A comprehensive study of ligand binding to the human PARP family of proteins reveals the molecular basis of inhibitor selectivity and promiscuity.
Inhibitors of poly-ADP-ribose polymerases (PARPs) are under evaluation in dozens of clinical trials for a wide variety of cancers, including breast, ovarian and pancreatic tumors1. But despite the promise of PARP inhibitors in drug development, little is known about their effects on specific members of the family of human PARPs. In this issue, Wahlberg et al.2 present the first large-scale profile of PARP-inhibitor selectivity, created by screening 185 small-molecule inhibitors against the catalytic domains of 13 human PARPs. Analysis of these binding data together with nine new crystal structures of catalytic domains bound to inhibitors revealed key binding interactions that control the degree of PARP-inhibitor selectivity.
ADP-ribosyltransferases, first discovered nearly 50 years ago3, transfer ADP-ribose groups from nicotinamide adenine dinucleotide (NAD+) onto various substrates4. Initially classified as PARPs, several of these enzymes were recently found to be mono-ADP-ribose transferases5. The human genome encodes 17 PARP family proteins, each containing a conserved catalytic domain. Although the substrates and function of most individual PARPs are unknown, the enzyme family has been implicated in a plethora of cellular processes, including chromosome stability, the DNA damage response and regulation of apoptosis, cell division, transcriptional regulation and differentiation. The PARP family also includes the tankyrases TNKS1 and TNKS2, which regulate the activity of the Wnt/β-catenin signaling pathway. Dysregulation of Wnt signaling has been implicated in the growth and progression of multiple human cancers6.
PARP inhibitors have been evaluated in many clinical and preclinical settings1, most recently in oncology. Owing to the key role of PARPs in DNA repair, PARP inhibitors display synthetic lethality in cells defective in homologous recombination, such as tumors with mutations in BRCA1 and BRCA2 (ref. 7). PARP inhibitors are also undergoing clinical trials as chemo- and radio-sensitizers. Finally, as PARPs can regulate NAD+ levels, PARP inhibitors are being investigated in various models of reperfusion injury, inflammation, cardiovascular disease and neurodegenerative disorders.
Wahlberg et al.2 began by producing the catalytic domains of 13 human PARPs through expression in Escherichia coli. They then measured binding between the 13 PARPs and a library of 185 small molecules using differential scanning fluorimetry, a technique applied previously to profile kinase inhibitors. The compound library included research reagents as well as inhibitors in clinical testing, such as olaparib7, rucaparib (AG-014699)8 and veliparib (ABT-888)9. Although it would have been preferable to test the effects of the compounds on PARP enzymatic activity, this could not be done as the substrates of most human PARPs are unknown. For a few PARPs with known substrates, the authors confirmed that compounds that bound also inhibited enzymatic activity.
A heat-map representation of the binding results yielded some notable insights (Fig. 1). First, most binding events involved PARP1–4. This is not surprising as many inhibitors in the library were designed to target PARP1, and PARP1 is closely related to PARP2–4. Second, several research reagents, such as 6(5H)-phenanthridinone and TIQ-A, bound to multiple PARPs. Third, a group of compounds, including XAV939, as previously reported, showed selectivity for tankyrases with only modest activity against other PARPs. Fourth, olaparib and veliparib, two compounds in clinical trials, both inhibited PARP1-4 but olaparib had additional activity against PARP12, 15 and 16 whereas veliparib did not. Finally, rucaparib, another compound in clinical trials, showed promiscuous binding, targeting nine PARPs, including TNKS1 and TNKS2. These subtle differences in selectivity may help to explain the clinical observations with these agents.
Wahlberg et al.2 next used crystal structures of PARP-inhibitor complexes to interpret the binding data and to suggest directions for the development of new inhibitors. They crystallized TNKS2 or PARP14 with one of nine compounds and analyzed the resulting structures together with nine other PARP-ligand crystal structures that they had solved previously. In all of the 18 structures, the inhibitors bound to the NAD+ binding pocket. Like nicotinamide, they interacted with the pocket through hydrogen bonding with nearby glycine and serine residues along with π-stacking to a conserved tyrosine. The smaller inhibitors tended to interact only with conserved regions of the protein, explaining their promiscuity. In contrast, larger, more-selective inhibitors interacted with the outer edges of the NAD+ pocket and a D-loop structure, both of which show greater sequence variability.
The structural studies also explain the selectivity of PARP1–4 and TNKS1,2 inhibitors (Fig. 1). Inhibitors of these two classes occupied distinct regions of the NAD+ binding pocket. PARP1–4 inhibitors are generally larger, hydrophilic compounds that occupied a polar site at the entrance to the pocket. In PARP1–4, the binding site was enlarged by a longer D-loop lid and contained a conserved asparagine residue that often interacted with inhibitors. In contrast, TNKS1,2 inhibitors occupied a narrower, hydrophobic binding pocket created by phenylalanine and proline side chains pointing into this region and a shorter D-loop lid.
Several other interesting compounds emerged from the structural analysis. Some compounds stabilized PARPs other than PARP1–4. For example, compound 98 made a unique tyrosine-triazole stacking interaction with PARP14 and this insight provides opportunities for further compound design.
The results of Wahlberg et al.2 suggest several avenues for future research. First, it will be important to evaluate the functional consequences of inhibitor binding. This will require the identification of substrates for all PARP family members and the development of biochemical assays to measure enzyme function, particularly for enzymes with mono ADP-ribosyltransferase activity for which no such assays currently exist. Second, identifying compounds that bind outside the NAD+ binding pocket would facilitate the design of new inhibitors with diverse selectivity profiles. For instance, a recent study identified a novel series of TNKS1 and TNKS2 inhibitors that bind in the adenosine binding pocket and lack the classical nicotinamide triple-H-bond network10. Third, these assays should be extended to full-length PARPs, not just catalytic domains as in the current study. Such experiments could be used to identify allosteric modulators and to study the effects of substrate binding or protein-complex formation on enzymatic activity. Although the differential scanning fluorimetry (DSF) results are encouraging, with data from representative compounds correlating well with results obtained by surface plasmon resonance, several PARP inhibitors may have been lost owing to false negatives as a result of the inability of DSF to identify weaker binders. Other label-free biophysical techniques that are more sensitive may offer further insights.
Although kinase inhibitor profiling is routine, the work of Wahlberg et al.2 provides the first comprehensive data set and structural evaluation for PARP inhibitors. In addition to profiling the selectivity of the current generation of inhibitors, it should prove valuable as a foundation for the design of new inhibitors targeting specific PARP family members.
Jagtap, P. & Szabo, C. Nat. Rev. Drug Discov. 4, 421–440 (2005).
Wahlberg, E. et al. Nat. Biotechnol. 30, 283–288 (2012).
Chambon, P., Weill, J.D. & Mandel, P. Biochem. Biophys. Res. Commun. 11, 39–43 (1963).
Hottiger, M.O., Hassa, P.O., Lüscher, B., Schüler, H. & Koch-Nolte, F. Trends Biochem. Sci. 35, 208–219 (2010).
Kleine, H. et al. Mol. Cell 32, 57–69 (2008).
Huang, S.M. et al. Nature 461, 614–620 (2009).
Fong, P.C. et al. N. Engl. J. Med. 361, 123–134 (2009).
Thomas, H.D. et al. Mol. Cancer Ther. 6, 945–956 (2007).
Penning, T.D. et al. J. Med. Chem. 52, 514–523 (2009).
Shultz, M.D. J. Med. Chem. 55, 1127–1136 (2012).
The author declares no competing financial interests.
About this article
Identification and quantification of DNA repair protein poly(ADP ribose) polymerase 1 (PARP1) in human tissues and cultured cells by liquid chromatography/isotope-dilution tandem mass spectrometry
DNA Repair (2019)
Parent-Metabolite Pharmacokinetic Modeling and Pharmacodynamics of Veliparib (ABT-888), a PARP Inhibitor, in Patients With BRCA 1/2 -Mutated Cancer or PARP-Sensitive Tumor Types
The Journal of Clinical Pharmacology (2017)
Expert Opinion on Therapeutic Patents (2013)
Nature Biotechnology (2012)