Chromatographic Separation and Identification of a Water-Soluble

Anal. Chem. 2003, 75, 4217-4222

Chromatographic Separation and Identification of a Water-Soluble Dendritic Methano[60]fullerene Octadecaacid Najla Gharbi,† Stephan Burghardt,‡ Michael Brettreich,‡ Christine Herrenknecht,† Sara Tamisier-Karolak,† Rene´ Victor Bensasson,§ Henri Szwarc,| Andreas Hirsch,‡ Stephen R. Wilson,⊥ and Fathi Moussa*,†

Service de Chimie Analytique, Faculte´ de Pharmacie, Universite´ Paris XI, Rue J-B Cle´ ment, 92 296 Chaˆ tenay-Malbry, France, Institut fu¨r Organische Chemie der Universita¨t Erlangen-Nu¨rnberg, Germany, Laboratoires de Biophysique et de Chimie des Substances Naturelles, Muse´ um National d’Histoire Naturelle, Paris, France, Laboratoire de Chimie Physique, UMR 8000 du CNRS, Baˆ timent 490, Universite´ Paris XI, Orsay, France, and C Sixty, 2250 Holcombe, Houston, Texas 77030

The chromatographic separation of a highly water-soluble dendritic monoadduct methano[60]fullerene octadecaacid (dendrofullerene) with octadecylsilica bonded phases has been studied. It has been found that the RP-HPLC behavior of this dendrofullerene obeys the general rules of stationary-phase and mobile-phase selection for controlling the separation of usually acidic compounds. An RP-HPLC-ESI-MS analysis confirms the identity of the dendrofullerene and allows characterization of the molecular weights of the main impurities contained in the sample. The described methods can control the synthesis and efficiently purify this fullerene derivative, which has been previously shown to be active against mutant infectious clones of HIV-1, which are resistant to AZT and 3TC, drugs that are widely used in AIDS therapy. Over the past decade, it has been shown that some watersoluble fullerene derivatives exhibit considerable biological activity1-4 including the inhibition of HIV protease,5,6 antimicrobial activity,7 neuroprotective property,8 and transport of radioactive metals for imaging or therapeutic applications in nuclear medicine.9 However, * Corresponding author. E-mail: [email protected]. Fax: (33) 1 44 73 53 09. † Universite´ Paris XI. ‡ Universita¨t Erlangen-Nu ¨ rnberg. § Muse´um National d’Histoire Naturelle. | UMR 8000 du CNRS. ⊥ C Sixty. (1) Kroto, H. W.; Heath, J. R.; O’Brien, S. C.; Curl, R. F.; Smalley, R. E. Nature 1985, 318, 162-163. (2) Hirsch, A. The Chemistry of the Fullerenes; Georg Thieme Verlag: Stuttgart, 1994. (3) Jensen, A. W.; Wilson, S. R.; Schuster, D. I. Biorg. Med. Chem. 1996, 4, 767-787. (4) Da Ros, T.; Prato, M. J. Chem. Commun. 1999, 8, 663-687. (5) Friedman, S. H.; DeCamp, D. L.; Sijbsma, R.; Srdanov, G.; Wudl, F.; Kenyon, J. L. J. Am. Chem. Soc. 1993, 115, 6506-6507. (6) Schuster, D. I.; Wilson, S. R.; Kirschner, A. N.; Schinazi, R. F.; SchlueterWirtz, S.; Barnett, T.; Martin, S.; Ermolieff, J.; Tang, J.; Brettreich, M.; Hirsch, A. Electrochem. Soc., Proc. 2000, 11, 267-269. (7) Bosi, S.; Da Ros, T.; Castellano, S.; Banfi, E.; Prato, M. Bioorg. Med. Chem. Lett. 2000, 10, 1043-1045. 10.1021/ac026419k CCC: $25.00 Published on Web 07/22/2003

© 2003 American Chemical Society

only a few studies of in vivo toxicity and the metabolic fate of this family of new compounds have been performed. To further investigate these drug candidates, especially their toxicity and pharmacokinetics, it is essential to use very pure and well-defined chemical compounds. Although several analytical methods including TLC, NMR, MS, IR, and UV-visible spectroscopy have been used to identify these water-soluble fullerene derivatives,8-11 methods for their purification have not yet been established. For this purpose, it is necessary to develop efficient methods of separation. The reversed-phase high-performance liquid chromatographic (RP-HPLC) behavior of unmodified fullerenes is welldetermined,12-14 but this it is not the case for their water-soluble derivatives. Among the variety of C60 derivatives synthesized to date, a highly water-soluble octadecaacid dendritic methano[60]fullerene monoadduct (dendrofullerene, DF, Figure 1) is one of the most promising compounds for medical applications. Activity of DF has been demonstrated in vitro.6 In addition, DF has a very low toxicity with respect to rodents and it is readily eliminated through the kidneys.15 We report here the systematic investigation of the separation of DF by RP-HPLC. We concentrated on systematic variations of solvent strength, pH, type of salt, ionic strength, flow rate, and column type. RP-HPLC-ESI-MS was used to identify DF and various impurities. (8) Dugan, L. L.; Turetsky, D. M.; Du, C.; Lobner, D.; Wheeler, M.; Almli, C. R.; Shen, C. K.-F.; Luh, T.-Y.; Choi, D. W.; Lin, T. S. Proc. Natl. Acad. Sci. U.S.A. 1997, 94, 9434-9439. (9) Cagle, D. W.; Kennel, S. J.; Mirzdah, S.; Alford, J. M.; Wilson, L. J. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 5182-5187. (10) Lamparth, I.; Hirsch, A. J. Chem. Soc., Chem. Commun. 1994, 1727-1728. (11) Brettreich, M.; Hirsch, A. Tetrahedron Lett. 1998, 39, 2731. (12) Diack, M.; Hettich, R. N.; Guiochon, G. Anal. Chem. 1992, 64, 2143-2148. (13) Jino, K.; Uemura, T.; Ohta, H.; Nagashima, H.; Itoh, K. Anal. Chem. 1993, 65, 2650-2654. (14) Baena, J. R.; Gallego, M.; Valcarel, M. Trends Anal. Chem. 2002, 21, 187198. (15) Gharbi, N.; Pressac, M.; Tomberli, V.; Da Ros, T.; Brettreich, M.; Hadchouel, M.; Arbeille, B.; Trivin, F.; Ce´olin, R.; Hirsch, A.; Prato, M.; Szwarc, H.; Bensasson, R.; Moussa, F. Electrochem. Soc., Proc. 2000, 11, 240-243.

Analytical Chemistry, Vol. 75, No. 16, August 15, 2003 4217

Figure 1. Chemical structure of dendrofullerene and 0.1 M NaOH titration curve ([DF] 3.952 mM in water).

EXPERIMENTAL SECTION Basic studies on HPLC separations were performed using a P4000 multisolvent delivery system coupled with a UV6000LP photodiode array detector (Thermo Separation Products, Les Ulis, France). Instrument monitoring and data acquisition were performed using ChromQuest Software (ThermoQuest, Les Ulis, France). The ESI-mass spectrometer (ESI-MS) used in this study was a Bruker Esquire LC (Bruker, Wissembourg, France). The HPLCESI-MS studies were performed with the same mass spectrometer coupled with a HP 1100 HPLC system (Hewlett-Packard, Les Ulis, France). Basic separations were carried out with a 4.0 mm i.d. × 125 mm Superspher 100 RP-18 column (5 µm, end-capped, Merck). The other HPLC columns used were a 9.0 mm i.d. × 250 mm Vydac 218TP column (polymeric ODS, 10 µm, 300 Å, Grace Vydac, Southboro, MA), a 4.6 mm i.d. × 150 mm X-Terra MS C18 column (5 µm, 120 Å), and a 3.9 mm i.d. × 150 mm Nova-Pak Phenyl column (4 µm, 60 Å) (both from Waters, Milford, MA). All chemicals were used as purchased without further purification. Mobile phases based on 0.1 M aqueous buffers were prepared daily. To obtain effective buffering (pH ) pKa (1), trifluoroacetic, trichloroacetic, oxalic, citric, and phosphoric acids were used as a function of the pH domain explored. In all cases, final pH was adjusted with ammonia. DF has been synthesized as previously described.11 DF stock solutions (2 g/L) were prepared monthly in a water-acetonitrile (50/50 v/v) mixture and stored at -20 °C until needed. After equilibration at room temperature, an appropriate aliquot of the stock solution was diluted in the mobile phase. RESULTS AND DISCUSSION Spectroscopic Properties and Titration of the Dendrofullerene.. The UV-visible spectrum of the DF sample in a mixture of water-acetonitrile (1/1, v/v) is characteristic of cyclopropyl-fused monoadducts of C60 as evidenced by the presence of the 426-, 490-, and 680-nm absorption bands.16 The ESI(16) Bensasson, R. V.; Bienvenue, E.; Fabre, C.; Janot, J. M.; Land, E. J.; Leach, S.; Leboulaire, V.; Rassat, A.; Roux, S.; Seta, P. Chem. Eur. J. 1998, 4, 270275.

4218

Analytical Chemistry, Vol. 75, No. 16, August 15, 2003

MS spectrum [m/z 2827 (M - H)-, m/z 1413.1 (M - 2H)2-] confirms its molar weight of 2828.74.11 NaOH titration of DF (Figure 1) led to three distinguished pH domains with significantly different behaviors, as previously discussed.17,18 Using NaOH titration, the sample purity was calculated to be 94%. HPLC Separation. Selecting the Chromatographic Conditions. DF is insoluble in common fullerene solvents,2 but it is highly soluble in water at pH 7.4 (34 mg/mL).11 In the protonated (acidic) form, DF is sparingly soluble in water, highly soluble in methanol and acetonitrile, and freely soluble in methanol or acetonitrile-water mixtures. For this reason, we chose to evaluate a low-pH buffer-acetonitrile mixture as a candidate for the mobilephase RP-HPLC separation. For a stationary phase, we first chose a Superspher RP18 column that is strongly hydrophobic (carbon content, 21%) and reasonably stable in acidic conditions. Under such conditions, the chromatogram of DF exhibits a main peak accompanied by several secondary peaks (Figure 2A). All the peaks exhibit the same UV-visible spectrum as that of the initial mixture (Figure 2B), which are characteristic of cyclopropyl-fused monoadducts of C60.16 The number and the relative areas of these peaks do not change when (i) varying the pH of the mobile phase between pH 1.5 and 7.4; (ii) keeping the DF powder at ambient temperature for 2 years (amber flask); (iii) keeping the DF solution at ambient temperature for 3 days; and (iv) heating the DF solution at 80 °C for 3 h. This indicates that the minor additional peaks do not correspond to anionic forms of DF or to degradation products. They are probably due to various cyclopropyl-fused monoadducts of C60 formed during DF synthesis. Controlling the Efficiency of the Separation. Figure 3A shows the evolution of the retention factors and the efficiency of the separation as a function of the concentration of acetonitrile in the mobile phase. The best separation in terms of efficiency, symmetry, and rapidity is obtained with 45% acetonitrile for DF. The same efficiency can be also obtained with 90% methanol. (17) Tamisier-Karolak, S. R.; Pagliarusco, S.; Herrenknecht, C.; Brettreich, M.; Hirsch, A.; Ce´olin, C.; Bensasson, R. V.; Szwarc, H.; Moussa, F. Electrophoresis 2001, 22, 4341-4346. (18) Brettreich, M. Dissertation. Friedrich-Alexander University of Erlangen, 2000.

Figure 2. Chromatogram (A) and extracted spectra (B) of a dendrofullerene sample (3.5 nM on column). Spectra of all the peaks eluting after 1.5 min, not shown here, are identical with those shown in (B). Chromatographic conditions: Superspher RP18 column, acetonitrile-0.1 M aqueous buffer pH 2.0 (50/50, v/v), 0.8 mL/min at 40 °C, detection, UV at 320 nm.

However, methanol is not a suitable solvent for DF because it is known to react with DF to give polymethyl esters.11 Thus, the acetonitrile-based mobile phases were used for further investigations. We also chose to work with 50% acetonitrile to reduce the run time. Figure 3B shows the influence of temperature on the retention characteristics of DF. As a general rule, retention in RPLC can be expected to decrease when the temperature increases. Increasing the temperature from 15 to 60 °C decreases the retention of DF by 0.18-0.83%/1 °C. This slow decrease is followed by a relatively rapid decrease between 60 and 69 °C reaching ∼2%/1 °C. However, it is worth noting that this minor retention change below 60 °C is accompanied by a relatively significant increase in the efficiency of the separation. From 10 to 40 °C, H decreases sharply from 1.7 to 3.45%/1 °C to reach a plateau between 40 and 50 °C, corresponding to ∼50% of the initial value. Figure 3C shows the effects of flow rate on the efficiency of the separation. As expected, the efficiency increases significantly when the flow rate decreases. Figure 3D illustrates the usefulness of using pH to control the retention and the efficiency of the separation. Varying the concentration or the type of the salt in the mobile phase resulted in minor retention characteristic variations as compared with changes in pH or solvent strength. DF is more strongly retained at the lowest pH, where it is more protonated and therefore more hydrophobic. The shape of the curve representing the variation of H closely resembles that of the first domain of the NaOH

Figure 3. Influence of the mobile-phase composition on retention of dendrofullerene (column, Superspher 100 RP18). (A) Influence of the concentration of acetonitrile. Mobile phase: acetonitrile-0.1 M buffer, pH 2.0, 0.8 mL/min at ambient temperature). (B) Influence of the temperature. Mobile phase: acetonitrile - 0.1 M buffer pH 2.0 (50/50, v/v), 0.8 mL/min. (C) Influence of the flow rate at 40 °C (see B for chromatographic conditions). (D) Influence of pH changes. Mobile phase: acetonitrile-0.1 M aqueous buffer (50/50, v/v), (see Experimental Section for buffer type), 0.8 mL/min at ambient temperature.

Analytical Chemistry, Vol. 75, No. 16, August 15, 2003

4219

Table 1. Influence of pH and Acetonitrile Concentration on Retention of Dendrofullerenea

a

pH

acetonitrile (%)

k

5 5 5.5 5.5 6 6 6.5 6.5 7 7 7.5 7.5

40 30 30 20 30 20 20 15 15 10 15 10

0.08 17.92 1.03 not eluted 0.08 27.39 1.97 23.32 5.49 29.41 4.41 24.68

H (µm) 420 74 460 620 80 900 100 290 150 370 290

See Figure 3D for chromatographic conditions.

titration curve (Figure 1). Between pH 1.5 and 2.5, the efficiency is unaffected by the variation of pH. At pH levels above 3, the efficiency decreases substantially. Since the pKa values of the two more acidic carboxylic groups of DF are ∼3.2, the broadening of the DF peak in this pH domain may be related to some ionization equilibrium leading to an unbalanced retention between the ionized species. The range above pH 4.5 is very critical. In this pH domain, DF is not retained when the concentration of acetonitrile is greater than 40%. Below this percentage, small variations of pH or mobilephase solvent strength drastically modify the ionization equilibrium and consequently the solubility of DF, which is reflected in dramatic changes in its retention characteristics (Table 1). In this pH domain, the difference between two successive pKa values decreases steeply (Figure 2). However, as the differences between the pKa values are very small (∼0.1)18 and the equilibrium rates are rapid, increasing pH did not allow us to separate the different anionic forms of DF. Increasing pH while maintaining the organic solvent at a constant concentration or maintaining pH while increasing the organic solvent concentration led only to a decrease in the retention factor concurrently with the broadening of the DF peak without peak splitting. To efficiently separate DF above pH 4.5 a gradient elution is required. Column-Type Effects on DF Separation. Since the molecular weight of DF is relatively high, the use of a column filled with large pore size particles (Vydac column) was expected to help enhance the selectivity and the efficiency of the separation. The chromatographic profiles, in terms of number of peaks and relative areas obtained by using the Vydac column with different mobilephase compositions, were similar to those obtained with the Superspher one. Using the Vydac column under the same mobilephase conditions as used with the Superspher column, the best DF separation was obtained with 40% acetonitrile (H ) 100 µm, k ) 8.09) and a flow rate set at 3.0 mL/min. Using a column with different stationary-phase chemistry, phenyl-bonded phase (NovaPak phenyl) instead of C18 chemistry under the same mobile-phase conditions, the best efficiency with a relatively large retention factor (H ) 57 µm, k ) 12.96) has been obtained with 35% acetonitrile. The chromatographic profiles obtained with this last column filled with 60-Å pore size particles are similar to those obtained with the first two columns and do not result in any size exclusion phenomena. 4220

Analytical Chemistry, Vol. 75, No. 16, August 15, 2003

The last type of column we considered was X-Terra MS C18 filled with hybrid particles, which are stable under extreme acidic and basic conditions. Using acetonitrile-aqueous buffer mixtures from pH 1.3 to pH 7.4 as mobile phases, we observed no significant differences in the chromatographic profiles of the DF separation with respect to those obtained with the Superspher column. As DF is not retained at a pH higher than 7.5, using the X-Terra column did not allow us to separate the fully ionized form of DF. Under the best chromatographic conditions obtained with the Superspher column (Figure 2), the X-Terra column is more retentive but much less efficient (k′ ) 7.23, H ) 66 µm) for the DF separation than Superspher. Regarding the efficiency of the separation of DF, the differences observed between the four columns can be explained by the different particle sizes and packing techniques as well as by the hydrophobicity indexes (HI) of the packing expressed as the ratio anthracene k/benzene k.19 The part played by the last parameter is particularly obvious when the HI of the Superspher column is compared to that of the X-Terra one. Although the Superspher column is less retentive for benzene (benzene k ) 0.89) than the X-Terra (benzene k ) 1.19), its HI (7.29) is higher than that of the latter (HI ) 5.89). Thus, the Superspher column is more selective for multiring compounds than the X-Terra one. In conclusion, systematic investigation of the RP chromatographic behavior of DF shows that the separation of this highly water-soluble dendritic methano[60]fullerene octadecaacid is similar to that of the usual acidic compounds. Thus, for maximum method robustness, it is best to work, as usually recommended for acidic compounds, at a pH value for which the compounds are fully ionized or ionization is suppressed.19 Since DF is not retained at pH levels higher than 7.5 and since the separation of the other ionized forms requires a gradient elution, it is simplest to work under acidic conditions. HPLC-ESI-MS. In an attempt to identify the additional peaks revealed by HPLC and in order to make this method applicable to the study of DF pharmacokinetics, HPLC-ESI-MS measurements have been performed. DF is not detected by the ESI mode under acidic or neutral conditions in the presence of relatively high concentration of buffer. To overcome this problem, it is necessary to perform a gradient elution permitting the removal of the major part of the buffer before the compounds reach the ionization source. Although the additional peaks are not so well resolved, this method allowed us to identify DF as the major peak eluting at 9.28 min (Figure 4A). [MW 2828, calculated from the m/z ) 1413.1 (M - 2 H)2- ratio of the parent ion] and to obtain the MW of the four main impurities (Figure 4B). The m/z ratios of the main impurities suggest that they probably correspond to some C60 derivatives formed during DF synthesis. These impurities may be attributed to an incomplete conversion of the parent ester form of DF to its acidic form and to a coupling reaction between C60 as well as some side products exhibiting structural defects formed during dendritic generation (11). The peak eluting at 9.223 min (Figure 4E)] [MW 3000, calculated from the m/z ) 1499.1 (M - 2 H)2- ratio of the parent ion, relative area 3%] probably corresponds to a triester form (tBu)3-D. The peaks eluting at 8.293 min (Figure 4C)] [MW 2554, calculated from the m/z ) (19) Walters, M. J. J. Assoc. Off. Anal. Chem. 1987, 70, 465-469.

Figure 4. (A) UV-visible chromatogram of a DF sample, with gradient elution (solvent A, 0.05 M ammonium formate pH 5.5; solvent B, acetonitrile; solvent C, water. Gradient program: at time 0-2 min, 80% A and 20% B; at time 15 min, 20% B and 80% C; at time 20 min, 20% B and 80% C. Flow rate: 1.0 mL/min at ambient temperature,detection, UV at 260 nm. (B) Reconstructed ESI-MS chromatogram (see A for chromatographic conditions). (C-F) Extracted spectra. Acquisition parameters: ESI, polarity negative, std/normal mode; CapExit, -100 V; trap drive, 75; skim 1, -60 V, S; scan range, 500-1900 m/z; accumulation time, 200 ms; summation, eight spectra.

1276.6 (M - 2 H)2- ratio of the parent ion, relative area 2%], at 9.47 min [MW 2810, calculated from the m/z ) 1404.2 (M - 2 H)2- ratio of the parent ion, relative area 5% (Figure 4A)] and at 11.93 min (Figure 4F) [MW 2560, calculated from the m/z ) 1279.2 (M - 2 H)2- ratio of the parent ion, relative area 0.5%] appear to correspond to DF compounds with dendrimers exhibiting structural defects (incomplete branches, etc). Among these, the impurities eluting after the DF and which are therefore less polar than it can also contain some ester groups, but their structure is difficult to deduce directly from their mass spectra. The method of separation proposed herein is now used to isolate these impurities in order to determine their accurate structures.

It is possible to estimate the purity of the DF sample from the UV-visible chromatogram obtained in Figure 4A because the differences between the MW of DF and those of the major impurities are relatively small ((6-7%). The UV-visible absorption spectra of the main impurities are all characteristic of those of cyclopropyl-fused monoadducts of C60. If we neglect the differences between the dilution factors of the chromatographic peaks, we can then consider that the response factors of the corresponding compounds are closely equivalent. These reasonable approximations allow deducing the purity of the DF sample (90 ( 1%) from the chromatogram (Figure 4A), which is closely equivalent to its relative area. Analytical Chemistry, Vol. 75, No. 16, August 15, 2003

4221

CONCLUSION In the first attempts of DF synthesis, the purity of the target compound (the acid) had not been completely verified. With the method described herein, a possibility of monitoring the hydrolysis is now available. By optimizing the conditions of the hydrolysis, a nearly quantitative conversion of ester to acid should be possible by varying the conditions. Our results show that it is possible to use RP-HPLC at a preparative scale to purify DF under acidic conditions. To obtain DF in a highly pure acidic form after HPLC separation, it is necessary to remove the buffer present in the DF fraction by flush chromatography on Si-C18, for example. This can be accomplished quickly in three steps: (1) loading the column with the DF fraction; (2) removing the salt by washing the column with a mixture of water-acetonitrile (80/20, v/v); and (3) eluting DF with a mixture of water-acetonitrile (5/95, v/v).

4222

Analytical Chemistry, Vol. 75, No. 16, August 15, 2003

Although it is possible to use unpurified DF as drug product, all impurities would need to be identified, tested for toxicity, and examined for metabolic fate. A more straightforward approach would be to purify DF using the method described in the present work. In conclusion, we show that the purification of DF exhibits the same difficulties and problems as standard (nonfullerene) dendrimers. ACKNOWLEDGMENT This work was performed under European Contract FMRXCT98-0192. We are grateful to Professor P. Potier for financial support. Received for review December 12, 2002. Accepted May 28, 2003. AC026419K