Altering Nitrogen Heterocycles of AZD2461 Affords High Affinity Poly(ADP-ribose) Polymerase-1 Inhibitors with Decreased P-Glycoprotein Interactions

Abstract

Poly(ADP-ribose) polymerase inhibitors (PARPi) are targeted therapeutics with enhanced selectivity and cytotoxicity in BRCA1/2 mutant cancer cells. AZD2461, a congener of FDA approved olaparib, is a potent PARPi with high affinity for PARP-1 and nonsubstrate for P-glycoprotein (P-gp), an attractive characteristic for cancer therapeutics. Analogues of AZD2461 were synthesized and profiled in BRCA1 functional and nonfunctional cell lines, revealing compounds (2, 3, and 5) of low cytotoxicity and excellent PARP-1 affinities (∼4–8 nM). In comparison to AZD2461, these agents were found to be less stimulating of P-gp, suggesting that these compounds may be excellent candidates for neurological applications where blood brain barrier penetrance is sought.

Introduction

Poly(ADP-ribose) polymerase-1 (PARP-1) is an essential nuclear protein that orchestrates the DNA damage response pathway, making it an attractive target for cancer therapy, although therapeutic indications in other disease states are on the horizon.¹⁻³ PARP-1-mediated cell death, parthanatos, has been defined as a hallmark signature of neuronal cell death in Parkinson’s disease.⁴ Additionally, PARP-1 has been shown to propagate the inflammatory cycle when hyperactivated by reactive oxygen species-induced DNA damage by rapidly catalyzing PAR which promotes NF-κB signal transduction.⁴ For these reasons, pharmacological inhibition of PARP-1 may serve as a therapeutic strategy for slowing disease progression of inflammatory related illnesses.

While there is a strong rationale for the development of PARP inhibitors (PARPi) as anti-inflammatory agents for neurodegenerative disorders, there lacks reports of drugs within this class that are central nervous system (CNS) penetrant and noncytotoxic. Currently, three PARP inhibitors, olaparib, rucaparib, and niraparib (Figure 1), are FDA approved to treat patients with ovarian or breast cancer expressing BRCA mutations or as maintenance therapy in platinum sensitive ovarian cancer patients.^1,5,6 However, these drugs are not CNS penetrant and are also cytotoxic because of the respective intrinsic anticancer mechanisms of each compound.³ Recently, we demonstrated how the cytotoxic properties of olaparib can be greatly reduced when replacing piperazine with a 2,6-diazaspiro[3.3]heptane core;⁷ however, the CNS uptake of this analogue is still under investigation. Therefore, it is important to understand why PARPis are not CNS penetrant and whether PARPis can be developed with reduced DNA damaging properties.

Figure 1.

Chemical structures of known PARPi.

In 2008, Jonkers and co-workers reported tumor-bearing mice to be nonresponsive to long-term treatment with olaparib because of upregulation of Abcb1a and Abcb1b, genes responsible for encoding P-glycoprotein (P-gp) drug efflux pumps.⁸ This acquired drug resistance was reversed through administration of tariquidar, a P-gp inhibitor, illustrating a potential strategy to combat P-gp-related resistant mechanisms observed with anticancer agents. AstraZeneca then developed AZD2461 (Figure 1), a structurally similar analogue of olaparib with lower enzyme–substrate affinity for P-gp.³ Oplustil O’Connor and co-workers identified AZD2461 to be less sensitive to drug resistance mechanisms than olaparib, as AZD2461 was more tolerable when combined with chemotherapeutics in mice, suggesting that this compound may be a promising anticancer agent in future clinical applications.⁹ Akin to veliparib and BGB-290,¹⁰ AZD2461 is considered a poor substrate for P-gp, a desirable characteristic for CNS penetrating drugs, and can also be evaluated in various neurological applications associated with PARP-1 hyperactivation such as neuroinflamation,¹¹ neurodegeneration,¹² neuroimaging,^13,14 and even drug addiction.¹⁵⁻¹⁷

Because of the unique pharmacological profile of AZD2461, we set out to investigate if incorporating other nitrogen-containing and cycloalkyl ring systems with a methoxy functional group into the phthalazine architecture would result in a PARPi with increased PARP-1 affinity, reduced cytotoxicity, or decreased P-gp activity. Here, we report the synthesis, PARP-1 binding profiles, cell kill properties in using BRCA1-functional, and nonfunctional cell lines, as well as P-gp interactions of AZD2461 analogues 2–10. Select compounds appear to behave less as P-gp substrates than AZD2461, affording potential therapeutics for neurological applications related to PARP-1 overexpression.

Results and Discussion

Compounds 2–10 were readily synthesized through amide coupling with commercially available precursor 1 and the respective amines, outlined in Scheme 1. It should be noted that compound 1 can also be accessed following previously reported literature conditions.¹⁸⁻²⁰ In most cases, the illustrated reaction conditions afforded good-to-moderate yields of the desired products. The trans and cis steroisomers (7 and 8, respectively) of compound 6 were also synthesized to examine the pharmacological properties of each isomer.

Scheme 1. Reagents and Conditions: 1, Amine, HOBt Hydrate, Triethylamine, Ethylcarbodiimide Hydrochloride, Tetrahydrofuran, 60 °C, 12 h.

Following our previously reported PARP-1 radioligand binding protocol,²¹ compounds 2–10 were evaluated for enzymatic inhibition with BRCA1 methylated ovarian cancer cells (OVCAR8), outlined in Table 1. In comparison to both AZD2461 and olaparib, a slight decrease in PARP-1 affinity was observed when incorporating 3-methoxyazetidine (2), asymmetric ring systems 3-methoxypyrrolidine (3), and 4-methoxyazepane (5) into the phthalazine ligand scaffold. PARP-1 inhibition was further reduced when examining cycloalkyl systems in scaffold 6–10, affording PARP-1 IC₅₀ values between 20 and 226 nM. Compound 6, identified by ¹H NMR (see Supporting Information) as a 1:5 mixture of stereoisomers 7 (trans) and 8 (cis), was the most promising PARPi in this cycloalkyl series, with a PARP-1 IC₅₀ of 43.8 nM. Thus, we evaluated each geometrical isomer of 6 and identified trans congener (7) be the more active species with a PARP-1 IC₅₀ value of 20.8 nM, compared with the 107.2 nM enzyme affinity of 8. Although the six-membered ring, 4-methoxypiperidine, in AZD2461 results in a potent PARPi (4, PARP-1 IC₅₀ = 2.8 nM), we found 4-methoxycyclohexan-1-amine to be a poor substitute, affording the least potent analogue in this study (10), with a PARP-1 IC₅₀ value of 226.3 nM.

Table 1. PARP-1 IC₅₀ Values for Olaparib and 2–10^a.

compound	PARP-1 IC50 (nM)	compound	PARP-1 IC50 (nM)
olaparib	1.7 ± 1.1	6	43.8 ± 1.0
2	8.2 ± 1.1	7	20.8 ± 1.1
3	4.2 ± 1.1	8	107.2 ± 1.1
4 (AZD2461)	2.8 ± 1.1	9	49.5 ± 1.1
5	4.7 ± 1.1	10	226.3 ± 1.1

^aPARP-1 IC₅₀ (nM) ± (SEM) data. Inhibition of PARP-1 enzymatic activity was analyzed using our previously reported radioligand-binding methodology.²¹ Dose–response curves were produced to calculate 50% maximum inhibition values (IC₅₀) where n = 3.

Computational modeling was conducted with olaparib, 4, 2, 7, and 8 inside the PARP-1 catalytic domain. These in silico studies indicated a clear trend between the binding profiles of the selected compounds, and the distance between the methoxy oxygen atom of the ligand and the ARG878 residue in the PARP-1 binding pocket (Table 2). For example, the oxygen atom in ligand 7, which exhibited a PARP-1 IC₅₀ value of 20.8 nM, had a calculated distance of 3.9 Å from ARG878, compared with the 5.4 Å value observed with the cis-isomer (8), a compound with a binding profile of 226.3 nM. The average binding energies of the compounds were also found to be consistent with the enzymatic inhibition data observed in vitro as well. This docking study provides insight into the contrasting PARP-1 binding properties obtained in vitro with 7 and stereoisomer 8, despite the similar ligand architectures of these compounds. Docking poses illustrated in Figure 2 further highlight the unique chemical space the nitrogen ring in ligand 8 occupies in the binding pocket of the enzyme, compared with the nitrogen rings of the more potent PARPi in this study which are found to be more proximal to ARG878.

Table 2. In Silico Binding Energies and Distance to ARG878 Residue of Olaparib, 4, 2, 7, and 8^a.

compound	average binding energy (kcal/mol) ± SD	distance to ARG878 (Å)
olaparib	–12.5 ± 0.2	2.9
4 (AZD2461)	–11.5 ± 0.8	3.1
2	–11.0 ± 0.7	3.4
7	–10.8 ± 0.7	3.9
8	–10.5 ± 0.8	5.4

^aAverage ligand binding energies and distance to ARG878 residue in the PARP-1 complex of selected compounds were calculated using PyMol modeling software.

Figure 2.

Distance between the oxygen atom of methoxy substituent of the ligand, and the ARG878 residue within the PARP-1 complex: olaparib, 2.9 Å; AZD2461, 3.1 Å; 2, 3.4 Å; 7, 3.9 Å; 8, 5.4 Å. Color label: olaparib (red); AZD2461 (cyan); 2 (marine); 7 (orange); 8 (green). (PDB ID 4ZZZ).

Anti-proliferative properties of olaparib, and compounds 2–10 were then analyzed in BRCA1-null/restored UWB1.289 isogenic cell lines (Table 3). In the BRCA1-null model, a highly sensitized cell line for PARPi, we found 2, 3, and 5 to be less cytotoxic than both AZD2461 and olaparib, despite the comparable PARP-1 IC₅₀ values of each agent. Surprisingly, all compounds reported showed decreased cytotoxicity to olaparib and AZD2461. This is consistent with the PARP-1 trapping theory where PARPis have differential ability to trap PARP-1 irregardless to high catalytic inhibitory properties.²²⁻²⁴

Table 3. EC₅₀ Values for Olaparib and 2–10 in UWB1.289 Cell Lines^a.

compound	UWB1.289 EC50 (μM)	UWB1.289 + BRCA1 EC50 (μM)
olaparib	0.90 ± 0.5	1.97 ± 0.4
2	2.12 ± 0.2	13.37 ± 0.7
3	1.50 ± 0.4	8.58 ± 1.1
4 (AZD2461)	0.89 ± 0.7	2.28 ± 0.3
5	2.84 ± 1.4	3.49 ± 0.2
6	10.12 ± 1.5	6.62 ± 4.3
7	1.94 ± 1.1	2.81 ± 1.4
8	4.35 ± 1.7	13.92 ± 1.1
9	1.32 ± 0.5	9.31 ± 1.0
10	2.42 ± 0.9	6.16 ± 1.3

^aEC₅₀ (μM) ± (SEM) data obtained for listed compounds in BRCA1-null and -restored cell lines.

After restoring the BRCA1 deficiency in the UWB1.289 cell line, all agents, including olaparib and AZD2461, were found to be much less cytotoxic when compared with the antiproliferative properties of these compounds in the in BRCA1-null model. The general increase in EC₅₀ values observed for these compounds, including olaparib and AZD2461, with UWB1.289 BRCA1-restored cells, highlights the synthetic lethality of PARPi in repair-deficient BRCA1 and BRCA2 mutated cancers.²⁵⁻²⁷

Overexpression of multidrug resistance pumps, such as the drug transporter P-gp, has been shown to be a possible PARPi resistance mechanism in cancer cells.³ This also prevents PARPis from being used as anti-inflammatory agents in neurodegenerative disorders. Because AZD2461 is described as a non-P-gp substrate, we evaluated each compound for P-gp activity to assess if decreasing or increasing the ring size in the AZD2461 ligand architecture could further alleviate interactions with the plasma membrane protein. This study was conducted by incubating each compound with recombinant human P-gp, an adenosine triphosphate (ATP)-dependent drug efflux pump. Compounds acting as substrates for P-gp, such as positive control Verapamil, result in lower signal intensity because of an increase of ATP consumption by P-gp ATPase (Figure 3). In contrast, inhibitors of P-gp, such as Na₃VO₄ (sodium orthovanadate), afford an increase in luminescence because of the light-generating reaction from the unmetabolized ATP and firefly luciferase. Our data suggest that olaparib acts more of a P-gp ATPase activator, than 4, coinciding with previous reports illustrating AZD2461 to be a poorer substrate of P-gp, when compared with the FDA-approved congener.

Figure 3.

Luminescence generated with olaparib, 2–10, Na₃VO₄, and Verapamil. Verapamil, a substrate for P-gp that stimulates P-gp ATPase activity resulting in decreased luminescence, was utilized as a positive control. Increase in luminescence results from the light-generating reaction from luciferase and unmetabolized ATP. Unconsumed ATP indicates a decrease in P-gp ATPase stimulation, rendering the compound as a P-gp inhibitor. Data are normalized to known P-gp inhibitor Na₃VO₄.

An increase in luminescence was observed with AZD2461 analogues containing nitrogen ring systems (2, 3, and 5), suggesting that these compounds can be suitable candidates for neuroinflammation investigative studies, where blood–brain-barrier (BBB) penetrance is sought. Compounds 3 and 5 were the weakest activators of P-gp ATPase, suggesting that these poor cytotoxic PARPi can be useful for neurological applications, such as neuroinflammation, where PARP-1 inhibition and brain penetrance are more desirable than cell kill. Compound 6 was identified as the only P-gp substrate in this study, resulting in luminescence that was comparable to known P-gp substrate Verapamil. Similar to 3, compound 9 also appeared to be a poor activator of P-pg ATPase, suggesting that P-gp interactions may be less-favorable with five-membered ring systems in the AZD2461 ligand architecture.

Conclusions

In summary, we identified three AZD2461 analogues which exhibited impressive PARP-1 IC₅₀ values of less than 10 nM (2, 3, and 5), which however exhibited poor antiproliferative effects in BRCA1-restored UWB1.289 cell lines, in contrast to olaparib and AZD2461. Nitrogen-containing ring systems in 2–5 afforded PARPi with increased PARP-1 affinity, compared to the decreased enzymatic inhibition observed with ligands containing cycloalkyl rings systems (6–10). In silico studies revealed a clear correlation among the binding profiles of the ligands and distance calculated between the methoxy oxygen atom in the ligand and the ARG878 residue in the PARP-1 binding pocket. These data provide reasoning to the contrasting PARP-1 IC₅₀ values obtained with 7 and geometrical isomer 8. Compounds containing five-membered methoxy rings (3 and 9) were found to be the least activating agents of P-gp ATPase in this study, suggesting that these PARPis may be less prone to drug efflux in cells. Furthermore, these compounds may also be interesting candidates for therapeutic applications in neuroinflammation or neurodegeneration where cytotoxicity is not desired and minimal P-gp interactions are needed to cross the BBB.

Glossary

Abbreviations

ATP

adenosine triphosphate

BBB

blood–brain-barrier

EDC·HCl

ethylcarbodiimide hydrochloride

HOBt

1-hydroxybenzotriazole

MDR

multidrug resistance

P-gp

permeability glycoprotein

TEA

triethylamine

THF

tetrahydrofuran

Supporting Information Available

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsomega.8b00896.

• Protocol for PARP-1 radioligand binding in live OVCAR8 cells, in silico studies, cell viability assays, and Pgp-Glo assay system (PDF)

• ¹H and ¹³C NMR spectrum, high-resolution mass spectrometry data, and full characterization of isolated compounds 2–10 (CSV)

Author Contributions

S.W.R. and L.N.P. contributed equally. The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

S.W.R. and L.N.P. conducted this research through the support of training grants 5T32DA028874-07 and T32GM008076, respectively.

The authors declare no competing financial interest.