Dynamic structures in DNA damage responses & cancer

A singular triumph of structural biology came from the application of diffraction physics to the elucidation of the DNA double helix: this atomic structural model based upon low resolution fiber diffraction solved the mystery of genetic replication essential to life by revealing DNA base pairing for accurate copying (Watson and Crick, 1953). A singular missed implication of the double helix was its critical value for the maintenance of genome integrity by allowing accurate DNA damage repair (Crick, 1974). A third element required for eukaryotic life, which has substantial genetic information, is the long-range coordinated regulation of replication, transcription, and repair events to avoid destructive interference and consequent loss of critical genetic information (Hopfner et al., 2000).

Within recent years, advances in biophysical analyses and their applications are addressing issues and uncovering challenges in DNA replication, repair, and pathway coordination by providing mechanistic knowledge connecting nucleic acid–protein complexes to cellular pathways for genome integrity. As a reduction in replication and repair fidelity is involved in cancer etiology, cancer prognosis, and a target for cancer therapy (Helleday et al., 2008), this biophysical data is timely and increasingly proving relevant to cancer biology and therapeutics. To provide insights into these advances, this volume of PBMB presents nine selections by internationally leading investigators outlining recent progress and unsolved problems concerning the structural and functional basis of genome integrity. Biophysical approaches are increasingly appreciated as essential to uncovering the mechanisms of damage recognition and removal, and coordination of events in the context of replication, transcription and epigenetic modifications. Although coordination likely occurs in most biological pathways, the great selective pressure to limit toxic and mutagenic loss of genetic integrity means that DNA damage responses are especially informative for biological studies of exquisite structural regulation of pathway handoffs and coordination.

Papers in this issue grew out of the International Workshop on Dynamic Structures in DNA Damage Responses & Cancer in 2014. Insights resulting from the interactive exchange of results and ideas at this conference are presented in a set of exemplary papers for this themed volume of Progress in Biophysics and Molecular Biology. This is an exciting and enabling period for advances in X-ray structural analyses. The free electron laser (Cohen et al., 2014) is enabling tiny crystals and time resolution in crystal structures. Recent analysis of the R-factor gap between crystallographic models and data has discovered untapped information on flexible regions and solvent in crystallographic data sets, suggesting an opportunity to substantially improve all existing macromolecular structures by providing more accurate models (Holton et al., 2014). Advances in X-ray scattering, including the development of a novel scattering invariant (Rambo and Tainer, 2013) and an objective measure of flexibility (Rambo and Tainer, 2011), promise to increase the accuracy of high throughput methods for analysis of macromolecular conformations and assemblies in solution. The many conformational states acting in genome integrity suggest different conformations can give different outcomes, as shown for Mre11 and Rad50 (Deshpande et al., 2014), implying that top down methods often cannot predict outcomes from gene networks alone and that bottom-up structural mechanisms can be critical to predictive understanding of DNA damage responses.

Crystallography provides precise structures but limited information about flexibility (Putnam et al., 2007), although temperature factors can identify functionally flexible regions when corrected for crystal contacts (Tainer et al., 1984). Unveiling the roles of flexibility in DNA damage responses is an ongoing process that would benefit from continued methods development. Flexibility for interaction and functional access to DNA ends makes sense geometrically, and is seen in systems such as the DNA ligase-PCNA interaction, where a linear ligase wraps around the DNA ends like a molecular watch band (Pascal et al., 2006). Yet, intra-molecular interactions of flexible regions can contribute to the stability of the folded region and can act in intermolecular interaction, making interpretations of structures and their functions challenging by single methods (Hegde et al., 2013). Thus, multiple biophysical techniques combined with molecular biology may be crucial for a predictive understanding of functional flexibility in macromolecular machines.

In this context there is also increasing evidence for the great value of specific inhibitors, which can be optimized from X-ray structures (Garcin et al., 2008) and developed from structural analyses of fragment libraries (Winter et al., 2012). Notably, only 1 out of 20 drugs for oncology makes it through clinical trials, and these typically cost a stunning 2 billion dollars (see http://csdd.tufts.edu/news/complete_story/pr_tufts_csdd_2014_cost_study (Fig. 1)). Therefore the cost of an average novel cancer drug including failures may exceed ~10–20 billion dollars (Kola and Landis, 2004). This current cost to success ratio for oncology is thus likely unsustainable for pharmaceutical companies, so a mechanistic understanding linking structures to pathways and outcomes, which may increase the percentage of drugs validated in clinical trials, will have extremely high impact for science and public health.

The genome instability associated with most cancers is an exploitable cancer target (Pearl et al., 2015). However, we need to understand mechanisms of genome stability and instability more completely before we can effectively exploit instability to improve therapeutic strategies and better avoid resistance and toxicity. The combination of NMR, X-ray scattering, and insightful analysis is revealing the nature of dynamic DNA processing machines that control genome stability and are often built from multiple, structurally-independent functional domains, as seen for RPA (Mer et al., 2000). An important conceptual insight from this work is that DNA repair complexes can coordinate activities by forming hubs or keystone complexes that contact DNA and multiple protein partners via modular interactions (Sugitani and Chazin, 2015). The combination of multiple moderate (uM) binding affinities provides sufficient affinity for specificity while allowing interface exchange and the handoff of DNA intermediates. DNA repair is essential to cell biology, so active site chemistry and core domains can be conserved among the three Domains of life. Yet, the greater selective pressure to coordinate repair steps and pathways to avoid loss of genetic information in eukaryotes with their larger chromosomes is evident in added regulatory interactions, conformations, and post-translational modifications in human DNA repair structures compared to those from bacteria and archaea, as seen in the flexible terminal extension of repair nuclease APE1 (Hopfner et al., 2000). Transitions of the repair machinery employ interface competition and multi-domain allostery. Technical advances in the production and analysis of dynamic functional assemblies are enabling the understanding of both their activities and coordination with other DNA transactions. These results highlighting hubs suggest new strategies and different classes of cancer drug targets than active sites. Notably, disrupting coordinating interactions and conformation can cause apoptosis in cancer cells. Normal cells are typically distinguished from cancer cells by retaining multiple layers of regulation that allow them to halt replication and transcription and to employ alternative repair pathways that may also be absent or defective in transformed cells.

The most common form of DNA damage is base damage including deaminations, alkylations, and oxidations where the chemical structure of the DNA is altered, so DNA base repair is central to all life, and structures are uncovering elegant mechanisms for this process (Hitomi et al., 2007). Deamination of the exocyclic amines in adenine, guanine and cytosine forms base lesions that can cause mutations due to non-canonical base pairs and misincorporation during replication. Both repair of deaminated DNA and processing of functionally-important deamination of RNA in humans is an important function for Endonuclease V (EndoV), as elucidated by Dalhus et al. (2015), who consider the structural basis for EndoV incision at deaminated adenines in DNA and RNA. In contrast to DNA glycosylases that remove a damaged base by excision of the nucleobase, EndoV incises DNA at the second phosphodiester 3′ to base lesions without removing the deaminated base. Structural investigation of this novel incision by EndoV shows that a wedge motif promotes DNA strand-separation and separate pockets recognize the flipped out base lesion and phosphate backbone. Human EndoV, but not the bacterial homolog, incises RNA substrates containing inosine, the deamination product of adenosine, which is a frequent RNA modification. Structural analyses of EndoV are increasing our knowledge of EndoV activity, thus providing comparative links between DNA repair and RNA processing and promising to address unanswered questions regarding RNA activities in the human enzyme.

Completion of DNA replication and repair pathways requires ligation of the DNA ends (Tomkinson et al., 2013). Although DNA ligase has conformational flexibility to wrap around the DNA (Pascal et al., 2006), the ligation reaction is completely blocked by modified DNA ends, such as the products and repair intermediates of DNA oxidation, alkylation, or the aberrant incorporation of ribonucleotides into genomic DNA. Aborted DNA ligation reactions create 5′-adenylated DNA and RNA-DNA damage that is reversed by aprataxin (Aptx), a member of the histidine triad (HIT) superfamily. Aptx acts in single-stranded DNA repair through its nucleotide-binding activity and its diadenosine polyphosphate hydrolase activity (Schellenberg et al., 2015). X-ray crystal structures of Schizosaccharomyces pombe Aptx (SpAptx) and human Aptx–substrate complexes provide insights into Aptx recognition of RNA-DNA, AMP lesions, and Zn cofactors (Schellenberg et al., 2015). These structural analyses combined with molecular biology data provide a critical molecular framework for a mechanistic knowledge of the Aptx deadenylation reaction and for the linkage of human APTX mutations to the neurological disorder Ataxia with Oculomotor Apraxia 1 (AOA1). A detailed understanding of the processing and control of DNA breaks to produce ends that can be rejoined by DNA ligase to reestablish the unbroken DNA double helix is critical to understanding DNA damage responses and regulated repair synthesis.

Acting in replication fork rescue and homologous recombination repair of DNA breaks, the human breast and ovarian cancer type 1 susceptibility (BRCA1) protein of 1863 residues has conserved RING domain and tandem BRCT domains at its termini. These domains recognize post-translational modifications and act in BRCA1 assembly into a stable heterodimer (Wu et al., 2015). The BRCA1 RING domain has E3 ubiquitin ligase activity. The BRCT domains bind phosphorylated proteins during the DNA damage response to coordinate repair and signaling. Importantly, informed structural analyses are revealing protein partners, activities, and mechanisms whereby mutational defects cause human disease. In particular, the sites at which BRCA1-interacting proteins bind have been identified. Furthermore, clinical mutations have been mapped, which target the RING domains, disrupt the BRCA1 BRCT domain interaction with phosphorylated proteins, and block the accumulation of BRCA1 at damage-induced foci. Another mutation abrogates the BRCA1 E3 ubiquitin ligase activity without damaging the RING structure. In addition, cancer patient mutations at conserved residues on the BRCA1 coiled-coil domain abrogate its interaction with partner and localizer of BRCA2 (PALB2; also known as FANCN) and compromise homologous recombination-mediated DNA repair. This analysis of interactions and mutations advances our understanding by examining the many structures of complexes for BRCA1-tandem-BRCT domains that include BACH1, BRCA1-interacting protein carboxy-terminal helicase 1, and CtIP among others (Wu et al., 2015).

DNA-PKcs, a PI3/PI4-family member, was discovered as multi-functional protein kinase (Lees-Miller and Anderson, 1989) that we now know functions for both DNA double-strand break repair and mitosis (Jette and Lees-Miller, 2015). It forms the catalytic subunit of the DNA-dependent protein kinase (DNA-PK) that functions with Ku70/Ku80 heterodimer in DNA double strand break repair and Non-Homologous End Joining (NHEJ). Breakthrough crystal structures of a large fragment of DNA-PK (Sibanda et al., 2010) and of the Ku-DNA complex (Walker et al., 2001) have opened the door to X-ray scattering analyses and provided the framework for DNA-PK structural and mutational analyses in mitosis (Jette and Lees-Miller, 2015) as well as repair (Williams et al., 2014). Furthermore, structures of the machinery that holds the DNA ends and channels the pathway to ligase reveal how DNA-PK may regulate the architecture acting in NHEJ coordination. DNA-PKcs is autophosphorylated on multiple sites in mitosis and dephosphorylated by protein phosphatase 6. DNA-PKcs is required for correct alignment of mitotic chromosomes on the metaphase plate for accurate mitosis. Emerging results suggest that the mechanism of activation of DNAPKcs in mitosis may be Ku-independent and distinct from that in NHEJ. It will be exciting to decipher the different molecular mechanisms for DNA-PKcs activities in mitosis and in NHEJ to better understand the functions for this DNA-dependent kinase that is far more highly expressed in humans than in other mammals.

XPB helicase is an essential enzyme component of the eukaryotic transcription factor complex TFIIH, and has dual roles in transcription and DNA repair (Fuss and Tainer, 2011). Structure-based analyses of XPB (Fan and Du Prez, 2015) reveal functional motifs and architectures for its activities in opening the DNA promoter to initiate RNA polymerase II transcription and in further opening dsDNA flanking a DNA lesion to initiate nucleotide excision repair (NER). Among the questions under current investigation involve the ability of NER to remove diverse, structurally-unrelated DNA helix-distorting lesions. The key role played by XPB is evident from the severe clinical consequences of inherited defects in XPB including cancer prone Xeroderma Pigmentosum (XP), defective development in Cockayne Syndrome (CS), their combination in XP and CS (XP/CS), and rapid aging in Trichothiodystrophy (TTD). The relationships of XPB defects and particular diseases are open questions under investigation. In the companion TFIIH helicase XPD, the structural implications of patient mutations are that XP results from defects in DNA and ATP binding that cause DNA repair defects, CS results from mutations causing defects in functionally important conformational dynamics, and TTD involves framework defects that weaken XPD assembly interactions with TFIIH (Fan et al., 2008; Kuper et al., 2012; Liu et al., 2008). A major ongoing question is how the XPB structures and defects match or differ from the mechanistic disease models proposed from XPD.

Mutational defects in the human Bloom syndrome helicase BLM cause tumors at early age in diverse tissues (Wu and Hickson, 2003). BLM is one of five human homologs of RecQ helicase from Escherichia coli. Whereas the Werner’s RecQ helicase includes a nuclease component activated by coiled-coil assembly (Perry et al., 2010), BLM pairs with the EXOI nuclease whose DNA complex has been characterized as a member of the FEN1 superfamily (Orans et al., 2011; Tsutakawa et al., 2011). The shared biochemistries of BLM and RecG suggest convergent evolution of cellular function: human BLM fulfills the genomic stabilization role of RecG (Bianco, 2015). Notably, expression of RecG in human BLM-deficient cells suppressed both elevated sister chromatid exchange and the gene cluster instability phenotypes of BLM-deficient cells. Yet, RecG expression has no impact on these phenotypes in human cells with functional BLM. The structures of BLM, its disease-causing mutations, and its interactions with EXO1 as well as their functional comparisons to RecG will be powerful next steps in defining BLM functions in genome integrity.

Although DNA replication is accomplished with high fidelity, as implied by the base-pairing feature of the DNA double helix, the mutation rate for every replication would be far too high without the post-replication correction from the DNA mismatch repair (MMR) pathway. Replication errors are minuscule for high fidelity replication due to MMR (10⁻⁹) when compared to the numbers of endogenous DNA damage alone (~70,000/cell/day). This is evident from the inhibition of MMR by cadmium that causes 2000-fold increases in mutation rates (Jin et al., 2003). High levels of DNA damage are reduced to low levels of mutations by DNA repair, so alteration of key DNA damage response pathways may prove even more important than direct DNA damage by mutagens and an understanding of repair mechanisms has immense biological and medical value (McMurray and Tainer, 2003). The coordinated actions of two ATPases (MutS and MutL) initiate the mismatch repair response and defects in the genes encoding for these proteins have been linked to sporadic and hereditary cancers. Although MMR has been studied for decades, the mechanism of strand discrimination has remained elusive in most organisms including humans. However, biophysical studies on the MutL ATPase and nuclease are revealing its roles in damaged strand discrimination and removal during mismatch repair (Guarné and Charbonnier, 2015). Biochemical, biophysical and structural analyses are showing how MutL aids in distinguishing the newly synthesized strand from its template and in marking it for removal. In general, MMR employs a surprisingly large number of conformational states as part of its activity, based upon X-ray scattering data on comprehensive conformational analyses and on the conformations of gold-nanocrystal labeled DNA bound to MutS, and to both MutS and MutL (Hura et al., 2013a, 2013b).

The most toxic and mutagenic DNA damage is DNA double-strand breaks (DSBs), which are repaired by NHEJ or homologous recombination repair (HRR). Whereas initiation of NHEJ involves the DNA-PK kinase noted above, initiation of HRR requires the heteromeric Mre11-Rad50-Nbs1 (MRN) (Lafrance-Vanasse et al., 2015). Structures of components and complexes for MRN are revealing its functions in pathway selection and coordination of events at collapsed replication forks, where Mre11 nuclease can excise stalled replication forks (Schlacher et al., 2011) that are unprotected by BRCA1 or BRCA2 (Schlacher et al., 2012). BRCA1, discussed above, interfaces with the MRN complex to protect stalled replication forks from unregulated nuclease activity. Notably, the dynamics of Rad50 ATP binding and hydrolysis can also control the Mre11 nuclease and thus DNA end processing versus end tethering (Deshpande et al., 2014), as can the phosphoprotein binding subunit Nbs1, which binds across the side of Mre11 opposite from its DNA binding face to allosterically regulate Mre11 activity upon Nbs1 binding via FHA and BCRT domains to phophoprotein partners (Williams et al., 2009). Furthermore the conformational sculpting of the DNA needed for the Mre11 endonuclease and exonuclease activities can be targeted in chemical knockdowns that reveal licensing and committed steps in HRR (Shibata et al., 2014). In the future it will be interesting to understand how the distinct Mre11 and Rad50 DNA binding sites cooperate to bind and then process DNA substrates and which conformations and interactions may make the optimal targets for possible cancer therapeutics.

What are the current challenges in DNA repair machines and where should we look for the next breakthroughs? The ongoing experimental advances in crystallography and X-ray scattering noted above promise to provide improved models incorporating flexibility and bound solvent, which will be important for overall progress. Computational advances will also be enabling. DNA is a polyanion, and an appropriate positive patch on the repair enzyme surface can guide flexible enzyme-substrate recognition while allowing flexibility of the enzyme binding site (Mol et al., 2000). Such positive patches are for example seen on both Mre11 and Rad50 (Hopfner et al., 2001) and were predictive of subsequent DNA-complex crystal structures (Rojowska et al., 2014; Williams et al., 2008). Indeed such electrostatic recognition allows regulation of DNA repair processes by DNA mimicry (Mol et al., 2000; Putnam et al., 1999) that may act in many DNA repair pathways (Putnam and Tainer, 2005). In general, electrostatic potential gradients can increase the rate of interactions (Getzoff et al., 1983) as well as properly orient two macromolecules for productive collisions (Roberts et al., 1991). Improvements in computational methods to include more accurately electrostatic forces in intermolecular interactions will likely be of great value for understanding DNA damage response pathway steps and regulation.

In biology, the rate limiting steps of enzymatic pathways are typically key points for regulation and pathway connections. Rate limiting steps in DNA repair are often conformational changes and product release rather than chemistry, as seen for the human DNA repair nuclease APE1 (Mol et al., 2000). The DNA conformational change of nucleotide or base flipping discovered in DNA base repair initiation by uracil-DNA glyocosylase (Slupphaug et al., 1996) are seen in most of the repair pathways (Huffman et al., 2005). Blocking base flipping provides a means to inhibit repair nucleases, such as Mre11. Analogously, blocking ATP-mediated conformational change, as seen for Rad50 (Deshpande et al., 2014), can inhibit specific functions of multi-functional complexes without having to target the ATP site with its risks of cross-reactivity and difficulty of competing with mM ATP concentrations in cells.

Where this has been analyzed carefully, DNA repair machines reveal a stunningly large number of conformational states. Indeed, DNA repair complexes can be considered as analog computers that respond to cell cycle states, such as replication and transcription as well as to the nature of the DNA lesion, as seen for the ATPase complexes of MutS-MutL and Mre11-Rad50 (Hura et al., 2013a,b; Williams et al., 2010). DNA sculpting, which allows DNA repair nucleases to measure twice-cut once (Tsutakawa et al., 2014), can be a target for inhibitor design as seen for the DSB repair nuclease Mre11 (Shibata et al., 2014). Chemical compounds to knock-down specific activities based upon structures are likely to be of tremendous value for understanding the effect of chemotherapeutics in different cell types, resistance mechanisms, and synthetic lethality where cancer cells may have much greater susceptibility to knocking-down a given repair step than normal cells. Evidence for the clinical importance of missing repair activities comes from the extraordinary responders to clinical trials, e.g. defects in Rad50 allowed one cancer patient to be cured despite the failure of the clinical trial for other patients (Al-Ahmadie et al., 2014). Allosteric conformations and modular binding sites are becoming attractive new targets for inhibitor design to dissect functions and leverage synthetic lethality.

From the conference discussions and papers presented here, it is evident that the three requirements for life that are the focus of this special volume of PBMB (replication, repair, and repair coordination with transcription and replication) merit substantial investment and intense investigation to link dynamic structures to functions and examine their relevance to cancer etiology and future therapies. Taken together these papers on dynamic structures in DNA damage responses and cancer will therefore be of substantial interest to researchers as well as funding agencies in considering priorities to bridge gaps in current knowledge. For example, the dynamic conformational control of pathways and coordination evident from cross-genomic comparisons coupled to advanced biophysical methods is surprisingly complex and has been under appreciated for cell biology and under utilized in therapeutic approaches. The approval of the Poly ADP-Ribose Polymerase (PARP) inhibitor Olaparib to treat ovarian cancer patients with BRCA1 or BRCA2 mutations provides the first cancer drug directed at genetic instability. This critical first step forward in targeting a defect in DNA break repair also underscores how much remains to be done in developing the detailed knowledge of structure and dynamics expected to improve specificity and to reduce therapeutic failures and resistance (Pearl et al., 2015). From the International Workshop on Dynamic Structures in DNA Damage Responses & Cancer, we know that targeting the DNA damage response is far from intractable, and there are many promising and druggable targets to be explored with advanced structural methods. Major overall goals for structural biophysics going forward will be the mechanistic dissection of the dynamic conformations coordinating pathways, and of the multiple activities underlying the functions for DNA repair machines. Achieving these goals will be aided by the advancement of specific chemical knockdowns, which are a major means for the manipulation of specific biochemical activities for intracellular targets. Such small molecule inhibitors promise both to help define the multiple roles of these molecular machines in replication, repair and coordination needed for life and to provide leads for novel advanced cancer therapeutics.