of the kinase inhibitor to the ATP binding site of synapsin I modifies its molecular interactions with the F-actin-based cytoskeleton.

Synapsin I is closer to Pim-1 than to other protein kinases

Staurosporine is a pan-kinase inhibitor exhibiting not only nanomolar affinities to Pim-1 but also to a wide array of protein kinases [42]. To ascertain whether the ATP-binding site of synapsin I is equally close to all known ATP-binding sites or specifically related to Pim-1, we computed with SiteAlign [41] the distance between the ATP-binding site of bovine synapsin I (1aux) and 978 ATP-binding sites from the Protein Data Bank. The 978 ATP-binding binding sites were extracted from the sc-PDB database [39] and feature a total of 433 unique proteins among which 110 are protein kinases. 113 entries describing 46 different protein kinases present a binding site distance below 0.20, the previously-determined computed threshold for discriminating similar from dissimilar binding sites [41]. This result suggests that the ATP site of synapsin I is similar to that of many other protein kinases. When looking at the top 25 ranked entries (Table 1), Pim-1 binding sites to various ATP-competitive inhibitors are the most numerous (8 times), but other serine/threonine protein kinases (e.g. casein kinase II) also share strong binding site similarities with the ATP site of synapsin I. A statistical analysis of binding site distances (SiteAlign d2 distance) to that of synapsin I was undertaken by computing, for each single protein present at least in 5 copies in the sc-PDB, the area under the ROC curve [40] in a simple binary classification system (similar, dissimilar) is calculated. Briefly, each of the 978 ATP-binding sites are ranked by decreasing distance to that of synapsin I and the rank distribution of every binding site sharing the same protein name (true positives are presumed similar to the reference) are compared to the ranks of all other active sites (true negatives are presumed dissimilar to the reference, see full results in Supplementary Table S2).

Table 1
The 25 ATP-binding sites of protein kinases closest to that of bovine synapsin I (PDB entry 1aux).

Areas under the curve (AUC) vary from very high (e.g. Pim-1; AUC  = 0.85) to significant (e.g. casein kinase II or CKII, cyclin-dependent kinase 2 or CDK2; AUC >0.65) to random (protein kinase A or PKA, Heat shock protein-90α or HSP-90α; AUC ≈0.50) and to even worse than random (Checkpoint kinase 1 or CHK1, DNA topoisomerase II, Diacylglycerolkinase or DGK; AUC <0.50; Fig. 4). We therefore predicted that inhibitors of protein active sites presenting a high ROC AUC value (Pim-1 kinase, CDK2, CKII) should cross-react with synapsin I whereas inhibitor of other proteins presenting a close-to-random of even lower AUC value (PKA, HSP-90α, CHK1, DNA topoisomerase II, DGK) should not. One high-affinity inhibitor (Fig. 5) of each of these 8 representative targets was purchased and tested for in vitro binding to bovine synapsin I. It is important to point out that these commercially available inhibitors and/or very close analogs have all been co-crystallized in the ATP-binding site of their respective target.

Figure 4
Similarity of ATP-binding proteins to bovine synapsin I.
Figure 5
Chemical structures of compounds tested for binding to bovine synapsin I.

In qualitative agreement with the predictions, the Pim-1 kinase inhibitor quercetagetin was the most potent binder to synapsin I (IC500.15±0.08 µM; maximal inhibition of 88.3±4.9%; Fig. 6A,B). (R)-roscovitin and compound 70159800251 still compete with ATP for binding to bovine synapsin I but with a higher IC50 value (1.0 and 0.5 µM, respectively) and a lower level of maximal inhibition (ca. 70%, Fig. 6B). Among other inhibitors tested here, other protein kinase inhibitors (PKA and CHK1 inhibitors) did not show any significant binding to synapsin I (Fig. 6A).

Figure 6
Binding of distinct ATP-competitors to the ATP-binding site of bovine synapsin I.


Comparing at high-throughput 3-D features of protein cavities is likely to play an increasing role to guide the functional annotation of novel genomic structures as well as predicting target-based selectivity profiles for pharmacological ligands. Despite several sequence and fold-independent computational methods have been proposed to detect remote binding site similarities among unrelated proteins [37], there are very few in silico studies guided by binding site comparisons that successfully predicted either the function of a protein from its 3-D structure or assigned a novel macromolecular target to an existing ligand. The SOIPPA method [74] was successfully used to predict remote binding site similarities between the binding site of selective estrogen receptor modulators (SERMs) at the ERα receptor and the Sarcoplasmic Reticulum Ca2 ion channel ATPase protein (SERCA) transmembrane domain [75], thus explaining known side effects of SERMs. Likewise, the NAD binding site of the Rossmann fold and the S-adenosyl-methionine (SAM)-binding site of SAM-methyltransferases were found to be similar and consequently permitted the prediction of the cross-reactivity of catechol-O-methyltransferase (COMT) inhibitors (entacapone, tolcapone) with the M.tuberculosis enoyl-acyl carrier protein reductase (InhA) [76]. The same approach predicted two unexpected targets for a known inhibitor of Trypasonoma brucei RNA editing ligase [77]. Last, the PocketPicker algorithm [78] was used to detect a cavity on the surface of a APOBEC3A structure, a protein which is able to inactivate retroviral genomes. Encoding the pocket as correlation vectors enabled the comparison of a set of 1300 ligand-binding sites from the PDBBind dataset [79]. Among top scoring entries were only nucleic acid-binding pockets [80]. Point mutation of the cavity-lining residues effectively led to mutants with a reduced antiviral activity. The pocket was shown to recognize the small 5.8S RNA [80] as a preliminary step to inactivate retroviral particles.

The main reason for the paucity of predictive reports is that many 3-D site comparison tools are extremely sensitive to atomic coordinates and thus better suited to detect global than local similarities [81]. In this context, a true advantage of the SiteAlign algorithm, used in the current study to detect local similarity between ATP-binding sites of Pim-1 kinase and synapsin I, is that cavity descriptors are assigned to Cα carbon atoms thus rendering the method fuzzy enough to be relatively insensitive to variations in 3-D coordinates (e.g. rotameric state and orientation of a side chain). A good illustration of this feature is exemplified by the matching of Pim-1 to synapsin I binding sites (Fig. 2C) which clearly shows a good fit of several pairs of residues (Lys67 vs. Lys269, Glu121 vs. Glu305) with significantly different side chain orientations. Recently-described alignment-independent binding site comparison methods (PocketMatch [82], FuzCav [70]) focusing on protein C-α atoms also found a significant similarity score between 1yhs and 1aux active sites (PocketMatch PMscore  = 50.79; FuzCav score  = 0.160). Conversely, two state-of-the-art full atom-matching methods (SitesBase [83], SiteEngine [84]) failed in finding the same binding sites similar (SiteEngine Match score  = 18.71; no output for SitesBase). SiteEngine notably fails to fit the pair of glutamic acid residues (Glu121 vs. Glu305, recall Fig. 2C) diverging in the orientation of their side chains. A detailed binding site representation at the atomic level is therefore detrimental to the detection of remote similarity between synapsin I and Pim-1 kinase.

The herein predicted remote similarity between ATP-binding sites of Pim-1 kinase and synapsin I could be experimentally validated by an in vitro competition assay. The affinity of staurosporine, a pan-kinase inhibitor, to bovine synapsin I is about 0.3 µM (Fig. 3). It is thus comparable to that observed for most serine/threonine protein kinases [42]. The polypharmacological profile [85] of staurosporine may be attributed to its lipophilicity and the tendency to recognize apolar surface patches. The hypothesized binding mode to synapsin I is however remarkably reminiscent from that seen in protein kinases with a bidentate hydrogen bond to main chain atoms of a hinge region accompanying apolar interactions (Fig. 2D). Binding to synapsin I could be verified for other inhibitors (roscovitin, quercetagetin, 70159800251) targeting different protein kinases (cyclin-dependant kinases, Pim-1, casein kinase II), still with submicromolar affinities (Fig. 5). One might argue that a remote similarity among ATP-binding sites is trivial and that the herein presented data are not surprising. Interestingly, the ATP-binding site of synapsin I was predicted to be much more distant from that of other serine/threonine protein kinases (e.g. Chk1, PkA; Fig. 4) and this assumption could be verified in vitro by testing inhibitors of the latter proteins for binding to synapsin I (Fig. 6). Other ATP-competitors binding to a kinase (DGK) or two other targets (DNA topoisomerase II, HSP-90α) did not either bind to synapsin I (Fig. 6A). Our data therefore pinpoints a binding site similarity between synapsin I and some serine/threonine protein kinases but not all of them.

Off-targets for protein kinase inhibitors outside the protein kinome (e.g. glycogen phosphorylase, malate dehydrogenase, HSP-90β) have already been discovered by proteomics [86], [87], [88] although no inhibition constants have been reported. Notably, roscovitin binds to pyridoxal kinase at the pyridoxal but not the ATP-binding site [32]. We herewith supplement the list of off-targets for four protein kinase inhibitors with synapsin I. Binding affinities in an ATP-γ35S competition assay are remarkably high (submicromolar) and comparable to those seen for protein kinases.

Since some protein kinase inhibitors tested here actively and directly compete for the binding of ATP to synapsin I and modify the interactions of synapsin I with the actin-based cytoskeleton, it is tempting to speculate that at least some of the effects of protein kinase inhibitors on neurotransmitter release and presynaptic function are attributable to a direct binding to synapsin I and reveal new potential targets for the action of protein kinase inhibitors on synaptic transmission and plasticity.

Materials and Methods

Virtual screening of sc-PDB binding sites

Protein-ligand binding sites were retrieved from the 2006 release of the sc-PDB (, a database of 6 415 druggable protein-ligand binding sites [39] from the Protein Data Bank. A binding site is described by any amino acid for which at least one heavy atom is closer than 4.5 Å from any heavy atom of the bound pharmacological ligand. The full sc-PDB dataset was screened for similarity to the staurosporine-binding site of the Pim-1 kinase (PDB entry 1yhs) using standard settings of the SiteAlign v4.0 program. Algorithmic details of SiteAlign have been described elsewhere [41]. Briefly, eight topological and physicochemical attributes are projected from the Cα-atom of cavity-lining residues to an 80 triangle-discretized polyhedron placed at the center of the binding site, thus defining a cavity fingerprint of 640 integers. 3-D alignment is performed by moving the sphere within the target binding site while keeping the query sphere fixed. After each move, the distance of the newly described cavity descriptor is compared to that of the query, the best alignment having the minimal distance between both cavity fingerprints. Two distances are used in SiteAlign. The d1 distance is suited to measure global similarities and is a sum of normalized distances between the 8 descriptors on all indexed triangles with non-null values for either the query or the target. Previous benchmarking studies suggest that a d1 distance of 0.60 is a good threshold for discriminating similar from dissimilar binding sites [41]. The d2 distance is suited to measure local similarities and is a sum of normalized distances between the 8 descriptors on all indexed triangles with non-null values for both the query and the target. Previous benchmarking studies suggest that a d2 distance of 0.20 is a good threshold for discriminating similar from dissimilar binding sites. In the current screen, sc-PDB entries were ranked by increasing d2 distance to the 1hys query. To avoid false positives [41], the d2 distance was post-processed and set to 1.0 if the corresponding d1 distance is higher than 0.60.

Screening for binding site similarity to the ATP-binding site of bovine synapsin (PDB entry 1aux) was done as described above. The dataset of 978 ATP-binding sites was selected from the current sc-PDB database as previously described [41], [70]. SiteAlign screen was performed using standard settings and entries were ranked by increasing d2 distance with the same post-processing as previously described (d2 = 1 if d1 >0.6). A ROC score (area under the ROC curve) is computed from the distance table with an in-house Pipeline Pilot workflow [89] for each occurrence of a protein name represented by at least 5 entries in the dataset (n = 43). The higher the ROC score for a particular protein, the more similar its protein-ligand binding sites to that of the synapsin I reference. A ROC score of 0.5 indicate a random distribution of d2 scores for a particular protein.

Automated docking of staurosporine to bovine synapsin I

3-D atomic coordinates of staurosporine were obtained by Corina v3.5[90] from a 2-D Marvin sketch [91]. Hydrogen atoms were added using standard topological rules in Sybyl v.8.0 [92] and coordinates were saved in mol2 format. Standard settings of the Gold v4.1 program [93] were used to dock staurosporine to the ATP-binding site of bovine synapsin (PDB entry 1aux) whose coordinates were retrieved from the sc-PDB databank [30]. The cavity was defined as any protein atom present in a 10 Å-radius sphere centered on the center of mass of the sc-PDB binding site. The side chain of Lys67 was considered flexible during the docking by explicit definition of 27 rotameric states from the standard Gold rotamer library. Poses were scored with the Goldscore fitness function.

Comparison of bovine synapsin I and human Pim-1 ATP-binding sites with other binding site matching methods

ATP-binding sites of 1aux (bovine synapsin I) and 1yhs (human Pim-1 kinase) were retrieved from the sc-PDB website ( and compared with the in-house FuzCav algorithm with default parameters [70]. Web interfaces to SitesBase [83] (, SiteEngine [84] ( and PocketMatch [82] ( were used to compare the same entries. Active site detection was here achieved by specifying the chemical component HET code of the co-crystallized ligands (SAP for 1uax, STU for 1yhs).

ATP-γ35S binding assays

Synapsin I was purified from bovine brain [71] and stored in liquid nitrogen in 200 mM NaCl, 25 mM TrisCl, pH 7.4. Synapsin I (500 nM) was incubated with 200 nM ATP-γ35S (Perkin Elmer, Waltham, MA) in 50 mM HEPES-NaOH pH 7.4, 25 mM NaCl, 0.5 mM CaCl2 and 2 mM MgCl2 for 1 h at room temperature in the absence or presence of increasing concentrations (0.1–30 µM) of either cold ATP, staurosporine (Sigma, Milan, Italy), (R)-roscovitin (Caiman, Ann Arbor, MI), quercetagetin (Calbiochem, San Diego, CA), 70159800251 (Otava, kiev, Ukraine), H-89 (LC Laboratories, Woburn, MA), 4072–7730 (ChemDiv, San Diego, CA), R59022 and novobiocin (MP Biochemicals, Illkirch, France) and last CCT018059 (SPI-Bio, Montigny Le Bretonneux, France). ATP-γ35S binding was quantified as previously described [94]. Briefly, aliquots of the samples were spotted onto squares of phosphocellulose paper (Upstate/Millipore, Billerica, MA). The paper squares were extensively washed with deionized water for 30 min, air-dried and analyzed for radioactivity by using the Perkin Elmer Cyclone Plus Phosphor Imager. After subtraction of the background values (samples with no synapsin I), data from individual competition curves were fitted with a sigmoidal dose-response function (f = min +(max–min)/(1+10∧((logEC50-x)*Hillslope)) using the Sigmaplot 8.0 software (SPSS Inc., Chicago, IL) to yield IC50 and maximal inhibition values. Data in the plots are the means ± sem of at least 5 independent experiments.

Actin Bundling Assays

Actin was purified from acetone powder of rabbit skeletal muscles [95], [96] and stored in liquid nitrogen in in 2 mM Tris pH 8, 0.2 mM ATP, 0,2 mM CaCl2, 0.125 mM β-mercaptoethanol and 0.005% NaN3 (G-buffer). Before the experiments, both G-actin and synapsin I were prespun for 1 h at 4°C at 300,000× g to remove large aggregates. G-actin was polymerized at room temperature for 1 h in the presence of 100 mM KCl, 1.2 mM MgCl2. Synapsin I (final concentration, 0.5 µM) was preincubated with increasing concentrations (1–20 µM) of staurosporine for 1 h at room temperature in 200 mM NaCl, 25 mM TrisCl pH 7.4. Actin bundling was assessed by incubating the synapsin/staurosporine samples with F-actin (final concentration, 5 µM) under polymerization conditions (100 mM KCl, 1.2 mM MgCl2 in G-buffer) for 1.5 h at room temperature followed by low-speed centrifugation (10,000× g for 15 min) to separate actin bundles (Bahler & Greengard, 1987). Pellets were solubilized in sample buffer [97] and analyzed by sodium dodecylsulfate polyacrylamide gel electrophoresis (SDS-PAGE) using 10% acrylamide in the resolving gel. Gels were fixed, stained with Coomassie Blue and destained. Densitometric analysis of the actin bands was carried out by using the ImageQuant system (GE Healthcare) followed by densitometric analysis of the fluorograms and by data interpolation into a standard curve of purified G-actin run in parallel with the unknown samples.

Supporting Information

Table S1

(0.04 MB DOC)

Table S2

(0.06 MB DOC)


We acknowledge the supercomputing facilities of the CINES (Montpellier, France) and IN2P3 (Villeurbanne, France) for allocation of computing time to D.R. (Project x2010075024).


Competing Interests: The authors have declared that no competing interests exist.

Funding: The project was supported by the French Ministry of Research and Technology (Ph.D. grant to C.S.), the Ministry of the University and Research (PRIN 2008 grant), The Compagnia di San Paolo Torino and Telethon-Italy (GCP09134 grant to F.B.). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.


1. Wermuth CG. Similarity in drugs: reflections on analogue design. Drug Discov Today. 2006;11:348–354. [PubMed]
2. Schuster D, Laggner C, Langer T. Why drugs fail—a study on side effects in new chemical entities. Curr Pharm Des. 2005;11:3545–3559. [PubMed]
3. Shoshan MC, Linder S. Target specificity and off-target effects as determinants of cancer drug efficacy. Expert Opin Drug Metab Toxicol. 2008;4:273–280. [PubMed]
4. Klabunde T, Evers A. GPCR antitarget modeling: pharmacophore models for biogenic amine binding GPCRs to avoid GPCR-mediated side effects. Chembiochem. 2005;6:876–889. [PubMed]
5. Yang L, Chen J, He L. Harvesting candidate genes responsible for serious adverse drug reactions from a chemical-protein interactome. PLoS Comput Biol. 2009;5:e1000441. [PMC free article] [PubMed]
6. Dessailly BH, Nair R, Jaroszewski L, Fajardo JE, Kouranov A, et al. PSI-2: structural genomics to cover protein domain family space. Structure. 2009;17:869–881. [PMC free article] [PubMed]
7. Nair R, Liu J, Soong TT, Acton TB, Everett JK, et al. Structural genomics is the largest contributor of novel structural leverage. J Struct Funct Genomics. 2009;10:181–191. [PMC free article] [PubMed]
8. Daniel M, Stuart L, Christopher C, Stuart W, Adam N. Synthesis of Natural-Product-Like Molecules with Over Eighty Distinct Scaffolds. Angew Chem Int Ed Engl. 2009;48:104–109. [PMC free article] [PubMed]
9. Nielsen TE, Schreiber SL. Towards the optimal screening collection: a synthesis strategy. Angew Chem Int Ed Engl. 2008;47:48–56. [PMC free article] [PubMed]
10. Berman HM, Westbrook J, Feng Z, Gilliland G, Bhat TN, et al. The Protein Data Bank. Nucleic Acids Res. 2000;28:235–242. [PMC free article] [PubMed]
11. Schneider M, Lane L, Boutet E, Lieberherr D, Tognolli M, et al. The UniProtKB/Swiss-Prot knowledgebase and its Plant Proteome Annotation Program. J Proteomics. 2009;72:567–573. [PMC free article] [PubMed]
12. Wang Y, Xiao J, Suzek TO, Zhang J, Wang J, et al. PubChem: a public information system for analyzing bioactivities of small molecules. Nucleic Acids Res. 2009;37:W623–633. [PMC free article] [PubMed]
13. Paolini GV, Shapland RH, van Hoorn WP, Mason JS, Hopkins AL. Global mapping of pharmacological space. Nat Biotechnol. 2006;24:805–815. [PubMed]
14. Caron PR, Mullican MD, Mashal RD, Wilson KP, Su MS, et al. Chemogenomic approaches to drug discovery. Curr Opin Chem Biol. 2001;5:464–470. [PubMed]
15. Ong SE, Schenone M, Margolin AA, Li X, Do K, et al. Identifying the proteins to which small-molecule probes and drugs bind in cells. Proc Natl Acad Sci U S A. 2009;106:4617–4622. [PubMed]
16. Bajorath J. Computational analysis of ligand relationships within target families. Curr Opin Chem Biol. 2008;12:352–358. [PubMed]
17. Rognan D. Chemogenomic approaches to rational drug design. Br J Pharmacol. 2007;152:38–52. [PMC free article] [PubMed]
18. Shoichet BK. Virtual screening of chemical libraries. Nature. 2004;432:862–865. [PMC free article] [PubMed]
19. Ekins S, Mestres J, Testa B. In silico pharmacology for drug discovery: methods for virtual ligand screening and profiling. Br J Pharmacol. 2007;152:9–20. [PMC free article] [PubMed]
20. Keiser MJ, Roth BL, Armbruster BN, Ernsberger P, Irwin JJ, et al. Relating protein pharmacology by ligand chemistry. Nat Biotechnol. 2007;25:197–206. [PubMed]
21. Keiser MJ, Setola V, Irwin JJ, Laggner C, Abbas AI, et al. Predicting new molecular targets for known drugs. Nature. 2009;462:175–181. [PMC free article] [PubMed]
22. Peltason L, Bajorath J. SAR index: quantifying the nature of structure-activity relationships. J Med Chem. 2007;50:5571–5578. [PubMed]
23. Mestres J, Martin-Couce L, Gregori-Puigjane E, Cases M, Boyer S. Ligand-based approach to in silico pharmacology: nuclear receptor profiling. J Chem Inf Model. 2006;46:2725–2736. [PubMed]
24. Nettles JH, Jenkins JL, Bender A, Deng Z, Davies JW, et al. Bridging chemical and biological space: “target fishing” using 2D and 3D molecular descriptors. J Med Chem. 2006;49:6802–6810. [PubMed]
25. Nidhi, Glick M, Davies JW, Jenkins JL. Prediction of biological targets for compounds using multiple-category Bayesian models trained on chemogenomics databases. J Chem Inf Model. 2006;46:1124–1133. [PubMed]
26. Martin YC, Kofron JL, Traphagen LM. Do structurally similar molecules have similar biological activity? J Med Chem. 2002;45:4350–4358. [PubMed]
27. Chen YZ, Zhi DG. Ligand-protein inverse docking and its potential use in the computer search of protein targets of a small molecule. Proteins. 2001;43:217–226. [PubMed]
28. Li H, Gao Z, Kang L, Zhang H, Yang K, et al. TarFisDock: a web server for identifying drug targets with docking approach. Nucleic Acids Res. 2006;34:W219–224. [PMC free article] [PubMed]
29. Paul N, Kellenberger E, Bret G, Muller P, Rognan D. Recovering the true targets of specific ligands by virtual screening of the protein data bank. Proteins. 2004;54:671–680. [PubMed]
30. Kellenberger E, Foata N, Rognan D. Ranking targets in structure-based virtual screening of three-dimensional protein libraries: methods and problems. J Chem Inf Model. 2008;48:1014–1025. [PubMed]
31. Zahler S, Tietze S, Totzke F, Kubbutat M, Meijer L, et al. Inverse in silico screening for identification of kinase inhibitor targets. Chem Biol. 2007;14:1207–1214. [PubMed]
32. Tang L, Li MH, Cao P, Wang F, Chang WR, et al. Crystal structure of pyridoxal kinase in complex with roscovitin and derivatives. J Biol Chem. 2005;280:31220–31229. [PubMed]
33. Do QT, Renimel I, Andre P, Lugnier C, Muller CD, et al. Reverse pharmacognosy: application of selnergy, a new tool for lead discovery. The example of epsilon-viniferin. Curr Drug Discov Technol. 2005;2:161–167. [PubMed]
34. Cai J, Han C, Hu T, Zhang J, Wu D, et al. Peptide deformylase is a potential target for anti-Helicobacter pylori drugs: reverse docking, enzymatic assay, and X-ray crystallography validation. Protein Sci. 2006;15:2071–2081. [PubMed]
35. Muller P, Lena G, Boilard E, Bezzine S, Lambeau G, et al. In silico-guided target identification of a scaffold-focused library: 1,3,5-triazepan-2,6-diones as novel phospholipase A2 inhibitors. J Med Chem. 2006;49:6768–6778. [PubMed]
36. Ferrara P, Gohlke H, Price DJ, Klebe G, Brooks CL, 3rd Assessing scoring functions for protein-ligand interactions. J Med Chem. 2004;47:3032–3047. [PubMed]
37. Kellenberger E, Schalon C, Rognan D. How to measure the similarity between protein ligand-binding sites? Current Computer-Aided Drug Design. 2008;4:209–220.
38. Rognan D. Structure-Based Approaches to Target Fishing and Ligand Profiling. Mol Inf. 2010;29:176–187.
39. Kellenberger E, Muller P, Schalon C, Bret G, Foata N, et al. sc-PDB: an annotated database of druggable binding sites from the Protein Data Bank. J Chem Inf Model. 2006;46:717–727. [PubMed]
40. Triballeau N, Acher F, Brabet I, Pin JP, Bertrand HO. Virtual screening workflow development guided by the “receiver operating characteristic” curve approach. Application to high-throughput docking on metabotropic glutamate receptor subtype 4. J Med Chem. 2005;48:2534–2547. [PubMed]
41. Schalon C, Surgand JS, Kellenberger E, Rognan D. A simple and fuzzy method to align and compare druggable ligand-binding sites. Proteins. 2008;71:1755–1778. [PubMed]
42. Fabian MA, Biggs WH, 3rd, Treiber DK, Atteridge CE, Azimioara MD, et al. A small molecule-kinase interaction map for clinical kinase inhibitors. Nat Biotechnol. 2005;23:329–336. [PubMed]
43. Manning G, Whyte DB, Martinez R, Hunter T, Sudarsanam S. The protein kinase complement of the human genome. Science. 2002;298:1912–1934. [PubMed]
44. Johnson LN. Protein kinase inhibitors: contributions from structure to clinical compounds. Q Rev Biophys. 2009;42:1–40. [PubMed]
45. Grant SK. Therapeutic protein kinase inhibitors. Cell Mol Life Sci. 2009;66:1163–1177. [PubMed]
46. De Camilli P, Benfenati F, Valtorta F, Greengard P. The synapsins. Annu Rev Cell Biol. 1990;6:433–460. [PubMed]
47. Greengard P, Valtorta F, Czernik AJ, Benfenati F. Synaptic vesicle phosphoproteins and regulation of synaptic function. Science. 1993;259:780–785. [PubMed]
48. Fdez E, Hilfiker S. Vesicle pools and synapsins: new insights into old enigmas. Brain Cell Biol. 2006;35:107–115. [PubMed]
49. Sudhof TC, Czernik AJ, Kao HT, Takei K, Johnston PA, et al. Synapsins: mosaics of shared and individual domains in a family of synaptic vesicle phosphoproteins. Science. 1989;245:1474–1480. [PubMed]
50. Kao HT, Porton B, Hilfiker S, Stefani G, Pieribone VA, et al. Molecular evolution of the synapsin gene family. J Exp Zool. 1999;285:360–377. [PubMed]
51. Esser L, Wang CR, Hosaka M, Smagula CS, Sudhof TC, et al. Synapsin I is structurally similar to ATP-utilizing enzymes. EMBO J. 1998;17:977–984. [PubMed]
52. Brautigam CA, Chelliah Y, Deisenhofer J. Tetramerization and ATP binding by a protein comprising the A, B, and C domains of rat synapsin I. J Biol Chem. 2004;279:11948–11956. [PubMed]
53. Hosaka M, Sudhof TC. Synapsin III, a novel synapsin with an unusual regulation by Ca2+. J Biol Chem. 1998;273:13371–13374. [PubMed]
54. Hosaka M, Sudhof TC. Synapsins I and II are ATP-binding proteins with differential Ca2+ regulation. J Biol Chem. 1998;273:1425–1429. [PubMed]
55. Hosaka M, Sudhof TC. Homo- and heterodimerization of synapsins. J Biol Chem. 1999;274:16747–16753. [PubMed]
56. Lane-Guermonprez L, Morot-Gaudry-Talarmain Y, Meunier FM, O'Regan S, Onofri F, et al. Synapsin associates with cyclophilin B in an ATP- and cyclosporin A-dependent manner. J Neurochem. 2005;93:1401–1411. [PubMed]
57. Sakurada K, Kato H, Nagumo H, Hiraoka H, Furuya K, et al. Synapsin I is phosphorylated at Ser603 by p21-activated kinases (PAKs) in vitro and in PC12 cells stimulated with bradykinin. J Biol Chem. 2002;277:45473–45479. [PubMed]
58. Jovanovic JN, Benfenati F, Siow YL, Sihra TS, Sanghera JS, et al. Neurotrophins stimulate phosphorylation of synapsin I by MAP kinase and regulate synapsin I-actin interactions. Proc Natl Acad Sci U S A. 1996;93:3679–3683. [PubMed]
59. Jovanovic JN, Sihra TS, Nairn AC, Hemmings HC, Jr, Greengard P, et al. Opposing changes in phosphorylation of specific sites in synapsin I during Ca2+-dependent glutamate release in isolated nerve terminals. J Neurosci. 2001;21:7944–7953. [PubMed]
60. Onofri F, Messa M, Matafora V, Bonanno G, Corradi A, et al. Synapsin phosphorylation by SRC tyrosine kinase enhances SRC activity in synaptic vesicles. J Biol Chem. 2007;282:15754–15767. [PubMed]
61. Menegon A, Bonanomi D, Albertinazzi C, Lotti F, Ferrari G, et al. Protein kinase A-mediated synapsin I phosphorylation is a central modulator of Ca2+-dependent synaptic activity. J Neurosci. 2006;26:11670–11681. [PubMed]
62. Benfenati F, Valtorta F, Chieregatti E, Greengard P. Interaction of free and synaptic vesicle-bound synapsin I with F-actin. Neuron. 1992;8:377–386. [PubMed]
63. Benfenati F, Valtorta F, Rubenstein JL, Gorelick FS, Greengard P, et al. Synaptic vesicle-associated Ca2+/calmodulin-dependent protein kinase II is a binding protein for synapsin I. Nature. 1992;359:417–420. [PubMed]
64. Ceccaldi PE, Grohovaz F, Benfenati F, Chieregatti E, Greengard P, et al. Dephosphorylated synapsin I anchors synaptic vesicles to actin cytoskeleton: an analysis by videomicroscopy. J Cell Biol. 1995;128:905–912. [PMC free article] [PubMed]
65. Chi P, Greengard P, Ryan TA. Synaptic vesicle mobilization is regulated by distinct synapsin I phosphorylation pathways at different frequencies. Neuron. 2003;38:69–78. [PubMed]
66. Chi P, Greengard P, Ryan TA. Synapsin dispersion and reclustering during synaptic activity. Nat Neurosci. 2001;4:1187–1193. [PubMed]
67. Kao HT, Song HJ, Porton B, Ming GL, Hoh J, et al. A protein kinase A-dependent molecular switch in synapsins regulates neurite outgrowth. Nat Neurosci. 2002;5:431–437. [PubMed]
68. Bonanomi D, Menegon A, Miccio A, Ferrari G, Corradi A, et al. Phosphorylation of synapsin I by cAMP-dependent protein kinase controls synaptic vesicle dynamics in developing neurons. J Neurosci. 2005;25:7299–7308. [PubMed]
69. Hilfiker S, Benfenati F, Doussau F, Nairn AC, Czernik AJ, et al. Structural domains involved in the regulation of transmitter release by synapsins. J Neurosci. 2005;25:2658–2669. [PubMed]
70. Weill N, Rognan D. Alignment-free ultra-high-throughput comparison of druggable protein-ligand binding sites. J Chem Inf Model. 2010;50:123–135. [PubMed]
71. Bahler M, Greengard P. Synapsin I bundles F-actin in a phosphorylation-dependent manner. Nature. 1987;326:704–707. [PubMed]
72. Bahler M, Benfenati F, Valtorta F, Czernik AJ, Greengard P. Characterization of synapsin I fragments produced by cysteine-specific cleavage: a study of their interactions with F-actin. J Cell Biol. 1989;108:1841–1849. [PMC free article] [PubMed]
73. Valtorta F, Greengard P, Fesce R, Chieregatti E, Benfenati F. Effects of the neuronal phosphoprotein synapsin I on actin polymerization. I. Evidence for a phosphorylation-dependent nucleating effect. J Biol Chem. 1992;267:11281–11288. [PubMed]
74. Xie L, Bourne PE. Detecting evolutionary relationships across existing fold space, using sequence order-independent profile-profile alignments. Proc Natl Acad Sci U S A. 2008;105:5441–5446. [PubMed]
75. Xie L, Wang J, Bourne PE. In silico elucidation of the molecular mechanism defining the adverse effect of selective estrogen receptor modulators. PLoS Comput Biol. 2007;3:e217. [PMC free article] [PubMed]
76. Kinnings SL, Liu N, Buchmeier N, Tonge PJ, Xie L, et al. Drug discovery using chemical systems biology: repositioning the safe medicine Comtan to treat multi-drug and extensively drug resistant tuberculosis. PLoS Comput Biol. 2009;5:e1000423. [PMC free article] [PubMed]
77. Durrant JD, Amaro RE, Xie L, Urbaniak MD, Ferguson MA, et al. A multidimensional strategy to detect polypharmacological targets in the absence of structural and sequence homology. PLoS Comput Biol. 6:e1000648. [PMC free article] [PubMed]
78. Weisel M, Proschak E, Schneider G. PocketPicker: analysis of ligand binding-sites with shape descriptors. Chem Cent J. 2007;1:7. [PMC free article] [PubMed]
79. Wang R, Fang X, Lu Y, Wang S. The PDBbind database: collection of binding affinities for protein-ligand complexes with known three-dimensional structures. J Med Chem. 2004;47:2977–2980. [PubMed]
80. Stauch B, Hofmann H, Perkovic M, Weisel M, Kopietz F, et al. Model structure of APOBEC3C reveals a binding pocket modulating ribonucleic acid interaction required for encapsidation. Proc Natl Acad Sci U S A. 2009;106:12079–12084. [PubMed]
81. Wallach I, Lilien RH. Prediction of sub-cavity binding preferences using an adaptive physicochemical structure representation. Bioinformatics. 2009;25:i296–304. [PMC free article] [PubMed]
82. Yeturu K, Chandra N. PocketMatch: a new algorithm to compare binding sites in protein structures. BMC Bioinformatics. 2008;9:543. [PMC free article] [PubMed]
83. Gold ND, Jackson RM. Fold independent structural comparisons of protein-ligand binding sites for exploring functional relationships. J Mol Biol. 2006;355:1112–1124. [PubMed]
84. Shulman-Peleg A, Nussinov R, Wolfson HJ. Recognition of functional sites in protein structures. J Mol Biol. 2004;339:607–633. [PubMed]
85. Hopkins AL. Network pharmacology: the next paradigm in drug discovery. Nat Chem Biol. 2008;4:682–690. [PubMed]
86. Knockaert M, Wieking K, Schmitt S, Leost M, Grant KM, et al. Intracellular Targets of Paullones. Identification following affinity purification on immobilized inhibitor. J Biol Chem. 2002;277:25493–25501. [PubMed]
87. Brehmer D, Greff Z, Godl K, Blencke S, Kurtenbach A, et al. Cellular targets of gefitinib. Cancer Res. 2005;65:379–382. [PubMed]
88. Bach S, Knockaert M, Reinhardt J, Lozach O, Schmitt S, et al. Roscovitin targets, protein kinases and pyridoxal kinase. J Biol Chem. 2005;280:31208–31219. [PubMed]
89. Pipeline Pilot v7.5, Accelrys Limited, San Diego, CA
90. Molecular Networks GmbH, D-91052 Erlangen, Germany
91. ChemAxon Kft., Budapest 1037, Hungary
92. TRIPOS, Assoc., Inc., St-Louis, MO
93. The Cambridge Crystallographic Data Centre, Cambridge, CB2 1EZ, UK
94. McGuinness TL, Lai Y, Greengard P. Ca2+/calmodulin-dependent protein kinase II. Isozymic forms from rat forebrain and cerebellum. J Biol Chem. 1985;260:1696–1704. [PubMed]
95. MacLean-Fletcher S, Pollard TD. Identification of a factor in conventional muscle actin preparations which inhibits actin filament self-association. Biochem Biophys Res Commun. 1980;96:18–27. [PubMed]
96. Spudich JA, Watt S. The regulation of rabbit skeletal muscle contraction. I. Biochemical studies of the interaction of the tropomyosin-troponin complex with actin and the proteolytic fragments of myosin. J Biol Chem. 1971;246:4866–4871. [PubMed]
97. Laemmli UK. Cleavage of structural proteins during the assembly of the head of bacteriophage T4. Nature. 1970;227:680–685. [PubMed]

Articles from PLoS ONE are provided here courtesy of Public Library of Science