SUPPLEMENTARY MATERIAL
Materials and methods.
Extraction of interaction models from molecular dynamics trajectory
Construction and molecular dynamics simulations of CDK6-p16INK4a complex, as well as the determination of its interactions, were carried out as described in (Villacanas et al., 2002). During the last 500 ps of the production time (500 snapshots), the position of those atoms involved in the pharmacophoric features of p16INK4a (those atoms belonging to residues of p16INK4a that effectively interact with CDK6) was monitored. As a result, the mean positions of these atoms and their maximum deviations were calculated. If several atoms contribute to the same effective interaction, the centre of masses (CM) of the mean positions was calculated and it was treated as a single interaction point. Since all points coming from the production time must be referred to the same reference system, each snapshot was superimposed to the initial geometry (first snapshot). By carrying out this procedure for all the pharmacophoric features, a 3D picture of all the pharmacophoric points was obtained.
Computer screening procedure and compound selection
The distances from the selected 3D models and the nature of the interactions were introduced as input to search compounds matching this query. Catalyst (Accelrys, Inc., San Diego, CA, USA) software was used, together with the Available Chemical Database (ACD), the National Cancer Institute (NCI) database, the Maybridge database and the Derwent World Drug Database. The matches were filtered. In general, molecules with unusual bonds or with a molecular weight greater than 750 a.u. were not selected. Each conformation was docked into the CDK6 structure, obtained from CDK6-p16INK4a dynamics production. Pharmacophoric features were automatically assigned to the ligands and the distances between the pharmacophoric points for all the conformations were calculated. Next, we tested for a combination in each conformation that satisfied our pharmacophore hypothesis. If it did, a superimposition between the ligand and the 3D structure of our pharmacophoric hypothesis was carried out, optimizing the RMSD (root mean square deviation). Different steps were carried out in order to get the best-docked conformations for each hit. The maximum RMSD allowed was increased by 0.25 Å at every step, starting at 0.25 Å. The process was stopped when either the maximum RMSD was 3 Å or every molecule had at least one valid docking result. A visual structure analysis was carried out to reduce the number of final modeled complexes. Selected compounds were classified according to their structure.
p16INK4a competition assays
CDK6 immunoprecipitations were performed as described and non-esterified calcein 2 mM was added to some immunoprecipitation pellets. The pellets were kept at 4 ºC under gentle shaking for 30 min. 1 µM of p16INK4a was then added to both conditions (calcein 2 mM and control) and again they were kept shaking gently at 4 ºC for 30 min. After a fast spin, the supernatants containing p16INK4a not bound to CDK6 were collected and the pellets containing p16INK4a bound to CDK6 were washed three times in kinase buffer. All the samples were run in a 15% PAGE and p16INK4a and CDK6 immunoblottings were performed.
References
Villacanas, O., Perez, J.J., and Rubio-Martinez, J. (2002). Structural analysis of the inhibition of Cdk4 and Cdk6 by p16(INK4a) through molecular dynamics simulations. J Biomol Struct Dyn 20, 347-358.
Figure legends
Fig. SM1. CDK6 activity after treatment with the different compounds. CDK6 activity was determined on CDK6 immunoprecipitations in the presence of 1 mM of concentration of the different compounds to discriminate which structures presented inhibitory properties. A pGST-Rb (379-928) fusion protein was used as a substrate of the reaction and water as vehicle. Results are expressed as mean percentage with respect to a vehicle control. Mean + SD; n = 3. (**) indicate p < 0.01.
Fig. SM2. a Dose-effect curve of calcein on CDK2 and CDK1 activities. a CDK2 and cyclin B1 immunoprecipitations and kinase assays tested in the presence of increasing concentrations of non-esterified calcein (10 to 500 µM). Histone H1 was used as a substrate and p21Kip/Cip as a positive control of the inhibition. Mean + SD; n = 3. (*) indicates p < 0.05 and (**) indicates p < 0.01. b p16INK4a competition assay. p16-specific Western blot image shows how the presence of non-esterified calcein 2 mM diminishes the specific binding of p16 INK4A to CDK6, suggesting a competition for the same binding site. The experiment was performed three times. One representative example is shown. (SN: supernatant).