Inhibition of the HIF1α-p300 interaction by quinone- and indandione-mediated ejection of structural Zn(II)

Abstract

Protein-protein interactions between the hypoxia inducible transcription factor (HIF) and the transcriptional coactivators p300/CBP are potential cancer targets due to their role in the hypoxic response. A natural product based screen led to the identification of indandione and benzoquinone derivatives that reduce the tight interaction between a HIF-1α fragment and the CH1 domain of p300. The indandione derivatives were shown to fragment to give ninhydrin, which was identified as the active species. Both the naphthoquinones and ninhydrin were observed to induce Zn(II) ejection from p300 and the catalytic domain of the histone demethylase KDM4A. Together with previous reports on the effects of reated compounds on HIF-1α and other systems, the results suggest that care should be taken in interpreting biological results obtained with highly electrophilic/ thiol modifying compounds.

Introduction

In humans and other animals the hypoxia inducible factor (HIF) system plays a central role in the hypoxic response [1–4]. When oxygen becomes limiting, levels of the HIF-1α subunit rise, enabling its dimerization with the HIF-1β subunit. α, β-HIF activates gene expression that works to alleviate the effect of hypoxia in a context dependent manner [5]. HIF target genes, e.g. vascular endothelial growth factor (VEGF), are upregulated in many tumours, hence inhibition of HIF activity is a potential anti-cancer strategy [6–8]. The factors that regulates HIF target gene expression are still emerging, but it is clear that the transcriptional coactivator proteins p300/CREB(cAMP response element-binding protein)-binding protein (CBP) promote transcription of most, possibly all, HIF target genes [9,10]. Hence blocking the HIF-1α/p300(CBP) interactions is of most interest as an anticancer target [11,12].

The HIF-1α/p300 protein-protein interaction (PPI) is tight (K_D ≈ 7 nM) [13], involving the C-terminal transactivation domain (C-TAD) of HIF-1α/-2α isoforms binding to the CH1 (Cysteine/Histidine-rich 1) domain of p300/CBP (Figure 1) [14–16]. Large surface interactions, as are observed between the HIF-1α C-TADs and p300/CBP, represent one of the challenges in inhibiting PPIs [17,18]. Interruption of the HIF-1α/p300(CBP) interaction has shown to negatively regulate oncogene expression and tumor growth [19–22]. Thus, the therapeutic significance of the HIF system has stimulated further high-throughput- and natural product-screening approaches for its inhibition [23–31]. The screens have employed both cell-based and isolated protein approaches; the cell-based approaches have yielded compounds that act indirectly on HIF, affecting the stability of HIF system proteins or by binding the hypoxia response elements (HREs) in DNA. Disrupting binding of HIF-1α to HREs has been demonstrated [25,32,33], though selectivity of DNA binders remains a concern.

Figure 1.

View from an NMR structure a fragment of the HIF-1α C-terminal transactivation domain (C-TAD) (785-826) (red) complexed with the CH1 domain of p300 (323-423) (green) with structurally important p300 zinc atoms shown in magenta (PDB: IL3E) [14].

In pioneering work, Kung et al. used a competition ELISA assay, with a biotinylated HIF-1α C-TAD truncate (785-826) and a GST-tagged CH1-domain, to identify chetomin, one of the epidithiodiketopiperazine (ETP) class of natural products, as a HIF-1α/p300 inhibitor [34]. The core, electrophilic, ETP functionality has been shown to be sufficient for activity, with a number of analogues showing similar activity to the natural products [35–37]. Subsequent work determined that chetomin and other ETPs work, at least in part, by Zn(II) ejection from the CH1 domain of p300, thus disrupting its structure, ablating the interaction with HIF-1α [38]. Modes of action involving cysteine modification and/or zinc ejection are likely inherently unselective, with the ETPs, such as chaetocin, showing inhibition against thioredoxin reductase and a number of histone methyl transferases [39–41].

In a search for inhibitors for the HIF-1α/p300 interaction we conducted an HTS of 10,000 natural product-based structures using a similar ELISA competition assay (Figure 2). The results led to the identification of electrophilic inhibitors of the HIF-1α/p300 PPI.

Figure 2.

Inhibitors of the HIF-1α/p300 interaction identified in a natural product-like compound screen. IC₅₀ values (μM) are in parentheses; (a) benzoquinones; (b) indandiones.

Results and Discussion

The output of the screen led to the identification of two distinct compound classes that showed promising activity: benzoquinones 1-3, and 2,2-disubstituted indandiones 4-6 (Figure 2). The observation that a structurally diverse set of quinone derivatives displayed similar levels of activity suggested that the core quinone may be the active component. Indeed, we found that simple commercially available quinones were also active (Figure 3). The results showed that the benzoquinone core is sufficient for activity, with potency correlating well with reported oxidative potentials [42,43]. Thus, anthraquinone 7 (half wave potential (E_1/2) = -1.26 V) was much less active than naphthoquinone 8 (E_1/2 = -1.03 V) or 2,3-dichloro-5,6-dicyano-1,4-benzoquinone (DDQ) 9 (E_1/2 = -0.34 V) [44,45].

Figure 3.

Assays of commericially available quinones 7-12 and disulfide 13 for disrupting the HIF-1α (785-826)/p300 CH1 domain (323-423) binding. (a) tested compounds; (b) assay results (1% DMSO; triplicate, ±SD). * 7, 9, 10, 13 were tested at 100 μM, 8, 11, 12 were tested at 63 μM.

Dipyridyldisulfide 13, which contains a disulfide, as do the ETP inhibitors, was also weakly active (IC₅₀ ≈ 100 μM). Analogous aromatic compounds, which did not contain quinone functionality, displayed no activity (e.g. 8 compared to 11) although hydroquinones 10 and 12 showed modest activity. Hydroquinones could also be involved in a redox process, generating reactive quinones in situ. Spontaneous oxidation of hydroquinone and catechol by molecular oxygen has been observed to covalently modify DNA, suggesting that such redox cycles may be responsible for activity in our assay [46,47]. A common feature of all the active groups is the presence of electrophilic groups that can react with cysteines/thiols [48,49].

We then investigated the nature of the indandione inhibition. A range of 2-amino-, 2-imino- and 2-amidoindandiones 5, 14-20 were synthesized to investigate SAR (Figure 4a). The amino- and imino-derivatives 18-20 were less active than amido-derivatives 5 and 14-17 (Figure 4b). Notably, ninhydrin 23 the parent compound of the indandione derivatives, displayed similar potency to the amido-derivatives (Figure 4b, Supp. Info – Table S1), suggesting that ninhydrin may be the active component of the inandione compound class. Indeed, mass spectrometric and NMR analysis indicated that an aqueous solution of amido-compound 5 generates ninhydrin 23 (Figure 5a). The decay of 5 to picolinamide and 23 was monitored by ¹H NMR at pH 8 (D₂O, 10 mM phosphate buffer), indicating that 80% hydrolysis occurs after one hour (Figure 5b). Compounds which were structurally similar, but lacked the reactive C-2 centre; i.e. 2-amido-indoline 21 and indandione 22, were inactive (Figure 4b). We thus propose that all the apparently active indandione derivatives fragment to give ninhydrin 23, which is the active species.

Figure 4.

Assays with ninhydrin related compounds a) tested compounds included mono- (5, 14-17) and di-ninhydrin adducts (18) and related derivatives (19-22); (b) Inhibition data of tested compounds at three doses; (c) dose response curves for selected compounds (5, 14-17, 21 and 23). (1% DMSO; triplicate, ±SD).

Figure 5.

The ninhydrin adducts undergoes fragmentation in aqueous solution. Adduct 5 was dissolved in deuterated phosphate buffer (pH 8) and it’s stability was monitored by ¹H NMR (500 MHz). Increased appearance of picolinamide 24 signal reveals fragmentation.

To further investigate the mode of action of these electrophilic compounds we tested whether they caused Zn(II) ion ejection from jumonji domain 2A histone demethylase (KDM4A), for which treatment with other Zn(II) ejectors has been shown to be inhibitory [50–52]. In the catalytic domain of KDM4A a Zn(II) ion is bound to three cysteines and one histidine in an analogous fashion to the coordination observed in the CH1 domain of p300 (Figure 6b and c). Ebselen, a known zinc-ejector for KDM4A was used as a positive control, with the dye FluoZin-3™ (FZ-3) providing a measure of the unbound zinc concentration [50]. Compounds 8, 10, 16 and ninhydrin 23, which were active in the competition binding assay also caused Zn(II) ion loss from KDM4A in a dose and time dependent manner (Figure 7b). Despite being less effective than ninhydrin 23 in the competition-binding assay, quinone 8 and reduced quinone 10 showed comparable KDM4A activity to 23. Analogous studies on p300 yielded similar results, although the high basal levels of Zn(II), added to p300 such that it folds correctly, results in poorer resolution (Figure 7a, Supp. Info – Figure S2). The lack of selectivity observed by the quinones and indandiones identified in our initial screen suggest that they are likely not selective for different zinc binding sites [53,54]

Figure 6.

Proposed outline mechanism of electrophile-promoted Zn(II) ejection from p300 (a); Zn(II) binding sites in the CH1 domain of p300 (b) are structurally similar to those found in other proteins including the catalytic domain of KDM4A (c). PDB: p300: IL3E and KDM4A: 2PXJ respectively [14,52].

Figure 7.

Zinc ejector behavior of Ninhydrin 23, and benzoquinone 8 against a) p300 and b) KDM4A. Fluorescence-based assays for release of Zn (II) ions from the CH1 domain of p300 and the catalytic domain of KDM4A. Compounds show a dose- and time-dependent increases in fluorescence as Zn is released into the buffer. Established zinc ejector ebselen was used as a positive control for Zn(II) release.

When 8 and 23 were tested in a HeLa cell viability assay, significant dose-dependent cytotoxicity was observed after 48 h (Supp. Info - Figure S3). The inactive compound 21 was not cytotoxic under the tested conditions. Naphthoquinone 8, ninhydrin 23 and related compounds have been shown to form protein adducts resulting in nonspecific toxicity [55,56].

Conclusions

In conclusion, our results have validated the output of an HTS on the HIF-1α/p300 interaction which led to the identification of quinone and indandione inhibitors. Subsequent studies demonstrated that the core quinone and ninhydrin parent rings are sufficient for inhibition, which likely occurs via non-selective loss of zinc ions, leading to disruption of the domain fold. Whilst it is possible that appropriate derivatisation could enable selectivity to be achieved, the available evidence is that this will be non-trivial. Further, we note that related electrophilic and redox sensitive compounds have also been shown to inhibit the hypoxia system (Table 1). ETPs have been shown to have targets other than p300, including histone methyl transferases (HKMTs) and thioredoxin reductase (TrxR), where reaction with thiols is also proposed [34,40,41]. A variety of quinone containing compounds have also been suggested to inhibit HIF-1α either directly [29,31] or indirectly by interacting with HIF-1α stabilizing protiens [30,57–60]. The prevalence of potentially reactive inhibitors against p300, and hypoxia system proteins thioredoxin (Trx) and TrxR, might indicate that proteins involved in this cascade are particularly sensitive to electrophilic molecules.

Table 1.

Literature inhibitors of the HIF system and HIF system components. Potentially electrophilic functionality has been highlighted (bold).

INHIBITOR	REPORTED TARGET (PROPOSED)	REFERENCE
	HIF-1α /p300 KDM4A Others	55
	HIF-1α/p300 HKMTs TrxR	34, 40, 41, 42
	HIF-1α/p300	29
	HIF-1α/p300	31
	Trx-1 (indirectly on HIF-1α)	30
	HIF-1α/p300	57
	HSP90 (indirectly on HIF-1α)	58, 59

Whether the repeated identification of redox sensitive compounds in screens on the hypoxia system/ HIF components is more than coincidence is unknown at this stage. However, the development of such compounds into (selective) pharmaceuticals could be problematic, and it may of interest to configure (at least some of) the outputs of future screens to indentify such compounds [61,62].

Experimental

General information

Reactions were carried out under a nitrogen or argon atmosphere in oven-dried glassware at room temperature unless otherwise stated. Standard inert atmosphere techniques were used in handling all air and moisture sensitive reagents.

Anhydrous acetonitrile and dichloromethane (from commercial sources) were obtained by filtration through activated alumina (powder ~ 150 mesh, pore size 58 Å, basic, Sigma-Aldrich) columns, or were dried on an MB-SPS-800 dry solvent system. Other solvents and reagents were used directly as received from commercial suppliers. Petrol refers to distilled light petroleum of fraction (30 °C - 40 °C).

Flash column chromatography was carried out using VWR Kieselgel 60 silica gel (60-63 μm). Thin-layer chromatography was carried out using Merck Kieselgel 60 F254 (230-400 mesh) fluorescent treated silica, visualized under UV light (250 nm) and by staining with aqueous potassium permanganate solution.

¹H and ¹³C NMR spectra were recorded using a Bruker 500, 400 or 300 MHz spectrometer running Topspin™ software and are quoted in ppm for measurement against a residual solvent peak as an internal standard. Chemical shifts (δ) are given in parts per million (ppm), and coupling constants (J) are given in Hertz (Hz). The ¹H NMR spectra are reported as follows: δ / ppm (number of protons, multiplicity, coupling constant J / Hz (where appropriate), assignment). Multiplicity is abbreviated as follows: s = singlet, br = broad, d = doublet, dd = doublet of doublets, t = triplet, dt = doublet of triplet, q = quartet, dq = doublet of quartet, qn = quintet, sept = septet, m = multiplet. Compound names are those generated by ChemBioDraw™ (CambridgeSoft) following IUPAC nomenclature. However, the NMR assignment numbering used is arbitrary and does not follow any particular convention. Numbering of compounds is illustrated on the spectra themselves; vide infra. The ¹³C NMR spectra are reported in δ / ppm. Two-dimensional (COSY, HSQC, HMBC) NMR spectroscopy was used to assist the assignment of signals in the ¹H and ¹³C NMR spectra. IR spectra were recorded on a Bruker Tensor 27 FT-IR spectrometer from a thin film deposited onto a diamond ATR module. Only selected maximum absorbances (ν_max) of the most intense peaks are reported (cm⁻¹). High-resolution mass spectra were recorded on a Bruker MicroTof mass spectrometer (ESI) by the internal service at the Department of Organic Chemistry, University of Oxford. Melting points were recorded using a Leica Galen III hot-stage microscope apparatus and are reported uncorrected in Celsius (°C).

N-(2-Hydroxy-1,3-dioxo-2,3-dihydro-1H-inden-2-yl)picolinamide (5)

Ninhydrin (400 mg, 2.25 mmol) and picolinamide (274 mg, 2.25 mmol) were added to a mixture of acetonitrile (15 mL) and anhydrous MgSO₄ (150 mg) and the mixture was stirred at room temperature for 2 h. The mixture was filtered and washed with acetonitrile (15 mL). The solvent was removed from the filtrate in vacuo and the resulting residue was dissolved in dichloromethane (50 mL). The resulting solution was partitioned with water (50 mL). The product was crystallized from the aqueous phase as pale green crystals (27 mg, 0.10 mmol, 4 %); m.p 163-164; δ_H (400 MHz, d₆-DMSO): 9.10 (1H, s); 8.73 (1H, d, J 4.7); 8.10-8.03 (5H, m); 8.00 (1H, td, J 1.6, 7.7); 7.84 (1H, d, J 7.8); 7.68 (1H, ddd, J 1.1, 4.8, 7.6); δ_C (100 MHz, d₆-DMSO): 196.0; 163.6; 149.0; 147.6; 138.6; 138.1; 137.0; 127.5; 123.7; 121.9; 79.7; IR ν_max: 3291, 3020, 1748, 1710, 1657, 1360, 1096, 960, 736: HRMS (ESI) found 305.0529; C₁₅H₁₀N₂NaO₄ [M+Na]⁺ requires 305.0533.

N-(2-Hydroxy-1,3-dioxo-2,3-dihydro-1H-inden-2-yl)benzamide (14)

Ninhydrin (500 mg, 2.81 mmol) and benzamide (340 mg, 2.81 mmol) were added to a mixture of acetonitrile (15 mL) and anhydrous MgSO₄ (150 mg) and the mixture was stirred at room temperature for 2 h. The mixture was filtered and washed with acetonitrile (15 mL). The solvent was removed from the filtrate in vacuo and the resulting residue was dissolved in dichloromethane (50 mL). The resulting solution was partitioned with water (50 mL). The aqueous phase was washed with dichloromethane (3 × 50 mL). The solvent was removed from the combined organic phase in vacuo to yield the product as a white solid (264 mg, 0.93 mmol, 34 %); m.p 125 – 126; δ_H (400 MHz, d₆-DMSO): 9.82 (1H, s); 8.07-8.00 (4H, m); 7.96 (1H, s); 7.93-7.85 (2H, m); 7.58-7.42 (3H, m); δ_C (100 MHz, d₆-DMSO): 197.31; 166.9; 139.2; 137.5; 133.0; 132.1; 129.1; 128.7; 124.5; 81.4; IR ν_max: 3274, 1719, 1645, 1270, 1195, 1120, 967, 736. HRMS (ESI) found 304.0583; C₁₆H₁₁NNaO₄ [M+Na]⁺ requires 304.0580.

N-(2-Hydroxy-1,3-dioxo-2,3-dihydro-1H-inden-2-yl)propionamide (15)

The title compound (15) was prepared from ninhydrin (500 mg, 2.81 mmol) and propionamide (204 mg, 2.81 mmol) by following a procedure analogous to the one used for the synthesis of 14. The product was isolated as a cream solid (390 mg, 1.67 mmol, 60 %); m.p 153 – 154; δ_H (400 MHz, d₆-DMSO): 9.14 (1H, s); 8.03-7.95 (4H, m); 7.69 (1H, s); 2.13 (2H, q, J 7.6); 0.87 (3H, t, J 7.6); δ_C (100 MHz, d₆-DMSO): 197.7; 174.0; 139.3; 137.4; 124.3; 80.2; 27.4; 10.1; IR ν_max: 3375, 3133 (br), 1760, 1723, 1632, 1513, 1116, 965, 737. HRMS (ESI) found 256.0581; C₁₂H₁₁NNaO₄ [M+Na]⁺ requires 256.0580.

N-(2-Hydroxy-1,3-dioxo-2,3-dihydro-1H-inden-2-yl)nicotinamide (16)

Ninhydrin (500 mg, 2.81 mmol) and nicotinamide (342 mg, 2.81 mmol) were added to a mixture of acetonitrile (15 mL) and anhydrous MgSO₄ (150 mg) and the mixture was stirred at room temperature for 2 h. The mixture was filtered, washed with acetonitrile, concentrated and the resulting residue was dissolved in dichloromethane (50 mL). The resulting solution was partitioned with water (50 mL). The product was crystallized from the aqueous phase as pale green crystals (43 mg, 0.15 mmol, 5 %); m.p 199; δ_H (500 MHz, d₆-DMSO): 10.10 (1H, s); 9.04 (1H, d, J 1.7); 8.74 (1H, dd, J 1.6, 4.8); 8.24 (1H, dt, J 2.0, 8.0); 8.09-8.00 (5H, m); 7.51 (1H, dd, J 4.7, 7.8); δ_C (500 MHz, d₆-DMSO): 196.2; 164.6; 152.7; 148.9; 138.4; 136.5; 135.5; 126.7; 123.7; 123.4; 80.6; IR ν_max: 3344, 3270, 2980, 2696, 1753, 1716, 1191, 1124, 736. HRMS (ESI) found 305.0540; C₁₅H₁₀N₂NaO₄ [M+Na]⁺ requires 305.0533.

2-Chloro-N-(2-hydroxy-1,3-dioxo-2,3-dihydro-1H-inden-2-yl)acetamide (17)

Ninhydrin (500 mg, 2.81 mmol) and 2-chloroacetamide (260 mg, 2.81 mmol) were added to a mixture of acetonitrile (15 mL) and anhydrous MgSO₄ (150 mg) and the mixture was stirred at room temperature for 2 h. The mixture was filtered, washed with acetonitrile, concentrated and the resulting residue dissolved in dichloromethane (50 mL). The resulting solution was partitioned with water (50 mL). The aqueous phase was washed with dichloromethane (3 × 50 mL). The solvent was removed from the combined organic phase in vacuo, and the residue was washed with ether to yield the product as a pink solid (31 mg, 0.12 mmol, 4 %); m.p 140 – 143; δ_H (400 MHz, d₆-DMSO): 9.61 (1H, s); 8.07-7.98 (4H, m); 7.96 (1H, s); 4.10 (2H, s); δ_C (100 MHz, d₆-DMSO): 196.8; 166.7; 139.3; 137.8; 124.5; 80.4; 41.8; IR ν_max: 3364, 3157 (br), 2942, 1760, 1660, 1468, 1353, 1117, 740, 696. HRMS (ESI) found 276.0035; C₁₁H₈³⁵ClNNaO₄ [M+Na]⁺ requires 276.0034.

2,2-bis((3-Bromophenyl)amino)-1H-indene-1,3(2H)-dione (18)

According to literature procedure [63], 3-bromoaniline (0.28 mL, 2.66 mmol) was added to a solution of ninhydrin (473 mg, 2.66 mmol) in water (5.0 mL). After stirring at room temperature for 1 h, the yellow precipitate was filtered and washed with cold water. The residue was recrystalised from hexane:chloroform (1:5) to give the title compound 18 as a red/brown crystalline solid (150 mg, 0.31 mmol, 12 %),m.p. 142 – 143; δ_H (400 MHz, d₆-DMSO): 8.71 (4H, m); 7.23 (2H, t, J 1.8); 7.13 (2H, s); 6.97 (2H, t, J 8.0); 6.83 (2H, dd, J 1.5, 8.2); 6.79 (2H, dd, J 0.9, 7.8); δ_C (100 MHz, d₆-DMSO): 194.9; 147.2; 139.0; 138.3; 131.1; 125.2; 122.5; 121.5; 118.5; 115.0; 73.8; IR ν_max: 3377, 1696, 1589, 1474, 1256, 1138, 961, 767. HRMS (ESI) found 484.9332; C₂₁H₁₄⁷⁹Br₂N₂O₂ [M+H]⁺ requires 494.9330.

2.7 2-((4-Morpholinophenyl)imino)-1H-indene-1,3(2H)-dione (19)

A solution of ninhydrin (662 mg, 3.71 mmol) in water (10 mL) was added dropwise to a suspension of 4-morpholinoaniline (662 mg, 3.71 mmol) in water (10 mL). After stirring for 1 h, the brown precipitate was filtered and washed with MeOH (15 mL). The residue was recrystalised from MeOH to give the title compound 19 as dark purple crystals. (215 mg, 0.67 mmol, 18 %); m.p 212 – 214; δ_H (400 MHz, d₆-DMSO): 7.99-7.94 (4H, m); 7.77 (2H, d, J 9.2); 7.03 (2H, d, J 9.2); 3.8 (2H, t, J 5.0); 3.40 (2H, t, J 5.0). δ_C (100 MHz, d₆-DMSO): 188.0; 153.3; 140.9; 138.3; 136.9; 136.8; 130.7; 124.5; 113.8; 66.7; 47.7; IR ν_max: 1716, 1675, 1483, 1160, 1114, 979, 827; HRMS (ESI) found 343.1042; C₁₉H₁₆N₂NaO₂ [M+Na]⁺ requires 343.1053.

(±)8-Chloro-4b-hydroxybenzo[b]indeno[2,1-e][1,4]oxazin-11(4bH)-one (20)

A solution of ninhydrin (1.48 g, 8.31 mmol) in water (20 mL) was added to a solution of 2-amino-4-chlorophenol (1.19 g, 8.31 mmol) in water (10 mL). A few drops of pyridine were added and the mixture was stirred for 1 h. The precipitate was filtered and concentrated in vacuo affording the crude product as a white solid (2.08 g). A 500 mg sample of the residue was recrystalised from MeOH to give the title compound 20 as yellow/green crystals (40mg, 0.14 mmol, 2 %); m.p 273 – 275; δ_H (500 MHz, d₆-DMSO): 8.70 (1H, s); 8.19 (1H, d, J 7.8); 8.05-8.01 (2H, m); 7.90 (1H, t, J 7.7); 7.68 (1H, d, J 2.6); 7.38 (1H, dd, J 2.5, 8.7); 7.30 (1H, d, J 8.7); δ_C (125 MHz, d₆-DMSO): 191.6; 159.4; 143.8; 141.1; 137.5; 136.1; 134.9; 134.44; 128.4; 126.8; 126.7; 124.9; 123.9; 119.6; 86.0; IR ν_max: 2862, 1740, 1675, 1440, 1217, 971, 826, 717. HRMS (ESI) found 285.0215; C₁₅H₈⁽³⁵⁾ClNO₃ [M+H]⁺ requires 285.0193.

N-(1,3-Dioxoisoindolin-2-yl)benzamide (21)

According to literature procedure [64], phthalic anhydride (500 mg, 3.37 mmol) and benzohydrazine (545 mg, 4.00 mmol) were added to acetic acid (20 mL) and the mixture was heated at 125 °C for 2 h. The reaction was cooled to room temperature and water (35 mL, kept at 0 °C) was added. The white precipitate was filtered, washed with cold water and concentrated in vacuo to give the title compound 21 (517 mg, 1.9 mmol, 57 %); m.p 214 – 215; δ_H (400 MHz, d₆-DMSO): 11.34 (1H, s); 8.04-7.96 (6H, m); 7.68 (1H, t, J 7.7); 7.58 (2H, t, J 7.7); δ_C (100 MHz, d₆-DMSO): 165.37; 165.38; 135.5; 132.8; 130.7; 129.5; 128.8; 127.8; 123.90; IR ν_max: 3232 (br), 1799, 1733, 1662, 1282, 1118, 878, 700. HRMS (ESI) found 265.0617; C₁₅H₉N₂O₄ [M-H]⁻ requires 265.0619.