The discovery of purine-based agents targeting triple-negative breast cancer and the αB-crystallin/VEGF protein–protein interaction

Oestrogen receptor-negative breast cancer, particularly subtypes such as triple-negative breast cancer (TNBC, around 10–15% of cases), are characterised by poor long-term survival, poor response to therapy and early progression to metastasis. Purine-based compounds represent a privileged scaffold in anticancer drug design, with several clinically approved and experimental agents in clinical development comprising a purine core structure. In this study, a series of new purine-based compounds were synthesised; seven of the new analogues were found to significantly reduce the in vitro viability of TNBC cell lines (MDA-MB-231 and MDA-MB-436) with IC50 values of ≤50 μM. In previous work, we have proposed a new concept for targeting angiogenesis driving TNBC progression, by disrupting the protein–protein interaction between the molecular chaperone αB-crystallin (CRYAB) and VEGF. Since previous clinical studies applying anti-VEGF therapy to TNBC patients have met with limited success, we were interested to test our most promising purine analogues against CRYAB/VEGF, using a custom-designed cell-based CRYAB/VEGF165 interaction assay platform. Analogues 4e and 4f significantly reduced the interaction between CRYAB/VEGF165, and compound 4e (100 μM) was also found to decrease the levels of soluble VEGF expressed by MDA-MB-231 cells by 40%. In conclusion, these promising early activity profiles warrant further investigation to validate this concept.


Pl e a s e n o t e:
C h a n g e s m a d e a s a r e s ul t of p u blis hi n g p r o c e s s e s s u c h a s c o py-e di ti n g, fo r m a t ti n g a n d p a g e n u m b e r s m a y n o t b e r efl e c t e d in t his ve r sio n.Fo r t h e d efi nitiv e ve r sio n of t hi s p u blic a tio n, pl e a s e r ef e r t o t h e p u blis h e d s o u r c e.You a r e a d vis e d t o c o n s ul t t h e p u blis h e r's v e r sio n if yo u wi s h t o cit e t hi s p a p er.
Thi s v e r sio n is b ei n g m a d e a v ail a bl e in a c c o r d a n c e wit h p u blis h e r p olici e s. S e e h t t p://o r c a .cf. a c. u k/ p olici e s. h t ml fo r u s a g e p olici e s.Co py ri g h t a n d m o r al ri g h t s fo r p u blic a tio n s m a d e a v ail a bl e in ORCA a r e r e t ai n e d by t h e c o py ri g h t h ol d e r s .

Introduction
Triple-negative breast cancer (TNBC) represents an important and unmet clinical challenge, accounting for around 10-15% of all newly diagnosed breast cancer cases.TNBC is classified as a subpopulation of invasive breast cancer characterised by the lack of expression of the oestrogen receptor (ER), progesterone receptor (PR) and human epidermal growth factor receptor 2 (HER2) (Fosu-Mensah et al. 2015;Lehmann et al. 2011;Mayer et al. 2014).As a subset of breast cancer cases, TNBC is also a "disease of the young", as it frequently affects premenopausal women of minority background, including black and Hispanic women.It is an aggressive disease associated with high histological grade, increased risk of recurrence and metastatic spread and poor prognosis.
Currently, the standard of care treatment for TNBC patients involves the use of cytotoxic chemotherapy (e.g., taxanes).However, this treatment lacks long-term efficacy.
Angiogenesis, the development of new blood vessels from existing vasculature, is essential for early stage tumourigenesis and metastasis.This process is regulated by several pro-angiogenic factors, including vascular endothelial growth factor (VEGF).TNBC is highly vascularised as a result of upregulation of intra-tumoural VEGF levels, which in turn is associated with poor prognosis.In this regard, anti-angiogenic therapies were evaluated clinically as potential targeted therapies in TNBC albeit not successfully in terms of long-term survival.Bevacizumab is an anti-angiogenic antibody that has been used in combination with chemotherapy and was first investigated as a therapeutic option due to enhanced angiogenesis in TNBC presentations (Steeg 2006;Kim et al. 2004).This therapy showed promising results in early clinical trials, however, meta-analysis of clinical trial data showed no beneficial effect on overall survival rates (O'Shaughnessy et al. 2009;von Minckwitz et al. 2012;Bear et al. 2012).Most patients were non-responsive to the therapy, while patients who showed initial response eventually developed resistance and relapsed disease.Moreover, in 2011 the FDA revoked its approval of bevacizumab for breast cancer therapy due to reported adverse side effects (Goozner 2011;Ranpura et al. 2011;O 'Reilly et al. 2015).There are currently no molecularly targeted therapies that are effective against TNBC in the clinical setting, and therefore there is a high unmet medical need for the development of new effective and efficacious therapies for this poor prognosis patient group (Fosu-Mensah et al. 2015;Chen et al. 2014).
Given the high priority afforded to the further study and treatment of this disease, TNBC has been the subject of intense gene array profiling studies.These studies identified new molecular targets including but not limited to basal epithelial proteins, such as cytokeratin 5 and 6, CK14, CK17, P-cadherin, p53 mutations, epidermal growth factor receptor (EGFR) and αB-crystallin (Lehmann et al. 2011).Ultimately, these studies will provide potential therapeutic targets for drug discovery programmes.Since many of the newly identified TNBC therapeutic targets have established therapies in other disease settings, there are presently numerous clinical studies underway to assess their efficacy in TNBC.
An interesting TNBC-associated target under investigation within our laboratory is αB-crystallin (CRYAB, HspB5).CRYAB is a member of the mammalian small heat-shock protein family that function as molecular chaperones to modulate cell proteostasis, and consequently, cell survival by suppressing aggregation of denatured proteins in stressinduced environments (Tsang et al. 2012;Arrigo et al. 2007;Garrido et al. 2012).CRYAB is ubiquitously expressed in many tissues with varying functions, and upregulation of this protein is observed in many disease pathologies, including neuropathy, myopathy, ischaemia/reperfusion, cataract and various solid tumours (gliomas, prostate, ovarian, colon, liver, head and neck cancer) (Ruan et al. 2011).CRYAB represents a promising therapeutic target owing to its diverse functions as a molecular chaperone.Studies in breast cancers have provided compelling clinical data that showed that CRYAB is constitutively expressed in the most aggressive phenotypes of the disease, including TNBC.Accumulating evidence has revealed that overexpression of CRYAB in TNBC contributes to tumour progression and correlates with the poor survival rates observed in this aggressive tumour type.This can be attributed to its critical role as a metastatic and anti-apoptotic regulator.Thus, CRYAB has been evaluated as a potentially important biomarker in TNBC (Tsang et al. 2012;R u a ne ta l .2011).Our previous structure-based molecular docking studies identified 3-methylglutamic acid as a small molecule with the potential to inhibit CRYAB/VEGF interaction.3-Methylglutamic acid was shown to disrupt CRYAB/VEGF 165 interaction and elicit both in vitro and in vivo antitumour, anti-proliferative and anti-angiogenic effects through downregulation of VEGF signalling (Chen et al. 2014).
In recent years, our laboratory has focused on the discovery of new drug candidates for potential application in the TNBC setting (Chen et al. 2014;Haynes et al. 2016).As part of these efforts, we were attracted to the possibilities associated with purine-based compounds, given their long and distinguished history in cancer drug development, including clinically approved anti-leukaemia drugs such as fludarabine (Rodriguez 1994) and clofarabine (Ghanem et al. 2010).Purines represent a privileged heterocycle in drug design, given the ability of purine analogues to bind tightly within hydrophobic folds and inhibit a range of key enzyme targets, such as members of the diverse protein kinase, sulfatase and phosphorylase families.For example, roscovitine (Selciclib) is a cyclin-dependent kinase inhibitory purine analogue in clinical development in a range of cancer, infectious disease and anti-inflammatory settings (De Azevedo et al. 1997).The specific sub-class of thiopurine analogues, such as 6-mercaptopurine and thioguanine, are also used as standard cancer chemotherapeutic agents, principally as anti-leukaemic drugs (Sahasranaman et al. 2008).Figure 1 shows the chemical structures of selected purine analogues in clinical use or advanced development.

Medicinal Chemistry Research
During our previous work on the identification of 3methylglutamic acid in disrupting CRYAB/VEGF 165 interaction (Chen et al. 2014), we noted that the thioguanine core structure was also able to potentially bind within the protein-protein interface binding pocket.In this paper, we extend our studies on the identification of novel chemical scaffolds active in models of TNBC through the synthesis and in vitro antitumour evaluation of a series of substituted thiopurines, and test their potential as CRYAB/VEGF inhibitors.

Materials and methods
All solvents and reagents were used as obtained from commercial sources unless otherwise indicated.Final purification of target compounds was carried out by silica gel column chromatography using CH 2 Cl 2 /MeOH as eluent.Compound purity was established using 1 Ha n d 13 C NMR, alongside accurate mass spectrometry. 1H NMR and 13 C NMR spectra were recorded in the appropriate deuterated solvents with a Bruker Avance DPX500 spectrometer operating at 500 MHz.Chemical shifts (δ) are reported in parts per million (ppm) using the following abbreviations: s, singlet; d, doublet; t, triplet, q, quartet, quin, quintet; sex, sextet; sep, septet; m, multiplet, b r ,b r o a d ,a p p .t ,a p p a r e n tt r i p l e t .A l lc o u p l i n gc o n s t a n t s are reported in Hz.Mass spectra were recorded with a Bruker microTOF mass spectrometer using electrospray ionisation (ESI) conditions, or provided as a service at the National Mass Spectrometry Facility at Swansea University, UK.Synthetic procedures and analytical/spectroscopic data for intermediate compounds can be found in Supplementary Information.

General methods for the alkylation of purines
Alkylation method 1 Substituted purine compound (1.0 eq) and potassium carbonate (1.1 eq) were dissolved in DMF (5 mL) at r.t. for 5 min.Bromoethane (1.2 eq) was added dropwise, then the reaction stirred for 3-18 h, after which the reaction was quenched with cold water resulting in precipitation.The precipitate was filtered and washed with water.If no precipitate formed, the product was extracted with chloroform or ethyl acetate, dried over sodium sulphate and concentrated under reduced pressure.

Alkylation method 2
Substituted purine compound (1.0 eq) were dissolved in 1 M potassium hydroxide (1.1 eq).Bromoethane (1.2 eq) was added dropwise, then the reaction stirred for 3-5h.On completion of the reaction, the product was precipitated with acetic acid, filtered, washed with water and dried.

Alkylation method 3
Substituted purine compound (1.0 eq) and 1,8-diazabicyclo [5.4.0]undec-7-ene (1.1 eq) were dissolved in DMF at r.t. for 5 min.The alkyl halide (1.2 eq) was added dropwise, then the reaction stirred for 3-5 h.The reaction was then quenched with cold water resulting in precipitation.The precipitate was filtered, washed with water and dried prior to purification by column chromatography.

Biological evaluation
MTT assay (Mosmann 1983) Breast cancer cell lines MDA-MB-231 and MDA-MB-436 were maintained in Delbecco's Modified Eagle's Media (DMEM) supplemented with 10% FCS.Human microvascular endothelial (HMVECs) were maintained in microvascular endothelial cell basal medium 131 with growth supplement (Gibco).MCF10A cells were maintained in mammary epithelial cell growth medium (Lonza) optimized for the growth of mammary epithelial cells in a serum-free environment.All cells were incubated at 37 °C in 5% CO 2 .Cells were seeded in a 96well plate at a density of 5 × 10 3 cells/ml and incubated for 24 h at 37 °C and 5% CO 2 .The cells were then treated with different concentrations of the test compounds and controls and incubated for 72 h.After this period, 20 µL of sterile MTT solution in PBS (5 mg/mL) was added to each well.The plates were then incubated for 5 h at 37 °C and 5% CO 2 .Following this, 100 μL of acidified isopropanol was added (0.02% 2M HCl) and the mixture was allowed to stand until the formed formazan crystals had dissolved.The absorbance was measured at 570 nm (BioTeK), and the cell viability was expressed as a percentage of the absorbance of non-treated cells.The experiments were performed in three independent repeats, and the results were expressed as the mean ± the standard error of the mean (SEM).

Cytotoxicity assay
(Essen Bioscience 2017) MDA-MB-231, MDA-MB-436, HMVECs and MCF10A cells were seeded (100 μLp e r well) at a cell density of 5000 cells/well into a 96-well plate a n di n c u b a t e da t3 7°Ci n5 %C O 2 within the IncuCyte ZOOM® system for 24 h.Contrast images were captured every 2 h and analysed using an integrated confluence algorithm.The test compounds and controls were prepared in the media containing the IncuCyte™ Cytotox Reagent at 3x the final concentration, then 50 μLo ft h i ss o l u t i o nw a s added to each well to obtain a dilution of 1:3 with a final assay volume of 150 μL( final assay concentration of Cytotox Reagent 250 nM).Images of the plate were captured at 10x objective, two images per well with phase and fluorescence scanning every 2 h until completion of the assay (72 h).The data were analysed using integrated software.

PathHunter TM cell-based CRYAB/VEGF 165 interaction assay
The cells were maintained in a PathHunter Assaycomplete TM culture media and incubated at 37 °C and 5% CO 2 until confluent.The U2OS cells were detached with Assaycomplete TM cell detachment reagent, suspended in Assaycomplete TM cell plating reagent and counted.In total, 20 µL of the cell suspensionwastransferredtoeach well of a 384-well white-walled, clear bottom plate at the indicated cell densities (5000, 10000, 20000 cells/ml).

Medicinal Chemistry Research
The seeded cells were incubated for 24 h at 37 °C and 5% CO 2 , 5 µL of the test compounds (10 μM) dissolved in the vehicle (2% DMSO) were added to the appropriate wells and incubated at room temperature for 90 min.After incubating, 25 µL per well of PathHunter Detection Reagents were added to the plates, incubated for 60 min at room temperature in the dark and the plate read using the POLARstar Omega luminescence reader.

ELISA assay
VEGF 165 secreted by MDA-MB-231 cells were assayed using Quantio® ELISA assay kit (R&D Systems).The assay was performed according to the manufacturers' instructions.MDA-MB-231 cells were grown in a T25 flask at 37 °C and 5% CO 2 until 80% confluence.The cells were then treated with the test compounds dissolved in DMSO (0.1%).The cell supernatant was harvested by centrifugation at 2000 × g for 15 min at 6, 12 and 24 h.VEGF 165 concentrations were measured in triplicate in each sample.The results are expressed as VEGF 165 pg/ml for at least three separate experiments.

Molecular modelling
All molecular modelling studies were performed with Molecular Operating Environment (MOE) 2015.10 modelling software.The crystal structure of the ACD domain of CRYAB was obtained from the Protein Data Bank (PDB code: 2KLR).Hydrogen atoms were added to the crystal structure minimised with MOE until a gradient of 0.05 Kcal mol −1 Å −1 was reached, using the MMFF94x forcefield.The partial charges were automatically calculated.To simplify calculations and analysis within this hydrophobic active site, no water molecules were considered.The docking site was defined by selecting the important amino acid residues (Tyr122, Arg 123 and Arg 149) and extending to 4.5 Å to include other β7a n dβ9 amino acid residues.Ligand docking was performed using MOE default settings.After docking, all docked molecules were visually inspected.

Design of novel analogues and synthetic chemistry
We began our synthetic studies around a dithiopurine-8-one heterocyclic core, proposing the introduction of substituents to improve lipophilicity and enhance cellular penetration.We decided to explore alkylation chemistry to derivatise the core dithiopurine-8-one heterocyclic structure particularly at the sulphur and/or the N7/N9 positions, as well as introducing functional flexibility at position C8. Figure 2 shows the proposed structural changes and exploratory chemistry leading to five small series of new molecules for in vitro anticancer evaluation.
There is clear evidence in the literature that reaction conditions play a crucial role in determining the selectivity of alkylation of purines and pyrimidines.Therefore, various conditions reported in the literature were examined to ascertain the optimal reactions conditions required to obtain selective alkylation at the appropriate positions of the purinone core (Pathak et al. 2004;Cheng and Robins 1958;Robins 1958;Biagi et al. 1996;Corder et al. 2013;Johnston et al. 1958;Robins 1957).Scheme 1 outlines the synthetic route employed in the synthesis of Series 1 analogues 4a-g, characterised by derivatisation of the sulphur groups at positions 2 and 6 with various alkyl groups.The purine core structure 2 was prepared via a thermal cyclisation reaction between the commercially available pyrimidine 1 and urea (Levin et al. 1960).As suggested in the literature, it was expected that under alkaline conditions, alkylation of compound 2 would selectively produce the exclusively Salkylated product (Joule et al. 1995;Montgomery et al. 1966).In addition, it has been reported that low reaction temperatures favour S-over N-alkylation.However, it was noted that when using K 2 CO 3 or KOH as a base, the trisubstituted product (3) was produced regardless of the stoichiometry of the reagents.The best conditions found to facilitate selective S-alkylation to provide the double alkylation products (4a-g) involved the non-nucleophilic base, 1,8-diazabicyclo[5.4.0]undec-7-ene (DBU) in DMF (Biagi et al. 1996), in yields of 29-74% following purification by column chromatography.The synthetic route employed in the synthesis of Series 2 analogues 6a-g is outlined in Scheme 2. A thermal cyclisation reaction between pyrimidine (1) and thiourea resulted in the 8-thiopurine (5), which was followed by alkylation reactions with various alkyl halides as described in Scheme 1.Under all the investigated reaction conditions, the products of the alkylation reactions were the trisubstituted purines similar to the products obtained under the previously described alkaline conditions (K 2 CO 3 or KOH), in low-to-moderate yields of 17-54% (Scheme 1).The acidity of the thiopurine NH group adjacent to the pyrimidine N3-atom provides a rationale for alkylation at the N9-position under basic conditions.Although trisubstituted purines have been reported in the literature, they are usually constructed from alkylated pyrimidines cyclised with alkoxides and not in a one-pot synthesis as we achieved (Yang et al. 2005;Villatoro et al. 2015;Ibrahim and Legraverend 2009).
Scheme 4 was employed to give access to selectively Nalkylated purines 13a-g that were deemed necessary for SAR studies.Selective N-alkylation was readily achieved by a reaction between 2-amino-6-chloropurine 8 and various alkyl halides to afford compounds 9a-g/10a-g as a mixture of regioisomers, with the N9-isomer (9a-g) being the major product as previously reported in the literature (Chhabra et al. 2013).Both isomers were successfully separated by silica gel chromatography and fully characterised by NMR.The 1 H NMR spectrum showed the alkyl peaks of the N7 product to be shifted downfield due to the deshielding effect of the bulky chloro group.The N9 products were then taken forward in a multi-step synthesis to give the desired N-alkylated purinones 13a-g by introducing a keto-group at position 8. Compounds 9a-g were hydrolysed in aqueous HCl to afford the intermediates 11ag, which were then brominated to afford the 8-bromopurine intermediates 12a-g (Michael et al. 1993).Finally, the brominated purines were transformed to the desired 8oxopurines 13a-g through hydrolysis of the bromo group, in low to good yields (6-85%) (Verones et al. 2010).

Biological evaluation-anti-proliferative activity
The analogues were tested for their in vitro inhibitory effect on the viability of the invasive TNBC cell lines MDA-MB-231 and MDA-MB-436, using the MTT endpoint assay (72 h).Additionally, endothelial cell lines, MCF10A and human microvascular endothelial (HMVEC-d) cell lines were used to assess the effects of the analogues on endothelial cells.Table 1 details the result of this in vitro screen expressed as IC 50 values from dose-response studies.
In general, several analogues from Series 1 (4a-g), Series 2 (6a-g)a n dS e r i e s5( 16a-g) showed the most potent anti-proliferative effects, significantly reducing the viability of one or both TNBC cells with IC 50 values in the low micromolar range.It should be noted that active analogues were also generally more active as growth inhibitors in the endothelial cell lines HMVEC and MCF10A, perhaps not surprisingly for these early stage lipophilic drug candidates.The selectivity profiles between the different cell types certainly underscore the importance of carrying out more detailed toxicity studies in the future.Notably, benzyl/phenethyl substituted purinone/thiopurinone analogues 4e, 4f, and 6e, were the most potent in TNBC cell lines, suggesting that this group may be important for biological activity.Across all series, it was also evident that the chain length of the alkyl substituents has an influence on biological activity.For analogues with aliphatic substituents, it was observed that isobutyl > propyl > ethyl > methyl in terms of their anti-proliferative activity within cell lines examined.For example, ethyl substituted analogue 4a showed IC 50 values of 97 and 100 μM against MDA-MB-231 and MDA-MB-436 cells, respectively, compared with the propyl substituted analogue 4c, which showed an IC 50 value of 29 μM against both TNBC cell lines.However, for analogues with alkyl benzyl substituents, an increase in the chain length of the alkyl benzyl group appears to have a detrimental effect on biological activity.For example, compound 4f with a phenethyl group showed an IC 50 value of 3-4 μM in TNBC cell lines, while increasing the chain length to a phenylpropyl group as seen for compound 4g resulted in a complete loss of activity (IC 50 > 100 μM).
Overall, analogues from Series 3 and 4 showed no inhibitory effects against either TNBC cell line apart from compound 15f (Series 4).This suggests that the substitution at positions 2 and 6 and the keto group at position-8 may contribute to the observed biological effect.As noted above, the most potent analogues across all series also showed potent growth inhibitory effects on the epithelial and endothelial cell lines MCF10A and HMVEC-d, respectively, without the desired selectivity for TNBC cell lines.

Cytotoxicity evaluation
Compounds that demonstrated the most potent inhibitory effect in the anti-proliferative assay (IC 50 ≤ 50 µM) were evaluated for their ability to induce cytotoxicity against the cell lines under investigation.The in vitro cytotoxicity assay was performed using the cyanine nucleic acid dyes developed by Essen BioScience with their IncuCyte™ live-cell imaging system that enables real time detection of cell death.The cells were co-cultured with the IncuCyte™ Cytotox reagent, and test compounds 4c, 4e-f, 6d-e, 15f and 16d were added at different concentrations over a period of 72 h.Images were captured every 2 h for the duration of the assay and the cytotoxicity effect was quantified in real-time with the integrated analysis software.Generally, all the tested compounds showed a significant increase in the cell death characteristics of a cytotoxicity effect compared with the no-drug control, with a few notable exceptions (Fig. 3).All the seven compounds induced a cytotoxic effect at a concentration of 100 μM against the TNBC cells, while cytotoxic effects were detected at a concentration of 50 μM against endothelial cells.These results appear to correlate with the observed anti-proliferative activity and thus, provided confidence for these analogues to undergo further evaluation, notably in our custom cell-based CRYAB/VEGF 165 interaction assay to probe their ability to disrupt this interaction.

PathHunter TM cell-based CRYAB/VEGF 165 interaction assay
The PathHunter TM cell-based CRYAB/VEGF 165 interaction assay is a commercially available, custom cell-based assay designed by DiscoverX.The assay utilizes the enzyme fragment complementation (EFC) technology for studying protein-protein interactions.For the purposes of this research, the assay was designed to monitor the interaction between CRYAB and a cytosolically localised VEGF 165 .The two candidate proteins were fused to complementation fragments of the β-galactosidase (β-Gal) enzyme.CRYAB was co-expressed with the larger sequence encoding the majority of enzyme, termed enzyme acceptor (EA), while VEGF 165 was fused to a small peptide epitope called Prolink (PK).The two separate enzyme fragments are inactive; however, after the constructs were transduced into U2OS cells, the recombined active enzyme is generated as a result of CRYAB/VEGF 165 protein-protein interaction.The disruption of the CRYAB/VEGF 165 interaction was quantified by measuring the chemiluminescence signal produced as a consequence of the conversion of a non-luminescent substrate to a luminescent product by the active enzyme.
The U2OS cells were seeded at three cell densities (20,000, 10,000 and 5000 cells/well), allowed to adhere to the wells and challenged with different concentrations of the test compounds.Tunicamycin was used as a positive control for this assay because of its ability to elicit the unfolded protein response and thus inhibit angiogenesis (Banerjee et al. 2011).After 90-min incubation period, the nonluminescent substrates were introduced into the cells and    VEGF 165 ELISA measurement As previously discussed, one of the proposed mechanisms by which CRYAB promotes angiogenesis is by preventing the proteolytic degradation of misfolded or unfolded VEGF resulting in the upregulation of VEGF and maintenance of VEGF signalling (Ruan et al. 2011).Hence, we postulated that disrupting the interaction between CRYAB/VEGF would affect the levels of VEGF produced by breast cancer cells.The concentration of VEGF secreted by MDA-MB-231 cells was measured by the ELISA assay.MDA-MB-231 cells treated with 100 μM of compound 4e showed a significant reduction in VEGF levels compared with the control.A concentration of 160 pg/ml of soluble VEGF was detected by the assay compared with the control (260 pg/ml), representing a 40% decrease in VEGF concentration (Fig. 5).No significant reduction was observed in cells treated with compound 4f.This result supports our hypothesis that the disruption of the interaction between CRYAB/VEGF can lead to inhibition of VEGF production by breast tumour cells.

Molecular modelling studies
Previous work has reported sequence studies for the interaction of important regulatory proteins with CRYAB (Ghosh et al. 2007).Typical of most proteins of the Hsp family, the structure of CRYAB consists of an N-terminal region of ~65 amino acids, preceding a conserved α-crystallin domain (ACD) domain made up of 84 amino acids, and finally a C-terminal region composed of 26 amino acids.The ACD domain consists of eight β strands (β2t oβ9) and helical sequences with strong interactions to regulatory proteins.The aforementioned sequence studies demonstrated that the β strands (β3, β7, β8 and β9) play a crucial role in the interactions between many growth factors and CRYAB, including VEGF 165 .Additionally, an inhibitory competition study with decoy peptides corresponding to CRYAB and VEGF 165 confirmed the significance of β7 and β9 in the protein-protein interaction between CRYAB and VEGF 165 (Chen et al. 2014).Fortunately from the standpoint of computational drug design, a good crystal structure of the α-crystallin domain (ACD) of human CRYAB obtained from solid-state NMR, small angle X-ray scattering, and computational modelling studies has been published (PDB code: 2KLR) (Jehle et al. 2010).On this basis, for docking studies a hydrophobic structural pocket was selected that was formed by discontinuous segments of β7 and β9 and containing the amino acids Try122/Arg123 (β7) and Arg149 (β9).In a series of docking simulations carried out with Molecular Operating Environment (MOE 2017), all the analogues and one additional monosubstituted purinone, 6-mercapto-2-(methylthio)purin-8-one, were docked into the selected structural pocket to probe their interaction with the target site.
Figure 6 shows the predicted binding pose of 6-mercapto-2-(methylthio)purin-8-one.The methythio side group spatially occupies the structural pocket, albeit partially.Crucially, one of the nitrogen groups of the imidazolidine ring forms an H-bond with Tyr122.Moreover, there were other hydrophobic interactions between the purine ring and surrounding amino acids, including Asn78 of the β3 strand.Comparatively, 4e (Fig. 7) also retained the key interaction with Tyr122 and an additional interaction with β7 through a H-bond between a thiol group and Ile124.Compound 4f (Fig. 8) interacts with β7 through hydrophobic interaction between the purine core and Arg123.Furthermore, one of the nitrogen groups of the imidazolidine ring of both analogues forms a H-bond with Lys121 also of β7.Finally, it was noted that the benzyl/phenethyl rings penetrates further and occupies more of the binding pocket compared with 6mercapto-2-(methylthio)purin-8-one.Interestingly, even the aromatic ring of the benzyl/phenethyl groups does not fully occupy this pocket hence the presence of bulky substituents on the ring could promote better insertion, as well as new interactions that could improve on the biological activities of the analogues.Also noticeable is the fact that none of the compounds interact with β9; this may be indicative of its distance (26 amino acids away) from β7.Therefore, introducing bulky substituents to the aromatic ring could bring this group closer to β9 and enhance the interaction between the molecules and CRYAB.

Discussion
The biological and docking studies have highlighted structural features that could be responsible for the observed in vitro activities of the synthesised analogues and will inform future endeavours to design more efficacious CRYAB inhibitors.These observations are summarised below.

Effects of S-alkyl substituents
To improve on the lipophilicities of the purine core structure, various alkyl substituents were introduced to the thio-and/or nitrogen groups.It was observed that across all series, analogues with aromatic phenalkyl substituents demonstrated the most potent biological effects in the various assays.However, the chain length of the phenalkyl group was found to be a limiting factor, that is, an increase in chain length led to a decrease in activity.Furthermore, compounds 4e and 4f,w i t ht h i sb e n z y l / phenethyl functionality were able to disrupt the interaction between CRYAB/VEGF 165 in the protein-protein interaction assay.In addition, compound 4e also decreased the levels of VEGF secreted by MDA-MB-231 cells.In our docking studies, the benzyl/phenethyl group was also found to form key interactions with the selected structural pocket on CRYAB confirming the significance of this group.

Effects of N-substitutions
No improvement in activity was observed for the series of analogues with N-substitutions (Series 3).Additionally, the docking studies showed no interactions between this substitution and CRYAB (data not shown).

Effects of removing the thio group at position-2
Removal of the thio-group at position-2 as seen with Series 3 and 4 analogues could be responsible for the loss of activity in both the various in vitro assays.Crucially, in the docking studies, it appears that this substitution interacts with the structural pocket of CRYAB indicating that this substitution contributes to the biological activities of the analogues.

Functional flexibility at position-8
As Series 4 analogues are effectively non-toxic against the control cell lines, it can be deduced that this substitution may contribute to the observed toxicity of the analogues.However, the toxicity effect may not be limited to this substituent but also the substituted thiol group at position-2, since analogues without this substitution (Series 3 and 4) showed little or no cytotoxic effect.
CRYAB confers protection against apoptosis induced by chemotherapeutic agents by inhibiting caspase-3 activation, sequestration of the anti-apoptotic protein Bcl-2 a n dp r e v e n t i o no fB c l -2t r a n s l o c a t i o ni n t ot h em i t o c h o ndria.In a study where breast epithelial carcinoma cells were treated with the anticancer agent vinblastine, the Medicinal Chemistry Research phosphorylated form of CRYAB was found to inhibit apoptosis by interacting with Bcl-2, while nonphosphorylated CRYAB induced resistance against vinblastine (Launay et al. 2010).The silencing of CRYAB with antisense or RNAi moleculesreversedthiseffectand sensitised tumour cells to chemotherapeutic and apoptotic agents (Lee et al. 2012).Angiogenesis, a segment of the metastatic cascade and the role of CRYAB as a facilitator of this phenomenon is of particular interest to our group.In the tumour microenvironment, CRYAB modulates angiogenesis through its effect on both cancer and endothelial cells.Studies have shown that CRYAB can prevent proteolytic degradation of VEGF by binding to and correcting unfolded/misfolded VEGF leading to preservation of intracrine VEGF signalling.Inhibition of CRYAB leads to a decrease in endogenous VEGF levels, providing clear evidence that targeting CRYAB may represent a novel approach to suppress angiogenesis.It should be noted that CRYAB is also expressed at high levels in the the heart and cardiomyocytes, conferring protection against stress (such as an ischaemic event) by inhibiting apoptosis (Pereira et al. 2014;Gonçalves et al. 2016;Morrison et al. 2003).Therefore, direct inhibition of CRYAB could lead to unwanted cardiotoxicity.Consequently, we propose a novel strategy of targeting angiogenesis by disrupting the interaction between CRYAB and VEGF with small molecules that competitively bind to the interacting surface.This approach could produce a therapeutic effect while potentially avoiding the unwanted side effects associated with direct inhibition of CRYAB, and serve as a novel anti-angiogenic therapy in the treatment of TNBC.

Conclusions
Even though protein-protein interactions (PPI) are involved in many fundamental biological processes, designing small molecules to target PPI's has been deemed challenging due to the frequent occurrence of large and featureless protein interfaces.Over the last two decades, as our understanding of the interfaces have progressed, PPI's have become attractive molecular targets in drug development.Although the interfaces of PPI's are extensive, studies have shown that not all residues of these interfaces contribute towards the stability of the binding between target-and client proteins, but only the so-called "hotspots" (a number of conserved amino acid residues) are critical for interaction.Thus, most of these "hotspots" have become targets in the drug development process and the last few years have seen a number of small molecule PPI inhibitors enter clinical trials (White et al. 2008;A r k i ne ta l .2014).
Following this trend, we propose a novel approach of supressing angiogenesis in TNBC with small molecules that target the PPI between CRYAB and VEGF.There is a critical need for targeted therapies in TNBC since the only established treatment for this aggressive disease is chemotherapy, which has been shown to be ineffective in terms of medium-to long-term survival.Angiogenesis is one of the main targets in TNBC due to its role in disease progression, however, most anti-angiogenic therapies have shown mixed and sometimes disappointing results.In the tumour microenvironment, CRYAB is a chaperone protein that has been shown to bind to and correct unfolded/misfolded VEGF promoting angiogenesis.This protein is overexpressed in TNBC and has been found to correlate with poor prognosis in this disease setting, making it an attractive target for the development of new anti-angiogenic therapies.Several studies looking at the interaction between CRYAB and regulator proteins identified a "hotspot" on the surface of CRYAB where VEGF interacts.Aided by molecular docking methods, small molecule hit compounds were identified which bind to this "hotspot" and could potentially inhibit this PPI.
In this paper, we report a proof-of-concept study devised to validate this target.The work started with the synthesis of novel purine-based compounds for structural activity relationship studies.The newly synthesised analogues were then subjected to biological evaluation, firstly, in a cell viability assay where several of the analogues were found to significantly reduce the viability of MDA-MB-231 and MDA-MB-436 TNBC cell lines.The most potent of these analogues were then tested for their ability to disrupt the CRYAB/VEGF interaction in a custom designed cell-based CRYAB/VEGF 165 PPI assay.The PPI assay showed that compounds 4e and 4f successfully disrupted the interaction between CRYAB and VEGF 165 .Additionally compound 4e (100 μM) was found to decrease the amount of VEGF secreted by MDA-MB231 TNBC cells by 40%.Molecular docking studies also suggested that these two analogues interfere with the target "hotspot" through key interactions with specific amino acids.In conclusion, for this work to progress to pre-clinical trials as a viable therapeutic option in TNBC, further design, synthesis and in vitro studies are required to validate this novel concept.Additional studies on lead compound stability and other pharmacokinetic properties will also be required prior to evaluation within relevant in vivo models and identification of a new preclinical drug candidate.

Fig. 2
Fig. 2 Target molecule series derived from the dithiopurin-8-one core structure

Fig. 3
Fig. 3 Cytotoxic effect of selected analogues against TNBC and control cell lines.Data are expressed as mean ± standard error of the mean (SEM), n = 3

Fig. 4
Fig. 4 Concentration-response curves of analogues 4e and 4f and controls in the PathHunter TM cell-based CRYAB/VEGF 165 interaction assay, indicating chemiluminescence quantification relative to the vehicle control; data are expressed as mean ± standard error of the mean (SEM), n = 3, ***P < 0.001

Fig. 6 a
Fig. 6 a Predicted binding mode of 6-mercapto-2-(methylthio) purin-8-one (orange) in selected docking site of CRYAB (grey), and b ligand interaction diagram (the compound is shown in orange while protein residues are in green)