A Responsive Magnetic Resonance Imaging Contrast Agent for Detection of Excess Copper(II) in the Liver In Vivo

Abstract

The design, synthesis, and properties of a new gadolinium-based copper-responsive magnetic resonance imaging (MRI) contrast agent is presented. The sensor (GdL₁) has high selectivity for copper ions and exhibits a 43% increase in r₁ relaxivity (20 MHz) upon binding to 1 equiv of Cu²⁺ in aqueous buffer. Interestingly, in the presence of physiological levels of human serum albumin (HSA), the r₁ relaxivity is amplified further up to 270%. Additional spectroscopic and X-ray absorption spectroscopy (XAS) studies show that Cu²⁺ is coordinated by two carboxylic acid groups and the single amine group on an appended side chain of GdL₁ and forms a ternary complex with HSA (GdL₁–Cu²⁺–HSA). T₁-weighted in vivo imaging demonstrates that GdL₁ can detect basal, endogenous labile copper(II) ions in living mice. This offers a unique opportunity to explore the role of copper ions in the development and progression of neurological diseases such as Wilson’s disease.

Graphical Abstract

1. INTRODUCTION

Copper is the third most abundant transition metal in the body and a required dietary nutrient. The average healthy human has a total of ~110 mg of copper in tissues.^1–4 Copper is typically bound to specific proteins and enzymes where it plays fundamental catalytic and structural roles.^5–7 Copper has also been associated with signaling events in the brain.^8–10 In a biological environment, copper is present in two oxidation states, the cuprous (Cu⁺) and cupric (Cu²⁺) ions.¹¹ Typically, total extracellular Cu²⁺ can vary widely from nM to μM, while intracellular Cu⁺ can vary from μM to mM.^12–14 Due to its redox properties, copper homeostasis is tightly regulated in cells and disruption of this is associated with a number of neurodegenerative diseases including Alzheimer’s, Parkinson’s, prion, familial amyotrophic lateral sclerosis, Menkes and Wilson’s diseases.^15–18 For instance, genetic mutations of copper-transporting proteins ATP7A and ATP7B result in afflictions of systemic brain copper deficiency in Menkes disease and hyperaccumulation of hepatic copper ions in Wilson’s disease, respectively.^19–22 Accumulation of copper in the liver of Wilson’s disease patients ranges from a few micromolar to several millimolar.²³ Although copper ion homeostasis and the impact of abnormal copper levels on physiology have been widely studied, details about the functional role of copper ions in various tissues in vivo remain insufficiently understood due to lack of real-time copper imaging techniques in live animals.^24,25

Magnetic resonance imaging (MRI) is a powerful medical diagnostic technique that allows noninvasive, three-dimensional visualization of tissue with high spatial and temporal resolution. MRI is largely based on detection of water and fat protons so that image contrast among tissues reflects differences in proton content, cell density, and water perfusion and diffusion. Image contrast can be altered by the use of paramagnetic inorganic complexes that shorten the T_1,2 relaxation times of water molecules in various compartments. These agents are commonly known as contrast agents (CAs). Among all paramagnetic complexes designed for use as CAs, the Gd³⁺-based agents have proven to be the safest and most versatile agents for clinical use over the past ~30 years. The efficiency of an agent per unit concentration is commonly reported as r₁ (T₁) or r₂ (T₂) relaxivity.^26–28 Notably, the design of contrast agents that alter the T₁ of water protons in response to a given analyte is of major importance. Many responsive probes have been reported including sensors for metal ions,⁹ enzyme activity,²⁹ pH,³⁰ pO₂,³¹ and temperature.³² One of the first reports of a copper-activated MR sensor was based on a GdDO3A (1,4,7,10-tetraazacyclododecane-1,4,7,-triacetic acid) derivative having a iminodiacetate pendant arm for Cu²⁺ recognition. This derivative displayed a 41% increase in r₁ relaxivity upon binding Cu²⁺.³³ A later paper reported a GdDO3A derivative having a quinolone-based pendant arm,³⁴ and it displayed a 71% increase in r₁ upon addition of Cu²⁺. Neither of these agents were examined in vivo. Derivatives have also been designed to show greater selectivity for Cu⁺ over Cu²⁺.³⁵ In addition to MRI agents, other imaging approaches for imaging copper have included positron emission tomography (PET), optical imaging, and bimodal imaging techniques.^36,37 The positron emitter, ⁶⁴Cu, was successfully used to image greater uptake and accumulation of copper in livers of Wilson’s disease mouse models.²⁴ Similarly, a near-infrared fluorescent sensor for detection of Cu⁺ ions has been shown to be capable of monitoring fluctuations in exchangeable copper stores in living cells and mice under basal conditions, as well as in situations of copper overload or deficiency.³⁸

Our group recently reported several MR zinc-sensors capable of detecting the release of intracellular stores of zinc into extracellular space in prostate³⁹ and pancreas of live mice.⁴⁰ Divalent zinc ions released by cells in these organs are immediately chelated by a zinc-responsive MR agent, and the resulting binary complex then forms a ternary complex with serum albumin which results in reduced molecular motion and an increase of r₁.^41,42 Copper is also known to bind to albumin and other less abundant proteins in plasma.^43–47 Hence, we hypothesized that the key to an effective copper detection in vivo by MRI might be to design a Cu-responsive agent that also forms a ternary complex between Cu²⁺ ions, the sensor, and albumin, similar to the Zn-sensor designs.

Herein, we report the synthesis of a novel copper-responsive MRI contrast agent having a bis(benzoic acid)methylamine recognition motif (GdL₁) and the physicochemical properties of the resulting GdL₁–Cu²⁺ complex and the ternary complex formed with human serum albumin (HSA). X-ray absorption spectroscopy (XAS) of GdL₁ and some structural analogues was used to interrogate the Cu²⁺ binding site in this system. Finally, GdL₁ was injected into mice to detect extracellular Cu²⁺ in the liver by MRI. A comparison of GdL₁ (high Cu²⁺ affinity) with GdL₂ and GdL₃ (lower Cu²⁺ affinity) (Figure 1) indicated that only GdL₁ detects extracellular copper in mouse liver.

Figure 1.

Gadolinium-DO3A-based copper responsive (GdL) agents.

2. RESULTS AND DISCUSSION

Design and Synthesis.

The structure of the Cu²⁺-responsive agent reported here consists of a GdDO3A moiety with a bis(benzoic acid)methylamine side chain as a potential chelator for Cu²⁺. This design was motivated by our previous MR-responsive Zn²⁺ sensor scaffold where the ion of interest initiates formation of a ternary complex between the agent and serum albumin.^41,42 Although Cu²⁺ has a preference for nitrogen donor atoms, the coordination rigidity provided by the bis(benzoic acid)methylamine could potentially favor coordination by geometrical stabilization of tetragonal or square pyramidal structures typical for Cu²⁺.^48,49 This structural feature precludes the possibility of binding with the more abundant biological ions like Ca²⁺ and Mg²⁺. To evaluate the impact of repositioning of the carboxylate groups on the aromatic side chain and lowering the charge on the carboxylate groups, GdL₂ and GdL₃ were also studied for comparison. The synthetic details of all three contrast agents are described in the Supporting Information.

Water Proton Relaxivity Measurements in the Presence of Various Metal Ions.

The longitudinal relaxivity (r₁) of GdL₁ (4.7 ± 0.1 mM⁻¹ s⁻¹ at 20 MHz) was unchanged upon addition of Ca²⁺, Mg²⁺, Cu⁺, or Fe³⁺ ions (Figure 2 and Figure S4). However, addition of Zn²⁺ increased r₁ to 5.3 mM⁻¹ s⁻¹ (a 12% increase), while addition of Cu²⁺ increased r₁ to 6.7 mM⁻¹ s⁻¹ (a 43% increase). The background relaxivity due to the weak paramagnetism of Cu²⁺ and Fe³⁺ was subtracted from the r₁ values shown in Figure 2 and reported in Table 1.⁵⁰ This suggests that GdL₁ has some selectivity for Cu²⁺ over Zn²⁺ in agreement with the Irving–William series and Pearson’s hard–soft acid base (HSAB) theories.⁵¹ Even though the origin of this r₁ enhancement is unclear from these data alone, one possibility is that the linker side arm with the anionic carboxyl groups on the bis-benzoic acid motif may form a hydrogen bond with the single exchanging inner-sphere water molecule on the Gd³⁺ ion and this interaction is reduced when Cu²⁺ or Zn²⁺ binds to GdL₁. This could in principle alter the water exchange rate in this system and result in a small change in r₁. A second contributing factor might be that GdL₁ experiences relatively slower molecular rotation (τ_R) upon binding to Cu²⁺, and this could result in a slight increase in r₁ relaxivity. These two possibilities were examined in more detail below.

Figure 2.

20 MHz relaxivity (r₁) of GdL₁ in the presence of various Mⁿ⁺ ± HSA. The white bars reflect r₁ after addition of 0.5 mM Mg²⁺, Ca²⁺, Fe³⁺, Zn²⁺, Cu⁺, or Cu²⁺ to 0.5 mM solutions of GdL₁. The black bars reflect r₁ after subsequent addition of 0.6 mM HSA to the GdL₁–Mⁿ⁺ solutions. The data were collected in 0.1 M MOPS buffer (pH 7.4) at 37 °C.

Table 1.

20 MHz Relaxivity Values for GdL_1–3 ± Cu²⁺ and ± HSA and K_d Values for Binding of GdL_1–3–Cu²⁺ with HSA^a

in the absence of 0.6 mM HSA					in the presence of 0.6 mM HSA
	r1 (mM−1 s−1)				r1 (mM−1 s−1)
GdLx	no Cu2+	1 equiv of Cu2+*	% increase in r1	Kd (GdL–Cu2+)b (μM)	no Cu2+	1 equiv of Cu2+	% increase in r1	Kd (GdL–Cu2+–HSA)c (μM)
GdL1	4.7 ± 0.1	6.7 ± 0.1	43%	84 ± 10	6.1 ± 0.1 5.7 ± 0.1d	22.6 ± 0.2 15.4 ± 0.2d	270%	45 ± 3.1
GdL2	4.9 ± 0.2	5.5 ± 0.1	12%	895 ± 32	6.5 ± 0.2	14.5 ± 0.1	123%	59 ± 5
GdL3	4.8 ± 0.1	5.4 ± 0.2	12%	352 ± 9	6.3 ± 0.2	12.0 ± 0.2	90%	60 ± 10

^aAll experiments were performed in 0.1 M MOPS buffer (pH 7.4) at 37 °C.

^bK_{d(GdL–Cu²⁺)} was determined by fluorescence titrations.

^cK_{d(GdL–Cu²⁺–HSA)} was determined by proton relaxation enhancement titrations.

^dValues measured in the presence of 0.6 mM mouse albumin.

The binding stoichiometry between GdL₁ and Cu²⁺ was determined to be 1:1, as reported by the method of continuous variations (see the Job plot, Figure S9) and by an inflection point⁵² in the relaxivity data (Figure 3). This stoichiometry was assumed in all calculations of dissociation constants (K_d). The increases in r₁ of GdL₂ and GdL₃ were considerably lower upon addition of Cu²⁺ (Figure 3), suggesting that either these two complexes have a weaker affinity for Cu²⁺ or the resulting GdL_x–Cu²⁺ complexes have quite different water exchange properties.

Figure 3.

20 MHz r₁ relaxivity of GdL_1–3 as a function of added Cu²⁺. The concentration of GdL_1–3 was 0.5 mM in 0.1 M MOPS buffer (pH 7.4). The data were collected at 37 °C.

Binding Experiments in the Absence of HSA (GdL_x–Cu²⁺).

The equilibrium dissociation constants (K_d) between the three GdL_x complexes and Cu²⁺ were determined by fluorescence spectroscopy by performing titrations in which Cu²⁺ was added to a buffered solution containing GdL_x. Addition of Cu²⁺ results in quenching of the intrinsic fluorescence of the benzoic acid moieties of GdL_x (Figure S11).^53,54 The resulting binding curves were fit to a 1:1 binding model to give the K_d values reported in Table 1. These data indicate that GdL₁ has the highest affinity for Cu²⁺ (84 ± 10 μM), followed by GdL₃ (352 ± 9 μM) and GdL₂ (895 ± 32 μM). This suggests that the position of the carboxyl groups (meta versus para) and charge of the complexes are both important for Cu²⁺ binding.

Water Proton Relaxivity Measurements in the Presence of Various Metal Ions and HSA.

HSA, the most abundant protein in serum (~600 μM), plays a key role in the transport of metal ions, fatty acids, and other hydrophobic molecules including many drugs. HSA has two Cu²⁺ binding sites, the N-terminal site (NTS) and the multimetal binding site (MBS).^55,56 It was reported that Cu²⁺ has a significantly higher affinity for the NTS site (~1 pM)⁵⁷ than the MBS site (~10 nM).⁵⁸ Thus, the NTS site is considered to be the only site in HSA to be occupied by Cu²⁺, since the concentration of HSA is much higher than the biological concentration of free Cu²⁺ ions.^55–59 We recently demonstrated that analogous Gd-based MR contrast agents responded to an increase in free Zn²⁺ ions from pancreatic β-cells⁴⁰ and epithelial prostate cells stimulated by an increase in plasma glucose.³⁹ This functional response was shown to reflect the formation of a ternary GdL_x–Zn-albumin complex at the MBS site A.⁶⁰ This previous data suggested that perhaps Cu²⁺ could also be detected in vivo in those situations where excess free Cu²⁺ ions in extracellular spaces might be available for binding to a contrast agent. This motivated further relaxometric studies to determine the magnetic contributions of all of the GdL_x with Cu²⁺ in the presence of physiological levels of HSA. As summarized in Table 1, the r₁ values for all three complexes increase slightly in the presence of HSA alone (likely reflecting a slight increase in viscosity) but increase substantially after the addition of Cu²⁺ ions. This suggests that the GdL_x complexes experience slower molecular rotation by the formation of a GdL_x–Cu–HSA ternary complex. This is particularly true for GdL₁ where r₁ increases from 6.1 ± 0.1 to 22.6 ± 0.2 mM⁻¹ s⁻¹ (a 270% increase). As shown in Figure 2 (and Figure S5), addition of Mg²⁺, Ca²⁺, and Fe³⁺ does not result in an increase in r₁ in the presence of HSA, while Zn²⁺ ions do to a lesser extent, about 2-fold lower than the increase in r₁ induced by Cu²⁺.

Binding Experiments in the Presence of HSA.

Additional proton relaxation enhancement (PRE) titrations were carried out to examine the binding interactions and to quantitatively evaluate the binding constants for each GdL_x complex with HSA in the presence of 1 mol equiv of Cu²⁺. A fitting of these data (Figure S13 and Table S2) to a 1:1 binding model gave the K_d values reported in Table 1. GdL₁ showed the highest binding affinity to HSA–Cu²⁺ (K_d = 45 ± 3.1 μM), while the binding affinities of GdL₂ and GdL₃ were surprisingly weaker only by ~30%. This demonstrates significant differences in binding affinity between the different GdL_x and Cu²⁺ are leveled upon formation of the ternary GdL_x–Cu²⁺–HSA complexes. These data alone suggest that HSA plays a significant role in stabilizing the binding interactions between GdL_x and Cu²⁺ ions. To confirm the formation of the LnL_x–Cu²⁺–HSA ternary complex, samples of LaL₁ (a diamagnetic analogue), Cu²⁺, and HSA (from an EPR experiment, see below) were passed through a size exclusion chromatography column, and the eluent peaks were separately analyzed for Cu and La by ICP-MS. Those results (Figure S12 and Table S1) showed that ~63% of the total La eluted from the column in the form of a ternary LaL₁–Cu²⁺–HSA complex, confirming the formation of a stable Cu²⁺ mediated ternary complex.

Kinetic Inertness.

The kinetic stability of a GdL_x complex is also an important factor to consider when developing MRI probes. Previous studies reported that Cu²⁺ could displace Gd³⁺ from a complex by transmetalation.^61,62 This possibility was examined by challenging GdL₁ with 3 mol equiv of Cu²⁺ in 0.03 M phosphate buffer (pH 7.2). Under these conditions, if transmetalation occurred, any unchelated Gd³⁺ would then precipitate from the solution as an insoluble phosphate, a process that can be monitored by relaxometry. R_1obs values measured over the time (Figure S16) show that complexes are kinetically inert, even in the presence of a 3-fold excess of Cu²⁺. In addition, LC-MS data showed that in the presence of a 1 mol equiv of Cu²⁺ no metal transmetalation was observed at room temperature after 7 days in MOPS buffer.

Experiments to Identify the Cu²⁺ Donor Atoms in GdL₁.

The X-band EPR spectrum of LaL₁–Cu²⁺ exhibited an unusual axial spectrum (devoid of well-defined hyperfine features and g_⊥ ≅ 1.99) both in the absence and in the presence of HSA (Figure S19). The g_⊥ values shown by these spectra significantly deviated from the typical values of an axial EPR spectrum for a type-2 Cu²⁺ complex. This suggests that the Cu²⁺ centers in both complexes are electron poor, likely due to the strong electron withdrawing effect of the lanthanum ion in the complex. In comparison, the X-band EPR spectrum of HSA–Cu²⁺ exhibited a typical type-2 square pyramidal geometry very similar to previously reported EPR spectra in the literature.⁶³ However, the broadened hyperfine features of the Cu²⁺ EPR spectra after addition of LaL₁ precluded a detailed structural analysis of the copper center.

XAS Studies.

Copper K-edge X-ray absorption spectroscopy (XAS) studies were also performed to identify the Cu²⁺ donor atoms in GdL₁–Cu²⁺ and GdL₁–Cu²⁺–HSA. The XANES spectrum of GdL₁–Cu²⁺ (Figure 4) is characterized by an intense absorption feature at 8987–8988 eV with a broad low energy tail in the region below 8985 eV (normalized absorption of approximately 0.5 at 8988 eV) arising from a 1s → 4p transition characteristic of Cu²⁺ complexes. The presence of the first major inflection point at 8986 eV and the absence of lower energy features (normalized absorption 0.15 at 8984 eV and first inflection point 8984 eV) is typical of classic tetragonal Cu²⁺ complexes with nitrogen and oxygen ligands. The complex also presents a weak 8979 eV peak (more visible in the first derivative spectra) corresponding to the 1s → 3d transition possibly reflecting a less centrosymmetric nature of the center and thus a significant degree of distortion from planarity.⁶⁴ The XANES spectrum of GdL₁–Cu²⁺–HSA is nearly identical to the spectrum of Cu²⁺–HSA, suggesting a very similar coordination environment in the two complexes. The spectra are characterized by an intense absorption feature at 8987–8988 eV arising from a 1s → 4p transition. Additional features for the Cu²⁺ site include a lower-energy feature with normalized absorption of approximately 0.25 at 8984 eV and the first inflection point determined in the first derivative spectrum at 8982 eV, about ~1 eV higher than the one observed in Cu⁺ complexes. Possible photoreduction of Cu²⁺ was prevented experimentally by collecting the spectra at different locations in the frozen sample in each scan. Also, the XANES spectra were quite similar to the one observed for the Cu²⁺–DAHK peptide complex representing the N-terminal Cu²⁺ binding site in HSA, thus supporting the same coordination environment as in the full-length protein.⁵⁵

Figure 4.

XANES spectra for GdL₁ in the presence of Cu²⁺ and HSA.

Additional information on coordination environment and ligand metal distances were obtained by copper K-edge extended X-ray absorption fine structure (EXAFS). The experimental copper EXAFS spectra are presented in Figure S20 together with best fits and the corresponding EXAFS Fourier transforms. The spectrum of GdL₁–Cu²⁺ could be fitted with two ligand shells indicative of a Cu complex coordinated by three N/O ligands at 1.99 Å and a N/O ligand at 2.51 Å (Table 2). The XANES and EXAFS results are consistent with the formation of a distorted tetragonal complex in which two O donor atoms from the two carboxylates plus one N atom from tertiary amine of the bis(benzoic acid)methylamine moiety and fourth donor O atom contributed by a solvent water molecule to the Cu²⁺ ion.

Table 2.

Structural and Coordination Parameters Obtained from Fitting Cu K-edge EXAFS^a

complex	N	bond	R (Å)	σ2 (Å2)	F-factor
GdL1–Cu(II)	3	Cu–N/O	1.994(3)	0.0008	0.488
	1	Cu–N/O	2.51(1)	0.0001
Cu(II)–HSA	3	Cu–N/O	1.991(2)	0.0016	0.398
	1	Cu–N/O	2.278(7)	0.0013
	1	Cu–N/O	2.515(7)	0.0008
GdL1–Cu(II)–HSA	3	Cu–N/O	1.954(6)	0.0054	0.576
	1	Cu–N/O	2.33(1)	0.0037
	1	Cu–N/O	2.86(1)	0.0001
GdL1–Cu(II)–HSA	4	Cu–N/O	1.965(6)	0.0054	0.587
	1	Cu–N/O	2.31(2)	0.0037
	1	Cu–N/O	2.87(1)	0.0001

^aCoordination numbers are indicated by N, interatomic distances R are given in Å (the values in parentheses are the estimated standard deviations), Debye–Waller factors σ² (the mean-square deviations in interatomic distance) are given in Ų, and the fit-error function F is defined $K_{\mathrm{d}\left(\text{GdL}-{\text{Cu}}^{2+}-\text{HSA}\right)}$, where χ^(k) are the EXAFS oscillations and k is the photoelectron wave-number.

In the HSA-Cu²⁺ complex, the EXAFS data were best fit with three coordinating shells around Cu²⁺ with three Cu–O/N bonds at 1.99 Å and an additional (likely equatorial) N/O bond at 2.28 Å. In addition, a third shell corresponding to a N/O ligand at 2.51 Å was obtained in the fit, suggesting the presence of an axial ligand. This analysis was in agreement with a previously reported square pyramidal Cu²⁺–HSA coordination at the N-terminal site (also known as ATCUN) with high affinity for Cu²⁺.⁵⁵ Copper is coordinated to four nitrogen donors in the NTS site consisting of Asp, Ala, and His amide nitrogen atoms and the His side chain in an equatorial position and a water molecule or N-terminal amine nitrogen in the axial position.^55,59 For the GdL₁–Cu²⁺–HSA ternary complex, the EXAFS data could be best fitted with three or four N/O ligands (resulting in similar F-values) at 1.96 Å and two additional N/O ligands at 2.33 and 2.86 Å. The analysis predicts the presence of a distorted square pyramidal/octahedral coordination around the Cu²⁺ in the ternary complex with four equatorial ligands with short bond distances and one or two axial ligands with longer bond distance (Table 2). This coordination anticipates the replacement of a loosely bound axial ligand in the Cu²⁺–HSA complex by the chelating sites in the bis(benzoic acid)methylamine moiety of GdL₁. It should be noted that the bond distances for the distorted square pyramidal Cu²⁺–HSA complex are distinctly different from the bond distances of the distorted square pyramidal ternary complex. Despite being difficult to unambiguously distinguish the coordination of one or two axial ligands by XAS, the EXAFS analysis confirms small differences in the coordination shells between GdL₁–Cu²⁺–HSA and HSA–Cu²⁺. This distorted octahedral or square pyramidal coordination of Cu²⁺ in the ternary complex compared to the distorted square pyramidal coordination of Cu²⁺–HSA supports the formation of a copper-mediated complex between HSA and GdL₁.

Molecular Modeling.

Molecular models of the Cu centers in GdL₁, HSA, and the ternary complex were generated on the basis of the coordination geometry and bond lengths for N and O atoms obtained from the EXAFS experimental data (Table 2) using standard MM+ methods. The energy-minimized models are presented in Figure 5. The model of GdL₁–Cu²⁺ reflects a distorted tetragonal geometry around the Cu²⁺ center as predicted by XAS. The geometry of the NTS Cu²⁺ site in HSA reflects a square pyramidal geometry similar to the previously reported crystal structures,⁵⁵ while the Cu²⁺ center in the ternary complex is a distorted octahedral geometry. These models support the EPR and XAS data by predicting only small differences in the coordination geometry of GdL₁–Cu²⁺–HSA compared to that of HSA–Cu²⁺.

Figure 5.

MM+ minimized structures of (A) the domain structure of albumin (PDB ID code 1AO6): domains I and II are colored green (residues 1–373), and domain III is colored in yellow (residues 380–571); long chain fatty acid sites (FA), Sudlow’s drug binding sites, Cu²⁺ binding NTS site, and zinc binding site A (MBS/site A) are also shown. (B) GdL₁–Cu²⁺ complex. (C) HSA–Cu²⁺ complex (CCDC-809109).⁵² (D) HSA–Cu²⁺–GdL₁ distorted square pyramidal complex. (E) HSA–Cu²⁺–GdL₁ distorted octahedral complex consistent with all NMR, XAS, and EXAFS data. Hydrogen atoms and other sites of HSA have been removed to simplify visualization. Only residues at the NTS site in HSA are included. These figures were generated using Hyperchem7.5.

In Vivo Imaging of Free Copper Pools in Living Mice.

Most dietary absorbed copper is transported to the liver via enterohepatic circulation where serum albumin acts as a transporter protein to maintain total exchangeable forms of copper in the μM range.^10,65–70 Hence, the liver plays a key role in copper homeostasis by facilitating copper storage and incorporating copper into ceruloplasmin and other copper binding proteins. Either elevated or reduced copper in the liver has been associated with neurological disorders and acute liver diseases. Therefore, a noninvasive method to image those abnormalities in copper levels in vivo is of broad interest.

To examine whether GdL₁ can detect and respond to changes in extracellular copper in vivo, T₁-weighted MR images of C57BL/6 mice were collected at 4.7 T (Figure 6). This imaging field was chosen from the equipment available to us because the r₁ differences between GdL₁ versus GdL₁–Cu–HSA, although smaller at 4.7 T versus 0.47 T (20 MHz), remain significantly different. After i.v. injection of a bolus of 0.1 mmol/kg of GdL₁, the average gain in signal intensity throughout the liver of a healthy mouse was ~25% when compared to the precontrast images. This enhancement returned to baseline after reaching maximum intensity after ~6 min postinjection, reflective of relatively fast excretion of GdL₁ characteristic of most low molecular weight Gd-based extracellular agents. In a separate group, mice were pretreated with the copper chelator, ATN-224 (5 mg/kg), 2 h prior to injection of GdL₁.^38,70 In those animals, the average MR liver enhancement after injection of 0.1 mmol/kg of GdL₁ was lower, ~11% (p-value = 8.6 × 10⁻³). This suggests that pretreatment with the copper chelator removed some of the excess Cu²⁺ prior to injection of GdL₁. The tissue distribution of Cu and Gd in the same mice used in the imaging experiments was determined by ICP-MS analysis (Figure 6C). These results confirmed that the higher MR signal intensities directly correlated with higher copper levels in the liver of healthy mice, while the Gd content was identical in both ATN-224 treated and nontreated animals.

Figure 6.

(A) In vivo MRI images of wild type mouse (n = 3) pre- and postinjection of GdL₁ (0.1 mmol/kg) without (top) or with (bottom) pretreatment with ATN-224 (5 mg/kg in 50 μL). All images were obtained at 4.7 T. (B) The average MRI signal intensity of mouse liver 6 min after injection of either GdL₁ or Gadavist in control mice (black bars) versus mice pretreated with ATN-224 (white bars). The columns on the right show average liver signal intensities at 13 min after injection with Multihance. The data were compared using a two-tailed student t test. *p < 0.05 (n = 3); error bars reflect ± SD. (C) Total Cu and Gd (μg/g of tissue) in various tissues collected from control mice (black bars) and from mice pretreated with ATN-224 (white bars) 6 min after the injection of GdL₁. Tissue copper levels relative to tissue wet weight were determined by ICP-MS. The data were compared using a two-tailed student t test. *p < 0.05 (n = 3); error bars reflect ± SD.

Mice were also imaged using two different Gd-based agents as controls, Gadavist (an extracellular agent) and Multihance (a hepatobiliary agent). After injection of an equivalent amount of Gadavist, the signal intensity of liver gained intensity as expected for a typical extracellular agent but only by ~8%. Like GdL₁, the signal gain in liver reached a maximum at ~6 min and then returned to preinjection baseline values at about the same clearance rate as GdL₁. In mice pretreated with ATN-224, the liver enhancement was unchanged, consistent with a lack of affinity of Gadavist for Cu²⁺. Given that the possibility of a small amount of GdL₁ may clear via hepatobiliary excretion (a complete biodistribution study has not been done), it was important to perform a similar set of control experiments using a known hepatobiliary agent, Multihance, to determine whether ATN-224 treatment might alter liver function. Those imaging results are also presented in Figure 6B. Since the clearance of Multihance in liver was significantly slower than either GdL₁ or Gadavist, the signal intensity data shown here reflect the maximum values at 13 min rather than those at 6 min. As shown, liver image enhancement resulting from the passage of Multihance through the liver was identical in untreated mice versus mice pretreated with ATN-224, showing that liver function is unaltered by ATN-224. Therefore, the decreased intensity we observed when using GdL₁ must reflect a decrease in freely available Cu²⁺ in liver.

3. CONCLUSIONS

In this study, we investigated whether a new macrocyclic gadolinium complex, GdL₁, could act as a Cu²⁺-responsive MRI contrast agent. We also examined the physical-chemical properties of the ternary GdL₁–Cu²⁺–HSA complex that resulted in a magnified longitudinal r₁ relaxivity (20 MHz) of 22.6 mM⁻¹ s⁻¹. Our results showed that the observed r₁ enhancement due to the slow tumbling of the ternary complex was sufficient to allow detection of μM levels of freely available Cu²⁺ in the liver by T₁-weighted MR imaging, even at 4.7 T. After injection of GdL₁ into healthy untreated mice, the liver was nicely enhanced at 6 min and image contrast returned to background levels after ~20 min. However, when mice were treated with ATN-224, the MR signal gain in liver images was ~50% less compared to control animals. The lower contrast enhancement observed in the liver of mice pretreated with ATN-224 paralleled the reduction in total liver copper as detected by ICP-MS. On the basis of the ICP results, one can estimate the concentration of GdL₁ in liver at 6 min and ask the question of whether the decreased levels of Cu as detected by ICP-MS are consistent with the imaging results. The total amount of Gd in liver at 6 min was 40 μg/g of wet tissue. If one assumes that GdL₁ remains largely extracellular and that the extracellular fraction of wet liver is ~24–26%,^71,72 then the GdL₁ concentration can be estimated as ~1 mM (40 μg/g divided by 157 g/mol = 0.25 μmol/g = 0.25 μmol/0.26 mL = 1 mM). This same calculation for Cu gives a total [Cu] of 0.36 mM in liver of control mice (6 μg/g) and 0.18 mM in livers after treatment with ATN-224 (3 μg/g). Although not all of the Cu²⁺ is extracellular, a 50% change in available Cu would easily be detected using GdL₁ even if r₁ is no higher than 9 mM⁻¹ s⁻¹ at 4.7 T. One would expect to see even more dramatic changes in MRI signal intensity in liver if these experiments had been performed at typical clinical imaging fields, 1.5 or 3 T.

The second goal of this study was to identify the Cu²⁺ donor atoms on GdL₁, the donor atoms in the ternary GdL₁–Cu²⁺–HSA complex, and the location of the GdL₁–Cu²⁺ binding site in HSA. The fact that Cu²⁺ has only one high-affinity site in HSA, the N-terminal site,^55,57 it is reasonable to assume that GdL₁ also binds at this site by contributing donor atoms to Cu²⁺. This model is consistent with the changes in the Cu²⁺ coordination sphere as reported by EPR and X-ray absorption spectroscopy (XAS) data. The combined results indicate that the Cu²⁺ binds to GdL₁ via a single tertiary N atom and two carboxylate O atoms on GdL₁ and a single water molecule to form a distorted tetragonal complex. The Cu²⁺ center in the GdL₁–Cu²⁺-HSA ternary complex is most consistent with coordination by four equatorial nitrogen donor atoms from the protein and one or two axial O/N donors from GdL₁, resulting in a distorted octahedral/square pyramidal geometry. The slight coordination changes in the Cu²⁺ center with GdL₁ in the presence of HSA result in a stable ternary complex that results in a surprisingly high r₁ relaxivity at 20 MHz.

In summary, our study shows that GdL₁ can be used as a sensor of excess freely available Cu²⁺ ions in tissue. In the presence of HSA, the freely available Cu²⁺ forms a stable ternary complex GdL₁–Cu²⁺–HSA that magnifies the r₁ relaxivity to such an extent that in vivo detection of exchangeable Cu²⁺ MR imaging was possible. To our knowledge, this is the first time that extracellular copper levels in the liver could be detected and with a remarkable statistical difference. Although this work has not included a mouse disease model to validate the results obtained here, there is enough evidence to suggest that this sensor can be used in mouse models with known abnormal levels of copper. The total serum copper levels can be markedly elevated in acute liver failure due to its release of excess copper ions from liver tissue stores. This results in elevated total serum Cu²⁺ not bound to ceruloplasmin referred to as “free copper”.⁷³ For example, Wilson’s disease patients reported the significantly higher concentrations of serum nonceruloplasmin copper (>4.0 μM) in the blood⁷⁴ and hepatic copper content of >250 μg/g of dry liver weight.^75,76 Similarly, deficiency of copper has been reported in a variety of genetic, neurological, cardiovascular, and metabolic diseases.^9,77 Furthermore, it has recently been shown that elevated serum and tumor copper levels are linked to the progression of cancer malignancy⁷⁸ and also plays an important role in the regulation of sleep-related and arousal behaviors.⁷⁹ Therefore, we believe that the observations reported here using GdL₁ will catalyze discoveries of Cu²⁺-responsive MRI agents for imaging acute liver conditions such as that found in Wilson’s diseases or elevated copper levels in other disease conditions.