Synthesis of an azido-tagged low affinity ratiometric calcium sensor

Changes in high localised concentrations of Ca2+ ions are fundamental to cell signalling. The synthesis of a dual excitation, ratiometric calcium ion sensor with a Kd of 90 μM, is described. It is tagged with an azido group for bioconjugation, and absorbs in the blue/green and emits in the red region of the visible spectrum with a large Stokes shift. The binding modulating nitro group is introduced to the BAPTA core prior to construction of a benzofuran-2-yl carboxaldehyde by an allylation–oxidation–cyclisation sequence, which is followed by condensation with an azido-tagged thiohydantoin. The thiohydantoin unit has to be protected with an acetoxymethyl (AM) caging group to allow CuAAC click reaction and incorporation of the KDEL peptide endoplasmic reticulum (ER) retention sequence.


Introduction
Free Ca 2þ is an important cell messenger involved in numerous processes such as muscle contraction, cell division and cell death. 1 The effects of changing the Ca 2þ concentration, [Ca 2þ ], can have a wide reach, extending both within and among cells. 1 [Ca 2þ ] can also affect a range of different activities by acting selectively as a highly localised signal operating in subcellular regions. The free cytosolic [Ca 2þ ] is typically in the region of 100 nM at rest and increases to an averaged peak cellular value of <1 mM when the cell is activated. 2 However, to selectively perform multiple functions, cells localise signals to certain regions by creating high local [Ca 2þ ] in the range of tens to several hundred micromolar. 3 These localised concentrations of Ca 2þ enable the control of specific cellular activities, such as ion channel and transcription factor activation. In addition to high local concentrations in the cytoplasm, organelles such as the mitochondria, sarcoplasmic reticulum, endoplasmic reticulum and Golgi apparatus each store Ca 2þ at concentrations (hundreds of micromolar) far above the cytoplasmic average. 4 Understanding of how [Ca 2þ ] selectively controls cell function remains preliminary because of the difficulties in studying Ca 2þ signals in specific cell regions. Therefore, there have been efforts to develop sensors to study localised and dynamic calcium concentrations. 5e7 As part of these efforts, we set out to develop a fluorescent sensor that would detect changes in the high [Ca 2þ ] found in subcellular stores. We wanted a sensor that absorbed and emitted in the visible rather than UV region of the spectrum so that its use would disturb the system minimally, and that was ratiometric so that quantification was straightforward. Above all, we wanted to incorporate a tag so that bioorthogonal chemistry could be used to attach a targeting group or to attach the sensor to biomolecules.
We chose to develop a sensor using the highly selective octadentate binding of the Ca 2þ ion by a 1s,2-bis(ortho-aminophenoxy) ethane-N,N,N 0 ,N'-tetraacetic acid (BAPTA) ligand, 8,9 which has become a powerful tool for studying changes in [Ca 2þ ]. 10e18 Recent work has focused on developing sensors that absorb and emit in the low-energy red and near infrared regions of the spectrum, 14e17 both to limit damage and to allow the use of several different fluorophores in the same sample (multiplexing). 18 In line with this, we decided that our BAPTA-based sensor should absorb and emit in the red region of the spectrum.
We considered an ideal Ca 2þ sensor would be fluorescent both with Ca 2þ bound and in its unbound state, but with different absorption or emission wavelengths. This would allow the [Ca 2þ ] to be determined directly using the ratio of the two forms using the binding dissociation constant of the sensor (K d ). The use of such socalled ratiometric probes offers many advantages over traditional off/on probes such as correction for artefacts e.g., photobleaching and variation in probe loading. 19 However, the development of ratiometric Ca 2þ sensors has lagged behind the development of the traditional off/on probes. Indeed, most biological studies using this technique 20,21 rely on the original two sensors, Fura-2 and Indo-1, 8,9 which both absorb in the UV region of the spectrum. There are recent promising ratiometric probes developed by Liu et al., 13 but the area is underdeveloped. On the other hand, we and a few others have used Fura-Red 22,23 as a ratiometric Ca 2þ sensor that is excited by visible light. It is a dual excitation ratiometric probe so that [Ca 2þ ] can easily be determined from the ratio of the emissions at 640 nm when excited at 436 nm (Ca 2þ -bound sensor) and when excited at 472 nm (free sensor). 23 Fura-Red's very large Stokes shift of w200 nm also allows the simultaneous use of other dyes such as Fluo-4. 24 Given these excellent properties of Fura-Red, we decided to adjust its binding affinity so that it could detect changes when [Ca 2þ ] is high. The binding dissociation constant (K d ) should be midway between the starting and final [Ca 2þ ] in the process under study to maximise the observable response of the sensor. Most of the commonly-used Ca 2þ sensors have K d values within the nanomolar range (e.g., Fura-Red has a K d ¼380 nM 22,25 ), which makes them suitable for measuring global cytosolic [Ca 2þ ] fluxes. 10,11 We reasoned that a low affinity Fura-Red derivative, NitroAzidoFuraRed ( Fig. 1), could be prepared that would have an azido tag for bioorthogonal chemistry. 26 Fura-Red's fluorescent properties arise from photoinduced charge transfer because the fluorophore is in direct conjugation with the BAPTA unit and that conjugation changes when the nitrogen lone pair turns out of plane during binding. 18 We reasoned that the binding affinity would be decreased by incorporating an electron-withdrawing group on the A ring, which would increase the conjugation with the amino group. 18 This in turn would disfavour the rotation out of plane that is required for binding to Ca 2þ and so lower the binding affinity, raising K d . 11 We wished to generate a ratiometric probe with a K d in the 50e100 mM range that could be used to investigate intra-organellar [Ca 2þ ]. Having considered the binding affinities of known Ca 2þ sensors 10,11,27 and correlations with Hammett s constants, 28 we decided to incorporate a nitro group in the BAPTA core. 29 We also decided to incorporate an azido tag in Nitro-AzidoFuraRed because it would offer a universal site of attachment by bioorthogonal chemistry 26 that could be used to incorporate a targeting group or biomolecule using copper-catalysed or strainpromoted azideealkyne cycloaddition (CuAAC 16,27,30 or SPAAC 31 ). The azido group would be attached through the thiohydantoin unit so that it would be distal from the BAPTA binding site to minimise potential interference with Ca 2þ binding. 16,27,30,31 In summary, we had designed NitroAzidoFuraRed to be a lowaffinity, ratiometric, Ca 2þ sensor excitable with visible light and displaying a large Stokes shift, which would have the potential for attachment to targeting groups and biomolecules. Herein, we show how it was synthesised and present its calcium ion binding properties. We also provide the first examples of CuAAC on thiohydantoin derivatives.

Results and discussion
There have been no published syntheses of low affinity Fura-Red indicators. Our approach was to construct the BAPTA core 8 by adapting the route of Grynkiewicz et al. 9 Starting from hydroquinone 1, benzyl protection gave bisether 2, which was then converted into the nitro derivative 3. Selective deprotection of the ortho-benzyl group was achieved using aluminium trichloride instead of TFA to give phenol 4 in good yield. 32 Coupling phenol 4 with known bromide 6, prepared from phenol 5, 33 gave bis-nitro compound 7 in quantitative yield. Reduction of the nitro groups in the presence of the benzyl ether was achieved using iron rather than Pt/H 2 . The use of acetone as a co-solvent was necessary for solubility and critical to the success of this reaction. The resulting diamine was then alkylated to give the tetraethyl ester 8 in excellent yield. Selective formylation of the more electron rich ring gave aldehyde 10, which was deprotected by hydrogenolysis to yield phenol 11. Construction of the benzofuran unit by alkylation with the diethyl acetal of bromoacetaldehyde followed by acid-induced cyclisation 34 proved capricious, so we decided to investigate allylation of the phenol, followed by oxidative cleavage and cyclisation. This route to benzofuran-2-ylcarboxaldehydes is new and there was only one literature example 35 of an allyl group being cleaved in the presence of a tertiary amine. The literature example employed ozonolysis, but the use of osmium tetroxide/sodium periodate followed by acid proved satisfactory. Nitration of the A ring has to be carried out after construction of the BAPTA unit and deactivation of the B ring by the aldehyde to take advantage of the directing effect of the amino group, 14,29,36 and experimentation showed that it was best performed on the allyl ether 12. Oxidation of the nitrated compound 13 and cyclisation then gave the desired benzofuran 14 in modest yield over the two steps (Scheme 1).
The final part of the synthesis required the incorporation of the thiohydantoin unit (Scheme 2). 3-Azidopropylamine 15 was prepared by the literature method, 37 converted into an isothiocyanate and reacted with the methyl ester of glycine to give thiohydantoin 16. 38 Knoevenagel condensation 39 with aldehyde 14 completed the fluorophore 17 and saponification of the esters gave NitroAzidoFuraRed.
With the desired compound in hand we next investigated the optical properties of the new sensor. As expected, the UV/Vis absorption spectra of the Ca 2þ free and Ca 2þ bound Nitro-AzidoFuraRed are very similar to Fura-Red showing a w25 nm blue shift in the absorption maximum upon Ca 2þ binding (Fig. 2). The fluorescence spectra of the Ca 2þ free and Ca 2þ bound Nitro-AzidoFuraRed are also like Fura-Red showing the expected emission maximum at w630 nm.
Having established that the probe possessed the desired optical properties, we turned our attention to the binding properties. The excitation spectra of NitroAzidoFuraRed were obtained with different concentrations of Ca 2þ and showed a good isobestic point at 452 nm (Fig. 3). The K d of the probe was found to be 90 mM using this titration and the ratio of fluorescence intensities at 630 nm when excited at 420 nm and 485 nm, following the method of Grynkiewitz et al. 9 With a sensor with desirable properties in hand, we tested the conjugation reaction. A model thiohydantoin derivative 18, prepared from thiohydantoin 16, was used to test the CuAAC with alkyne-tagged phenylalanine derivative 20 (Scheme 3). However, mixtures immediately turned red upon mixing the thiohydantoin     with copper(I) ions and the click reaction was ineffective. This is consistent with the reported formation of a thiohydantoin-copper complex. 40 The tetracarboxylate BAPTA-based fluorophores are cell-impermeable and are generally caged as acetoxymethyl (AM) esters so that they can cross the plasma membrane. 41 The AM esters are hydrolysed rapidly by esterases inside cells to give the active Ca 2þ sensors. This strategy has been used for FuraRed, 11 which also includes N-acetylmethyl caging of the thiohydantoin group. Since NitroAzidoFuraRed is designed to be used in cells, we investigated whether AM-protection of the model thiohydantoin would allow CuAAC. Happily, conversion of thiohydantoin 18 into the AM derivative 19 was followed by smooth CuAAC to give the triazole adduct 21, a process that occurred without the dramatic colour change observed previously.
We then decided to illustrate the conjugation reaction for NitroAzidoFuraRed itself. Proteins to be retained in the endoplasmic reticulum (ER) of cells carry the peptide KDEL motif so that they are recovered from other parts of the cell and returned to the ER. 42 Inclusion of this retention sequence has been used as a way of localising compounds and nanoparticles in the ER. 43 Therefore, we exemplified the conjugation of NitroAzidoFuraRed by AM protection followed by CuAAC with an FFKDEL peptide functionalised at the N-terminus as the amide of 4-pentynoic acid. Under the conditions tested above, this gave the penta-AM protected triazole in 42% yield.

Conclusion
In conclusion, we have prepared a Ca 2þ sensor that is ratiometric, absorbs and emits at long wavelength, has a large Stokes shift, is tuned to low affinity for detection of changes in [Ca 2þ ] in intracellular stores and has an azido tag for conjugation to allow localisation in cells. Nitration to introduce the binding-modulator had to be carried out after the synthesis of the BAPTA unit and the introduction of an electron-withdrawing aldehyde, but before introducing the benzofuran moiety. The benzofuran was then constructed using a novel allylationeoxidationecyclisation. Finally, the azido-tagged thiohydantoin was introduced by Knoevenagel reaction. We then demonstrated the first CuAAC of a thiohydantoin derivative, showing that AM protection is critical to its success, and demonstrated its potential in bioconjugation by attaching an FFKDEL peptide, which incorporates the ER protein retention sequence.

General
All reactions under an inert atmosphere were carried out using oven-dried or flame dried glassware. Solutions were added via syringe. Dichloromethane and acetonitrile were dried where necessary using a solvent drying system, Puresolv TM, in which solvent is pushed from its storage container under low nitrogen pressure through two stainless steel columns containing activated alumina and copper. Reagents were obtained from commercial suppliers and used without further purification unless otherwise stated. Flash column chromatography was carried out using Fisher matrix silica 60 or using a Biotage Isolera one automated system. 1 H and 13 C NMR spectra were obtained on a Bruker AVIII/500 spectrometer operating at 500 and 125 MHz, respectively or a Bruker AVIII/400 spectrometer operating at 400 and 100 MHz, respectively. All coupling constants are measured in Hz. DEPT was used to assign the signals in the 13 C NMR spectra as C, CH, CH 2 or CH 3 . EI and CI mass spectra were obtained using the (M Station) JEOL JMS-700 spectrometer. ESI spectra were collected on a Bruker MicroTOF-Q. Infrared (IR) spectra were obtained on a Shimadzu FTIR-8400S spectrometer using attenuated total reflectance (ATR) so that the IR spectrum of the compound (solid or liquid) could be directly detected (thin layer) without any sample preparation.  31 Nitric acid (70%) (19.5 ml, 437.4 mmol, 2.0 equiv) was added slowly to a suspension of hydroquinol 2 (63.5 g, 218.7 mmol, 1.0 equiv) in glacial acetic acid (400 ml) at 0 C. The solution was stirred at 0 C for 3 h, poured into water (w1500 ml) and the resulting precipitate filtered off. The precipitate was washed with water (1500 ml), dissolved in CHCl 3 (300 ml), dried over magnesium sulfate and concentrated under vacuum to give nitro compound 3 as a yellow solid. (73.04 g, 100%). (3.99 g, 28.7 mmol, 1.00 equiv), 1,2-dibromoethane (7.5 ml, 87 mmol, 3.0 equiv) and K 2 CO 3 (4.36 g, 31.6 mmol, 1.10 equiv) were combined in DMF (6.0 ml) and stirred at 120 C for 3 h under argon.