, Volume 21, Issue 4, pp 469–490

Efficiency, thermodynamic and kinetic stability of marketed gadolinium chelates and their possible clinical consequences: a critical review


    • Guerbet, Research Division
  • Jean-Marc Idée
    • Guerbet, Research Division
  • Christelle Medina
    • Guerbet, Research Division
  • Caroline Robic
    • Guerbet, Research Division
  • Monique Sabatou
    • Guerbet, Research Division
  • Claire Corot
    • Guerbet, Research Division

DOI: 10.1007/s10534-008-9135-x

Cite this article as:
Port, M., Idée, J., Medina, C. et al. Biometals (2008) 21: 469. doi:10.1007/s10534-008-9135-x


Gadolinium-based contrast agents are widely used to enhance image contrast in magnetic resonance imaging (MRI) procedures. Over recent years, there has been a renewed interest in the physicochemical properties of gadolinium chelates used as contrast agents for MRI procedures, as it has been suggested that dechelation of these molecules could be involved in the mechanism of a recently described disease, namely nephrogenic systemic fibrosis (NSF). The aim of this paper is to discuss the structure-physicochemical properties relationships of marketed gadolinium chelates in regards to their biological consequences. Marketed gadolinium chelates can be classified according to key molecular design parameters: (a) nature of the chelating moiety: macrocyclic molecules in which Gd3+ is caged in the pre-organized cavity of the ligand, or linear open-chain molecules, (b) ionicity: the ionicity of the complex varies from neutral to tri-anionic agents, and (c) the presence or absence of an aromatic lipophilic residue responsible for protein binding. All these molecular characteristics have a profound impact on the physicochemical characteristics of the pharmaceutical solution such as osmolality, viscosity but also on their efficiency in relaxing water protons (relaxivity) and their biodistribution. These key molecular parameters can also explain why gadolinium chelates differ in terms of their thermodynamic stability constants and kinetic stability, as demonstrated by numerous in vitro and in vivo studies, resulting in various formulations of pharmaceutical solutions of marketed contrast agents. The concept of kinetic and thermodynamic stability is critically discussed as it remains a somewhat controversial topic, especially in predicting the amount of free gadolinium which may result from dechelation of chelates in physiological or pathological situations. A high kinetic stability provided by the macrocyclic structure combined with a high thermodynamic stability (reinforced by ionicity for macrocyclic chelates) will minimize the amount of free gadolinium released in tissue parenchymas.


Magnetic resonance imagingContrast agentsGadoliniumOsmolalityViscosityRelaxivityThermodynamic stabilityKinetic stabilityNephrogenic systemic fibrosis


Over recent years, there has been a renewed interest in the physicochemical properties of gadolinium chelates used as contrast agents for magnetic resonance imaging (MRI) procedures.

It has been suggested (Morcos 2007a, b; Perazella 2007) that dechelation of these molecules could be involved in the mechanism of a new disease, nephrogenic systemic fibrosis (NSF). NSF was first recognized in 1997 in 15 dialyzed patients and described in 2000 (Cowper et al. 2000). This rare and highly disabling disorder is characterized by extensive thickening and hardening of the skin associated with skin-coloured to erythematous papules that coalesce into erythematous to brawny plaques with a peau d’orange appearance.

NSF always occurs in patients with severe or end-stage renal failure (Cowper and Boyer 2006). Importantly, the systemic nature of NSF has recently been emphasized by several authors, who have reported that a considerable subset of patients have involvement of multiple organ systems, especially lungs, but also heart, skeletal muscles, oesophagus and diaphragm, which can eventually lead to death (Galan et al. 2006; Ting et al. 2003).

In 2006, two European teams independently suggested a link between the administration of gadolinium chelates and development of NSF in patients with renal failure (Grobner 2006; Marckmann et al. 2006). In these studies, a direct temporal relationship was found between injection of a Gd3+ chelate and subsequent onset of NSF. The vast majority of published cases clearly appeared to be associated with the open chain, nonionic Gd3+ chelate gadodiamide (Gd-DTPA-BMA). As of January 17, 2007, 85 of the more than 100 cases reported to the US Food and Drug Administration MedWatch reporting system were associated with Gd-DTPA-BMA, 21 were associated with Gd-DTPA, 6 were associated with Gd-DTPA-BMEA, one was associated with Gd-BOPTA (although this same patient had also received Gd-DTPA-BMA 5 days after his Gd-BOPTA MR examination), and none was associated with Gd-HP-DO3A (Kanal et al. 2007).

These compounds are administered intravenously and excreted unchanged by glomerular filtration. The plasma elimination half-life is about 1.5 h in healthy human volunteers, but is dramatically increased in patients with renal insufficiency and may then exceed 30 h (Bellin et al. 2003; Joffe et al. 1998). Among the numerous hypotheses proposed to explain this disease, the most widely supported (Morcos 2007a, b; Perazella 2007) postulates that, in the setting of renal failure, the remanence of a Gd3+ chelate in the body is dramatically increased, thus enhancing the chance for dissociation of Gd3+ from its chelate especially in the case of nonionic, linear open-chain molecules such as Gd-DTPA-BMA.

The objective of this review is to exhaustively and critically review the available data regarding all of the physicochemical properties of marketed gadolinium chelates in terms of their biological consequences. For this purpose, the physicochemical characteristics (osmolality, viscosity) of the pharmaceutical injected solution, the efficiency (and consequently the dose administered), and thermodynamic and kinetic stability assessed by in vitro or in vivo studies of all marketed gadolinium chelates will be compared and discussed. Structure-physicochemical properties relationships will be reviewed in order to underline the main structural parameters governing each physicochemical characteristic. Importantly, the concepts of kinetic and thermodynamic stability will be discussed, as this remains a somewhat controversial topic, especially to predict the amount of free gadolinium which may be released in vivo following dechelation of gadolinium chelates.

Chemical structure

The contrast of an MR image results from a complex interplay of various factors such as T1 and T2 relaxation rates, proton density, and MRI sequences. MRI contrast agents are used in MR imaging as they modify the longitudinal (T1) and transverse (T2) relaxation rate of water protons present in the tissues. The efficacy of an MRI contrast agent is measured in terms of relaxivity, i.e. the longitudinal and transverse relaxation rate of the water protons observed at a millimolar concentration of contrast agent. MR contrast agents are composed of paramagnetic ions that increase the relaxation rates of protons in the body. In terms of contrast agent design, attention has been essentially focused on gadolinium (III) both because of its high paramagnetism (due to the seven unpaired electrons of Gd(III)) and its very slow electronic relaxation rate (due to the asymmetric S-state of Gd3+) and nine contrast agents based on gadolinium have now been marketed: Table 1 and Scheme 1.
Table 1

General characteristics of currently marketed gadolinium chelates used for magnetic resonance imaging (Idée et al. 2006; Caravan et al. 1999; Brücher and Sherry 2001)












Generic Name

Gadopentetate dimeglumine



Gadobenate dimeglumine

Gadoxetic acid disodium salt

Gadofosveset trisodium salt

Gadoterate meglumine



Trade Name




















Chemical structure




















Dissociated particles per molecule










Log P BuOH/H2O










Concentration (M)










Standard dose (mmol/kg)










Osmolality at 37°C (mOsm/kg H2O)










Osmotic loadb (mOsmol/l)










Relaxivity (r1/r2) mM−1s−1 at 37°C, 1.5 T in waterc










Viscosity (mPa.s) at 37°C











Free DTPA 0.2% (1 mmol/l)

Ca-DTPA-BMA (Na+ salt) 5% (25 mmol/l)

Ca-DTPA-BMEA (Na+ salt) (50 mmol/l)

No formulation

Ca-EOB-DTPA (trisodium salt)d

Fosvest ligand (0.325 mmol/l)e

No formulation

[Ca-HP-DO3A]2 (Ca2+ salt) 0.1% (0.5 mmol/l)

Ca-BT-DO3A (Na+ salt) (1 mmol/l)

Log Ktherm










Log Kcond










aDose for liver imaging: 0.05 mmol/kg

b\( {\text{osmotic load (mOsm/l)}} = \frac{{{\text{dose (mmol/kg)}}*70}} {{{\text{V}}_{{{\text{distribution}}}} (l)}}{\text{* number dissociated ions}} \) (values are calculated on the assumption that the agents distribute homogeneously in the interstitial space (10.5 l for a patient weighing 70 kg)

cGuerbet measurement on commercial solution; Incertainty on relaxometric measurement: ±0.3 mM−1 s−1 for relaxicity measurements

d(concentration not disclosed)

eSteger-Hartmann et al. (2006)

fCaravan et al. (2001)

gMoreau et al. (2004) and Guerbet calculations

hUggeri et al. (1995)

iSchmitt-Willich et al. (1999)

jBellin et al. (2003)
Fig. 1

Dechelation kinetics at pH 1.2, 37°C, of marketed macrocyclic gadolinium contrast agents: percentage of remaining gadolinium chelate over time
Scheme 1

Structure of currently marketed gadolinium chelates used for magnetic resonance imaging

However, free gadolinium is highly toxic (Idée et al. 2006), as the ionic radius of Gd3+ (107.8 pm) is close to that of Ca2+ (114 pm) and this element is an inorganic blocker of many types of voltage-gated calcium channels at nano- to micromolar concentrations. Gadolinium inhibits the activity of some enzymes (Ca2+-activated-ATPase in the sarcoplasmic reticulum of skeletal muscle fibres, some dehydrogenases and kinases, glutathione S-transferases, etc.) (Evans 1990; Itoh and Kawakita 1984) and has a remarkable capacity to depress the reticuloendothelial system (Evans 1990). Gadolinium inhibits phagocytosis in normal and activated Kupffer cells and increases the expression of several cytokines.

Gadolinium must therefore be chelated with a coordinating ligand as chelation of gadolinium by appropriate highly kinetic and thermodynamically stable polyamino-polycarboxylic ligands dramatically reduces its acute toxicity. For example, intravenous LD50 of Gd-DOTA in mice was 10.6 mmol/kg, whereas LD50 of GdCl3, tested under similar conditions, was only 0.35 mmol/kg (Bousquet et al. 1988).

In order to understand the key features involved in the physicochemical properties of gadolinium chelates, several chemical structural aspects must be taken into account:
  • The nature of the chelating moiety: either macrocyclic chelates such as Gd-DOTA, Gd-HP-DO3A, Gd-BT-DO3A where Gd3+ is “caged” in the pre-organized cavity of the ligand which perfectly matches the size of Gd3+, or “ linear open-chain” chelates such as Gd-DTPA or Gd-DTPA-BMA, Gd-DTPA-BMEA, Gd-BOPTA, Gd-EOB-DTPA. All these ligands form octa-coordinated gadolinium chelates, and the coordination sphere is completed with a water molecule in the inner sphere, achieving a total coordination of nine. As discussed below, this structural difference between macrocyclic and linear open-chain chelates mainly has an impact on the stability of the gadolinium chelate.

  • The ionicity of the chelate: some ligands such as Gd-DTPA, Gd-BOPTA, Gd-EOB-DTPA chelates are di-ionic whereas Gd-DOTA is a mono-ionic chelate. MS325 is the only tri-ionic agent. In the pharmaceutical formulation, the charge of the ionic chelates is neutralized at physiological pH by a stoichiometric quantity of protonated counter cation, meglumine, except for Gd-EOB-DTPA and MS325 marketed as a disodium and trisodium salts, respectively. On the contrary, ligands such as HP-DO3A, BT-DO3A, DTPA-BMA, DTPA-BMEA form nonionic complexes with Gd(III), as the charges on the ligand and metal ion are equal (Chang et al. 1992).

  • The hydrophilicity of the chelate: the high solubility of gadolinium chelates correlates with a pronounced hydrophilicity of all gadolinium chelates as demonstrated by their negative log P values measured between butanol and water. However, ionic chelates are significantly more hydrophilic (log P one order of magnitude higher) than nonionic chelates (Caravan et al. 1999) Gd-DOTA, Gd-HP-DO3A, Gd-BT-DO3A, Gd-DTPA, Gd-DTPA-BMA and Gd-DTPA-BMEA are considered to be nonspecific extracellular agents which means that, after IV injection, they are distributed non-specifically in plasma and interstitial fluid of the body and are eliminated almost exclusively by renal glomerular filtration with a elimination half-life of about 90 min in patients with normal renal function but which can be dramatically increased in patients with renal failure (Joffe et al. 1998). However, three linear open-chain gadolinium chelates are characterized by the presence of aromatic residues on the scaffold of the DTPA ligand: MS-325, Gd-EOB-DTPA and Gd-BOPTA. The introduction of an aromatic lipophilic group on the DTPA backbone decreases the overall hydrophilicity of the chelates, as demonstrated by lower log P values in comparison to Gd-DTPA (log P for MS325 = −2.11 vs. −3.16 for Gd-DTPA) and modifies the biodistribution and pharmacokinetics of these contrast agents, as the presence of the aromatic group can lead to hepatocellular uptake and excretion into the bile duct, gallbladder and intestines (Gd-EOB-DTPA and Gd-BOPTA) (Vander Elst et al. 2001) or can cause strong but reversible albumin binding thereby confining the contrast agent to the vascular compartment, allowing blood pool imaging (MS325) (Caravan et al. 1999).

Physicochemistry of the pharmaceutical solution: osmolality and viscosity

Nonionic chelates were originally designed to lower the osmolality of 0.5 M injected solutions which is higher than that of plasma, as, based on the experience of X-ray iodinated contrast agents, it was hypothesized that lowering osmolality could reduce adverse events such as sensation of warmth or heat at the injection site or vascular endothelium damage (Cohan et al. 1991). However, all marketed gadolinium chelate formulations are hypertonic in comparison to blood (the osmolality of blood and body fluids is about 280–300 mOsmol/kg) and the global increase in plasma osmolality after injection of the standard clinical dose is insignificant, as demonstrated by the value of osmotic load (Table 1) and does not have impact on the body’s osmotic balance. As osmolality is a colligative property, and therefore depends on the total number of particles in solution, regardless of their charge, Gd-BT-DO3A solutions formulated at 1 M do not conserve the characteristic of reduced osmolality in comparison to 0.5 M nonionic gadolinium chelate solutions. Interestingly, the experimental value of osmolality of gadolinium chelates is higher than predicted by the theory, which has been explained by a high hydration number of the gadolinium chelates, reducing the amount of free water and increasing the apparent concentration of the gadolinium chelate (Gries and Miklautz 1984). The negative charge of ionic chelates may result in a slower rate of diffusion into disk cartilage which contains negatively charged proteoglycans compared to a nonionic chelate, as demonstrated in an MR study in rabbits after intravenous administration of equimolar doses of Gd-DTPA and Gd-DTPA-BMA (Ibrahim et al. 1994).

The gadolinium concentration has an obvious impact on viscosity, as the 1 M Gd-BT-DO3A solution is more viscous than 0.5 M solutions of gadolinium chelates, which are more viscous than 0.25 M MS-325 and Gd-EOB-DTPA solutions. The introduction of an aromatic group into the structure of Gd-BOPTA also significantly increases the viscosity of the pharmaceutical solution in comparison with the other 0.5 M formulations of gadolinium chelates. However, the viscosity of all these pharmaceutical solutions does not preclude bolus IV administration.


According to the Solomon Bloembergen-Morgan equations (Caravan et al. 1999), various structural parameters influence the relaxivity of gadolinium chelates such as correlation time, with contributions derived from the rotational correlation time, the electronic relaxation time and the water residence life time but also the distance between Gd3+ and the coordinated water hydrogen.

The NMRD profiles of the marketed gadolinium chelates recorded at 37°C in water were recently published (Laurent et al. 2006; Adzamli et al. 1999; Vander Elst et al. 1997; Muller et al. 1999). When measured in water, the r1 and r2 relaxivities obtained for non-protein-binding chelates are fairly similar, in the field range from 0.47 T to 4.7 T. Gd-DOTA is nevertheless characterized by a higher low field relaxivity (0.005–0.05 T) in comparison to other non-protein-binding chelates due to its longer electronic relaxation time related to the high symmetry and rigidity of this chelate (Benmelouka et al. 2006). The water residence time of bisamide chelate (Gd-DTPA-BMA and Gd-DTPA-BMEA) has also been demonstrated to be longer than that of other chelates (Caravan et al. 1999; Adzamli et al. 1999). However, the relaxivity of these chelates at 37°C is not limited by this increase of the water residence time. The relaxivities obtained in water for protein-binding chelates are slightly higher than those of non-protein-binding chelates, partly due to the higher molecular weight of these molecules which lengthens the rotational correlation time, but it has also been proposed that the distance between Gd3+ and the coordinated water hydrogen could be reduced due to the lipophilic residue grafted on the C4 position of the DTPA backbone of EOB-DTPA-Gd and MS325 (Laurent et al. 2006; Muller et al. 1999; Vander Elst et al. 1997).

In a recent comparative study, the r1 and r2 relaxivities in bovine plasma of all commercially available gadolinium chelates were measured at 37°C (Rohrer et al. 2005). When measured in bovine plasma, the r1 and r2 relaxivities obtained for non-protein-binding chelates are again fairly similar in the field range from 0.47 to 4.7 T (Rohrer et al. 2005) and higher than those measured in water due to a slightly higher viscosity of plasma resulting in lengthening of the rotational correlation time. Due to the reversible interaction of the gadolinium chelate with plasmatic proteins, which further lengthens the rotational correlation time of the bound chelate, the relaxivities of protein-binding chelates are significantly increased in the presence of plasma.

However, it is noteworthy that the value of the relaxivities of these protein-binding chelates reported in the literature is highly dependent on the experimental conditions used, especially the concentration of contrast agent (typically from 0.25 to 0.5 mmol/l) and plasma protein (Cavagna et al. 1997; Caravan et al. 2002, De Haen et al. 1999) and are not easily comparable to the well-defined relaxivity of non-protein-binding chelates. Physicochemical measurements of relaxivity in plasma performed in vitro at a fixed concentration of gadolinium therefore do not necessarily accurately reflect the efficiency of the contrast agent in vivo (De Haen et al. 1999; Port et al. 2005; Corot et al. 2003), as, in vivo, the proportion of free and bound forms of protein-binding contrast agents varies according to the pharmacokinetic profile, and the relaxivities of protein-bound and free contrast agents are different. The concept of dynamic relaxivity was therefore introduced (Port et al. 2005) to compare the in vivo efficiency of protein-binding chelates in a protocol mimicking an angiographic procedure. A significant variation of dynamic relaxivities between the bolus and post-bolus phase was demonstrated for Gd-BOPTA, as its relaxivity varies from 6.7 to 5.2 mM−1 s−1 and for MS325 (relaxivity variation from 17.3 to 8.6 mM−1 s−1).

Chelate stability

  • The integrity of the gadolinium chelate must be maintained in vivo in order to ensure good tolerability. Dissociation of Gd3+ from an MRI contrast agent is undesirable, as both the free metal and unchelated ligands are generally more toxic than the chelate itself. This stability issue is discussed, in particular, in the context of NFS (Morcos 2007a, b; Perazella 2007). Two sometimes confusing concepts have been proposed to describe the stability of gadolinium chelates: thermodynamic and kinetic stability.

Thermodynamic stability

Two thermodynamic concepts have been proposed in the literature to explain the thermodynamic stability of Gd chelates:
  1. (1)

    Thermodynamic stability constants log Ktherm and log Kcond reflecting the affinity of gadolinium for its ligand (at high basic pH and 7.4, respectively) These two thermodynamic constants are useful to quantify the amount of free Gd3+ or free ligand in equilibrium in a water solution.

  2. (2)

    Calculation based on thermodynamic models in an attempt to predict the amount of free Gd3+ released from gadolinium chelates in the presence of endogenously available ions, e.g., Ca2+, Cu2+, Zn2+, Fe3+, H+, OH, CO32− and PO43− etc, once thermodynamic equilibrium has been achieved.


Thermodynamic stability constant Ktherm and conditional stability constant Kcond

By definition, as gadolinium is chelated, a thermodynamic equilibrium exists between the metal [M], the ligand [L] and the chelate [ML]:
$$ [{\text{M}}]{\text{ }} + {\text{ }}[{\text{L}}]{\text{ }} \rightleftarrows [{\text{ML}}] $$
The stability of gadolinium chelates is expressed in terms of log Ktherm (sometimes called log K or log KGdL), where Ktherm is the thermodynamic stability constant, defined as:
$$ K_{{{\text{therm}}}} = [{\text{ML}}]/([{\text{M}}]*[{\text{L}}]) $$

As the value of log Ktherm is valid at high basic pH, the conditional thermodynamic stability constant log Kcond (sometimes called log K′ or \( {\text{log }}K^{\prime }_{{{\text{GdL}}}} \)) is calculated at pH 7.4 on the basis of log Ktherm values and protonation constants of the ligand. Consequently, log Kcond describes the position of the equilibrium at physiological pH 7.4. It should be noted that the values for log Ktherm and log Kcond reported in the literature can differ slightly as they depend on the methods used to determine these constants (potentiometric measurement or competition experiments) and the conditions used (ionic force, nature of the electrolyte, temperature) (Moreau et al. 2004).

The conditional thermodynamic stability constant is defined as:
$$ K_{{{\text{cond}}}} = K_{{{\text{therm}}}} *[{\text{L}}]/{\text{L}}_{T} $$
where LTis the total concentration of the unchelated ligand, i.e. {L + [HL] + [H2L] + ···} where [HL], [H2L] are the protonated forms of the free ligand species.

As the thermodynamic stability constants reflects the affinity of gadolinium for its ligand (at high basic pH for log Ktherm and pH 7.4 for log Kcond), the higher the thermodynamic stability constants (log Ktherm or log Kcond), the more stable is the chelate and the less free gadolinium ion and free ligand are present when given enough time to reach thermodynamic equilibrium.

As there is a significant proton competition at pH 7.4 depending on the basicity of the ligand, which differs between ligands, the conditional thermodynamic stability constant log Kcond at pH 7.4 is always substantially lower than the overall thermodynamic constant log Ktherm as shown in Table 1.

Thermodynamic stability constants and conditional stability data are undoubtedly useful in terms of formulation, where formulated solutions have weeks to equilibrate at the pH of the formulation (the shelf-life of marketed gadolinium contrast agents is 3 years) (Tweedle 1992). Because of their relatively low conditional stability, pharmaceutical solutions of some gadolinium chelates include a variable amount of free ligand, or sodium or calcium salt (Table 1). These excipients are intended to ensure the absence of free Gd3+ cations in the pharmaceutical solution throughout their shelf-life (Idée et al. 2006); in the case of Gd-BT-DO3A, the calcium chelate was introduced to avoid certain metallic impurities appearing during the sterilization process (Schmitt-Willich 2007).

However, the need for free ligand or calcium chelate in the formulation of less stable gadolinium chelates can have an impact in terms of in vivo biocompatibility. Corot et al. (1998) demonstrated that Gd-DTPA and Gd-DTPA-BMA significantly inhibit angiotensin-converting enzyme (ACE) activity at the clinical dose of 0.1 mmol/kg, whereas no significant effect was observed for the two macrocyclic chelates Gd-DOTA and Gd-HP-DO3A. This effect were mainly due to the presence of free ligand and calcium chelate in the formulation of Gd-DTPA and Gd-DTPA-BMA, respectively, because the addition of Zn2+ in the same quantities suppressed their inhibitory effects. Puttagunta et al. (1996a) compared the effects of a clinical dose (0.1 mmol/kg) of Gd-DTPA, Gd-HP-DO3A and Gd-DTPA-BMA on serum and urine levels of Cu2+ and Zn2+ in 31 healthy volunteers (by using inductively-coupled AES). No significant difference was observed between the three chelates with respect to serum and urine Cu2+ levels (although a slight increase in Cu2+ excretion was found for the three molecules). Conversely, Gd-DTPA-BMA induced a drop in serum Zn2+ concentration (but not significant, due to the large inter-individual variability). Gd-DTPA-BMA caused the greatest increase in urinary zinc excretion (post-injection urinary zinc amount: 27.4 ± 9.2 μmol), followed by Gd-DTPA (5.9 ± 1.9 μmol) and Gd-HP-DO3A (1.2 ± 0.5 μmol). The excess ligand present in the formulation of the contrast agents (Table 1) could at least partly explain the high urinary zinc excretion reported. In the case of Gd-DTPA-BMA and Gd-HP-DO3A, Zn2+ may have replaced Ca2+ on the Ca-ligand in excess and would subsequently be excreted in urine in the form of Zn-DTPA-BMA or Zn-HP-DO3A. For Gd-DTPA, the additional free DTPA may have chelated blood Zn2+. More recently, the influence of gadolinium chelates on urinary excretion of Zn2+ and Cu2+ (assayed by atomic absorption spectrometry) was studied in a prospective randomised trial where patients were divided into three groups receiving either Gd-DTPA-BMA, or Gd-DTPA or Gd-DOTA. A fourth group of patients, undergoing MRI but without gadolinium chelate, was used as a control (Kimura et al. 2005). The linear open chain chelate Gd-DTPA-BMA caused the greatest increase in urinary zinc excretion among the three agents tested, whereas the macrocyclic agent Gd-DOTA had no effect on zinc excretion and Gd-DTPA induced an intermediate effect. No significant difference in Cu2+ excretion was observed between the three gadolinium chelate. This result could be explained by a higher blood concentration of Zn2+ (10–50 μmol/l) vs. Cu2+ (1–10 μmol/l) (Mann 1993). This study is fully consistent with the data reported in rats by Corot et al. (1998), where an increase in urinary Zn2+ excretion was observed after injection of Gd-DTPA and Gd-DTPA-BMA due to the presence of free ligand and calcium chelate in the respective formulations.

Structure/thermodynamic relationships

All teams that developed contrast agents for imaging applications endeavoured to increase log Ktherm and log Kcond to reduce the risk of dissociation of the chelate into free gadolinium and free ligand. However, it is difficult to achieve high thermodynamic stability with the gadolinium ion because its valency f orbitals are inner orbitals with no directing effects. The stability of lanthanide chelates is therefore essentially derived from electrostatic interactions between the gadolinium ion considered to be a “spherical” hard acid ion and the donor groups of the chelating agent (such as carboxylate considered to be a hard base). Since chemical bonds in gadolinium chelates are predominantly ionic, stable chelates are formed with multidentate polyaza-carboxylate chelates and three main structural factors influencing the thermodynamic stability of gadolinium chelates have been described:
  1. (1)

    The basicity of the polyaza-carboxylate scaffold, which can be evaluated by calculating the sum of the protonation constants of each donor atom of the ligand. This overall basicity (ΣpKa) is directly correlated with the intensity of electrostatic interactions between the metal and the donor atoms of the ligand, and therefore with the stability of the chelate. This assertion is supported by a linear correlation between log Ktherm vs. ΣpKa observed for linear open chain and also macrocyclic polyamino-carboxylates, which form five-membered chelate rings (Kumar et al. 1994, 1995).

  2. (2)

    The number of five-membered rings (N–Gd–N and N–Gd–O) (Mann 1993) formed by the chelate between the metal and the various donor atoms of the ligand as five-membered rings minimizes the steric strain in the chelate (Meyer et al. 1988).

  3. (3)

    The macrocyclic effect (Desreux 1980), which is related to the cavity size of the chelate ring and also to the preorganization, rigidity and conformation of the ligand. Macrocyclic ligands appeared particularly well suited for chelation of “spherical” gadolinium because these ligands feature an internal cavity with oxygen and/or nitrogen binding sites and are able to wrap themselves around the gadolinium. NMR experiments have demonstrated that DOTA ligand is a highly preorganized structure encouraged by the stability of the [3333] conformation of the macrocycle (Desreux 1980). To illustrate this macrocyclic effect, the spider-like DOTA ring very tightly incorporates the gadolinium ion, as demonstrated by molecular calculation of binding cavities using Connolly surfaces (Fossheim et al. 1991) and forms a tightly packed chelate with marked rigidity (Kumar 1997).

These structural aspects have major consequences on the thermodynamic stability of the available chelates:

(1) Macrocyclic chelates are characterized by quite different thermodynamic constants which can be correlated with their ionic or nonionic chemical structure, as the electrostatic interactions between the metals and ligands would be drastically decreased if the macrocyclic rings were substituted with nonionizable pendant arms. This point is clearly illustrated by the fact that the stability constants of Gd-DOTA are 2–4 orders of magnitude higher than those of Gd-HP-DO3A and Gd-BT-DO3A. This difference is a direct consequence of the higher negative charge of the DOTA4− ligand, which results in a stronger electrostatic metal ligand interaction than that in nonionic macrocyclic chelates. The difference of stability between the two nonionic macrocyclic chelates, Gd-HP-DO3A and Gd-BT-DO3A, is due to a weaker metal-alcoholic OH interaction in Gd-BT-DO3A. As coordination of the bulky group containing the three OH groups of Gd-BT-DO3A is sterically hindered, the coordinated alcoholic OH group is more acidic in Gd-BT-DO3A (pKOH = 9.48) than in Gd-HP-DO3A (pKOH = 11.36) and this factor leads to a significant loss of thermodynamic stability for Gd-BT-DO3A (Toth et al. 1996).

(2) The linear open-chain ligands do not benefit from the macrocyclic effect and the number of five-membered rings (N–Gd–N and N–Gd–O) is reduced (seven) in comparison to the macrocyclic series (eight). These two factors explain the difference in terms of thermodynamic stability between the ionic linear open-chain chelates (Gd-DTPA, Gd-EOB-DTPA, MS325, Gd-BOPTA) and the macrocyclic ionic chelate Gd-DOTA. The BOPTA, EOB-DTPA and MS325 ligands possess the same number and type of donor atoms as DTPA and the resulting stabilities of these linear open-chain gadolinium chelate are fairly similar. As nonionic donor atoms are less basic and consequently bind more weakly than ionic donor atoms, replacement of the two anionic carboxylate groups of Gd-DTPA by a nonionic amide group coordinating gadolinium via the neutral carbonyl oxygen in Gd-DTPA-BMA and Gd-DTPA-BMEA results in a marked (threefold) decrease in the stability of gadolinium chelates (Brücher and Sherry 2001).

Equilibrium calculation and thermodynamic transmetallation: is thermodynamic selectivity useful?

Blood and other body fluids are relatively complex milieux containing a very large number of ligands or endogenous metal ions. Consequently, various competitive reactions can occur in vivo. For example, in vivo, endogenous cations (Fe3+, Ca2+, Zn2+, Cu2+) can react with gadolinium chelates by displacing Gd3+ in a metal-metal transmetallation exchange. The free Gd3+ is then complexed by metal-binding proteins, endogenous ions (citrate, glutaminate) or endogenous anionic precipitants (CO32−, OH, PO43−), whereas the free ligand reacts with various endogenous metal ions (Idée et al. 2006). In order to understand the in vivo fate of gadolinium chelates, some authors (Cacheris et al. 1990; Sarka et al. 2000; Jackson et al. 1990; Chang 1993) have tried to quantify the release of free Gd3+ using a broad thermodynamic approach using calculations based on a simplified biospeciation plasma model. For this purpose, they defined the thermodynamic selectivity constant (Ksel) which addresses the possibility of thermodynamic transmetallation according to the following equilibrium:
$$ {\text{GdL}} + [{\text{M}}]^{{2 + }} \rightleftarrows {\text{Gd}}^{{3 + }} + [{\text{ML}}]^{ - } $$
where M is the endogenous bivalent cation which exchanges with Gd3+ in the chelate.

In fact, different and confusing definitions of the selectivity constant have been reported in the literature (Caravan 1999).

One definition assumes that the logarithmic selectivity constant S (sometimes called log Ksel), of the ligands for Gd3+ over other metal ions is the difference between log Ktherm and the stability constant of the endogenously available metal ion chelate, log Ktherm (ML) (where M = Ca2+, Cu2+, Mn2+, Fe3+ and Zn2+), individually (Table 2) (Kumar 1997; Kumar et al. 1995). This selectivity constant can also be determined at pH 7.4 (called S*), (De Haen et al. 1999) and can be calculated using log Kcond and log Kcond (ML) (where M = Ca2+, Cu2+, Mn2+, Fe3+ and Zn2+). Another definition of selectivity constant log Ksel (sometimes called log \( K^{\prime }_{{{\text{sel}}}} \) (Kumar 1997; Kumar et al. 1995) was also introduced and represents an overall selectivity, which is a hybrid calculated constant that takes into account all of the competition reaction (Gd3+, H+, OH, Ca2+, Zn2+, Cu2+, etc) for the ligand studied. \( K^{\prime }_{{{\text{sel}}}} \) is defined according to equation 5 (where K1, K2, K3 are the consecutive ligand protonation constants) with Ca2+, Zn2+ and Cu2+ concentrations corresponding to physiological plasmatic concentration i.e. 2.5 mM, 1 μM and 50 μM, respectively (Kumar et al. 1995):
$$ K^{'}_{{{\text{sel}}}} = K_{{{\text{(ML)}}}} {\left( {\alpha _{{\text{H}}} ^{{ - 1}} + \alpha _{{{\text{CaL}}}} ^{{ - 1}} + \alpha _{{{\text{CuL}}}} ^{{ - 1}} + \alpha _{{{\text{ZnL}}}} ^{{ - 1}} } \right)} $$
$$ \begin{aligned}{}\alpha _{{\text{H}}} ^{{ - 1}} = & 1 + {\text{ K}}_{1} [{\text{H}}^{ + } ] + {\text{K}}_{1} {\text{K}}_{2} [{\text{H}}^{ + } ]^{2} + {\text{K}}_{1} {\text{K}}_{2} {\text{K}}_{3} [{\text{H}}^{ + } ]^{3} + \cdot \cdot \cdot \\ \alpha _{{{\text{CaL}}}} ^{{ - 1}} = & 1 + {\text{K}}_{{{\text{CaL}}}} [{\text{Ca}}^{{2 + }} ] \\ \alpha _{{{\text{CuL}}}} ^{{ - 1}} = & 1 + {\text{K}}_{{{\text{CuL}}}} [{\text{Cu}}^{{2 + }} ] \\ \alpha _{{{\text{ZnL}}}} ^{{ - 1}} = & 1 + {\text{K}}_{{{\text{ZnL}}}} [{\text{Zn}}^{{2 + }} ] \\ \end{aligned} $$
Table 2

Thermodynamic stability constant log Ktherm (ML) and log Kcond (ML) of several chelates, according to the cation (Mn2+, Ca2+, Cu2+, Zn2+, Fe3+) (De Haen et al. 1999; Caravan et al. 1999)

Ligand L

Log Ktherm (CaL)

Log Ktherm (CuL)

Log Ktherm (ZnL)

Log Kcond (MnL)

Log Kcond (CuL)

Log Kcond (ZnL)

Log Kcond(FeL)

















































With the initial goal to try to predict relationships between thermodynamic constants and toxicity, Cacheris et al. (1990) used an eight-component plasma model to try to simulate the equilibria occurring in plasma and incorporating the endogenous cations H+, Ca2+, Cu2+ and Zn2+ (iron was not considered), citrate, amino acids and albumin. According to Cacheris’ calculations based on the selectivity constant log \( K^{\prime }_{{{\text{sel}}}},\) Gd-DTPA-BMA releases approximately one half as much as Gd3+ as Gd-DTPA and Gd3+ release was predicted to mainly result in Gd-citrate (phosphate was not taken into account although GdPO4 precipitates) (Jackson et al. 1990). According to their biospeciation model, the overall selectivity constants log \( K^{\prime }_{{{\text{sel}}}} \) were situated in the order Gd-DTPA-BMA > Gd-DTPA which parallels the order observed for LD50 for these two linear open-chain chelates (and two other linear open-chain chelates). In contrast, no correlation was observed between Ktherm values and their rodent LD50 values (intravenous LD50, injected in divided doses in the case of Gd-DTPA-BMA, in contrary with the classical method and therefore not comparable with other published LD50). Consequently, Cacheris et al. suggested that an increase in selectivity for Gd3+ over endogenous cations substantially contributes to the high LD50 of Gd-DTPA-BMA (it should be noted that other parameters can also contribute to acute toxicity, such as osmolality, viscosity, injection rate, strain of animals) and, strictly speaking, only compounds that have been tested under rigorously similar conditions should be compared in terms of LD50. Furthermore, LD50 values reflect immediate toxicity and therefore cannot be used to compare molecules in terms of their potential to trigger long-term toxic effects such as those supposed to be associated with dechelation. Despite this calculated increase in selectivity of Gd-DTPA-BMA, in the same article, the authors attempted to minimize transmetallation of this gadolinium chelate. For this purpose, they studied the influence of the formulation of Gd-DTPA-BMA on acute toxicity LD50 (once again determined after injection in divided doses). Acute toxicity of non-formulated Gd-DTPA-BMA was found to be 14.8 mmol/kg in mice. The addition of 5 mol% Na[Ca-DTPA-BMA] to the Gd-DTPA-BMA solution dramatically reduced this toxicity (LD50 of the formulated solution = 38.3 mmol/kg) (Cacheris et al. 1990). The authors proposed a possible mechanism: displacement of non-toxic Ca2+ from Na[Ca-DTPA-BMA] by endogenous Zn2+ which, under these conditions, is no longer available for Gd3+ transmetallation. This hypothesis was confirmed in a subchronic toxicity study of Gd-DTPA-BMA (Cacheris et al. 1990), in which animals receiving a dosage of 5 mmol/kg showed signs of zinc depletion. Following injection of Na[Ca-DTPA-BMA] to the rats, 7% of the initial molecule was excreted in the urine in the form of Zn2+ complex (and 1% in the form of Cu2+ complex) (Okazaki et al. 1996), thereby confirming that the previously hypothesized displacement probably occurs in vivo and that despite high overall selectivity constants log \( K^{\prime }_{{{\text{sel}}}} \) (100 times higher than for Gd-DTPA), Gd-DTPA-BMA is transmetallated in vivo by Zn2+.

The experimental results reported by Puttagunta et al. (1996b) contradict the prediction of Cacheris’ biospeciation model and demonstrate that the selectivity between Gd3+ and Zn2+ in an in vitro model is substantially different from that calculated from thermodynamic data. In an in vitro transmetallation assay, in which solutions of gadolinium chelates were mixed with an equimolar solution of zinc citrate in 1:1 ratio at pH 7.4, these authors found that the selectivity of Gd-DTPA-BMA for Zn2+ was more than tenfold lower than that calculated by Cacheris (Puttagunta et al. 1996b). This contradiction can be explained by the fact that the equilibrium model proposed by Cacheris neglected the formation of binuclear chelates. In reality, DTPA-BMA (like DTPA (Tweedle et al. 1988) presumably forms binuclear complexes with Zn2+, which decrease the selectivity for Gd3+ over Zn2+ (Brücher 2002). Moreover, an additional limitation in the thermodynamic model of blood pool plasma used by Cacheris is that it does not consider iron as a gadolinium competitor and, above all, phosphate as a gadolinium precipitant.

In another study, Sarka et al. (2000) used a simplified plasma model in which only competition between Gd3+ and Zn2+ was considered to study the release of gadolinium from Gd-DTPA. This equilibrium calculation indicated partial displacement of Gd3+ by Zn2+ and a large amount of Gd3+ displaced in the form of Gd-citrate (9.2%), as only 90.7% of gadolinium was still chelated in the form of Gd-DTPA.

The practical application of this type of equilibrium model based on selectivity constants is limited by several main factors at the present time (Sarka et al. 2000; Bianchi et al. 2000):
  • The ability to simulate all of the equilibria that are expected to occur in vivo and then to define a relevant plasma model.

  • In addition, these thermodynamic models do not consider the kinetic aspects related to dechelation and transmetallation. As mentioned in the initial paper (Cacheris et al. 1990; Chang 1993), the biospeciation model is not adapted to gadolinium chelates with high kinetic stability such as Gd-DOTA as the thermodynamic equilibrium cannot be reached due to the kinetic inertia of this macrocycle chelate. Therefore, according to the authors themselves, log \( K^{\prime }_{{{\text{sel}}}} \) has no physiological meaning for Gd-DOTA. Numerous experimental data have now demonstrated that the kinetic stability of gadolinium chelates is also an important parameter to predict the fate of gadolinium chelates in vivo. It has been suggested that the relative tolerance of gadolinium chelates can be explained not only by selectivity constants, but also by kinetic inertia, acid-assisted dissociation and metal exchange rates (Kumar et al. 1995) and that, like plasma equilibrium models, kinetic models taking into account excretion rates and distribution rates in the extracellular space of gadolinium chelates need to be developed in order to predict the amount of free gadolinium that dissociates in vivo (Brücher 2002).

Kinetic stability

High thermodynamic stability constants measured in water are neither necessary nor sufficient conditions for in vivo stability (Tweedle et al. 1991), as kinetic stability is also clearly an important parameter for understanding relative in vivo dissociation by describing:
  • Either the dechelation kinetics corresponding to the reaction:
    $$ {\text{GdL}} \rightleftarrows {\text{Gd}}^{{3 + }} + {\text{L}} $$
This reaction can occur via two mechanisms: spontaneous or proton-assisted dissociation of the Gd3+ chelate possibly catalysed by endogenous metals (Sarka et al. 2000; Robic et al. 2007, unpublished results)
  • or the transmetallation kinetics of the gadolinium chelate by endogenous metals, particularly Ca2+, Cu2+ and Zn2+) corresponding to the reaction:
    $$ {\text{GdL + [M}}]^{{2 + }} \rightleftarrows {\text{Gd}}^{{3 + }} + [{\text{ML}}]^{ - } $$
This metal exchange reaction can occur via two mechanisms (Brücher 2002)
  • Spontaneous or proton-assisted dissociation of the Gd3+ chelate followed by rapid reaction of the free ligand with endogenous metal ions.

  • Direct attack of the endogenous ion on the chelate via formation of a dinuclear intermediate GdLM followed by the formation of ML.

Once the importance of kinetic stability of the Gd3+ chelates was recognized, various types of studies were performed to describe the structural aspect driving the dechelation and transmetallation kinetics, which are significantly different according to each marketed gadolinium chelate.

Kinetic stability: in vitro studies

The dissociation rates of gadolinium chelates are low at pH 7.4 and the chelates dissociate much more rapidly in acidic solutions (Brücher 2002). Consequently, the rate of spontaneous or proton-assisted dissociation of the Gd3+ chelate was studied by measuring the rate of Gd3+ dissociation in the presence of 0.1 M HCl (pH 1.0) at 25°C (Tweedle 1992). In this type of acidic medium, dissociation of the chelate is a pseudo-first order reaction, with:
$$ {\text{Dissociation rate}} = k_{{{\text{obs}}}} {\text{ }}[{\text{GdL}}] $$
$$ k_{{{\text{obs}}}} = k_{0} + k_{1} [{\text{H}}^{ + } ] + k_{2} [{\text{H}}^{ + } ]^{2} $$
Neglecting k0 and k2, the kobs constant has been determined for several chelates by calculating the slope of the curves ln[GdL]= f(t) by Tweedle (1992). The results obtained demonstrated that the half-lives dramatically differed and were considerably longer for macrocyclic chelates (T1/2 > 1 month for Gd-DOTA, 3 h for Gd-HP-DO3A) than for linear open chain chelates (T1/2 = 10 min for Gd-DTPA and ∼35 s for Gd-DTPA-BMA). However, these authors reported different kobs constant for the same experiment in successive publications (Tweedle 1992; Kumar et al. 1993; Wedeking et al. 1992; Kumar and Tweedle 1993) that were also different to that measured by Pulukkody et al. (1993). Schmitt-Willich reported Bayer Schering’s dissociation half-life T1/2 values at pH 1.0 for Gd-HP-DO3A and Gd-BT-DO3A, but unfortunately not for Gd-DOTA (Schmitt-Willich 2007). As these T1/2 values depend on the experimental conditions used (pH, temperature, ionic strength, metal concentration) and the assay method (free gadolinium, free ligand, proton consumption due to protonation of the released ligand), in an attempt to be exhaustive, we measured the kobs and T1/2 values of the three macrocyclic compounds by using a strict comparative protocol in our laboratory (arzenazo dosage) (Table 3). This comparative study demonstrated that, under the experimental conditions used (pH 1.0 and 1.2, temperature 37°C and 25°C), the kinetic stability of macrocyclic gadolinium chelates can be classified in the following order: Gd-DOTA > Gd-BT-DO3A > Gd-HP-DO3A and that, as expected, the kinetics are more rapid at 37°C in comparison to 25°C. This is clearly illustrated by Fig. 1 showing the decrease of the remaining percentage of chelate as a function of time at pH 1.2, 37°C. Under these experimental conditions, the T1/2 for all linear open-chain chelates was <5 s (Table 3), clearly demonstrating their lower kinetic stability in comparison to that of macrocyclic chelates.
Table 3

T1/2 values of gadolinium chelates


T1/2 pH = 1.2, 37°C (h)

T1/2 pH = 1.0, 37°C (h)

T1/2 pH = 1.0, 25°C (h)













Linear open-chain chelates



<5 s

The kinetic stability of linear open-chain chelates such as MS325 (Caravan et al. 2001), Gd-DTPA (Sarka et al. 2000), Gd-DTPA-BMEA (Rothermel et al. 1997) and Gd-DTPA-BMA (Sarka et al. 2002) was assessed by characterizing the rates of the exchange reactions with Eu3+. From a mechanistic point of view, the dissociation rate of the bisamide linear open-chain gadolinium chelates Gd-DTPA-BMA and Gd-DTPA-BMEA is higher because Gd3+-amide oxygen bonds are weaker than Gd3+-carboxylate oxygen bonds (White et al. 1991) and consequently the protonated intermediate, which is important for the rate-controlling step of dissociation, can be more easily formed (Sarka et al. 2002).

Various other studies have addressed the kinetic stability of the three macrocyclic gadolinium chelates (Kumar et al. 1993; Wang et al. 1992), but only two studies have used the same protocol for Gd-DOTA (Toth et al. 1994), Gd-BT-DO3A, and Gd-HP-DO3A (Toth et al. 1996). The kinetic stability of the three macrocyclic gadolinium chelates was studied under the same conditions by UV spectroscopy by following the exchange reactions between high concentration of gadolinium chelates and very low concentration of Eu3+. Under these conditions, the exchange rate does not depend on the europium concentration and exchange only occurs via proton-assisted dissociation of the gadolinium chelate. Interestingly, the dissociation rate was studied in the pH range 3.2–5.3 (at 37°C for Gd-DOTA and at 25°C for Gd-BT-DO3A and Gd-HP-DO3A), which is obviously more relevant in terms of physiology compared to the studies discussed above, performed at pH 1 or 1.2. Under these conditions, the exchange is considered to be a pseudo-first order reaction characterized by the constant kobs:
$$ {\text{d}}[{\text{GdL}}]/{\text{d}}t = k_{{\rm obs}} [{\text{GdL}}] $$
In the pH range studied, the first order constant kobs is directly proportional to the H+ concentration
$$ k_{{\rm obs}} = {\text{ }}k_{0} + k_{1} {\text{ }}[{\text{H}}^{ + } ] $$
The rate constants k0 and k1 are characteristic of spontaneous and proton-assisted dissociation of gadolinium chelates (Brücher 2002). The dissociation of these macrocyclic chelates, which is much less important than that of linear open-chain chelates (Varadarajan et al. 1994), is also a proton-assisted process and the observed linear dependence of the exchange rate (kobs) on the H+ concentration has been interpreted in the same way for the three macrocyclic chelates. In weakly acidic solutions (3.2 < pH < 5), the exchange rate is linearly proportional to the concentration of a monoprotonated species. In this monoprotonated chelate, the protonation of a carboxylate oxygen reduces the kinetic stability of the chelate, but, prior to dissociation the proton is transferred to a nitrogen atom of the ligand and Gd3+ steps out of the coordination cage. This proton transfer, which proceeds via an intramolecular rearrangement of the protonated chelate, is probably the rate-controlling step and very slow due to the rigid structure of macrocyclic chelates. After or during this rearrangement, a second proton is immediately protonated to another nitrogen atom forming a diprotonated intermediate, while the Gd3+ ion leaves the coordination cage due to the electrostatic repulsion between the protons and Gd3+. As previously shown at pH 1 and 1.2, the data obtained between pH 3.2 and 5 (Table 4) demonstrate that the sequence of kinetic stability of the gadolinium macrocyclic chelates is as follows: Gd-DOTA > Gd-BT-DO3A (the dechelation kinetics are accelerated when the temperature increases: Table 3).
Table 4

Rate constants characterizing the proton-assisted dissociation of the macrocyclic chelates

Gadolinium chelates

Temperature (°C)

k0 (s−1)

k1 (M−1 s−1)




5 × 10−10

2 × 10−5

Toth et al. (1994)



−7 × 10−10

2.6 × 10−4

Toth et al. (1996)



2 × 10−10

2.8 × 10−5

Toth et al. (1996)

Several studies have been performed to determine the dechelation or transmetallation kinetics in physiological milieu (or a milieu mimicking biological fluids).

As described above, in vivo, several endogenous cations such as Fe3+, Cu2+, Zn2+ or Ca2+ can compete with the Gd3+ cation for the ligand and a number of other anions such as PO43−, CO32−, OH, can compete with the ligand for the Gd3+ (Tweedle et al. 1988). An in vitro test was performed, in which precipitated GdPO4 was measured 15 min after mixing 25 mM of macrocyclic and linear open chain gadolinium chelates with 25 mM of endogenous Cu2+ or Zn2+ cations in a phosphate medium (at pH 7.0 and at room temperature) (Tweedle 1992; Tweedle et al. 1991). The formation of insoluble GdPO4 is the driving force of this in vitro stress (Chang 1991). The experimental data showed that macrocyclic chelates, Gd-DOTA and Gd-HP-DO3A, are kinetically inert unlike linear open-chain chelates. Dechelated Gd3+ was not observed for the macrocyclic chelates Gd-DOTA and Gd-HP-DO3A for Cu2+ or Zn2+ cations, whereas a significant release of Gd3+ was observed for the linear open-chain chelates Gd-DTPA-BMA and Gd-DTPA (35% and 25% for Cu2+ and 25% and 21% for Zn2+, respectively).

In another in vitro transmetallation assay, solutions of various gadolinium chelates were mixed with an equimolar solution of zinc citrate in a 1:1 ratio at pH 7.4 (Puttagunta et al. 1996b). Kinetic stability was higher for the macrocyclic molecule Gd-HP-DO3A than for Gd-DTPA. The lowest kinetic stability in the presence of Zn2+ was observed for Gd-DTPA-BMA. In a second experiment, the transmetallation kinetics were assessed by measuring the changes in the concentration of Gd-chelates at various time-points following mixing with equimolar zinc citrate. Interestingly, formulated Gd-DTPA-BMA demonstrated the immediate appearance of Gd citrate when mixed with equimolar Zn2+ citrate (Puttagunta et al. 1996b).

Magerstät et al. (1986) performed serum stability studies for DOTA, DTPA chelates of 153Gd incubated at 37°C. Loss of Gd3+ from the chelates was evidenced by measuring the radioactivity in the precipitates. Over the 150-h observation period, Gd-DOTA lost almost no radioactivity, whereas Gd-DTPA lost about 10–20% of free radioactive gadolinium. The observed difference between the two chelates in this study was most likely due to the difference in terms of dissociation kinetics between the macrocyclic and linear open-chain gadolinium chelates.

Sarka et al. assessed the kinetic stability of linear open-chain chelates Gd-DTPA (Sarka et al. 2000), and Gd-DTPA-BMA (Sarka et al. 2002) in an in vitro transmetallation assay by characterizing the rates of the exchange reactions with endogenous cations or Cu2+ or Zn2+. Under physiological conditions, the exchange reaction mainly occurs via direct attack of Cu2+ and Zn2+ on the chelate for Gd-DTPA, but proton-assisted dissociation is still important for Gd-DTPA-BMA. During direct attack of Cu2+ and Zn2+, a dinuclear intermediate is formed, in which the functional groups of the ligand are gradually and slowly transferred from Gd3+ to the attacking metal ion step by step. Although the plasma Cu2+ concentration is very low, Cu2+ has been found to be kinetically very active as Cu2+ is 10–80 times more effective than Zn2+ in this exchange reaction. The authors concluded that the contribution of proton-assisted dissociation is about tenfold higher for Gd-DTPA-BMA than for Gd-DTPA, but Zn2+-assisted dissociation is about ten-fold faster for Gd-DTPA than for Gd-DTPA-BMA, which compensates for the difference observed for proton-assisted dissociation. Consequently, when all possible exchange reactions are taken into account (i.e. proton-assisted dissociation and Cu2+-, Zn2+-assisted dissociation), the two chelates Gd-DTPA and Gd-DTPA-BMA have similar overall dissociation rates. To explain these results, the authors hypothesized that although the binding strength of the amide group is weaker than that of a carboxylate, its reactivity with H+, Zn2+ or Cu2+ is lower. However, these conclusions are in contradiction with the results obtained in vivo and by the in vitro relaxometric method (see below).

The kinetic stability of Gd-DOTA, Gd-HP-DO3A, Gd-DTPA, Gd-DTPA-BMA and Gd-EOB-DTPA in solutions containing metabolites such as adenosine triphosphate (ATP), phosphocreatine (PCr), inorganic phosphate (Pi) found at the relatively high concentrations observed in the cytosol was studied by measuring the longitudinal P-31 relaxation rate enhancements of phosphate groups (Vander Elst et al. 1994, 1997). No kinetic dechelation of gadolinium ion was observed for macrocyclic chelates, but the effects observed for the two linear open-chain gadolinium chelates indicate that ATP, which is known to be an extremely good chelator of metals, competes with DTPA, DTPA-BMA and EOB-DTPA for chelation of Gd3+ ions, indicating a poor kinetic inertia of these chelates. Under these experimental conditions, Gd-DTPA-BMA was shown to be less stable than Gd-DTPA which was also less stable than Gd-EOB-DTPA. Gd-BOPTA was not studied in this protocol. However, another study demonstrated that ATP interacts with Gd-BOPTA at physiological pH (Bianchi et al. 1999).

An easy and reliable in vitro relaxometric method has recently been validated to quantitatively evaluate the release of Gd3+ from gadolinium chelates (Laurent et al. 2001) in the presence of Zn2+ at pH 7.4. This method is based on the following principle: if dechelation of a soluble gadolinium chelate by diamagnetic Zn2+ ions were to occur in a phosphate-buffered solution, then the released Gd3+ would react to form precipitated GdPO4 which has a negligible influence on the relaxation rate of water. Therefore, a decrease of relaxation rates R1p(t) would occur, with time and the magnitude of this decrease would be related to the proportion of precipitated GdPO4. The time-course of the R1p(t)/R1p(t = 0) ratio of macrocyclic and linear open-chain chelates is presented in Fig. 2 as recently published (Laurent et al. 2006) and completed by values obtained with MS325, Gd-EOB-DTPA and Gd-DTPA-BMEA under strictly similar conditions (Robic et al. 2007, unpublished results). As previously proposed (Laurent et al. 2006), 3 classes of gadolinium contrast agents can be determined using a “long time index” equal to the ratio of the paramagnetic relaxation rates after 50 h (Fig. 2):
  • Macrocyclic chelates, Gd-DOTA or Gd-HP-DO3A and Gd-BT-DO3A, characterized by very high kinetic inertia stability: long-term index >0.95.

  • Ionic linear open-chain chelates, Gd-DTPA, EOB-DTPA, Gd-BOPTA and MS325, with moderate kinetic inertia leading to significant dechelation in the presence of Zn2+ under these pseudo-physiological conditions (long-term index: 0.49 to 0.85).

  • Nonionic linear open-chain chelates Gd-DTPA-BMA and Gd-DTPA-BMEA which exhibit poor kinetic stability and the highest degree of dechelation (long-term index <0.3) of all commercially available gadolinium chelates.

Whether the reaction studied under these experimental conditions corresponds to transmetallation or dechelation catalysed by Zn2+ is still an unresolved question that is currently under investigation.
Fig. 2

Time-course of the R1p(t)/R1p(t = 0) ratio for Gd-DTPA, Gd-DTPA-BMA, Gd-DTPA-BMEA, Gd-BOPTA, Gd- EOB-DTPA, MS325, Gd-DOTA, Gd-BT-DO3A, Gd-HP-DO3A. Concentration of Gd chelates and ZnCl2 are 2.5 mM in phosphate buffer at pH 7.0 and 310 K (B0 = 0.47 T)

Kinetic inertia: structural aspect driving the kinetic stability or structure/stability relationships

Overall, these physicochemical and in vitro studies demonstrate that:
  • All linear open-chain gadolinium chelates are significantly less kinetically stable than macrocyclic polyaminocarboxylates.

  • The order of kinetic stability of macrocyclic gadolinium chelates is Gd-DOTA > Gd-BT-DO3A > Gd-HP-DO3A.

The low kinetic stability of linear open-chain chelates has been explained by the rapid unravelling of these chelates due to their conformational mobility (Kumar 1997). As donor atoms derived from nonionic amides are less basic, they bind more weakly to gadolinium than ionic carboxylate donor atoms (Brücher and Sherry 2001; Brücher 2002) probably introducing more flexibility and conformational mobility in gadolinium chelates and explaining the significant loss of kinetic stability for Gd-DTPA-BMA and Gd-DTPA-BMEA. The introduction of a functional group on the DTPA scaffold (EOB-DTPA, Gd-BOPTA and MS325) appears to slightly improve kinetic inertia. This may be explained by the steric effect of the bulky substituents which slightly hinders unwrapping of the ligand around the metal ion (Caravan et al. 2001).

The kinetic inertia of macrocyclic chelates probably stems from their tight packing and their high conformational rigidity (Tweedle et al. 1991; Meyer et al. 1998; Wang et al. 1992). Their exceedingly slow dissociation rate is attributed to the peculiar structure of macrocyclic chelates: because of the high rigidity of the tetraaza ring, the inversion rate of nitrogen atoms and the dissociation rate of the gadolinium-oxygen bonds are considerably reduced by the steric requirements of the macrocyclic ligand. The dissociation of macrocyclic chelates probably involves the formation of a protonated species which undergoes a slow rearrangement into an intermediate chelate to transfer the proton to a nitrogen ring atom. Proton transfer via this mechanism is very slow, as entry of a proton into the macrocyclic cage is markedly hindered because of the rigid structure of the macrocyclic chelate. The acid-catalyzed reaction is therefore much less effective than for linear open chain chelates (Brücher and Sherry 2001). Gd-DOTA is the most kinetically stable chelate, as ionic carboxylate binds more strongly to gadolinium and therefore more efficiently limits the conformational motions of the chelate than the alcohol chain of Gd-HP-DO3A or Gd-BT-DO3A (Meyer et al. 1998; Desreux and Barthelemy 1988). Pollet and Marx (2007), using ab initio molecular dynamic simulation in aqueous solution at ambient temperature, recently demonstrated that the hydroxypropyl arms of Gd-HP-DO3A which break the symmetry hemisphere, is dynamically flexible in contrast with the four Gd-DOTA carboxylates. The proton-assisted dissociation rate of Gd-BT-DO3A is about tenfold lower than that of Gd-HP-DO3A. In this case, the greater kinetic inertia of Gd-BT-DO3A may be due to the large size of its functional group, containing three alcoholic OH groups. The bulkiness of the “butrol” group could slow down and restrict the rearrangement and dissociation of the protonated chelate for steric reasons (Toth et al. 1996).

Kinetic inertia of gadolinium chelates: in vivo studies

The relevance of in vitro studies demonstrating the significant difference in kinetic inertia between various marketed gadolinium chelates can be challenged, as the experimental conditions are very different from in vivo conditions, which is why many papers have addressed this issue of the in vivo inertia of gadolinium chelates.

A study demonstrated higher levels of 153Gd in bones with Gd-DTPA than with Gd-DOTA 7 days and more after intravenous injection in mice (0.25 mmol/kg) (Wedeking and Tweedle 1988). The authors attributed this difference to either a different uptake and/or washout of intact gadolinium chelates by bone or bone marrow or higher in vivo dechelation for Gd-DTPA than for Gd-DOTA.

In vivo dissociation of gadolinium chelates was subsequently measured by quantifying the residual gadolinium in various organs of mice at different residence intervals (5 min to 14 days) following intravenous injection of 0.4 mmol/kg of radiolabelled gadolinium chelates (Tweedle 1992, Wedeking et al. 1992). At residence intervals > 60 min, significant differences were observed, especially concerning long-term distribution of 153Gd in the liver and femur. Significant amounts of residual gadolinium were found in the case of linear open-chain chelates, Gd-DTPA-BMA and to a lesser extent Gd-DTPA, indicating in vivo dissociation of gadolinium chelates, whereas the macrocyclic chelates Gd-DOTA and Gd-HP-DO3A resulted in much lower residual gadolinium levels. The amount of residual gadolinium at 7 days did not correlate with the thermodynamic conditional constant but a correlation was observed between acid dissociation rates measured at pH 1 and long-term deposition of 153Gd in the whole body, liver and femur of mice (Wedeking et al. 1992). This correlation probably confirms the relevance of measuring acidic dissociation rates even though the experimental conditions are very different from physiological pH. Nevertheless, in another study (McMurry et al. 1998), no correlation was demonstrated between the release of lanthanides and the acidic dissociation of several chelates based on the DTPA scaffold. This indicates that more detailed kinetic studies, taking into account the major endogenous ions such as Cu2+ and Zn2+, are probably needed to understand the release of free lanthanides in vivo. It should also be noted that the amount of residual gadolinium in the whole body found by Wedeking et al. (1992) was also correlated with the order of stability assessed by the relaxometric method in the presence of Zn2+ cations developed by Laurent et al. (2006) under experimental conditions much closer to physiological conditions (pH 7.4).

Subsequently, Tweedle et al. (1995) injected 153Gd-labelled formulated solutions of Gd-DTPA, Gd-DOTA, Gd-HP-DO3A and Gd-DTPA-BMA to mice and rats as well as the unformulated form of Gd-DTPA-BMA, and measured the residual 153Gd at various time-points up to 14 days after intravenous administration (0.1 mmol/kg). They found that unformulated Gd-DTPA-BMA showed considerably greater residual 153Gd at 7 and 14 days in whole body, liver and bone than the other chelates. The addition of 25 mmol/l Na[Ca-DTPA-BMA] (in the commercially available solution) to the Gd-DTPA-BMA solution reduced the residual 153Gd in the mouse whole body (from 1% to 0.3% of the injected dose), thus suggesting that Gd-chelate dissociation is actually responsible for the higher levels of residual 153Gd. Furthermore, this study clearly showed that the formulated solutions of the linear open chain chelates Gd-DTPA and especially Gd-DTPA-BMA led to higher amounts of residual 153Gd in whole body, liver and femur than those of the macrocyclic chelates Gd-HP-DO3A and Gd-DOTA at late time-points and the authors concluded that the order of residual whole body gadolinium at 14 days (lowest to highest) was: Gd-HP-DO3A # Gd-DOTA ≤ Gd-DTPA << Gd-DTPA-BMA (Tweedle et al. 1995; Tweedle 2007; Port et al. 2008).

Similar studies were also conducted (Harrison et al. 1993) and tissue distribution of 157/153Gd chelates was measured at 5 min and 24 h after an intravenous injection of 0.1 μmol/kg, (much lower than the dose commonly used for MRI) in mice. At 24 h post-injection, <0.05% I.D was detected for Gd-DOTA in the skeleton whereas the percent I.D. retained in the skeleton of injected mice was higher for Gd-DTPA and Gd-BOPTA (0.3% I.D and 0.22%, respectively). In this study, the percent of gadolinium retained in the skeleton was greater than that reported for a high dose in mice at 24 h (Wedeking et al. 1992). An explanation for this difference proposed by the authors is related to the susceptibility of linear open-chain chelates to cation (particularly zinc and copper) promoted dissociation. The amount of chelate dissociated would be limited by the amount of cations present in the plasma so that a smaller fraction of the high dose would dissociate in the study performed at 0.4 mmol/kg of radiolabelled gadolinium chelates (Tweedle 1992; Wedeking et al. 1992).

In vivo gadolinium release was evaluated for Gd-BOPTA, Gd-DTPA-BMA and Gd-BT-DO3A estimating gadolinium content following single or repeated intravenous administrations to rats at a dose of 1 mmol/kg (cumulative dose: 6 mmol/kg). Under these experimental conditions, the order of residual gadolinium found in these organs was Gd-DTPA-BMA > Gd-BT-DO3A > Gd-BOPTA (Bussi et al. 2007). These results can be explained by the dual (urinary and biliary) route of elimination of Gd-BOPTA considering that biliary excretion of Gd-BOPTA is significantly higher in rats than in humans (Lorusso et al. 1999; Spinazzi et al. 1999).

The human relevance of these animal data was demonstrated by Gibby et al. (2004) and White et al. (2006) who measured Gd3+ levels in bone collected after administration of a clinical dose of either Gd-HP-DO3A or Gd-DTPA-BMA in patients undergoing total hip arthroplasty. Bone tissue retention of Gd3+ was significantly higher in patients who received Gd-DTPA-BMA than in those who received the macrocyclic agent Gd-HP-DO3A, as demonstrated by inductively coupled plasma (ICP) atomic emission spectroscopy (Gibby et al. 2004) and, more recently, by ICP mass spectroscopy. In the latter study, approximately 4 times more Gd3+ remained in femoral head bone after Gd-DTPA-BMA than after Gd-HP-DO3A (White et al. 2006). However, the analytical method used is unable to define the status of gadolinium (chelated or precipitated as GdPO4).

Conclusion: what is the possible in vivo and clinical relevance of thermodynamic and kinetic data?

As mentioned above, thermodynamic stability has a practical relevance in terms of formulation, as formulated solutions have a shelf-life of several years to equilibrate at neutral pH (marketed gadolinium contrast agents have shelf-lives between 2 and 3 years).

In vivo, under physiological or pathological conditions, the amount of free gadolinium released will depend on either the dissociation rate (i.e. kinetic stability) and thermodynamic stability when the system cannot reach a thermodynamic equilibrium (kinetic phase) or mainly thermodynamic stability when the system is able to reach a thermodynamic equilibrium (thermodynamic state).

These points are illustrated in Fig. 3
  • Red and blue curves illustrate the percentage of free gadolinium released by two gadolinium chelates with the same thermodynamic stabilities, but different dissociation kinetics. At the thermodynamic plateau, the same amount of free gadolinium is released, but in the kinetic phase (i.e. time dependant part of the curves), the slower the dissociation kinetics, the less free gadolinium is released.

  • Red and black curves illustrate the percentage of free gadolinium released by two gadolinium chelates with the same kinetic constant, but different thermodynamic stabilities (log Kcond). At the thermodynamic plateau and kinetic phase, the amount of free gadolinium is significantly different and the amount of free gadolinium released decreases as log Kcond increases.

  • Red and pink curves illustrate the percentage of free gadolinium released by two gadolinium chelates with different dissociation kinetics and different thermodynamic stabilities. At both the thermodynamic plateau and the kinetic phase, the amount of free gadolinium is significantly different and less free gadolinium is released for more kinetically stable (low dissociation rate) and more thermodynamically stable (high log Kcond) chelates.

No validated mechanism for NFS has yet been identified, but several hypotheses have been proposed to explain this highly disabling disease. They are summarized in Table 5. One of the most widely accepted hypotheses (Perazella 2007; Morcos 2007a, b) postulates that, in the setting of renal failure, remanence of a Gd3+ chelate in the body is dramatically increased, thereby increasing the chance for dissociation of Gd3+ from its chelate especially in the case of nonionic, linear open chain molecules such as Gd-DTPA-BMA. This would allow increased tissue exposure. This is the only hypothesis consistent with the higher incidence of NSF cases associated with Gd-DTPA-BMA.
Fig. 3

Calculated percentage of free gadolinium released by various gadolinium chelates assuming a first-order mechanism

Table 5

Current hypotheses for the pathophysiology of NSF

• Possible involvement of an infectious agent or toxic contaminant, although none have been identified to date (Evenepoel et al. 2004);

• Treatment with calcineurin inhibitors such as cyclosporin which are known to increase the level of TGF-β (Khanna et al. 1997; Inigo et al. 2001). TGF-β may mediate the florid fibroblastic proliferation which characterizes NSF;

• Involvement of antiphospholipid and anticardiolipin antibodies (a marker of small-vessel thrombotic vasculopathy) and the HLA-A2 allele (Evenepoel et al. 2004; Mackay-Wiggan et al. 2003);

• Early dermal infiltration with bone marrow-derived CD45RO+/CD34+ circulating fibrocytes in response to macrophages that have phagocytosed Gd3+ from Gd3+ chelate following dechelation (Morcos 2007a, b; Perazella 2007);

• Immune response of the host to noxious stimuli involving dendritic cells and synthesis of TGFβ (as part of the response of the dendritic cells to this noxious agent). The TGFβ produced by these dendritic cells would, in turn, precipitate both the fibrotic process and the expansion and enhancement or initiation of antigen-presenting functions of further dendritic cells establishing a vicious circle (Mendoza et al. 2006);

• Resistance to erythropoietin (Swaminathan et al. 2006);

• Metabolic acidosis as a cofactor (Dharnidharka et al. 2006).

Consequently, the question to be raised is then which criterion, thermodynamic or kinetic stability, is more important to ensure the lowest Gd release in vivo?

Due to their relatively rapid dechelation or transmetallation kinetics, it has been argued that linear open-chain gadolinium chelates could reach a thermodynamic equilibrium (in contrast with macrocyclic chelates characterized by high kinetic stability) and in that case conditional stability and selectivity could be important physicochemical criteria to predict the amount of gadolinium released (as shown by the plateau in Fig. 3). However, this speculation should be relativised as, in physiological situations, fast rate of diffusion and fast rate of excretion of the gadolinium chelate do not guarantee that thermodynamic equilibrium is reached in vivo (Tweedle et al. 1991; Cacheris et al. 1990; Brücher 2002).

The situation is even more complex in vivo in pathological situations such as renal failure where the biodistribution is modified and excretion is slowed and where gadolinium chelates can be sequestrated in compartments characterized by a special biological composition in terms of proteins, ions and pH, etc as suggested in NSF (Thakral et al. 2007; Broome et al. 2007). In the case of compartmentalization, it can be speculated that a pseudo-equilibrium can be reached and consequently more thermodynamically stable chelates should be preferred in order to minimize any free gadolinium release.

Beyond this speculative debate on the possible achievement of a pseudo-equilibrium in vivo, it appears that high kinetic stability (i.e. macrocyclic chelates in the order of kinetic stability: Gd-DOTA > Gd-BT-DO3A > Gd-HP-DO3A) combined with high thermodynamic stability (i.e. macrocyclic chelates in the order of thermodynamic conditional stability: Gd-DOTA > Gd-HP-DO3A > Gd-BT-DO3A) will minimize the amount of free gadolinium released in vivo.


The authors thank Prof. R.N. Muller and his group for the experimental data shown in Fig. 2 concerning Gd-BT-DO3A, Gd-DTPA, Gd-BOPTA.

Copyright information

© Springer Science+Business Media, LLC. 2008