A novel copper-chelating strategy for fluorescent proteins to image dynamic copper fluctuations on live cell surfaces† †Electronic supplementary information (ESI) available. See DOI: 10.1039/c4sc03027c

A strong but selective copper-binding tripeptide was employed to develop a highly sensitive and selective copper(ii) protein reporter.


Introduction
Copper is an essential metal ion in most aerobic organisms, participating in many critical biological processes including energy generation, oxygen transport, cellular metabolism and signal transduction. 1 However, when unregulated, copper can become toxic by generating reactive oxygen species (ROS) via Fenton-type reactions. 2 The adverse effect of dysregulated copper homeostasis in humans is well illustrated in Menkes and Wilson's diseases, which are related to copper deciency or overload, respectively. 3 To address copper transport and traf-cking in living systems, a variety of synthetic and genetically encoded probes have been developed with a focus mainly on visualizing intracellular copper, 4 where Cu + is the major form due to reduction upon cellular uptake and the reducing environment of the cytosol. 5 In contrast, extracellular copper trafficking has been insuf-ciently investigated due to a lack of appropriate tools for monitoring in living systems. In the more oxidizing extracellular media, Cu 2+ is favored over Cu + . 6 This oxidized form of copper has long been linked to many neurodegenerative diseases such as Alzheimer's disease, prion diseases, and Parkinson's disease through its participation in the aggregation of brogenic proteins. 6 The protein-Cu 2+ interactions are also directly involved in a catalytic cycle of ROS production, which exerts oxidative damage to cells. 7 Moreover, exogenous copper plays a key role in neuronal signaling pathways by modulating the synaptic plasticity. 8 Therefore, monitoring dynamic changes in the copper concentration in the extracellular environment will be critical for understanding the mechanisms of Cu 2+ release and its physiological functions in the brain. In particular, imaging Cu 2+ uctuations on specically targeted cell surfaces will greatly improve our insight into cellular responses to outside copper and copper-based cell-to-cell communication. Although a few cellular targeting strategies have been devised for synthetic metal reporters using organelle-targeting moieties, 9 they cannot direct the probes to specic cells of interest. Furthermore, mislocalization of small-molecular probes inside cells might lead to controversy in data interpretation. 10 In comparison, genetically encodable protein reporters have major advantages for the cell-specic and surface-specic detection of target molecules.
Several genetically encoded reporters have been applied for imaging copper in cellular context. 11 However, they are limited by relatively small responses (15-50%), 11b-e low copper selectivity, 11a,11e or reversed responses in mammalian cells compared with the response in vitro and in living bacteria. 11b Another challenge posed by imaging the extracellular pool of copper in particular is to provide the necessary binding affinity for Cu 2+ sensors to compete with copper-binding molecules in the extracellular medium. While extracellular Cu 2+ is present at micromolar concentrations (10-25 mM in blood plasma, 0.5-2.5 mM in cerebrospinal uid (CSF), and 30 mM in the synaptic cle), 6 the metal ion is rather bound to highly abundant copperbinding proteins or small molecular ligands than present as a free ion. 12 For example, human serum albumin (HSA), found at $600 mM in plasma, binds Cu 2+ with a picomolar binding affinity. 13 The HSA concentration in CSF is relatively low ($3 mM) but represents 35-80% of the total CSF proteins. 14 Developing protein-based Cu 2+ sensors with strong but selective copper-binding properties will require potent Cu 2+chelating strategies within proteins. Previously reported sensors have employed simple histidines or a metal-chelating unnatural amino acid (3,4-dihydroxy-L-phenylalanine) to bind Cu 2+ since the metal ion prefers to bind nitrogen or oxygen donors. 11a,15 Unfortunately, reported binding affinities (K D 0.2-100 mM) of these sensors under physiologically relevant conditions have been still insufficient for imaging extracellular Cu 2+ , and it has not been demonstrated whether they could detect changes in copper concentration under conditions where competing copper-binding biomolecules are present.
Here, we report a new genetically encoded uorescent reporter that binds Cu 2+ with a K D of $8 nM while maintaining its selectivity for the metal ion. Green uorescent protein (GFP) was engineered to contain a natural short but strong Cu 2+binding tripeptide, the amino terminal copper-and nickelbinding (ATCUN) motif, which is also the major copper-binding site of albumins in many species including humans. 16 The location of this copper binding sequence in GFP was carefully optimized to provide maximum uorescence quenching, strong Cu 2+ binding, and stable protein folding in cells. The resulting reporter was sensitive enough to detect Cu 2+ even in the presence of equimolar HSA in vitro. Furthermore, our reporter was stably displayed on cell surfaces and showed specic and reversible responses to extracellular Cu 2+ .

Results and discussion
The ATCUN motif consists of three amino acids (Xaa-Xaa-His) with a free N-terminus, where Xaa is any amino acid that can provide an amide nitrogen. As depicted in the X-ray crystal structure, this motif binds Cu 2+ by forming a square-planar complex with the a-NH 2 -terminal nitrogen, the two amide nitrogens, and the imidazole nitrogen in histidine. 17 Although the peptide binds Cu 2+ (K D $ 1.18 Â 10 À16 M) 18 and Ni 2+ (K D $ 10 À16 M) 19 with similarly high affinities, it is highly context dependent, such that the ATCUN in HSA was reported to bind Ni 2+ with a nearly ve orders of magnitude weaker binding affinity than Cu 2+ . 20 Considering its small size and strong binding affinity for Cu 2+ , we hypothesized that the ATCUN motif can serve as a powerful Cu 2+ -binding site for protein sensors.
To place the ATCUN motif in close proximity of the GFP chromophore, thereby inducing maximum uorescence quenching by Cu 2+ binding, the protein was circularly permuted to generate new N-and C-termini at the beginning of b-strand 7. Circular permutations on b-strand 7 have shown relatively minimal effects on the GFP uorescence. 21 The ATCUN motif (Gly-Gly-His) was then inserted into the N-terminal region in a manner in which the histidine at the third position replaced histidine 148 near the phenolic ring of the chromophore ( Fig. 1 and Table 1). To provide an a-NH 2 -terminal nitrogen for ATCUN in the circularly permuted GFP, a free N-terminus at the rst glycine of the motif was precisely generated by intein-based protein cleavage. 22 N-terminal sequencing of the intein-cleaved protein conrmed proper insertion of the ATCUN motif ( Fig. S1 †). For fabrication of the sensor protein, we employed superfolder GFP (OPT) 23 for higher stability in the assay buffer (100 mM HEPES, pH 7.0, 300 mM NaCl) and enhanced folding in E. coli as well as in mammalian cells (data not shown).
Fluorescence spectra of the circularly permuted OPT (cpOPT) with the ATCUN insert were acquired by exciting at 480 nm and collecting the emission signals at 500-600 nm. Upon the addition of Cu 2+ , the uorescence intensity of the sensor protein clearly decreased in a concentration-dependent manner (Fig. 2a). The decrease in the uorescence signal (at 510 nm) was up to <5% of the apo-protein uorescence (95% quenched). Compared with wild-type OPT (wtOPT) and cpOPT, the ATCUNfused cpOPT (GCS-1) exhibited a considerably higher degree of quenching over a wide range of Cu 2+ concentrations (Fig. 2b). Although cpOPT displayed noticeable uorescence quenching at high Cu 2+ concentrations (>10 mM), introduction of the ATCUN motif increased the Cu 2+ sensitivity of the protein. In addition, longer incubation of the ATCUN-fused cpOPT at low Cu 2+ concentrations further enhanced uorescence quenching, which was likely due to the rather slow Cu 2+ binding by ATCUN in cpOPT (Fig. 2c). We named the sensor protein as geneticallyencoded copper(II) sensor (GCS)-1 (see Table 1 for a summary of N-terminal sequences in the GCS proteins).
To further optimize the Cu 2+ responses, the location of the ATCUN motif in cpOPT was varied around H148 in GCS-1. ATCUN insertion at the À1 position of H148 ((À1)GCS-1, 1) resulted in a sensor protein with Cu 2+ binding responses similar to those of GCS-1 but a poor protein yield (data not shown). In contrast, cpOPT with the ATCUN motif at the +1 position of H148 (GCS-2) showed rapidly saturated uorescence quenching even at low concentrations of Cu 2+ (Fig. 2b and c), possibly indicating a higher Cu 2+ binding affinity. Although the degree of quenching was slightly lower for GCS-2 (<15% of apoprotein uorescence, >85% quenched), the binding response of GCS-2 was signicantly faster (within 10 s) than that of GCS-1 (>20 min, Fig. S2 †). The improved response rate of GCS-2 might be due to an increased exibility of the ATCUN motif in the protein. Unlike GCS-1, the ATCUN motif in GCS-2 does not include H148 which is known to interact with neighboring bstrands. 23a Thereby the ATCUN tripeptide of GCS-2 can be more open for Cu 2+ binding, exhibiting a faster association. On the other hand, H148 is also in a hydrogen-bonding network that stabilizes the chromophore. 24 Conformational changes of H148 and subsequent effects on the chromophore by Cu 2+ binding might be less signicant in GCS-2 than in GCS-1. The absorbance spectrum of GCS-1 was altered more strongly by Cu 2+ binding than that of GCS-2 (Fig. S3 †), possibly indicating stronger inuences on the chromophore and thereby a higher degree of quenching response in GCS-1.
To demonstrate that the ATCUN motif is indeed the major Cu 2+ binding site in GCS-2, we constructed two GCS-2 variants, G-GCS-2 and GCS-2-G ( Table 1). The ATCUN motif in G-GCS-2 was disrupted by adding a glycine residue to the N-terminus of GCS-2. Upon the addition of Cu 2+ , G-GCS-2 showed a severely diminished quenching response (Fig. 2d). On the other hand, the uorescence intensity of GCS-2-G, in which the ATCUN motif is intact but H148 was mutated to glycine, was still reduced to <30% of the apo-protein uorescence (>70% quenched) under the same conditions, supporting that the ATCUN motif is primarily responsible for Cu 2+ binding (Fig. 2d). In addition, titration with Cu 2+ revealed a 1 : 1 binding stoichiometry between GCS-2 and Cu 2+ (Fig. S4 †), suggesting the presence of a single dominant metal binding site (such as the ATCUN tripeptide) in GCS-2. However, it should be noted that H148 of GCS-2 also plays a partial role in Cu 2+ binding and the resulting uorescence quenching since the Cu 2+ -mediated response of GCS-2-G was smaller than that of GCS-2.
The metal selectivity of both GCS proteins was subsequently investigated by monitoring changes in uorescence intensity in the GCS proteins when other biologically relevant metal ions (Cr 3+ , Mn 2+ , Fe 3+ , Co 2+ , Ni 2+ , Zn 2+ , Ag + , Cd 2+ , Hg 2+ , Pb 2+ , K + , Ca 2+ , and Mg 2+ ) were present. Most of these metal ions had no effect on the uorescence quenching by Cu 2+ (Fig. 3a and b). Although Ni 2+ caused a small uorescence change for GCS-2, subsequent Cu 2+ addition readily induced uorescence quenching (Fig. 3b). The ATCUN motif in our GCS sensor proteins likely binds to Ni 2+ more weakly than to Cu 2+ , which is similar to the ATCUN motif in HSA. 20 The effects of Cu + were examined in a buffer containing a 50-fold molar excess of L-ascorbic acid. L-ascorbic acid had no inuence on the uorescence intensities of the GCS proteins (Fig. S5 †) or Cu 2+ binding of the ATCUN motifs. 25 Cu + addition showed no change in uorescence for both GCS proteins (Fig. 3c). The reversibility of the uorescence quenching of the sensor protein was also tested, and uorescence signals from the Cu 2+ -bound GCS proteins were recovered upon addition of metal-chelating ethylenediaminetetraacetic acid (EDTA, Fig. 3d).
The Cu 2+ -binding affinities of the GCS proteins were determined by measuring uorescence changes upon titration with copper. 15b All assays were performed at relatively high buffer and salt concentrations (100 mM HEPES, pH 7.0, 300 mM NaCl), which might reect various extracellular media. 26 The dissociation constant for GCS-2 was calculated to be 7.93 AE 3.20 nM (Fig. 4a), which is the strongest Cu 2+ binding among the currently available protein-based Cu 2+ sensors (Table S1 †). The use of a natural copper(II)-specic tripeptide provides a strong chelating strategy for robust Cu 2+ capture near the GFP chromophore, presenting high sensitivity for the metal ion while not compromising the metal selectivity of the GCS proteins. GCS-1 was found to bind Cu 2+ more weakly (94.04 AE 19.33 nM) as discussed above (Fig. S6 †).  To test whether the binding affinity of the GCS proteins for Cu 2+ is high enough to monitor the metal ion even in the presence of other copper-binding molecules, the GCS responses to Cu 2+ were measured in the presence of various amounts of HSA. GCS-2 showed Cu 2+ binding properties comparable to HSA, where >40% of the added Cu 2+ was bound to GCS-2 in the presence of equimolar HSA (Fig. 4b). Considering that the reported dissociation constants for HSA are in the picomolar range, 13 Cu 2+ binding to GCS-2 should be weaker by at least three orders of magnitude. This discrepancy is likely due to the different experimental methods used for K D measurements, components/concentration of buffers, pH, or ionic strengths. 27 Nevertheless, under our conditions, an optimized ATCUN tripeptide in GCS-2 showed nearly as strong binding as the ATCUN motif in HSA. In contrast, the uorescence intensity of GCS-1 began to decrease only aer the concentration of Cu 2+ was high enough to ll in most of the ATCUN sites in HSA (Fig. 4c). Therefore, GCS-1 was not able to effectively compete with HSA. We also performed competing experiments with an amyloid beta peptide (Ab40), another copper-binding molecule (K D 0.47-30 mM) which plays a key role in the pathogenesis of Alzheimer's disease. 28 Ab40, which has a rather weak Cu 2+ -binding ability, was outcompeted by both GCS proteins (Fig. S7 †). Taken together, our data show that GCS-2 (K D $ 8 nM) is a highly sensitive reporter that can efficiently compete with other highaffinity copper-binding molecules in extracellular media. In comparison, GCS-1 (K D $ 94 nM) might be useful for applications in which low-affinity reporter proteins are required.
Finally, we used the GCS proteins to detect dynamic Cu 2+ uctuations on the surfaces of live mammalian cells. A signal peptide for the secretory pathway and the transmembrane domain of the platelet-derived growth factor receptor were fused to GCS-2 for display on the extracellular plasma membrane. The signal peptide is cleaved during sensor protein synthesis in the endoplasmic reticulum, generating the ATCUN motif precisely at the N-terminus for effective Cu 2+ binding. HeLa cells expressing the GCS-2 construct were monitored with a total internal reection uorescence (TIRF) microscope for selective visualization of the cell surface uorescence. Upon Cu 2+ addition (50 mM), the GCS-2 uorescence intensity on the cell surface was rapidly decreased within $1.5 min, and it was saturated to 47% (average from 13 cells in the same well, 53% quenched) of the initial signal aer 3 min (Fig. 5). EDTA was then added to the cells, and the GCS-2 uorescence intensity was recovered to 97% within 4 min, which is in agreement with the in vitro results. GCS-1 that was expressed on HeLa cell surfaces provided uorescence changes similar to those of GCS-2 against extracellular Cu 2+ (data not shown).  To conrm that the decrease in uorescence of GCS-2 was specic for Cu 2+ binding, Zn 2+ , another crucial metal ion in the brain, was added to the cells (50 mM) before Cu 2+ treatment.
Interestingly, the uorescence intensity of several cells of the 13 monitored cells was slightly increased (<115%) by Zn 2+ (Fig. 6). This could result from structural stabilization of the b-barrel of   the GCS-2 protein under relatively high zinc concentrations as recently reported for a red uorescent mKate2 mutant. 11a Zinc(II)-mediated changes in protein distribution in the plasma membrane might also affect the uorescence intensity of GCS-2 as observed using TIRF imaging. Nevertheless, all cells exhibited rapid uorescence quenching upon Cu 2+ addition (Fig. 6). The copper(II)-specic uorescence decrease for GCS-2 was further conrmed by its co-expression with mCherry, a red uorescent protein, on the cell surfaces. Addition of 10 mM CuCl 2 decreased the uorescence intensity of GCS-2 but not that of mCherry, which again shows the applicability of GCS-2 for specic Cu 2+ imaging on the surface of live cells (Fig. 7). Moreover, the uorescence signals of cpOPT proteins on cell surfaces were not altered by 10 mM Cu 2+ (Fig. S8 †).

Conclusions
The GCS proteins demonstrate that the natural and small Cu 2+ binding ATCUN motif can be successfully utilized as a strong and specic copper chelating site for genetically encoded Cu 2+ reporters. The tight Cu 2+ binding of GCS-2 allows for reliable Cu 2+ monitoring in the presence of HSA, one of the major Cu 2+ binding biomolecules in extracellular environments. The sensor proteins also demonstrate high selectivity for Cu 2+ compared with other biologically relevant metal ions. Imaging Cu 2+ on the surface of live cells with GCS-2 showed a quick and reversible response, which demonstrates the possibility of monitoring oscillating Cu 2+ levels in living systems. The copper-chelating strategy described herein has multiple advantages. Taken from one of the major players in extracellular copper trafficking, the ATCUN motif may enable noninvasive detection of dynamic changes in copper concentration in biological systems. The motif is composed of only three amino acids, creating minimal disruption in the uorescent protein's structure. To expand the applicability of GCS sensors to intracellular Cu 2+ imaging, strategies to generate ATCUN motifs and protect these motifs from various post-translational modications 29 inside cells should be developed. Nonetheless, we envision that GCS proteins can be employed to monitor Cu 2+ release on neuronal cells. Specically, copper release in the synaptic cle which is induced by chemical or biological cues could be detected by GCS proteins expressed on pre-or postsynaptic cell surfaces. Further studies will be aimed at improving the membrane targeting efficiency and applications of red uorescent proteins that enable Cu 2+ imaging in live animals.