C.2. Cell Signaling and Regulation
Machinery in Response to DNA Damage
C.2.1. Investigators/areas of scientific expertise: R. D. Wood (PI) (Pharmacology, Pitt & Mol Oncology, UPCI); B.
Rajasekaran & V. Rapic-Otrin (Molecular Gen & Biochem,
Pitt); J. Rosenberg (Co-PI)
(Biological Sci, Pitt); B. Ermentrout & C. Chow (Math, Pitt); I. Bahar
(Comp Biol & Bioinformatics, Pitt); J. Stiles (PSC/CMU); S. Ta’asan (Mathematics, CMU); J.
Madura (Chem & Biochem, Duquesne)
C.2.2. Specific Aims
Numerous individual proteins and
molecular assemblies are involved in the activation and regulation of cellular
responses to DNA damage such as DNA repair, cell cycle arrest, or
apoptosis. Many of these responses are
mediated by mechano-chemical interactions, ranging from relatively simple
structural changes, through allosteric effects,
phosphorylation/dephosphorylation, proteolytic cleavage, to assembly into
multimeric machines and interaction cascades.
The molecular mechanism of these actions is in general not known.
Similarly, although it is believed that the spatial and temporal localization
of many of these interactions is critical, the effects are not understood even
at a qualitative level. The specific
aims of this development project (DP2) are to design, develop, and implement
computation tools to assist in the experimental analysis and overall
understanding of these phenomena at different levels, focusing on (i) the
structure-based dynamics of selected complexes implicated in DNA lesion
recognition, repair and signaling, (ii) the spatio-temporal localization of the
complexes acting as sensors of damage and initiators of repair, and (iii) the
kinetics of DNA-damage signaling pathways activated by ATM (178).
C.2.3. Background and Significance
DNA
repair and damage response mechanisms play a vital role in maintaining the
integrity of the genome against thousands of mutagenic and cytotoxic
modifications daily induced in the DNA sequence (179;180). The cellular response to DNA damage involves a concerted
interplay of multiple processes: cell cycle arrest, activation of DNA repair mechanisms,
control of the length and composition of telomeric chromatin, coordinated
translocation of proteins towards the damaged site, activation of transcription
or replication genes, or induction of apoptosis.
How DNA damage transmits signals to the cell cycle checkpoint machinery or to
the apoptotic pathways, or what are the factors that control the choice between
the temporary inhibition of DNA synthesis (while repair takes place), versus
the promotion of an apoptotic response are issues that remain to be elucidated (181). We will focus here on the regulation of the G1/S and G2/M
checkpoints in response to IR or UV damage.
C.2.3.1.
Multiprotein assemblies initiate and regulate DNA lesion recognition and
repair. The DNA
damage-signaling network couples a diversity of proteins - acting as sensors, transducers and effectors. Mutations in these give
rise to complex phenotypes. Several inherited human syndromes are
associated with the failure to sense DNA-damage, and damage-sensing proteins or
mechanisms are still unclear. UV-DDB protein, a mammalian protein having a high
affinity for UV-damaged DNA, and defective in one form (XP-E) of the skin
cancer prone disorder xeroderma pigmentosum (XP), is implicated in
nucleotide excision repair (NER), although the precise structure and function
of its two subunits, p48 and p127, are not known (138). Likewise, the XPC-hHR23B complex is proposed to be the
initial damage recognition factor in the global genome NER(139;182), although its mechanism of action and assembly with other
NER proteins, including the transcription/repair factor TFIIH, remain unclear (183). A major question is whether the multiprotein machines
initiating or regulating NER and homologous recombination (HR) form by
consecutive assembly of individual components at the site of DNA damage, or
whether they are assembled complexes, or factories, pre-existing in the cell (184-186). Fluorescence
measurements give a strong indication of sequential assembly for NER (187), while the idea of a massive cellular complex, BASC
(BRCA1-associated genome surveillance complex), has also been promoted (188). Another question regarding the mechanism of damage sensing
is whether there is a random diffusion of recognition components around the
nucleus, or a systematic screening and repair in nuclear factories devoted to
this purpose.
C.2.3.2.
ATM-activated pathways act as transducers of DNA damage signals. ATM (the product of the mutated gene in human autosomal recessive disorder ataxia telangiectasia) and ATM related proteins (ATR) are the homologs of
Rad3 in humans. They play a central role in triggering cellular responses to
DNA damage and to developmentally programmed DNA alterations (178;181;189).
ATM is particularly important in response to oxidative stress. Its substrates
include; c-Abl tyrosine kinase which regulates apoptosis (190;191) through phosphorylation of RNA polymerase II
(RNAP II) (192); the
tumor suppressor protein p53; the breast cancer tumor-suppressor gene product
BRCA1 (193), and the checkpoint kinases Chk1 and Chk2 that can trigger apoptosis by activating p53 (194) or BRCA1 (195) or can execute cell cycle arrest by promoting the binding of the nuclear export protein 14-3-3z to the dual
specificity phosphatase CdC25 (196) thus preventing the interaction of Cdc25 with
Cdc2 (197). These
proteins are engaged in closely intertwined, complementing or competing,
networks of interactions. See for example the ATM-signaling pathways in http://www.biocarta.com/pathfiles/atmPathway.asp,
or the KEGG pathways (198). More detailed ‘molecular interaction maps’ of cell cycle
control and DNA repair can be found in the studies of Kohn(199).
C.2.3.3.
The same protein can trigger multiple, sometime opposing, responses. ATM can selectively regulate p53-dependent cell cycle
checkpoints as well as apoptotic pathways (200;201). Likewise, the notion that p53 functions mainly to induce
apoptosis is challenged by some experiments (202;203). BRCA1, another example, is
involved in cell cycle arrest (204) following DNA damage, as well as induction of
apoptosis (205), while it takes part, along with BRCA2 (206), in the HR multiprotein complexes (207), and
is part of a chromatin remodeling complex (208). The balance between the
regulatory activities of these key proteins may indeed shift in one direction
or another depending on the connectivity of the network in which they
participate. And for a given network, variations in the strength and duration
of stimuli, changes in local concentration, diffusivity and chemical
reactivities - which depend on the type, developmental stage, and environment
of the cell - can lead to unique responses, or synergistic effects, not
apparent from the examination of the isolated pathways.
Unique
opportunities. This DP2
permits two different prominent groups in the School of Medicine at Pitt,
specialized in DNA damage recognition, repair and signaling (labs of Wood and
Rajasekaran) to join efforts for the first time. Other newly initiated
collaborations include those between these biophysical and the computational
groups (Rosenberg, Bahar, Meirovitch, Ermentrout, Madura, Stiles and Ta’asan)
across the four institutions. Rosenberg, a computationally oriented X-ray
crystallographer specialized in protein-DNA interactions, and Ta’asan, an
expert in fast numerical techniques for PDEs with focus on efficient
discretization methods, will serve as
the computational co-PI of DP2. The computational and biomedical expertise of
the investigators, together with the availability of experimental facilities at
the UPCI and the MGB, and the computational facilities at Pitt, CMU, PSC and
Duquesne, present a unique opportunity for productive interaction across the
four institutions.
C.2.4. Research Design and Methods
C.2.4.1.
Specific Aim 1. Structure-based analysis of dynamics of the complexes
implicated in DNA lesion recognition and signaling. On the experimental
side, the structures of several DNA repair proteins are now available (209). The PDB (210) presently contains ~170 structures related to DNA repair,
which can provide information on molecular interaction mechanisms, structural
changes upon substrate binding, and other allosteric effects, if systematically
analyzed. Whereas little information is available on NER or HR sensors for mammalian DNA, there are considerable
structural data accumulated for simpler organisms. For example, RecA of E coli functions as a sensor for SOS
response, avoiding the conflict for binding space between sensor and repair
proteins. S. cerevisiae kinase
substrates Rad1, Rad9, Hus1 form complexes with damaged DNA similar to those of
PCNA loaded onto primed DNA (211), acting as sensors that selectively bind Rad53 – Chk2 (212). These will be exploited for inferring interaction
mechanisms for their human homologs (BRCA1 and 53BP1). The affinity (binding
constant or free energies) of different sensor molecules for damaged DNA will
be measured with Biacore (SPR) (Cascio),
isothermal calorimetry or electrophoretic mobility, which will be compared with
the free energies from simulations. Site-directed mutations at residues
indicated by simulations to be critical for DNA damage recognition and binding
will be performed to test/refine the predictions/models.
Our
computational strategy is first to
characterize the dynamics of DNA
binding and collective interactions for repair/transcription proteins using
GNM-based low-resolution methods developed by Bahar and coworkers (see §
A.3.1). Of particular interest are
complexes of the order of 103 residues involved in NER and HR, such
as RNAP II (94), or TFIIH, where the GNM methodology is particularly
useful. Selected molecules or
substructures identified by this low resolution methodology to play a key
mechanical role on a global scale will be further investigated using higher
resolution (atomic level) simulations by Meirovitch, Rosenberg, and Madura. The
computational issues to be investigated reduce to the question: What is the minimum level of detail
(granularity) necessary to address the relevant biological questions? The
latter include: Are there common
patterns in the collective motions of these complexes and assemblies? Do they change in response to common forms of
DNA damage? How are they affected by
phosphorylation or proteolytic nicking? Are they sufficiently unique that they
could serve as signals? What triggers
the relevant conformational changes? Do
dynamic instabilities contribute to these changes? Can these induce an
amplification mechanism whereby small alterations (damage-related) in the DNA
lead to large-scale effects?
The results will be
accumulated in a newly created database of protein-DNA interaction mechanisms,
including a visual interface and interactive software. We note that while the
computations proposed at this developmental stage apply to currently known
structures, there is a clear future need for additional structural information e.g.
on the XPC-hHR23B complex, UV-DDB protein and the MUS308
helicase. Accordingly, one experimental
goal of DP2 for future study will be to obtain crystals of these proteins and
complexes. Future studies (as a
Center) will also include the computational characterization of these sensors
because they appear to have a high affinity for certain types of DNA lesions, -
using multiple sequence homology and sequence-match-structure protocols of threading,
as well as mechanistic considerations (DNA distortion, bending and unwinding)
of structure and dynamics. The low frequency modes determined at a given scale
will provide the basis for eliminating the uninteresting degrees of freedom and
for selecting the optimal mechanical modules while proceeding to larger
assemblies at a higher level. Results at a given scale will thus be used as
input in a hierarchical scheme for defining the most plausible coarse-grained
models at increasingly larger, but lower resolution, scales.
These studies should provide us with
information on (i) the molecular basis for the recognition of particular DNA
lesions, or for the sensory role of multiprotein complexes, and (ii) the
cooperative machinery, or allosteric effects, prompting the transmission of
signals from the damaged site.
C.2.4.2. Specific Aim 2. Simulation of the spatio-temporal
localization of the multiprotein complexes acting as sensors/initiators of
damage/repair
Experimentally, we will capitalize on ongoing studies
based on irradiating cells with UV light or X-rays, and measuring the DNA
damage repair at different locations on different chromosomes, or different
places along a chromosome. The existence
of assembled complexes or factories, as opposed to the sequential assembly of
NER proteins, will be investigated by tagging the required proteins with GFP
and measuring diffusion coefficients with fluorescence recovery after
photobleaching (FRAP) technique (187).
Binding preferences will be assessed by electrophoretic shift-mobility
measurements, and the association and dissociation kinetics of repair proteins
will be probed by measuring the time evolution of the fraction of bound forms
as described in our earlier work (139).
Computational. The questions of sequential assembly vs
pre-existing complexes, and random diffusion vs systematic screening are
ideally suited for the MC simulations described in § A.3.2, and the
mathematical approaches described in § A.3.3, to be conducted in coordination
as described in § A.4.4; here too the basic question
is: What is the minimal level of detail
needed to address the relevant biological questions? The “minimal set of proteins’ required for
repair of most lesions which we defined in our previous work (213) will be used as the components of our
system. This consists of 18 polypeptides, including the repair complexes
XPC-HR23B, followed by TFIIH, RPA, XPA, XPG and ERCC1-XPF. The concentrations,
diffusion coefficients and binding constants provided by the experiments
carried out at the UPCI and MGB labs will be used, and the unknown parameters
will be varied over suitable ranges. The model proposed by Volker et al. (214) (see fig 7 therein) schematically depicts
the type of process and ingredients that will be explored in our
microphysiological simulations of DNA lesion recognition. Here, the goal is to
develop the computational tools to the point that, in conjunction with the
ongoing experimental activities, they will facilitate the framing and testing
of specific hypotheses regarding the spatio-temporal organization and dynamics
of DNA-repair factories.
These studies are expected to bring a new
perspective to two controversial issues regarding NER and HR activation: (1)
the existence of pre-assembled repairosomes vs. the sequential assembly of
proteins following damage, and (2) the random vs systematic, or processive,
action of repair factories on the chromosomes, particularly regarding the
mechanism of the search for rare damaged sites.
C.2.4.3. Specific Aim 3: Mathematical modeling of
ATM-activated DNA damage signaling pathways. Figure
C.2.1 depicts a rough sketch of the network of interactions involved in
ATM-dependent signaling process, formulated on the basis of recent observations
(181;190-193;197;201;205;215-220). We propose to explore the kinetics of
this network, and its variations, which will be include alterations in network
connectivity (adding/deleting control motifs), the chemical modifications of
the displayed molecules, and the couplings to other subsystems (the Wee1 kinase
subsystem, for example) or external bath/source. We plan to genetically perturb
the ATM pathways by using mutant cells lacking different components (mouse
fibroblasts from knockout animals), or to put a controlled number of specific
breaks at specific sites on chromosomes to start the damage signaling response (206). It is conceivable to measure all of the
RNA and protein levels, and phosphorylation levels of proteins in this pathway,
as well as the eventual changes driving towards cell death or cell survival.
These will be compared with the predictions from our mathematical model, and we
will refine the model and parameters iteratively to obtain a realistic
description of the dynamics of ATM-activated pathways in response to radiation
induced damage.
On the experimental
side, we will focus on two particular aspects of the ATM-activated pathways:
(i) Determine the role of ATM-activated c-Abl in
triggering cell cycle arrest and/or cell death in response to IR signal.
Studies from Rajasekaran’s lab have established that c-Abl is a direct
downstream target of ATM in IR-induced signaling (190;192). To determine the role of c-Abl in triggering cell death
response, we propose to create specific point mutations in the ATM-binding and
ATM-phosphorylation sites in Abl.
Specifically, the S465A mutant of Abl will be used to test if the
overexpression of the ATM-refractile mutant of Abl blocks the apoptotic
response.
(ii) Establish
the functional significance of tyrosine phosphorylation of RNAP II in DNA
damage-induced responses. DNA damage signals such as IR cause tyrosine
phosphorylation of RNAP II in an ATM and Abl-dependent manner(221). To determine the significance of RNAP II modification in
DNA damage response, we will examine by flow cytometry (FACS) the induction of
apoptosis in cell lines deficient and proficient in RNAP II phosphorylation.
Additionally, lysates prepared from these cells will be analyzed for activation
of caspases, DNA fragmentation and examination for changes in cellular
morphology. To determine if tyrosine phosphorylation of RNAP II mediates DNA
repair, untreated and treated cells will be plated with limiting dilutions, and
percentage survival will be scored by the number of colonies formed. To
determine if c-Abl phosphorylation of RNAP II is involved in the upregulation
of DNA repair genes, the IR-induced differences in the concentrations of
proteins implicated in DNA damage recognition and signaling will be examined.
Computationally, our goal within the scope of the preliminary studies in
the pre-NPEBC stage is to interpret the results from the above experiments, to
predict at least probabilistic endpoints (cell survival or death) in response
to perturbations, and the changes in the fractions of proteins in different
forms, and suggest further experiments. The starting methodology for
constructing and solving our system will be essentially similar to the original
approach undertaken by Aguda, the most (if not only) comprehensive mathematical
model published to date on the kinetics of G2/M DNA damage
checkpoint dynamics (222). The system analyzed by Aguda is broken into four
subsystems: signal transduction subsystem composed of Chk1, ATM (or Rad3) and
p53, maturation-promotion factor subsystem (Cdc2/cyclinB complex and p21),
Cdc25 subsystem (Cdc25 and 14-3-3), and Wee1 subsystem (Wee1), all of which,
will be subject to a modular analysis using system dynamics tools and MC
simulations, as described in § A.4.4 and
already initiated in the lab of Ta’asan (Math, CMU). Most of these elements
reside in two or (active/inactive, complexed/uncomplexed
phosphorylated/dephosphorylated, cytoplasmic/ nuclear, etc.) and will be
conveniently considered as distinct compounds.
In the absence of space dependence, the kinetic equations describing the
time evolution of all concentrations can be solved using XPPAUT developed by
Ermentrout (http://www.math.pitt.edu/~bard/xpp/xpp.html)
(7). We will explore the change in
the dynamics of the network imparted by the coupling of the two (G1/S
and G2/M) checkpoint systems as well as by the contribution of the
additional components, such as c-Abl, BRCA1, also exploring the effect of
network connectivity which is an important determinant of outcome. A major
difficulty in this and all other mathematical models of this complexity is that
no sufficient data are usually available on the effective rate parameters (rate
constants, equilibrium constants, etc.). Furthermore, compartmentaliza-tion,
and translocation of the components can play a significant role in determining
the cell signaling dynamics, which will be indeed investigated within the scope
of DP3 (see § C.3). So, within the scope of DP2 the mechanistic,
or systems-structural aspects of the model (i.e. the definition of the
components and their connectivities) will probably be conceived or postulated
to a good approximation. The construction/refinement of this block diagram
should be easier with existing information (see for example refs 199 and 252) and increasing proteomics data, while the parametric or
quantitative aspects may remain unknown to a large extent. Yet, it is worth
analyzing the dynamics of postulated networks over reasonable variations in the
parameters, because it is still possible to extract useful information from the
systems’ structure alone, without (or with very limited amount of) quantitative
information on parameters, as recently stressed by Bailey (223).
