NMR Studies of the Metal-Loading Kinetics and ... - ACS Publications

454

Bioconjugate Chem. 1999, 10, 454−463

NMR Studies of the Metal-Loading Kinetics and Acid-Base Chemistry of DOTA and Butylamide-DOTA David A. Keire* and Mitsuo Kobayashi The Beckman Research Institute of the City of Hope, 1450 East Duarte Road, Duarte, California 91010-0269. Received October 30, 1998; Revised Manuscript Received January 19, 1999

The conjugation of a chelating agent to a protein via a covalent linkage has been previously reported to change the metal-binding characteristics of the chelator. A fundamental understanding of these binding changes would enable design of a new generation of metal-chelating agents for biological applications. To assess the effect of conjugation on the commonly used chelating agent 1 4,7,10tetraaazacyclododecane-N,N′,N′′,N′′′-tetraaacetic acid (DOTA), we synthesized a model protein conjugate, 1,4,7-tris(carboxymethyl)-10-(butylaminocarboxymethyl)-1,4,7,10-tetraaazacyclododecane (BD) and explored the metal-binding characteristics via NMR. The extent of ionization of the carboxylic acid groups and the two protonated macrocycle nitrogens of DOTA and BD were determined as a function of pH by chemical shift changes in proximal carbon-bonded protons. In addition to the expected sensitivity of the chemical shifts to titration of proximate acidic groups, BD resonances from carbonbonded protons 10-17 bonds distant from the deprotonation site also shifted significantly, indicating the presence of conformational changes. Furthermore, increased shielding of the amide and alkyl proton signals upon deprotonation of the carboxylic acid groups indicates the presence of pH-dependent hydrogen-bonded BD isoforms. On the basis of these NMR data, we propose new structures for the doubly protonated forms of DOTA and BD. To measure metal loading, the yttrium-loading rates (type I to type II) of DOTA and BD were determined by following the intensity of type I and type II proton signals as a function of time. The yttrium-loading rates of BD are approximately one-half those of DOTA at pHs between 4.6 and 6.5 and 37 °C. The loading rates measured as a function of pH indicate that while both the doubly protonated and singly protonated forms of DOTA are reactive to metal loading, only the singly protonated form of BD is reactive.

INTRODUCTION

The conjugation of metals to proteins via a chelating agent is of fundamental interest in the design of proteins with modified properties. Introduction of a metal center may be used to add radioactive, paramagnetic, or redox character to proteins that lack these properties. For example, radiometals attached to antibodies via a chelating agent have been used for the delivery of therapeutic radiation directly to tumor sites and in the development of new contrast reagents for magnetic resonance imaging [for review, see Alexander (1995)]. However, the conjugation of a chelating agent to protein surface amino groups alters the metal-binding characteristics of the chelating agent, and although a number of applications would benefit from greater understanding of how the metalbinding parameters are modified, very little data exist describing the nature of the changes that occur. To begin to address this lack of information, we compared the acid-base chemistry and the yttrium metal loading kinetics of 1,4,7,10-tetraazacyclododecane-N,N′,N′′,N′′′-tetraacetic acid, DOTA,1 to those of a model * To whom correspondence should be addressed. Phone: (626) 357-9711Ext3907.Fax: (626)301-8186.E-mail: [email protected]. 1 Abbreviations: DOTA, 1,4,7,10-tetraaazacyclododecaneN,N′,N′′,N′′′-tetraaacetic acid; butylamide-DOTA or BD, 1,4,7tris(carboxymethyl)-10-(butylaminocarboxymethyl)-1,4,7,10-tetraaazacyclododecane; DOTA-OSSu, N-hydroxsulfosuccinimide ester of DOTA; sulfo-NHS, N-hydroxsulfosuccinimide; EDC, 1-ethyl-3-[3-(dimethylamino)propyl]carbodiimide; MALDI-TOF, matrix-assisted laser-desorption ionization-time-of-flight mass spectrometry.

protein conjugate, butylamide-DOTA (or BD) by NMR methods. The alkylamide-substituted BD compound is clinically relevant because it mimics the lysine-side-chain protein-conjugate form of the chelating agent. Significant differences in pKas, kinetics, and structure are observed between the two compounds, and the importance of this to protein conjugate design is discussed. We chose DOTA for study because it is well-known to form very stable complexes with a wide variety of divalent and trivalent metal ions (Figure 1) (Loncin et al., 1986; Cacheris et al., 1987; Kumar et al., 1989; Broan et al., 1991; Clarke and Martell, 1991a,b). For example, DOTA binds Ca2+ and Y3+ with stability constants (log KLM ) [ML]/[M][L] at 25 °C) of 16.5 and 24.9 M-1, respectively (Alexander, 1995). The tetraaza macrocycle contains a rigid preorganized-binding pocket that allows the formation of highly stable metal chelates. However, at the same time, this rigid pocket leads to a rate of metal incorporation slower than more flexible (and less stable) macrocycles. For example, Gd3+ is incorporated into DOTA at a rate 100 times slower than it is into 1,4,7,10,13,16-hexaazacyclooctadecane-N,N′,N′′,N′′′,N′′′′,N′′′′′hexaacetic acid (Kodama et al., 1991). The effective ionic radius of the metal is also important to the rate of metal incorporation into DOTA. Larger ionic radii metals such as gadolinium (1.06 Å) and yttrium (1.02 Å) load slowly into DOTA because the binding requires a reorganization of the binding pocket. The metal-DOTA complex formation with these larger radii metals is a two-step process (Kasprzyk and Wilkins, 1982; Wang et al., 1992). Initially, a reversible type I

10.1021/bc980128u CCC: $18.00 © 1999 American Chemical Society Published on Web 03/19/1999

NMR Studies of DOTA and a Model Conjugate

Figure 1. The reaction scheme of the HDOTA3- with Y3+.

adduct forms (ca. k1 ) 50 M-1 s-1 for Gd3+), followed by a relatively slow forming type II complex (ca. k* ) 10-3 s-1 for Gd3+) that has the metal fully coordinated (Figure 1). The formation of the fully coordinated metal DOTA complex is highly pH dependent. The singly protonated form of DOTA (LH3-) is the most reactive species, and the doubly protonated LH22- form is 4 orders of magnitude less reactive (Wang et al., 1992). Desreux proposed that the slow kinetics of metal-DOTA complex formation is caused, in part, by the dynamics of conversion between isomers of metal-DOTA compounds (Desreux, 1980). Thus, the rate of metal-DOTA complexation is much higher at elevated temperature or with more flexible macrocycles because the rate of conformational change between isomers is more rapid (Desreux, 1980; Kodama et al., 1991). Despite a relatively slow metal complexation rate, DOTA has been actively studied for use as a metal chelator in biological systems due to its serum stability (Desphande et al., 1990; Li and Meares, 1993; Lewis et al., 1994). The metal-loading yield of free polyaza macrocycles normally exceeds 95% at room temperature in about an hour [e.g., Stimmel et al. (1995)], but it has been observed that upon conjugation to an antibody much lower metal-loading yields are realized. For example, Meares and co-workers incubated 88Y3+ with an antibodyDOTA conjugate containing a flexible linker for 3 h at room temperature and obtained a 26% radiolabeling yield (Li and Meares, 1993). More recently, two studies have shown that elevated temperature (>40 °C) and optimized pH and the buffer used can greatly improve the yttrium metal incorperation into DOTA-antibody conjugates (Lewis et al., 1994; Govindan et al., 1998). These studies suggest that the chelation properties of free DOTA are altered in protein-chelating agent conjugates and that these changes affect metal-loading efficiency. Currently, no accepted model explains the observed change in the metal-loading efficiency of DOTA upon conjugation to a protein. A number of possible factors have been proposed to explain the altered metal-loading rates, including the alkyl amide modification of DOTA in the conjugation reaction, competition from other protein metal binding sites, and steric hindrance caused by the conformation of the chelating agent at the protein surface. In this work, a butylamide-DOTA conjugate has been synthesized to explore the effects of an alkyl amide modification on DOTA metal loading kinetics and pKas. EXPERIMENTAL PROCEDURES

General. The trisodium salt of DOTA was purchased from Parish Chemical Co. (Orem, UT), and the sulfo-NHS and EDC were purchased from Pierce (Rockford, IL). The butylamine for the conjugation reaction and the yttrium chloride hexahydrate for the metal-loading experiments were procured from Aldrich (Milwaukee, WI). All glassware was acid washed and liberally rinsed with distilled deionized water (Thiers, 1957).

Bioconjugate Chem., Vol. 10, No. 3, 1999 455

Synthesis and Purification of BD. Synthesis of BD began with the addition of a solution of 60 mg (128 µmol) of DOTA and 56 mg (256 µmol) of sulfo-NHS in 2 mL of H2O to 0.982 mL (256 µmol) of a freshly prepared solution of EDC (50 mg/mL)1. This reaction mixture gave a theoretical concentration of the active ester of 12.7 mM. After 30 min, 25 µL (18.7 mg, 256 µmol) of neat butylamine was added to the reaction mixture, the pH adjusted to 8 with 220 µL of 1 M NaOH, and the solution left overnight. The next morning, the reaction was quenched by lowering the pH of the reaction mixture to 2 with the addition of 670 µL of 1 N HCl. The final mixture contained both doubly and singly modified forms of BD and unmodified DOTA that were separated by reversed-phase HPLC. The products of the reaction were purified using a Biocad Sprint chromatography system (Perseptive Biosystems, Farmingham, MA), equipped with a 4.6 mm × 10 cm POROS R2 stationary phase column (equivalent to a C18 column) and UV detection at 214 nm. The purification method used a flow rate of 1 mL/min for the wash step and a 4 mL/min flow rate for the linear gradient from 0 to 100% solvent B over 5 min (solvent A, 0.1% TFA; solvent B, 0.1% TFA/90% CH3CN), beginning 7 min after injection. This method yielded three peaks upon injection of the reaction mixture. The first peak contained unreacted DOTA and other starting material (eluted in the wash step), the second peak contained singly modified BD (also eluted in the wash step, baseline resolved from the first peak), and the third peak contained the doubly conjugated product (eluted in the gradient portion of the method). The peaks were collected, lyophilized, and identified by their mass-to-charge ratio (vida MALDITOF1). The mass-to-charge ratio of the peaks from the HPLC purification were measured on a Kratos Kompact MALDI III spectrometer, using R-cyano-4-hydroxycinnamic acid as the matrix. The spectra were collected in low-mass mode and calibrated with an external standard. A saturated solution of matrix was made in acetone and spotted on the MALDI stage in two 0.5 µL aliquots. The HPLC fraction of each of the peaks was mixed one-to-one with a 70% H2O/30% CH3CN/0.1% TFA solution, spotted on the MALDI stage in 1 µL aliquots, and air-dried. Massto-charge ratios (MH+) were observed for DOTA, BD, and doubly modified DOTA (2BD) of 406, 460, and 516 amu, respectively. NMR Measurements. NMR experiments were performed on a Varian Unity Plus 500 MHz spectrometer (Varian Instruments, Palo Alto, CA) with the probe air temperature regulated at 25-37 °C. NMR studies on DOTA and BD in solution were performed at 1-2 mM concentrations in 90% H2O/10% D2O solutions with the pH adjusted by addition of HCl or NaOH. All pH measurements were performed with an Orion model 601 pH meter equipped with a combination electrode for 5 mm NMR tubes (Ingold Electrodes, Wilmington, MA). Standard aqueous buffers were used for electrode calibration at pH 4 and 7. All chemical shifts were referenced to the methyl proton signal of sodium-2,2-dimethyl-2silapentane-5-sulfonate (DSS, at 0 ppm), which was added to the samples. The assignment of the pH-dependent chemical shifts for BD protons was aided by the observed magnitude of the chemical shift changes during the titration of proximate acidic groups. Those carbon-bonded protons closest to the deprotonation site generally have the largest increase in shielding and, concomitantly, the largest chemical shift changes to lower ppm values. The assign-

456 Bioconjugate Chem., Vol. 10, No. 3, 1999

Keire and Kobayashi

ments were confirmed by total correlation spectroscopy [TOCSY (Braunschweiler and Ernst, 1983)] spectra connectivities at selected pHs. Typically, the TOCSYs were performed with 2K points collected in t2 (32-48 transients/ increment), and 256 complex points in t1 (States et al., 1982). The transmitter channel was used for excitation and observation of the 6000 Hz proton frequency range as well as presaturation of the H2O signal. Data were processed with zero filling to 2K points in F1 and shifted sine-bell apodization functions in both F1 and F2. The DOTA and BD type I to type II rate constants (k*) were determined as a function of pH by measuring the intensity of the type I and type II forms as a function of time after the addition of excess yttrium. Typically, 700 µL of a 1 mM BD or DOTA solution was made in 90% H2O/10% D2O, the pH adjusted to the desired value with 1 M KOH and 1 N HCl, and a 1D spectrum acquired. Subsequently, 50 µL of a 120 mM yttrium chloride stock solution was added to the NMR tube (ca. 5 mM), and a series of spectra acquired as a function of time over 10 to 14 h. The peak heights of the signals from the acetate methylene group protons in the type I and type II forms were measured from each time-course data set and plotted. Acid-Dissociation Constants. Acid dissociation constants were determined by fitting NMR chemical shift titration data to monoprotic and diprotic acid models by nonlinear least-squares (Jardetzky and Roberts, 1981; Keire et al., 1992). For a monoprotic acid system, the observed chemical shift is given by the eq 1

δobs ) fHAδHA + fAδA

(1)

where fHA and fA represent the mole fractions, and δHA and δA the chemical shifts of the acid in its protonated and deprotonated states. The model equation for a monoprotic acid is obtained by expressing fHA and fA in terms of the acid dissociation constant, K.

δobs )

[H+] δHA + KδA [H+] + K

(2)

The following model equation for the diprotic acid is derived in the same way.

δobs )

[H+]2δH2A + [H+]K1δHA + K1K2δA [H+]2 + [H+]K1 + K1K2

(3)

The criteria used to select a model were the number of observed inflection points and comparison of the χ2 values obtained from fitting the data to monoprotic and diprotic models. The DOTA and BD k*s for yttrium were determined by the method of initial rates. Peak intensities were fit to a straight line to determine the initial rate. The fitting of the acid dissociation and rate constants was performed with Mathematica 3.0 (Wolfram Research, Champaign, IL). RESULTS

Synthesis of BD. The N-hydroxysulfo-OSSu active ester of the DOTA carboxylic acids was prepared by a water-soluble method commonly used in peptide synthesis and protein modification (Yamada et al., 1981; Staros et al., 1986; Gilles et al., 1990). A mole ratio of DOTA to sulfo-NHS to EDC of 1:2:2 was used in order to minimize the number of multiply activated DOTA carboxyl groups. The active ester form of DOTA reacts with the primary

Figure 2. The synthesis of butylamide-DOTA (BD)

amine group of the butylamine to form an amide bond in the conjugation reaction. The reaction produced two main products, BD and doubly modified BD (2BD), that were separated by reversed-phase HPLC and identified by their mass-to-charge ratios. The yield of the reaction after purification (based on the DOTA as the limiting reactant) was 38% BD (22.1 mg) and 14% 2BD (9.4 mg). NMR pH Titration of DOTA. DOTA and BD have multiple sites with Brønsted acid-base properties in aqueous solution. For DOTA, these are the four carboxylic acid groups and two protonated macrocycle nitrogen protons. Over the range pH 0-13 these groups undergo protonation/deprotonation reactions, and the chemical shifts of the carbon-bonded protons near sites of protonation or deprotonation are expected to change due to through-bond electrostatic effects (Jardetzky and Roberts, 1981). Thus, acid dissociation constants were determined for these groups from the dependence of the chemical shifts of the carbon-bonded protons on pH. The NMR spectrum of DOTA in aqueous solution consists of two signals, one for the acetate methylene group protons and the other for the macrocycle ethylene protons (Desreux et al., 1981). The signal for the protons on the macrocycle nitrogens are not observable because of exchange broadening (Brucher et al., 1991). Instead, the presence of nitrogen-bound protons in the macrocycle are inferred by the pH dependence of the chemical shifts of other nearby carbon-bonded proton signals at basic pHs. An abrupt change in the chemical shifts is observed at pH 11 is due to K+ complexation by the fully deprotonated form of DOTA (Clarke and Martell, 1991a). Macroscopic acid dissociation constants were determined for the various groups by fitting the NMR pH titration data to monoprotic and diprotic models (eqs 1-3). The criteria used to select a model were the number of observed inflection points and the fit of the data based on the χ2 value. For example, the chemical shifts of the DOTA ethylene proton signal over the range pH 0-7 was fit to a monoprotic model and the acetate chemical shifts were fit to a diprotic model (data not shown). Both the acetate methylene and the macrocycle ethylene signals were fit with a monoprotic model for the amine deprotonation over the range pH 7-11. The chemical shift values of the fully deprotonated and fully protonated forms of DOTA titratable groups (and the intermediate value in the diprotic model) are also obtained from the fits of the NMR pH titration data. The values obtained for the DOTA pKas and the titration shifts for the acetate and methylene protons are listed in Table 1. The measured macroscopic pKas are assigned to protonation of two trans amine nitrogens (designated G amines, 10.26), the two carboxyl groups near the non-



Table 1. Acid Dissociation Constants and Titration Shifts for DOTA in 0.1 M KCl at 25 °C protons

acetate shift (ppm)

pKa

ethylene shift (ppm)

pKa

G carboxylates

0.252

1.63

0.103

1.46

F carboxylates

0.126

4.57

G amines

0.368

10.43

0.324

10.08

calculateda pKa

1 M NaClb 25 °C

0.1 M KClc 25 °C

0.1 M KCld 25 °C

0.1 M KNO3e 25 °C

1.24 1.84 4.27 4.87 9.66 10.86

1.71 1.88 4.18 4.24 9.23 11.08

3.95 4.85 9.69 11.14

4.41 4.54 9.73 11.36

4.36 4.37 9.68 11.22

a Calculated titration constants from the relationship between macroscopic and microscopic pK s [see Edsall and Wyman (1958)] and a the average of the measured pKas. b Desreux et al. (1981). c Clarke and Martell (1991a). d Stetter and Frank (1976). e Delgado and Fausto Da Silva (1982).

protonated amines (designated F carboxylates, 4.57), and the remaining two carboxylic acid groups near the protonated amines (designated G carboxylates, 1.54). The individual pKas of the six DOTA protons were calculated using the relationship between macroscopic and microscopic protonation constants (Edsall and Wyman, 1958). The agreement among protonation constants determined for DOTA in Table 1 is reasonably good compared to those measured in other studies by potentiometric and NMR methods (Stetter and Frank, 1976; Desreux et al., 1981; Delgado and Da Silva, 1982; Clarke and Martell, 1991a). The value of the most basic amine pKa reported in Table 1 is lower than the other measured values presumably because of the effect of the K+ complexation with the DOTA4- form at pHs above 11. NMR pH Titration of BD. The conjugation of butylamine with DOTA active ester to form BD greatly increases the number of NMR signals present, whereas only two signals were observed in DOTA spectra (Figure 3). For example, in the conjugate, separate NMR peaks are observed for the acetate and ethylene signals directly adjacent and opposite to the site of the modification because the molecule is no longer symmetric; in symmetrical DOTA, these signals overlap. The separate signals allow the measurement of the microscopic protonation constants of the three carboxylic acid groups and the two macrocyle amines via fits of the pH-dependent chemical shifts of proximal carbon-bonded protons. These proton signals shift as a function of pH and the protonation state of BD. The 1D 1H NMR spectra of BD as a function of pH for the acetate signals is shown in Figure 3 (similar shifts were observed for the ethylene proton signals). The resonances were individually assigned by comparison with DOTA chemical shifts, their pattern of chemical shift changes with pH, relative intensities, and TOCSY spectra that indicated connectivities between connected ethylene protons or adjacent butylamide proton signals. For example, the methylene protons of the DOTA acetate arm that is involved in the conjugation to butylamine (Figure 4, E protons) exhibits the smallest chemical shift change with titration of the unconjugated carboxylic acids. Similarly, the methylene groups on the carboxylic acids not attached to a protonated nitrogen (Figure 4F) show the smallest perturbation with titration of the two macrocycle nitrogen protons. In another example, the BD F and G acetate methylene protons give rise to separate signals at low pH (Figure 3). Because the BD G protons are the most DOTA-like acetate methylene protons in BD, they were assigned based on the similarity in chemical shift with the DOTA G proton signal at low pHs (at pH 0.8 BD and DOTA G protons have chemical shifts of 3.935 and 3.992, respectively). In addition, the BD F proton signal can be differentiated from the G signal because it is larger (it arises from four identical protons versus 2 G protons). To follow the titration shifts of the ethylene signal chemical over the 0-13 pH range, TOCSY spectra were

collected at pH 1.24, 2.92, 3.14, 5.70, 7.0, and 9.60. In these experiments, TOCSY connectivities are observed in BD between the G′ and F′′ proton signals and the E′ and F′ proton signals. These connectivities, in combination with the pH-dependent shifts were used to assign the macrocycle ethylene proton resonances. The acetate methylene proton signals (labeled E, F, and G) shift and change shape as a function of the protonation state of BD (Figure 3). For example, the F and G acetate methylene signals at pH 1 are broad and have separate signals. Upon deprotonation of the G carboxylate, these signals move together and overlap, become sharp, and shift to lower chemical shift values. The F and G signals become nonequivalent again upon deprotonation of the F carboxylic acid groups. The increased shielding (shift to lower chemical shift value) of the methylene signals is expected due to the increased electron density in the adjacent deprotonated carboxylic acid group. Typically, in the absence of structural alterations, the closer that the carbon-bonded proton is to the site of the deprotonation, the greater the magnitude of the chemical shift change. By contrast, deprotonation events that lead to deshielding (movement to higher chemical shift values) for carbon-bonded protons more than 3-5 bonds distant from the site of deprotonation indicate the presence of conformationdependent through-space interactions. In particular, the formation of hydrogen bonds between amines and carboxylate groups can be identified from observation of deshielding chemical shift changes as a function of pH (Bundi and Wuthrich, 1979). The chemical shifts change because the N-H bond becomes strongly polarized when the amine or amide proton forms a hydrogen bond, the proton is deshielded and its resonance shifted to higher chemical shift values. Furthermore, the deshielding tends to occur upon titration of carboxylic acid groups because a deprotonated carboxylate forms a stronger hydrogen bond than the protonated form. Deshielding chemical shift changes attributed to the formation of hydrogen bonds and conformational changes are evident in the NMR pH titration data of BD (Table 2). For example, the E acetate methylene signal is a broad singlet at low pH and upon titration of the G carboxylic acid group shifts to a higher (deshielded) chemical shift value. This signal also sharpens to an AB pattern as the two geminal protons become magnetically nonequivalent (at pH 3.3 2JAB is -14.2 Hz and ∆v(va - vb) is 0.033 ppm). With the titration of the F carboxylate groups, the E acetate methylene proton signal is slightly shielded and the AB pattern is altered (at pH 5.5, ∆v is 0.101 ppm). Simultaneously, the amide, D, and C proton signal chemical shifts decrease upon G deprotonation and increase upon F carboxylate deprotonation (Figure 5). Because the E acetate methylene and butylamide protons (A, B, C, D, and the NH) are 10-17 bonds distant from the site of the G carboxylate deprotonation and 7-12 bonds distant from the site of F carboxylate deprotona-


Keire and Kobayashi

Figure 4. A plot of the chemical shifts of the BD E, F, and G acetate methylene proton signals as a function of pH.

Figure 3. The 2.5-4.3 ppm region of the 500 MHz 1H NMR spectra of 1 mM BD in 90% H2O/10% D2O at pH (A) 1.14, (B) 3.30, (C) 5.51, (D) 8.53, (E) 9.60, and (G)10.12. The signals from the various BD protons are labeled in the accompanying structure and the spectra.

tion, these shifts indicate that BD undergoes conformational changes and that these changes are dependent on the protonation state of the molecule. The effects of these structural alterations are evident in the NMR pH titration spectra (the observed chemical shifts, line shapes, and coupling constants are weighted averages of these parameters in the different isomers present at a particular pH). The BD acetate methylene and macrocycle methylene titration shift data were fit to monoprotic or diprotic models to determine the pKas of the various functional groups and titration shifts (Table 2). Of note, the fit of pH shifts resulting from titration of the F carboxylate protons revealed two values, 3.52 ((0.12) (three of the

eight values) and 4.53 ((0.24) (five of the eight values). In addition, the formation of a possible hydrogen bond with the pendant arm amide proton is indicated by its deshielding chemical shift change which occurs with the deprotonation of the second F carboxylate group (Table 2). Yttrium Metal Loading Rates in DOTA. The yttrium metal loading into DOTA over time was monitored by proton NMR (as shown in Figure 6). The rate of the formation of type I complexes is such that by the time the first NMR spectrum is collected after adding excess metal, most of the chelator is in the type I and type II forms. In the presence of the metal, the NMR signals broaden and shift, and, over a period of hours, several of the broad signals disappear. This reflects the transition from the type I to type II complex (Figure 6). The resonances in the yttrium-DOTA complex previously have been assigned and show that the DOTA protons become magnetically nonequivalent because of the rigid structure induced by metal binding (Broan et al., 1991). The signals labeled F,G in Figure 6A are the methylene protons of the F and G acetate groups of YDOTA- (an AB pattern with 2JAB of 16 Hz). Three multiplets, J, I, and K, are also observed with a signal intensity ratio of 4:8:4, respectively. Signal J is from the ring protons closest to the carboxyl group in the structure induced by metal complexation. The signals in Figure 6A are from the fully metal-loaded form of DOTA (type II complex) and the spectrum in Figure 6B represents a mixture of signals from the type I and type II forms. With the identity of the type II complex signals established, the broad signals observed in 1H NMR spectra (Figure 6B) of Y-DOTA complexes (that do not correspond to signals in Figure 6A) are assigned to the intermediate type I complex. These type I signals decrease as a function of time after mixing and as the type II signals increase. These data reveal the rate of transition from the type I to type II complex. To determine the initial rates, the intensity of the I and K resonances (at 2.74 and 2.42 ppm, respectively) was plotted as a function of time after adding metal. Table 3 shows the initial rates of yttrium metal loading over the pH range 4.5-6.5 at 37 °C. Yttrium Metal Loading Rates in BD. The 1H NMR spectrum of the fully metal-loaded (type II) complex of yttrium and BD (Y-BD) has a similar pattern of signals as Y-DOTA (Figure 7A). The resonances of the E, F, and G acetate and the J, I, and K methylene protons of Y-BD appear in the same order as the F, G, J, I, and K signals in Y-DOTA (compare spectra in Figures 6 and 7A).



Table 2. Acid Dissociation Constantsa and Titration Shiftsb for BD Protons in 0.1 M KCl at 25 °C Acid Dissociation Constants proton(s)

A

B

C

D

amide

E

E′

F

F′

G

G′

mean ((std dev)

G carboxylate F carboxylates

2.39

1.98 3.49

2.32

2.35

2.46

2.18

2.49

1.90

2.02 3.68

2.09 3.40

2.57

4.53 9.00 10.97

4.52 9.30 10.70

4.40

4.96 9.04 10.77

8.87

4.25 9.05 10.47

8.92 10.42

9.14 10.85

2.25 ((0.21) 3.32 ((0.12) 4.53 ((0.24) 9.05 ((0.21) 10.73 ((0.20)

E amine G amine

9.07 10.89

10.56

Titration Shifts (ppm) proton(s) G carboxylate F carboxylates E amine G amine

A 0.009 -0.006 0.009

B 0.012 0.007 -0.006

C

D

amide

E

E′

F

F′

G

G′

0.026 0.010 -0.011 0.021

0.085 -0.037 -0.040 0.051

0.161 -0.148

-0.088 0.016 0.402 0.111

-0.251

0.539 0.221 -0.137 0.167

0.349 -0.037

0.410 0.128 0.202 0.117

0.340

0.213 0.235

0.304 0.089

a The pK s were calculated from fits to monoprotic and diprotic models of the NMR pH titration data. b The titration shifts were calculated a from the chemical shift values obtained in the pKa fits (δHA - δA).

Figure 5. Plots of the chemical shifts of the A, B, C, D, and amide protons of the butylamide portion of BD as a function of pH. The labels are the same as in Figure 1.

However, the E, F, and G acetate methylene protons in Y-BD are split into AB patterns and have different chemical shifts (E being the most deshielded) because of the butylamide substitution. A similar pattern is observed for the macrocycle ring K protons, in which separate signals are observed for the proton next to the E methylene [K(E)], and the F and G methylenes

[K(F,G)]. Also, the butylamide D methylene signal is split into a multiplet (vida TOCSY). The formation of the multiplet patterns in Y-BD for the methylene protons is indicative of the formation of a rigid structure that causes the geminal proton pairs to become magnetically nonequivalent. For example, the splitting of the E and D methylene proton resonances into


Keire and Kobayashi

Figure 6. The 1.5-3.9 ppm region of the 500 MHz 1H NMR spectra of (C) a 2 mM solution of DOTA at pH 5, (B) the same solution as in C6 6 min after adding 27 mM Y3+, and (A) the same solution as in panel B acquired 5 h after adding metal. All solutions were 90% H2O/10% D2O and the probe air temperature was regulated at 25 °C. Spectra B and C were acquired and displayed with the same parameters and A is labeled as in the accompanying structure. The up and down arrows of spectrum B indicate signals from the type II and type I Y-DOTA complex, respectively, that increase and decrease as a function of time. Table 3. Rate Constants for Yttrium Metal Loading (Type I to Type II) in DOTA and BD as a Function of pH at 37 °C pH 4.6 4.6 5.7 5.6 6.7 6.5

n 8 4 4 4 4 3

DOTA, k1 (s-1) 0.84 ((0.07) ×

10-3

1.68 ((0.12) × 10-3 2.35 ((0.28) × 10-3

BD, k1 (s-1) 0.45 ((0.02) × 10-3 0.63 ((0.10) × 10-3 1.07 ((0.06) × 10-3

AB patterns suggests that metal coordination occurs with the butylamide carbonyl group, fixing the geometry of this portion of BD. At the same time, the A, B, and C proton signals are mostly unchanged, indicating that their environment is not significantly altered by metal binding (although the B proton signal is broadened in the type II complex). As with Y-DOTA, to determine the BD type I to type

Figure 7. The 0.7-4.4 ppm region of the 500 MHz 1H NMR spectra of (C) a 1 mM solution of BD at pH 7.1, (B) the same solution as in C 6 min after adding 5 mM Y3+, and (A) the same solution as in panel B acquired ca. 24 h after adding metal. All solutions were 90% H2O/10% D2O and the probe air temperature was regulated at 37 °C. Spectra B and C were acquired and displayed with the same parameters, and A is labeled as in the accompanying structure. The up and down arrows of spectrum B indicate signals from the type II and type I Y-BD complex, respectively, that increase and decrease as a function of time.

II metal loading rate constants, the intensity of three BD proton signals were measured as a function of time after addition of excess yttrium (Table 3). In Figure 7B, the increase in the F,G multiplet pattern at 3.6 ppm and the K(F,G) multiplet at 2.43 ppm and the decrease in the E proton AB pattern doublet at 3.95 ppm peak height as a function of time were plotted and the initial rates determined. The initial rates of yttrium metal loading (type I to type II) into BD as a function of pH at 37 °C are shown in Table 3 and are approximately one-half the rates for DOTA at the same pH. Metal-Loading Forms of DOTA and BD. Examining of the combination of the pKa and rate data reveals that although both the doubly and singly protonated species of DOTA are reactive to metal loading, only the



Figure 8. Combined plots of the type I to type II metal loading rate (k*) and the fraction in the singly protonated form of (top) DOTA and (bottom) BD.

singly protonated form of BD is reactive (Figure 8). For DOTA, over the 4.6-6.5 pH range, the change in k* increases more rapidly with pH than the fraction of the chelator in the singly protonated form (HDOTA3-) (Figure 8, top). This increase in k* is best explained by metal loading into both the H2DOTA2- and HDOTA3- forms (the singly protonated form is orders of magnitude more reactive but much less abundant then the doubly protonated form). In contrast, for BD, the change in k* is best fit solely by the increase in the fraction of the singly protonated form (Figure 8 bottom). DISCUSSION

The modification of a single pendant arm of DOTA to create BD significantly alters the structure and characteristics of the chelator. For example, the NMR pH titration on BD shows a 0.6 pKa unit decrease in one of the amine nitrogens relative to DOTA. The decrease in the pKa of the macrocycle amine is important because the singly protonated form is the most reactive to metal loading in both DOTA and BD. The decrease in the pKa value leads to an increase in the population of the singly protonated form of BD compared to DOTA (at pH 7 the HDOTA3- is 0.002% and HBD2- is 0.009%). Thus, if the macrocycle amine pKas were the only determinant in the metal-loading rate of BD, BD would metal-load more rapidly than DOTA. Because the yttrium metal loading rates of BD are approximately one-half the rates measured for DOTA at the same pH and temperature, other factors (e.g., the proton-transfer ability of the pendant arm groups) must be involved in the type I to type II transition. Evidence for the proton-transfer mechanism is observed in ab initio calculations on Y-H2DOTA2- that show that the pendant

carboxylate groups are near enough in space to form hydrogen bonds with the protonated nitrogens of the macrocycle and may act to facilitate the deprotonation of these nitrogens that occurs with metal loading [Jang and Goddard (1999) J. Am. Chem. Soc. (Manuscript submitted for publication)]. Consistent with the pendant group proton-transfer mechanism are the observations of Takenouchi et al. (1993a,b) who reported carboxymethyl-substituted derivatives of 6-(4-nitrobenzyl)-1,4,8,11-tetraazacyclotridecane-N,N′,N′′,N′′′-tetraacetic acid (p-NO2-Bz-TRTA) that showed a distance dependent rate enhancement for Y metal loading. p-NO2-Bz-TRTA derivatives that contained a (carboxymethyl)amino group appended with a methylene or ethylene spacer increased the Y-metal loading rate by a factor of 2 or had no effect, respectively. Thus, the longer spacer arm may not allow effective proton transfer from the macrocycle nitrogens. The type I to type II rate constants of DOTA metal loading measured in this work (Table 3) differ from the rate constants measured previously under different conditions. Kodama et al. (1991) reported a DOTA-Y3+ metal loading rate of 7.7 × 10-5 s-1 (pH 7.8, 25 °C, I ) 0.1 from NaClO4 and 25 mM HEPES buffer) compared to value obtained in this work of 2.35 × 10-3 s-1 at pH 6.7 and 37 °C (Table 3). Wang et al. (1992) and Wu and Horrocks (1995) measured the pH dependent kinetics of DOTA-Gd3+ (pH 3.8-5.8, 25 °C, I ) 1 from NaCl and buffer) and DOTA-Eu3+ (pH 4.5-5.7, 25 °C, I ) 0.1 from KCl and 20 mM buffer) complexation, respectively, and showed rate constants 4-28 times faster (at similar pHs) than the rates measured in this work at 37 °C. These differences are attributed to the sensitivity of DOTA


metal loading rates to pH, ionic strength, temperature, buffer, and the effective ionic radii of the metal used. DOTA Structure. The structure of DOTA in solution may also influence the kinetics of metal loading. Protonation-state dependent H-bonded forms of DOTA have been proposed to explain the low basicity of two of the macrocycle amine nitrogens (Micheloni et al., 1978; Desreux et al., 1981). Micheloni et al. (1978) first proposed that the two trans protonated macrocycle nitrogens formed intracation H-bonds with the unprotonated macrocycle nitrogens. Later, Desreux (1981) proposed a structure for H4DOTA that had H-bonds between the two deprotonated G carboxylate groups and the macrocycle G amine protons (a five-membered ring). For type II complex formation, the possible structures of the H2DOTA2- and HDOTA3- are the most important because they are reactive to metal loading. Ab initio calculations indicate that in H2DOTA2- H-bonds between the F carboxylate and G amine groups (an eightmembered ring) are more energetically favorable than H-bonds between the G carboxylate and G amine (Jang and Goddard, unpublished results). The NMR data on DOTA indicates that the H-bonded isoforms must be in fast exchange because only two singlets (from carbonbonded protons) are observed in the proton NMR spectra. Furthermore, the two macrocycle amine protons may not be observed because of exchange among the possible H-bonded forms between the amines, solvent water, and the macrocycle carboxylates. The exchange mechanism between -NH, -CO2-, and H3O+ or OH- could also serve as a means to remove protons from the macrocycle ring when metal loading in the type I to type II metal complex forms of DOTA. Structures of the Various Protonated States of BD. The observed deshielding titration shifts in the BD signals are attributed to the formation of hydrogenbonded isomers of BD that are protonation-state dependent. The NMR shift data indicates the presence of several pH-dependent hydrogen-bonded structures. The H-bond formation is accompanied by structural changes that affect the chemical shifts of the butylamide portion of BD (Figure 4). Further evidence for the structural alterations is seen in the change in the coupling patterns and line shapes of the E, F, and G protons (Figure 3). Deriving structure from the titration shift data is complicated by the simultaneous presence of multiple protonated forms of BD. The NMR titration shifts of BD represent a mixture of the effects of the structural changes and the redistribution of electron density caused by titration of proximate groups. These changes can only be interpreted based on a series of structures for each of the protonation states of BD. The BD signals in the NMR spectra represent a population-weighted average of the chemical shifts and line widths of the protonation-state dependent structures present at a particular pH. The NMR spectra at basic pHs show the macrocycle amine protons exist predominantly in the E-G configuration because of the large titration shift (0.4 ppm, Table 2) of the E acetate methylene proton upon deprotonation of the first macrocycle amine. The BD pKas have been calculated based on the assumption that the E-G isomer is also favored in the higher protonation states. Thus, the G carboxylate is deprotonated first because it is adjacent to the positively charged protonated macrocycle nitrogen (as in DOTA). However, the macrocycle amine protons can exist in two different trans isomers (E-G and F-F pairs) that have similar energy in the fully protonated form (Jang and Goddard, unpublished results). If the F-F isomer is

Keire and Kobayashi

present in appreciable amounts (in the fully protonated BD), the assignment of the first deprotonated carboxylate would be one of the two F groups. Further experiments are necessary to clarify the location of the macrocycle amine protons in different protonation states of BD. For BD, the protonated forms that are most relevant to metal loading are the H2BD- and HBD2-. The titration shift data show the amide proton is deshielded (-0.148 ppm, Table 2) upon deprotonation of the second F carboxylate group. This shift indicates the formation of a H-bond involving the amide proton (Bundi and Wuthrich, 1979). For H2BD-, we propose that two H-bonds are present, one between the F carboxylate and the butylamide amide (F-HN) and the other between the second F carboxylate and the G amine proton (F-G). The H2BDform may metal-load more slowly than H2DOTA2- because the BD amide carbonyl is not as facile a protontransfer mediator as the carboxylate groups of DOTA. CONCLUSIONS

The ideal chelating agent for biological applications would metal-load rapidly (

NMR Studies of the Metal-Loading Kinetics and ... - ACS Publications

Recommend Documents