Biologically targeted probes for Zn 2+ : a diversity oriented modular “ click-S N Ar-click ” approach † EDGE ARTICLE

We describe a one-pot strategy for the high yielding, operationally simple synthesis of ﬂ uorescent probes for Zn 2+ that bear biological targeting groups and exemplify the utility of our method through the preparation of a small library of sensors. Investigation of the ﬂ uorescence behaviour of our library revealed that although all behaved as expected in MeCN, under biologically relevant conditions in HEPES bu ﬀ er, a plasma membrane targeting sensor displayed a dramatic switch on response to excess Zn 2+ as a result of aggregation phenomena. Excitingly, in cellulo studies in mouse pancreatic islets demonstrated that this readily available sensor was indeed localised to the exterior of the plasma membrane and clearly responded to the Zn 2+ co-released when the pancreatic beta cells were stimulated to release insulin. Conversely, sensors that target intracellular compartments were una ﬀ ected. These results demonstrate that this sensor has the potential to allow the real time study of insulin release from living cells and exempli ﬁ es the utility of our simple synthetic approach.


Introduction
Zinc is now rmly established as an essential trace element, widely found in proteins where it undertakes both structural and catalytic roles. 1 In addition to these xed forms of biological zinc, it is known that pools of loosely bound zinc, oen referred to as "free" or "mobile", 2 are found in the pancreas, 3 central nervous system, 4 prostate, 5 intestine 6 and retina. 7 The biological trafficking of zinc has been linked to many disease states in these vital organs, including Alzheimer's disease, 8 epilepsy, 9 ischemic stroke, 10 infantile diarrhoea, 6 age related macular degeneration, 7 prostate cancer 11 and type 2 diabetes. 3 For example, polymorphisms in the SLC30A8 gene, encoding the secretory granule zinc transporter ZnT8 in pancreatic islet bcells, are directly associated with type 2 diabetes. 3 However, although the importance of mobile zinc in human health is now well known, 12 its role in the physiological changes leading to the onset and progress of these disease states is not yet fully understood, to a large extent due to our current inability to monitor changes in mobile zinc at high resolution in specic biological space. 13 There is therefore a major drive to develop molecular probes which are able to map Zn 2+ levels in cells and organs via the chelatable mobile pools of zinc present.
Small molecule uorescent probes for Zn 2+ consisting of a chelating unit, attached (directly or indirectly) to a uorescent moiety can deliver a rapid, strong output signal in response to binding of Zn 2+ at the chelation site, even at very low concentrations of the metal ion. 14 A wide range of these probes have now been reported that display an array of metal-binding and uorescent properties and which, in principle, can be further tuned through rational synthetic modication. 15 We previously reported probe 1, consisting of a metal chelating tetra-azamacrocycle, linked to a uorophore by a triazole moiety (Fig. 1a), 16 which shows a high selectivity for Zn 2+ over other biologically relevant metals in a suitable pH range. We demonstrated the utility of probe 1 in vivo in zebrash (Fig. 1b). 17 However, although 1 and other notable rst generation small molecule Zn 2+ probes display high selectivity for zinc, 15 they are limited when judged against the requirements of biological imaging, in particular the need to target specic cellular and sub-cellular domains for optimal in vivo imaging, as well as the need to tailor the metal binding unit in order to detect mobile zinc at a wide range of biologically relevant concentrations. 14d To overcome this, a limited number of biologically targeted small molecule uorescent probes for Zn 2+ have been developed. 18,19 However, these probes are typically time-consuming and challenging to prepare (Fig. 2), and this, combined with the unpredictable nature of biological systems that necessitates a heuristic approach to probe optimisation, presents a signicant barrier to their application. Alternatively, targeted recombinant probes have been shown to permit measurement of Zn 2+ at the subcellular level. 20 However, the deployment of these probes in living cells is oen limited by the need for virus-mediated or transgenic approaches to allow gene delivery. 19c To facilitate the rapid development of targeted small molecule probes for Zn 2+ they should ideally be accessible in a synthetically exible and succinct manner such that their properties can be optimised quickly and easily. Herein we describe such an approach; a diversity-oriented modular, onepot synthetic method for the realisation of biologically targeted Zn 2+ sensors and demonstrate its utility through the synthesis of a small library of sensors with a selection of metal binding and biological targeting units. The selectivity and sensitivity of the sensors were studied to conrm that the addition of the targeting moiety did not signicantly alter zinc binding and uorescent behaviour. We showcase the potential of our approach for the development of new sensors for biological applications through cell-resolution Zn 2+ imaging studies in living pancreatic islets.

Results and discussion
To develop an expedient, modular approach to targeted sensors we decided to build on the success of sensor 1, the synthesis of which employed a high yielding Cu-mediated alkyne-azide cycloaddition (CuAAC) reaction, oen simply called the "click" reaction. 21,22 The generality of this reaction, its operational simplicity and the availability of suitable alkyne and azide starting materials has rapidly elevated it to a method of choice for the synthesis of complex non-natural molecular targets 23 since it was rst disclosed independently by Sharpless and Fokin, and Meldal in 2002. With a view to increasing the operational simplicity still further, we decided to develop a onepot iterative methodology in which the targeting unit and the metal binding moiety are combined sequentially with a readily available uorophore core.
Development of a click-S N Ar-click approach for the synthesis of targeted probes for Zn 2+ One-pot iterative click approaches have previously been developed that exploit the differential reactivity of dissimilar azide or alkyne functionality, 24 or in which the second alkyne or azide component is formed in situ. 25 The latter typically rely on the use of masked alkyne units such as silyl-protected acetylenes, 26 or masked azides in the form of leaving groups at sp 3 centres. 27,28 Instead we decided to focus on a novel S N Ar approach that makes use of the unusual reactivity of the electron-decient naphthalimide core of 1. 29 Building block 2 was synthesised in high yield (96%) in one step from commercially available 4-bromo-1,8-naphthalic anhydride. We optimised our iterative approach using benzyl azide and phenyl acetylene as model substrates. Acetylene 2 readily participated in the click reaction with benzyl azide in NMP in the presence of CuI. Subsequent addition of NaN 3 to the reaction mixture led to the S N Ar displacement of the bromide leaving group to give the required azide functionality for the second click reaction. The second coupling step proved more problematic. Firstly, the NaBr by-product of the S N Ar reaction was revealed to inhibit the click reaction in NMP, leading to slower reactions and signicant levels of decomposition compared with the stepwise process. Rather than using silver salts to scavenge the bromide ions, this problem was overcome though the addition of EtOH as a co-solvent during the second click reaction, which presumably solvates the bromide anion. Secondly, even with this modication the nal click reaction proved sluggish at room temperature. This was overcome by the addition of NaOAc as a base to accelerate the reaction. 30 Combining these results led to conditions under which the click-S N Ar-click sequence could be performed in one-pot over a 40 h period at room temperature through simple sequential addition of reagents with purication only performed aer the nal step (Fig. 3). Gratifyingly, in addition to being operationally trivial, and time and reagent efficient, the isolated yield of the one pot procedure signicantly exceeded that of the stepwise process with model compound 4 produced in 85% yield compared with 41% with isolation and purication aer each step.
With an effective iterative procedure in hand we investigated its utility through the synthesis of targeted sensors for Zn 2+ based on metal binding units previously reported to confer a selective response to zinc 15,16 and a selection of readily available biological targeting units. Accordingly, starting from 2 we synthesised sensors 6-8 that are designed to interdigitate in the cell membrane through the lipophilic di-dodecylamide motif, 13 but in which the metal binding unit is varied between dipicolylamine (6), aminodiacetate (7) and cyclam (8). Having demonstrated the exibility of our approach with respect to the metal binding unit we investigated simple variations in the targeting unit through the synthesis of sensors bearing a phosphonium salt previously demonstrated to target the mitochondria (9) 19d and a simple bis-amide, which has been shown to accumulate in the lysosomal space (10). 19e Finally, given that a wide range of propargylated biological motifs are available, in particular propargyl glycosides, 31 we extended our iterative approach to azide functionalised building block 3 which is readily synthesised from 4-bromonapthalic anhydride (2 steps, 79% overall yield) and evaluated its behaviour under our optimised to conditions. Pleasingly, 3 was converted to model compound 5 by iterative one-pot reaction with two portions of phenyl acetylene in excellent yield. Using these conditions, sensor 11, which bears a representative protected glycoside, was produced in excellent yield, further extending the utility of our iterative click approach.
All sensors were isolated in good to excellent yield over 3 steps; 61-82% overall, which corresponds to an average of 85-92% for each chemical step. These results clearly show that our iterative method can be used to prepare sensors efficiently and rapidly by combining biological targeting groups bearing either an alkyne or an azide with uorescent cores 2 or 3 respectively, and metal binding domains functionalised with an acetylene handle.

Fluorescent properties of our targeted Zn sensors
The combinations of metal binding unit and uorophore in sensors 6-11 are based on previously reported selective sensors for Zn 2+ with an expected 1 : 1 binding stoichiometry. In order to conrm that our novel sensor constructs maintain these selective properties with the inclusion of a targeting unit we investigated their response to Zn 2+ in both MeCN and HEPES buffer (Fig. 4). Where necessary, the metal binding domains and biological targeting units were deprotected using standard procedures prior to testing (see ESI † for full details).
In all cases, the sensors responded to 1 equivalent of Zn 2+ in MeCN as expected with switch-on uorescent enhancements of between 2-and 4.5-fold observed, and little or no further enhancement with the addition of further equivalents of Zn 2+ (Fig. 4a). [13][14][15] Similarly, with the exception of sensors 6 (vide infra) and 7, which precipitated on addition of Zn 2+ , the probes responded as expected in HEPES buffer, producing similar or enhanced on/off ratios to those observed in MeCN (Fig. 4b).
Dissociation constants (K d ) were determined by non-linear regression analysis of titrated samples of sensors 8-11, in HEPES assuming a 1 : 1 binding stoichiometry. Although reasonable values of K d were obtained when compared to those reported for structurally similar ligands, 15r,16a,32 the goodness of t to a 1 : 1 binding model varied considerably (see ESI † for full details). Similar issues have recently been identied by Ceroni, Doddi and co-workers in their investigations of dendrimeric cyclam sensors for Zn 2+ , 33 and work is currently underway to determine the origin of these effects in our systems. In summary, these in vitro results clearly demonstrate that the selective switch on response of probes 6-11 to zinc is not affected by the inclusion of a biological targeting group.

Investigation of the anomalous behaviour of sensor 6
The behaviour of sensor 6 in HEPES buffer proved more complicated than that of the other sensors investigated; addition of further equivalents of Zn 2+ led to further signicant increases in uorescence up to a maximum value of 15-fold uorescence enhancement with 15 equiv. Zn 2+ (c.f. 5.1 for 1 equiv. of Zn 2+ ) at our standard sensor concentration of 50 mM. Along with the enhanced uorescence response, a red shi in the emission maximum was observed (23 nm and 31 nm at 1 and 15 eq. Zn 2+ respectively). Having observed that solutions of 6 in HEPES were slightly turbid and that their visual appearance changed on addition of Zn 2+ (Fig. 5a) we hypothesised that this unusual behaviour might be due to aggregation phenomena, driven by the surfactant-like nature (polar dipicolylamine head group, lipophilic aliphatic tail) of the sensor construct, which the addition of Zn 2+ could serve to modulate. 34 In keeping with this hypothesis, increasing the probe concentration to 100 mM led to an even more dramatic 35-fold increase in uorescence intensity in the presence of 15 eq. of Zn 2+ , while at a sensor concentration of 10 mM the response to excess Zn 2+ was reduced. Dynamic light scattering (DLS) experiments conrmed the presence of aggregates and, importantly, that the average size and polydispersity of the aggregates decreases signicantly on addition of Zn 2+ and that these aggregates remained even when probe concentration was signicantly reduced to 10 mM.

No such aggregation was observed in MeCN.
To examine this further, we investigated the effect of solvent composition on the uorescence of sensor 6, in mixtures of HEPES and MeCN. 34b In the absence of Zn 2+ , the uorescence output and emission wavelength of 6 remained largely unchanged as the v/v% of HEPES in MeCN was varied.
Conversely, in the presence of Zn 2+ a steady increase in uorescence output was observed up to 60% v/v HEPES, above which a signicant drop in the uorescence output was observed, which coincides with the red shi observed in the uorescence emission (Fig. 5b).
Thus, it appears that two factors play a role in the dramatic sensitivity of sensor 6: (i) as in all the other sensors synthesised, binding of Zn 2+ to the ligand unit deactivates the PET quenching process leading to a "switch on"; (ii) the aggregation of the unbound sensor initially serves to quench the emission but binding of Zn 2+ changes the aggregation such that uorescence is recovered. Importantly, when sensor 8b was examined by DLS, aggregation was not observed, highlighting the subtle role that small structural changes can have on sensor performance.
Finally, given the important role that aggregation plays in the performance of sensor 6 in vitro, and that the binding of other metal ions could possibly inuence aggregation signicantly leading to a loss of selective Zn 2+ sensing behaviour, we validated the selectivity of 6 for zinc ions. Pleasingly, only Cd 2+ and Hg 2+ produced a uorescent response with sensor 6, which is to be expected based on previous studies and is not biologically relevant. 15b All other metal ions tested led either to no change in uorescent output or quenching of the background uorescence. Importantly, probe 6 retains its switch on response for Zn 2+ in physiological buffer and over a physiological relevant range of pH.
In vitro application of sensor 6 to monitor exocytosis via the release of Zn 2+ in pancreatic islet cells The release of insulin from pancreatic b-cells is associated with the concomitant release of Zn 2+ -containing granules, which allows the changes in the concentration of Zn 2+ on the exterior of the plasma membrane to be used as a proxy for the release of insulin. This approach has been rened in recent work by Li and co-workers through the development of ZIMIR, a selective sensor for Zn 2+ that anchors on the extracellular face of the plasma membrane and can be used to monitor the response of the pancreatic cells to various stimuli with high spatial resolution. 18 The structure and in vitro behavior of sensor 6 bear remarkable similarities to ZIMIR in that both: (i) possess a surfactant-like structure which in the case of ZIMIR has been shown to localize the sensor to the extracellular face of the plasma membrane; (ii) display a large increase in uorescence output in the presence of excess Zn 2+ (70-fold in the case of ZIMIR, 35 fold in the case of sensor 6); (iii) demonstrate selectivity for Zn 2+ over other biologically available divalent metal ions. Given the utility of ZIMIR in monitoring the stimulated efflux of zinc, we decided to evaluate probe 6 in cellulo to see if it displayed similar membrane targeting behaviour.
Sensor 6 (30 mM) was incubated with pancreatic islet cells for 1 h, which were then washed under a steady ow of biological buffer for the remainder of the experiment to remove unbound sensor. In order to overcome the less favourable absorption/emission prole of sensor 6 compared with ZIMIR (l ex ¼ 493 nm, l em ¼ 518 nm vs l ex ¼ 347 nm, l em ¼ 443 nm), the cells were imaged using 2photon excitation (l ex ¼ 850 nm), which is fast-becoming the imaging modality of choice, to visualise the distribution of the 6 within the rst few islet layers. Aer incubation and irrigation with buffer, sensor 6 remained clearly localised on the cell surface, appearing to stain the plasma membrane in a manner similar to ZIMIR with little or no internalisation of the probe over the time scale of the experiment.
Having demonstrated that sensor 6 displays the expected cell-membrane targeting behaviour, we used epiuorescence microscopy to investigate the probe's response to changes in the extracellular concentration of Zn 2+ . Trivially, addition of extracellular Zn 2+ led to an increase in uorescence, conrming that 6 is indeed anchored on the extracellular face of the plasma membrane (Fig. 6b). More interestingly, when KCl was administered, to cause membrane depolarisation and release of insulin and Zn 2+ into the extracellular space, 35 sensor 6 exhibited a strong uorescence response (Fig. 6c) in a manner similar to that previously observed with ZIMIR. These results suggest that readily available sensor 6 could also be used to map the release of insulin at the cellular level but has the advantage of being signicantly less challenging to prepare than ZIMIR. Intriguingly, when the same experiment was attempted with sensor 8b, despite the structurally similarity to sensor 6 (polar metal binding unit, lipophilic tail), no increased uorescence response in the cell membrane was observed (see ESI †). This result, combined with the unexpected insolubility of the zinc complex of sensor 7, which could also be expected to act as a membrane targeting probe, demonstrates the "trial and error" nature of sensor design for biological applications which our diversity oriented approach can facilitate.
In order to further demonstrate that the behaviour of the membrane targeting probe 6 relies on it being localised on the extracellular face of the plasma membrane, and also that probes based on our new general sensor scaffold are capable of targeting intracellular organelles, we also examined probes 9b and 10b which were designed to target specic organelles in cellulo. Aer incubation, the cellular distributions of 9b and 10b were different to that of 6, and from one another ( Fig. 6a and b): both probes displayed an extranuclear localisation with a punctate appearance consistent with their presence in or on intracellular organelles. However, the appearance of islets incubated with sensor 10b, which is designed to target the lysosome, is qualitatively different to those incubated with mitochondria targeted probe 9b, with 10b localised in larger bodies. The selective and differential localisation of 9b and 10b was conrmed by colocalisation with Mitotracker and Lysotracker respectively (Fig. 7c). 36 In keeping with the complete internalisation of probes 9b and 10b, addition of extracellular Zn 2+ was shown to result in essentially no change in their uorescence response (Fig. 7d). Conversely, when pyrithione, which permits transport of Zn 2+ across the plasma membrane, and Zn 2+ were coadministered, both 9b and 10b showed a signicant increase in uorescence (Fig. 7e). Finally, we were also able to show that none of the probes were cytotoxic as incubation of mouse islets with probes 6, 9b and 10b does not signicantly induce cell death (necrosis) compared to DMSO-alone (see ESI †). Fig. 6 Fluorescence study of primary mouse islets of Langerhans incubated for 1 h with 6 (30 mM). (a) Two photon image (l ex ¼ 850 nm, l em ¼ 400-550 nm, max z-projection 80-100 mm); Response of sensor 6 over time to exogenous (b) 10 mM ZnCl 2 and (c) 20 mM KCl measured by epifluorescence microscopy. (d) Epifluorescence images demonstrating the response to Zn 2+ co-released with insulin from dense core secretory granules after stimulation with KCl (20 mM). Fig. 7 Epifluorescence study of sensors 9b and 10b (300 mM). Two photon image of sensor 9b (l ex ¼ 850 nm, l em ¼ 400-550 nm, max zprojection 80-100 mm); (b) as for (a) but sensor 10b. (c) comparison of co-localisation coefficients for sensors 9b and 10b with (i) Mitotracker and (ii) Lysotracker; (d) response over time of sensor 9b to exogenous (d) ZnCl 2 and (e) 100 mM ZnCl 2 co-administered with 5 mM pyrithione.

Conclusions
We describe here the development of a new range of targeted Zn 2+ sensors that can be efficiently synthesised in high yield in an operationally simple manner. We anticipate that this approach will signicantly accelerate the development of probes for application in biological systems. The importance of ease of synthesis in order to overcome the difficulties associated with their trial and error optimisation in cellulo was inadvertently exemplied through the disparate behaviour of sensors 6-8, all of which were originally anticipated to be suitable for staining the plasma membrane in biological systems. Testing of sensors bearing representative targeting units in pancreatic islet cells demonstrated not only their differential localisation but also their divergent responses to extracellular stimuli, suggestive of applications in the study of Zn 2+ trafficking in the subcellular space. In particular, sensor 6 allowed us to monitor insulin release in vitro in living pancreatic islet cells, using Zn 2+ as a proxy, in a manner analogous to ZIMIR. This is suggestive of great potential in the high-resolution imaging of insulin secretion, an important goal in the study of diabetes. Thus our modular synthetic approach will facilitate the optimisation and application of targeted Zn 2+ sensors for biology, allowing the manufacture of tailor-made probes, such 6, with applicability for the study of Zn 2+ -signalling and insulin exocytosis.