Computational studies on antineoplastic target involved in cell cycle regulation:

an homology model for CDK7/CycH complex and several MD simulation for

ligands binding the CDK2/CycE complex outside the main pocket.
A) CDK7/CycH
Building a consistent homology model structure as far as CDK7/CycH complex has been done
starting from the only crystallized structure founded for this kinase inside the online PDB database.
The crystal, in its inactive form, (PDBcode 1UA2) is resolved at 3.02 and cocrystalized with ATP.
It shows also the Thr170 residue phosphorilated. Is not clear if exists a fixed temporal sequence for
the Thr170s phosporilation, the opening of the Activation Loop (res 155-182) and the formation of
the active complex. It is also uknown how this sequence of events evolve during time. The highest
B-factor for 1UA2 has been founded in the region close the Activation Loop furthermore the main
sequence showed an interruption in the main ammynoacidic sequence between the coil connetting
the C-alpha-helix, a extremely conserved sequence in all CDKs, and the third B-sheet. The
connecting coil is, with high probability, involved in the CDK7/CycH formation and activation and
starts from residue 44 until residue 55. It preceeds the conserved sequence NRTARLE, forming the
core of the C-alpha-helix, and when activated the coil present the residue 170 phosporilated, as
reported before. Initially, all proteins belonging to CDKs family have been ranked starting from the
one with the highest level of homology using the online tools BLAST and FASTA.
The homology score has been evaluated using alternatively a BLOSUM80 and a PAM250 matrix
and the CDK2 (PDBcode 1W98) structure has been choosen as best template to model the missing
coordinates for the AA sequence and the ALoop in the active conformation. A secondary crystal
structure (PDBcode 1VYW) has shown the highest identity level (over 44,5%), and has been used
as a secondary template only for the that residues hard to model.
A third template (PDBcode 1KXU) with a resolution of 2.60 has been employed to model the
regulatory subunit of the complex.
The initial run has generated ten poor trial models, further refined with a more accurate modelling
protocol. To emprove the method, the second run has been planned increasing the number of
resulting structures from 10 to 30 full-hydrogen models. It has been also scheduled the repeating of
an energy refinement protocol 5 times more accurate than in the trail. It has been also choosen a
strong alghoritm of energy minimization and an higher level of differentiation between the resulting
generated models.
All the results has been processed with the PROCHECK suite and the best model showed 90,2% of
the amminoacydic sequence in the core region and 8,8% in the allowed region. The structure, using

the Amber9 package, has been solvatated with an 8.00 TIP3P periodic box and minimized in 4
steps with initial restraints to avoid unexpected movement due by energy of the system. The heating
process has been regulated using both the Andersen coupling scheme and the Langevin dynamic in
6 different steps, eachone covering 50K. The MD simulation has been divided into 5 run of 0.50 ns
with a timestep of 0.002 for 500000 fs using the SHAKE alghorithm and the PME method.
Analyzing the trajectories of the complex after the MD production, the homology model for
CDK7/CycH has presented a quite stable RMSD.
B) CDK2/CycE
Setting this system has been easier because the higher number of crystallized and co-crystallized
structures in the active conformation available in the PDB database. The preliminary process of
blind-docking, using the Autodock4 software and different DFT optimized molecules, has been
made using a grid of 120x100x120 2. The grid spacing was of 0.375 and the grid has been
centered on ALA151 residue of 1W98 structure to show the most suitable pocket-regions. In further
investigations the grid has been linked to the different weight-mass centers of each subunit,
increasing the number of points for the grid and lowering the spacing. This has been done to
investigate deeper the surface of the several pockets, that usually hosts the Activation Loop of
CDKs, positioned over the interface region. Two ligands has been founded to have an intersting
binding mode for the regolatory subunit of the Cdk2/CycE complex: a derivative of Meldrums
acid and a ter-phenylic molecule. The best docked pose for first one showed, after the clusterization
with a tolerance in RMSD of 2.0 , a binding energy of -7.24 kcal/mol, the second one instead
showed a binding energy of -7.06 kcal/mol. Both poses has been submitted to acqueos TIP3P
periodic solvent MD simulation to test the stability of the docking results. The MD simulationprotocol has been repeated two times because during the first simulation the counterions employed
to neutralize the protein-charge gave undesidered interaction with the molecule inside the bindingsite. It would be impossible to analyze the binding mode, the trajectory and the RMSD through
time, for this reason during the second run, the total charge has been rescaled on the entire catalytic
subunit, without making use of any ions, and after 5 ns of simulation it has been possible to find
that both molecules remain inside the pocket with a quite good RMSD. This suggest that they could
be used as disruptor of the complex CDK2/CycE.