3509 J. Sep. Sci. 2011, 34, 3509–3516 Joseph J. Pesek1 Maria T. Matyska1 Steven M. Fischer2 1 Department of Chemistry, San Jose State University, San Jose, CA, USA 2 Agilent Technologies, Santa Clara, CA, USA Received July 11, 2011 Revised August 30, 2011 Accepted August 31, 2011 Research Article Improvement of peak shape in aqueous normal phase analysis of anionic metabolites The problem of poor peak shape for multiply charged negative-ion analytes under aqueous normal phase (ANP) conditions is investigated. Because less than adequate efficiency and symmetry can occur with a variety of mobile phases, gradients and additives, and to varying degrees depending on the instrument, sources other than solute/stationary phase interactions are more likely the cause. Since it is known that many of these compounds can interact strongly with metal ions, addition of a chelating agent to the mobile phase and/or the sample solvent was tested. In particular, ethylenediaminetetraacetic acid (EDTA) is a compound that forms strong complexes with most di-and tri-valent metal ions and can be used to verify whether trace amounts of these species are the source of the problem. In addition, the retention of a number of anionic compounds was measured at various concentrations of ammonium acetate and formate with EDTA in the mobile phase. Keywords: Hydrophilic compounds / Metal ions / Silica hydride stationary phase DOI 10.1002/jssc.201100607 1 Introduction Over the last several years, silica hydride stationary phases have demonstrated capabilities that are superior than many existing materials and possess the potential to provide solutions for some of the most demanding biological, clinical and pharmaceutical analyses [1–19]. These columns are referred to as TYPE-CTM silica because they are fundamentally different from many current HPLC stationary phases that utilized ordinary silica. The essential difference between the two materials is that TYPE-CTM phases have a surface that is populated with Si-H (silica hydride) groups while Si-OH (silanols) groups dominate ordinary silica. While this might seem like a trivial difference, it has profound effects on the fundamental nature of the material and hence its chromatographic properties. Silanols are very polar and can often interact irreversibly with polar compounds, especially bases, while silica hydride is weakly hydrophobic and results in less strongly adsorptive properties that are advantageous for good chromatographic performance. Several prominent innovations have emerged from investigations using silica hydride-based stationary phases and aqueous normal phase (ANP) chromatography. One important feature is the dual retention capability of all silica hydride columns fabricated to date. Both hydrophobic and Correspondence: Professor Joseph J. Pesek, Department of Chemistry, San Jose State University, San Jose, CA 95192, USA E-mail: joseph.pesek@sjsu.edu Fax: 11-408-924-4945 Abbreviations: ANP, aqueous normal phase; hydrophilic interaction liquid chromatography & 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim HILIC, hydrophilic compounds can be retained efficiently when a silica hydride column is used. The elution order depends on the type of gradient used (RP gradient from high aqueous to high organic versus an ANP gradient from high organic to high aqueous). This retention behavior distinguishes ANP from hydrophilic interaction liquid chromatography (HILIC). While retention of polar compounds occurs at high organic content in the mobile phase ordinary reversed-phase retention can be easily obtained by using high-aqueous content mobile phases. This significant dual retention capability (either simultaneous or sequential) cannot be obtained using either standard reversed-phase or HILIC separation materials. Another characteristic is that the hydrophobic silica hydride surface absorbs less water than ordinary silica thus creating a more uniform solvent environment around the stationary phase and at the particle interface. The lack of a substantial water layer on the hydride surface (a single mobile phase environment) is likely responsible for the rapid equilibration of the stationary phase after gradient analyses and perhaps the higher efficiency observed in the ANP compared with HILIC [4]. Another feature of the silica hydride materials that has been established for a wide range of samples and stationary phases is the reproducibility of retention times from run to run. Reproducibilities for a particular analysis are usually in the range of 0.1–0.5% RSD, even for samples in physiological matrices. Another attractive characteristic of the silica hydride materials is their long-term durability. The bonding of an organic moiety to a silica hydride surface via hydrosilation results in a direct silicon–carbon bond. Additional correspondence: Maria T. Matyska E-mail: maria.matyska-pesek@sjsu.edu www.jss-journal.com 3510 J. Sep. Sci. 2011, 34, 3509–3516 J. J. Pesek et al. One problem that has been identified in our laboratory for highly negatively charged compounds is poor peak shape. Similar observations have also been reported for HILIC columns using a nano-column format [20]. Nucleotides and organic acids with more than one carboxylate group are the primary analytes affected. Originally, it was thought that this was just a result of the high negative charge of the analyte and the resulting poor exchange kinetics. However, it was shown in the HILIC study that the real cause was the presence of trace metals, particularly iron, that is often found in very low concentrations from a variety of sources within the chromatographic system and the solvents. Our work has confirmed this conclusion using several LC/MS and LC/MS/MS instruments. In addition, it appears that copper may also contribute to the poor performance of the highly negatively charged ions. The solution in the HILIC study was to add a trace amount of EDTA to the mobile phase. Thus, in order to make the ANP method suitable for every type of negatively charged species a remedy must be found to remove the trace amounts of iron and copper in the system. This study explores the use of EDTA as a possible solution to improve the peak shape of negatively charged analytes in the ANP mode in a standard analytical column format. 2 Materials and methods 2.1 Materials The silica hydride stationary phase used in this study was the Cogent Diamond Hydride (DH) material in 150  2.1 mm columns (MicroSolv Technology, Eatontown, NJ, USA). The phase contains a small amount of an organic moiety (2% carbon as reported by the manufacturer) on a silica hydride surface. The analytes and mobile phase additives used in this study were purchased from Sigma-Aldrich (Milwaukee, WI, USA) in the highest purity available. Mobile phase solvents used were HPLC grade. Figure 1. EIC of ATP in negative-ion mode at m/z 505.9885. (A) No addition of EDTA and (B) after injection of 100 mL of a 1-mg/mL solution of EDTA. Mobile phase: Solvent A, 50:50 MeOH/DI water with 0.05% formic acid. Solvent B, 90:10 acetonitrile/DI water with 10 mM ammonium acetate adjusted to pH 7. Injection vol. 5 1 mL. Gradient: 0.0–1.0 min at 100% B; 1.0–3.0 min to 90% B; 3.0–6.0 min at 90% B; 6.0–7.0 min to 80% B; 7.0–9.0 min at 80% B; 9.0–10.0 min to 50% B; 10.0–12.0 min at 50% B; 12.0–13.0 min to 30% B; 13.0–15.0 min at 30% B. Flow rate 5 0.4 mL/min. & 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.jss-journal.com J. Sep. Sci. 2011, 34, 3509–3516 2.2 Instrumentation The HPLC was an Agilent (Little Falls, DE, USA) 1200SL Series LC system, including degasser, binary pump, temperature-controlled autosampler and temperaturecontrolled column compartment. The mass spectrometer system was an Agilent (Santa Clara, CA, USA) Model 6220 MSD TOF with a dual sprayer electrospray source (ESI). 2.3 Methods Stock solutions of the analytes were made in deionized (DI) water in the range of 0.2–0.7 mg/mL. Sample solutions were Liquid Chromatography 3511 made by diluting the stock 1:100 in 50:50 acetonitrile/water containing the mobile phase additive used in the analysis. The flow rate was 0.4 mL/min. The column temperature was 201C. 3 Results and discussion Negatively charged analytes like ATP with more than a single ionizable group often appear as irregular peaks when eluted in the ANP mode. An example of a typical chromatogram for ATP is shown in Fig. 1A. At first this was believed to be caused by poor exchange kinetics between the stationary phase and the mobile phase solvent. A Figure 2. EICs for nucleotides utilizing addition of 10 mM EDTA to the B solvent. Gradient: 0.0–1.0 min at 100% B; 1.0–15.0 min to 0% B; 15.0–17.0 min at 0% B. Other conditions same as Fig. 1. Peak identification: 1, AMP; 2, ADP; 3, ATP; 4, NAD. Figure 3. EICs for nucleotides utilizing addition of 10 mM EDTA to the A solvent. All conditions and peak identification are the same as Fig. 2. & 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.jss-journal.com 3512 J. J. Pesek et al. number of approaches were attempted to try to remedy this situation including varying the pH of the mobile phase, utilizing different gradients and increasing the column temperature. None of these proved satisfactory in achieving both high symmetry and efficiency. In addition, the degree of asymmetry varied considerably between LC systems even under identical experimental conditions. Because low concentrations of metal ions, particularly iron, were suspected as the cause of poor peak shape for anionic metabolites in nano-LC [5], it was assumed that a similar problem might be present in typical analytical HPLC systems. In order to test this assumption, EDTA was introduced as part of either the mobile phase or the sample solvent. There are three possible means of utilizing EDTA in the LC system in order to remove potential trace metal ion contaminants: (i) injecting a significant volume of a high- J. Sep. Sci. 2011, 34, 3509–3516 concentration solution of EDTA in order to remove the trace metals in a single treatment; (ii) including EDTA at low concentration as part of the mobile phase; and (iii) preparing the sample with EDTA as part of the solvent. Each of these approaches was tested in order to determine which one would produce good peak shape and efficiency as well as have the least effect on sensitivity when using MS for detection. The first approach was tested by injecting 100 mL of a 1-mg/mL solution of EDTA into the column and then equilibrating the column for 15 min using the starting composition of the mobile phase for the gradient being used in the subsequent injection. The initial injection resulted in peaks of very low intensity. The chromatogram in Fig. 1B shows the result of the fifth injection after the EDTA treatment of the same ATP sample solution and using the same gradient as in Fig. 1A with the DH column. A very Figure 4. EICs for nucleotides and related compounds utilizing addition of 5 mM EDTA to Solvents A and B. Mobile Phase: Solvent A, 50:50 MeOH/DI water with 0.025% formic acid. Solvent B, 90:10 acetonitrile/DI water with 5 mM ammonium acetate adjusted to pH 7. Injection vol. 5 1 mL. Gradient same as Fig. 1. Flow rate 5 0.4 mL/min. Peak identification: 1–4 same as Fig. 2; 5, NADP; 6, UDP-galactose; 7, UDPglucose; 8, GTP; 9, galactose-1-phosphate. & 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.jss-journal.com J. Sep. Sci. 2011, 34, 3509–3516 Liquid Chromatography 3513 Figure 5. EIC at m/z 115.0037. (A) No EDTA in the mobile phase and (B) after injection of 10 mL of 3.4 mM EDTA. All conditions same as Fig. 1. Peak identification: 1, maleic acid and 2, fumaric acid. significant improvement is seen in the peak shape so that both retention time identification and quantitation are possible. The reproducibility of this data including peak intensity is quite good for approximately ten subsequent injections after which noticeable peak tailing is evident indicating that the presence of trace metal ions such as iron and copper has returned to levels that have measureable effects. Peak shape identical to that shown in Fig. 1B can be obtained by another injection of the high concentration EDTA solution. Similar behavior is observed for a number of other di- and tri-phosphate containing compounds. These results indicate that the source of peak tailing is external to the column and as reported in the nano-LC study is probably due to trace metal ions from either the instrument or the solvents used for preparing the mobile phase. However, from a practical aspect, the use of repeated injections of EDTA and equilibration of the column on a frequent basis does not seem desirable. The next approach tested was to place EDTA in the sample solution. At concentrations of 5–10 mM there was a & 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim noticeable improvement in peak shape (AS5) but not sufficient for most analytical determinations. At 10–20 mM there was significant improvement in peak shape (AS3) over that shown in Fig. 1A, but there was a noticeable decrease (450%) in the sample signal since it was determined that EDTA has retention comparable to ATP and several other phosphate containing species on the DH column. The retention of EDTA on the DH column was verified by monitoring the extracted ion chromatogram at m/z 291. Thus it was concluded that this was not a viable approach for improving peak shape for these anionic species. The third means of adding EDTA to the system is to place a low concentration in the mobile phase. There are several possibilities for accomplishing this. The first case involves adding EDTA to only Solvent B (the organic component). The results of this test are shown in Fig. 2 for several nucleotides with optimization of the gradient under these conditions. As can be seen, there are reasonably good peak shapes for some compounds but others were still not suitable for developing a good analytical method. Next, the www.jss-journal.com 3514 J. J. Pesek et al. J. Sep. Sci. 2011, 34, 3509–3516 Figure 6. EIC at 521.9834. (A) Mobile phase same as Fig. 1 except with 10 mM ammonium formate and (B) mobile phase same as Fig. 1. Injection vol. 1 mL and flow rate 5 0.4 mL/min. Gradient: 0.0 min 100% B; 0.0–10.0 min to 20% B. Peak identification: 1, UDP-galactose and 2, UDPglucose. EDTA was added only to the aqueous component of the mobile phase (Solvent A). The results for this approach are shown in Fig. 3 for a representative number of analytes. As can be seen, the peak shape has improved in comparison to the results obtained for EDTA in the B solvent. In all cases, the chromatographic peak shapes are satisfactory for most analytical methods. The final approach investigated was to add EDTA to both the A and B solvents. The results of this mobile phase composition are shown in Fig. 4. Under these conditions each of the species tested gives both good efficiency as well as peak symmetry. In addition, UDP-glucose and UDP-galactose are almost completely resolved. The concentrations of all the mobile phase constituents are relatively low: 5 mM for EDTA, 0.025% for formic acid and 5 mM for ammonium formate. Under these conditions, signal suppression in the MS is minimized which is beneficial for method development requiring low detection limits. The use of 10 mM EDTA caused measurable signal depression for many compounds. At 20 mM EDTA signal intensity for most compounds was significantly decreased (450%). The ionization technique used is gas-assisted & 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim electrospray. This technique is capable of producing charged aerosol at much higher flow rates than pure electrospray, but is subject to the effect of non-volatile material in the solvent. Ammonia acetate and formate buffers are essentially non-volatile in the time frame of the electrospray ionization and desorption process. The result is sensitivity decreases with increasing buffer concentration in a nonlinear fashion. EDTA’s effect on signal response arises from a different mechanism than the buffers. It is an intermediately strong acid resulting in charge competition with the analytes to be ionized by the electrospray process. The continuous addition of 5 mM of EDTA to the mass spectrometer ion source will not damage the instrument. The EDTA will deposit on the surfaces of the ion source but cause no damage. Occasional cleaning of the ion source will be required but the ‘‘volatile’’ buffers ammonium formate or ammonium acetate accumulate faster than the EDTA. For some compounds, the use of EDTA in the mobile phase is not necessary. Good examples are the isobaric diprotic acids, maleic and fumaric. The results of one gradient run in the ANP mode are shown in Fig. 5. Figure www.jss-journal.com J. Sep. Sci. 2011, 34, 3509–3516 Liquid Chromatography 3515 Figure 7. EICs for NAD and NADP. Gradient and other conditions same as Fig. 6. Peak identification: 1A, NAD in 5 mM ammonium acetate; 1B, NAD in 10 mM ammonium acetate; 2A, NADP in 5 mM ammonium acetate; 2B, NADP in 10 mM ammonium acetate. Figure 8. EICs for ATP and GTP. Gradient and other conditions same as Fig. 6. Peak identification: 1A, ATP in 5 mM ammonium acetate; 2A, GTP in 5 mM ammonium acetate; 1B, ATP in 10 mM ammonium acetate; 2B, GTP in 10 mM ammonium acetate. 5A is the EIC at m/z 115 for the two compounds using a mobile phase with 10 mM ammonium acetate as the additive. The column is then injected with 10 mL of a 3.4-mM solution of EDTA. Figure 5B shows the chromatogram that is obtained after the EDTA treatment. For all practical purposes, it is essentially the same as the one shown in Fig. 5A. Similar results are obtained for the mobile phase used in Fig. 4. Using a variety of gradients, it is possible to obtain good peak shape for both of these acids in ammonium acetate or formate buffers. The compounds that are affected by trace metals are those capable of coordination to multi& 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim valent cations. Phosphate-containing compounds can coordinate to multivalent cations and the level of impact is dependent on the number of phosphates in the molecule; 34241. The di-acids are not affected because they do not contain phosphate and do not have a geometry that supports cation coordination. The result is that the di-acids do not coordinate with multivalent ions while ATP strongly coordinates with them. In addition to the amount of EDTA, factors such as type of additive as well as its concentration were also evaluated with respect to retention and signal intensity. Figure 6 www.jss-journal.com 3516 J. Sep. Sci. 2011, 34, 3509–3516 J. J. Pesek et al. shows the effect of the type of additive on retention, resolution and signal intensity. In this case, the evaluation is done for UDP-glucose and UDP-galactose, two compounds that are difficult to resolve due to the minor structural differences between them. There are some differences observed in the results between the additives. There is slightly longer retention for ammonium acetate but slightly better resolution for ammonium formate. The signal levels are comparable since the longer retention in ammonium acetate results in somewhat broader peaks. Another example is shown in Fig. 7 for the separation of NAD and NADP. Retention is longer for both compounds in the 10 mM ammonium acetate mobile phase than in the 5 mM buffer. There is some decrease in signal intensity for NAD when a 10-mM solution of the additive is used in comparison with the 5 mM buffer. Figure 8 shows another comparison of experimental conditions. In this case using ammonium acetate as the buffer it can be seen that pH effects retention similarly to concentration. In this example both ATP and GTP elute at essentially the same retention time with the gradient used. The higher concentration mobile phase (10 mM ammonium acetate) also is adjusted to a higher pH than the lower concentration eluent (5 mM). Longer retention is observed in the higher concentration of ammonium acetate at elevated pH. In addition, there is a small signal increase at the higher ammonium acetate concentration for both compounds. 4 Concluding remarks Peak shape (efficiency and symmetry) can be significantly improved by the addition of EDTA at the low micromolar level to the mobile phase. Continuous treatment is more effective than a single injection of a high-concentration EDTA solution or addition of the chelating agent to the sample solvent. It is assumed that the cause of the poor performance is the presence of trace amounts of metal ions in the system due to the effect of EDTA and the variability of peak shape from instrument to instrument. The use of EDTA is in most instances a more practical method of removing trace metals rather than vigorous leaching of the instrument with an agent such as a strong acid. The main drawback to the use of EDTA is the suppression of the analyte signal because under the mobile phase conditions used it also has more than a signal negative charge. Therefore, investigations are continuing into other reagents that might be used to complex the trace metal ions in the system without having a significant impact on analyte sensitivity. The authors acknowledge the support of the National Science Foundation (Grant 0724218) and Agilent Technologies, & 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Santa Clara, CA for donation of the equipment used in this study. The authors have declared no conflict of interest. 5 References [1] Brown, L., Ciccone, B., Pesek, J. J., Matyska, M. T., American Lab. 2003, 35, 23–29. [2] Pesek, J. J., Matyska, M. T., LCGC 2006, 24, 296–303. [3] Pesek, J. J., Matyska, M. T., J. Sep. Sci. 2009, 32, 3999–4011. [4] Pesek, J. J., Matyska, M. T., Loo, J. A., Fischer, S. M., Sana, T. R.., J. Sep. Sci. 2009, 32, 2200–2208. [5] Pesek, J. J., Matyska, M. T., Fischer, S. M., Sana, T. R., J. Chromatogr. A 2008, 1204, 48–55. [6] Pesek, J. J., Matyska, M. T., Hearn, M. T. W., Boysen, R. I., J. Chromatogr. A 2009, 1216, 1140–1146. [7] Pesek, J. J., Matyska, M. T. in: Grushka, E., Grinberg, N. (Eds.), Advances in Chromatography, vol. 48, CRC Press, Boca Raton, FL 2010, pp. 255–288. [8] Pesek, J. J., Matyska, M. T., Duley, J., Zamzami, M., Fischer, S. M., J. Sep. Sci. 2010, 33, 930–938. [9] Young, J. E., Matyska, M. T., Pesek, J. J., J. Chromatogr. A 2011, 1218, 2121–2126. [10] Pesek, J. J., Matyska, M. T., in: Wang, P. G., He, H. (Eds.), Hydrophilic Interaction Chromatography (HILIC) and Advanced Applications, CRC Press, Boca Raton, FL 2011, pp. 1–26. [11] Watson, D. G., Parasitology 2010, 137, 1409–1423. [12] Dejaegher, B., Vander Heyden, Y., J. Sep. Sci. 2010, 33, 698–715. [13] Dell’mour, M., Jaitz, L., Oburger, E., Puschenreiter, M., Koellensperger, G., Hann, S., J. Sep. Sci. 2010, 33, 911–922. [14] Cheng, C., Chen, C.-S., Hsieh, P.-H., J. Chromatogr. A 2010, 1217, 2104–2110. [15] Callahan, D. L., De Souza, D., Bacic, A., Roessner, U., J. Sep. Sci. 2009, 32, 2273–2280. [16] Butterfield, T. R., Fischer, S. M., Rhee, K. Y., J. Sep. Sci. 2009, 32, 2262–2265. [17] de Carvalho, L. P. S., Zhao, H., Dickinson, C. E., Arango, N. M., Lima, C. D., Fischer, S. M., Ouerfelli, O., Rhee, K. Y., Chem. Biol. 2010, 17, 323–332. [18] Marrero, J., Rhee, K. Y., Schnappinger, D., Pethe, K., Ehrt, S., Proc. Natl. Acad. Sci. USA 2010, 107, 9819–9824. [19] De Carvalho, L. P. S., Fischer, S. M., Marrero, J., Nathan, C., Ehrt, S., Rhee, K. Y., Chem. Biol. 2010, 17, 1122–1131. [20] Myint, K. T., Uehara, T., Aoshima, K., Oda, Y., Anal. Chem. 2009, 81, 7766–7772. www.jss-journal.com