Anal. Chem. 2006, 78, 2532-2538
Overload for Ionized Solutes in Reversed-Phase High-Performance Liquid Chromatography David V. McCalley*
Centre for Research in Biomedicine, University of the West of England, Frenchay, Bristol BS16 1QY, UK
Overloading occurs for submicrogram quantities of ionized solutes particularly when using low ionic strength mobile phases at low pH (e.g., formic acid), even with highly inert silica RP-HPLC columns of normal dimensions. Much higher loads can produce a sharp L-shaped peak with retention above the column void volume, in line with the hypothesis that a small number of high-energy sites fill first and are rapidly overloaded, followed by a much larger number of weaker sites. However, charged acids and bases show identical overloading behavior; overloading is reduced as the mobile-phase ionic strength is increased. These findings raise questions about the physical nature of the strong sites. The rapid overloading of silica and purely polymeric phases could be explained by mutual repulsion of ionic species or their inability to fully penetrate the hydrophobic structure of the phase. However, these alternative hypotheses cannot readily explain the high total saturation capacities obtained using frontal analysis. Ion pairing with trifluoroacetic acid may reduce overload, while the effect is less important for formate or phosphate buffers. A surface layer of acetonitrile is not a prerequisite for rapid overloading, as shown by studies using purely aqueous buffers. For many years, problems have been experienced with the reversed-phase HPLC analysis of ionized solutes, due to the asymmetric peak shapes that are often obtained. Bases are of particular interest due to their importance as pharmaceuticals and compounds of biomedical significance. The problems are usually attributed to the detrimental interactions that can occur between ionized, underivatized silanols, which remain on the column surface, even after extensive silanization and end-capping, and the base which is usually charged under typical mobile-phase conditions (pH 2.5-7.5). However, difficulties persist on phases with very few ionized groups when operated at low pH, such as some highly inert “type B” silica-ODS phases, or purely polymeric poly(styrene-divinylbenzene) phases.1,2 It seems in these cases that phase overloading is the principal cause of poor peak shape. Overloading can be observed even with submicrogram quantities of ionized basic solutes with standard size analytical columns, and is especially serious when using weak acid buffers such as formic acid, suitable for mass spectrometry.3 * E-mail:
[email protected]. (1) McCalley, D. V. Anal. Chem. 2003, 75, 3404-3410. (2) Buckenmaier, S. M.; McCalley, D. V.; Euerby, M. R. Anal. Chem. 2002, 74, 4672-4681.
2532 Analytical Chemistry, Vol. 78, No. 8, April 15, 2006
The overloading mechanism of ionized solutes is not straightforward and may vary depending on the type of phase. Snyder, working with more active type A phases, proposed that ionized silanols provided strong retention sites for ionized bases that are filled rapidly, prior to the filling of high-capacity weak sites furnished by the ODS ligands. He showed by frontal analysis that the total saturation capacity of the phase was similar for ionized and neutral compounds, but the decline in efficiency for ionized bases with load was extremely rapid.4-6 While considerable developments in column technology have occurred since the publication of this work, the tailing that even now occurs on some type B phases, for very small solute mass, could still be attributed to the presence of a small number of ionized silanols; thus, their influence cannot be discounted on all modern phases. Ståhlberg, working specifically with anionic solutes to avoid the influence of ionized silanols, proposed that overloading was caused by mutual repulsion of ions of the same charge on the phase surface;7 this mechanism was also suggested as a possibility in Snyder’s work. The repulsion mechanism was used by McCalley to rationalize the overloading observed on purely polymeric phases2 and also independently by Neue.8 Gritti and Guiochon studied the overloading of a variety of charged and uncharged compounds, mostly using highly inert modern RP materials.9,10 They showed that these phases were still definitely heterogeneous, consisting of weak, intermediate, and strong sites, which could be accessed by different types of analyte. For example, the overloading by very small amounts of ionized bases could be explained by the overloading of the strongest sites, while the continuing high total saturation capacity of the phase for these analytes measured by frontal analysis was due to the filling of weak sites. However, the physical identity of these important strong sites has remained elusive. According to Gritti and Guiochon, they could be heterogeneities in the ODS layer induced by the presence of silanol groups; however, the measured adsorption energies of these strong sites appeared insufficient to indicate they were ionized silanols. This result agrees with previous suggestions that peak (3) McCalley, D. V. J. Chromatogr., A 2005, 1075, 57-64. (4) Eble, J. E.; Grob, R. L.; Antle, P. E.; Snyder, L. R. J. Chromatogr., A 1987, 384, 45-79. (5) Snyder, L. R.; Cox, G. B.; Antle, P. E. Chromatographia 1987, 24, 82-96. (6) Cox G. B.; Snyder, L. R. J. Chromatogr. 1989, 483, 95-110. (7) Ha¨gglund, I.; Ståhlberg, J. J. Chromatogr., A 1997, 761, 3-11. (8) Neue, U. D.; Wheat, T. E.; Mazzeo, J. R.; Mazza, C. B.; Cavanaugh, J. Y.; Xia, F.; Diehl, D. M J. Chromatogr., A 2004, 1030, 123-134. (9) Gritti, F.; Guiochon, G. Anal. Chem. 2005, 77, 1020-1030. (10) Gritti, F.; Guiochon, G. J. Chromatogr., A 2005, 1095, 27-39. 10.1021/ac052098b CCC: $33.50
© 2006 American Chemical Society Published on Web 03/10/2006
tailing could not be due to ionized silanols on some highly inert phases.1,2 Recent work has confirmed that overloading problems for bases reduce or disappear when analysis at high pH is used, where the solute is partially or completely deprotonated.2,8,11 Nevertheless, analysis at high pH remains something of a goal due to the relatively few phases stable under these conditions. There are also questions about the sensitivity of detection when high pH is used with positive electrospray MS. Thus, analysis at low pH is still by far the most widely used approach, where silanol ionization is suppressed. In the present study, overloading problems at low pH for some ionized compounds were explored, and the possible interpretation of these results using the various proposed theories is discussed. EXPERIMENTAL SECTION An Agilent 1100 HPLC system comprising autosampler, highpressure binary pump, heated column thermostat, and variablewavelength UV detector (1-µL flow cell 5-mm path length) was used in all experiments. A 0.01-cm-i.d. connection tubing of minimum length was used in order to limit extracolumn volume. Buffers were filtered using 0.2-µm membranes. The column temperature was maintained at 30 °C throughout. The column saturation capacity (ws) was determined from overloaded peaks where efficiency had decreased to 50% or less using the formula
W2base )
16t02(1 + k0)2 6t02k02wx + N0 ws
where Wbase is the peak width measured at base, wx is the sample mass of the overloaded peak, t0 is column void time, k0 is the retention factor, and N0 is the column efficiency for a small sample mass.12 While the base peak width was used for calculation of ws, for convenience and reproducibility of the measurement, the halfheight procedure 13 was employed for the calculation of N0 and for the values of column efficiency shown in Figure 1, Figure 8, and Table 1. The loading factor (Lf) was calculated from the mass of injected sample divided by ws multiplied by 100. Asymmetry factors (As) were calculated at 10% of the peak height from the ratio of the widths of the rear and front sides of the peak. All results were the mean of at least duplicate injections. The columns used were XTerra MS C18, 3.5-µm particle size, pore diameter 14 nm, surface area 176 m2/g, carbon load 15.6%, 15 cm × 0.46 cm i.d. (Waters, Milford, MA); a similar experimental hybrid C18 5-µm phase manufactured using pyridine as catalyst and XTerra RP18 with an embedded polar group, 5-µm particle size, pore diameter 14 nm, surface area 172m2/g, carbon load 14.6%, 25 cm × 0.46 cm i.d. Mobile-phase flow was 1.0 cm3 min-1. The columns were equilibrated for at least 10 h with each mobile phase in order to avoid problems of “slow equilibration” that can occur when analyzing ionizable compounds. Slow equilibration appears to involve small changes in column charge. As a result the retention of cationic and anionic solutes drifts in opposite directions, with the retention of ionized bases decreasing and of ionized acids (11) Davies, N. H.; Euerby, M. R.; McCalley, D. V. J. Chromatogr., A (in press). (12) Snyder, L. R.; Kirkland, J. J.; Glajch, J. L. Practical HPLC Method Development; Wiley: New York, 1997. (13) Neue, U. D. HPLC columns; Wiley-VCH: New York, 1997.
increasing with time.14 Retention times in the present study were stable after this period of equilibration. Ionic strength calculations were performed using the PHoEBuS program (Analis, Orleans, France) using correction of activity coefficients according to the Debye-Hu¨ckel equation. Acetonitrile (far-UV grade) formic acid, and trifluoroacetic acid were obtained from Fisher Scientific (Loughborough, U.K.), and all test solutes (highest available grade) were obtained from Sigma-Aldrich (Poole, U.K.). The hydrochloride salts of strong bases were used. Acetonitrile is flammable and toxic (Fisher MSDS 00170). It needs handling in a fume hood using appropriate protective equipment. Solvent wastes require specialist disposal (e.g., high-temperature incineration). RESULTS AND DISCUSSION Figure 1 shows efficiency plotted as a function of sample mass for six test solutes on XTerra MS using 28% 0.02 M formic acid in acetonitrile-72% aqueous 0.02 M formic acid at wwpH ) 2.7 ( s wpH ) 2.9 for 28% ACN). Efficiencies were calculated using the half-height method. For overloaded peaks that are not Gaussian in shape, this measurement is an approximation or “apparent efficiency”.15 However, it shows the same general trends for monitoring the progress of overload as efficiency calculated by the statistical moments procedure3 and is more reproducible at lower signal/noise values.16 The chemical structures of the solutes are shown in Figure 2. Three of the solutes in Figure 1 are fully charged (propranolol and nortriptyline are cations, 2-naphthalenesulfonic acid is an anion) whereas caffeine, phenol, and 3-phenylpropanol are uncharged under the experimental conditions. Caffeine is a strong hydrogen bond acceptor, and phenol has potential H-bond donor and acceptor properties. While a detailed study of these two compounds was not the primary objective of the present study, their overloading behavior is also of much interest. Gritti and Guiochon have shown considerable differences in the peak shape for caffeine as a function of sample load when using XTerra MS (a highly inert silica-organic hybrid) compared with Resolve C18 (a type A silica ODS phase containing larger numbers of acidic silanols). Rapid overload with increasing mass of caffeine was shown on Resolve but not XTerra.9 Possibly, overload of H-bond donor silanol sites by caffeine occurs on Resolve; few such sites exist on the hybrid XTerra phase giving a low contribution to retention. Despite these complications for some types of neutral compound, there is a striking distinction between the overloading behavior of the charged and uncharged compounds (note the logarithmic scale of the sample mass axis). Use of 0.02 M formic acid results in considerable overload with 0.1 µg (Lf ∼ 0.03%) or less of nortriptyline or propranolol. The similar overloading behavior of cationic and anionic species is clearly shown and is distinct from the also rather similar behavior of the group of uncharged compounds (at least on XTerra). Table 1 shows the saturation capacities (ws) of these and some additional solutes calculated from the Snyder equation. This relationship contains an empirical adjustment, such that it differs slightly from the equivalent expressions derived exclusively from (14) Marchand, D. H.; Williams, L. A.; Dolan, J. W.; Snyder, L. R. J. Chromatogr., A 2003, 1015, 53-64. (15) Golshan-Shirazi, S.; Guiochon, G. Anal. Chem. 1988, 60, 2364-2374. (16) Yau, W. W.; Rementer, S. W.; Boyajian, J. M.; de Stefano, J. J.; Graff, J. F.; Lim, K. B.; Kirkland, J. J. J. Chromatogr. 1993, 630, 69-77.
Analytical Chemistry, Vol. 78, No. 8, April 15, 2006
2533
Figure 1. Plot of column efficiency against sample mass for three neutral compounds (3-phenylpropanol, caffeine, phenol), and three charged compounds (propranolol, nortriptyline, 2-NSA) on XTerra MS. Mobile phase, 0.02 M formic acid in acetontrile-aqueous 0.02 M formic acid (overall concentration 0.02 M) wwpH 2.7 (28:72, v/v) except for caffeine (12.5:77.5, v/v). Flow rate, 1 mL min-1. Column temperature, 30 °C. Injection volume, 5 µL.
Figure 2. Chemical structures of test compounds.
theory.5,15-18 It gives an estimate of ws, compared with direct measurements using frontal analysis. However, it offers a less timeconsuming approach for larger numbers of determinations especially when comparisons are being drawn between substantially different values of ws, as in the current work. The ws calculated from column efficiency may correspond to the saturation of a small number of accessible or higher energy sites rather than the total capacity of the column as determined by frontal analysis;4,9-10 these sites are of primary importance for ionized solutes, because they appear to completely dominate the performance of the column. For the ionized solutes using 0.02 M formic acid, ws is invariably 100 000 µg. (17) Golshan-Shirazi, S.; Guiochon, G. Anal. Chem. 1989, 61, 462-467. (18) Knox, J. H.; Pyper, H. M. J. Chromatogr. 1986, 363, 1-30.
2534 Analytical Chemistry, Vol. 78, No. 8, April 15, 2006
Similar low values of ws for ionized solutes have been obtained using a variety of other modern RP columns with formic acid.1 The plots of column efficiency versus sample mass shown in Figure 1 provide a simple visual indication of the effects of overload, but do not take into account the effect of the different values of k obtained for the different solutes (see Table 1). For large k, the proportion of the injected solute mass held in the stationary phases is greater. Thus, the reduction in efficiency for the same quantity of solute should be greater for solutes with high k. However, the influence of these differences in k is relatively small for the solutes in Figure 1, which indicate the same trends as ws values. Figure 3 shows overlaid chromatograms for 0.01-5 µg of nortriptyline (Lf ) 0.003-1.3%) using the same conditions as Figure 1. While peaks for the smallest mass are reasonably symmetrical, broadening takes place with increasing sample mass leading to right-angled triangle shapes, giving results typically expected for a Langmuir isotherm. Already with 5 µg of nortriptyline (Lf ) 1.3%) the performance of the column is seriously compromised, yielding an efficiency of ∼10% of its small mass value. Results for still higher sample mass are shown in Figure 4. If a single-site overloading mechanism was operational, peaks should become increasingly broadened with sample load, with the solute front eventually approaching the void volume of the column as all of the retention sites are occupied. For nortriptyline, evidence of this behavior is shown for solute mass up to 50 µg (Lf ) 13%), with the peak maximum occurring substantially earlier than for small solute mass. However, increased loading to 200250 µg (Lf ) 50-63%) gave a sharp L-shaped peak at ∼3.7 min, with a long tail extending back to the retention time for the smallest mass of nortriptyline. This tail is concealed by the
Table 1. Column Performance Parameters for Test Solutes on Various Columnsa column XTerra MS 3.5 µm 15 cm × 0.46 cm
experimental hybrid, 5 µm 15 × 0.46 cm
XTerra RP18, 5 µm 25 × 0.46 cm a
solute
w wpH
I (mM/L)
k
N
As
ws (µg)
3-phenylpropanol phenol propranolol nortriptyline 2-NSA (Na salt) benzylamine (5% ACN) caffeine (12.5% ACN) p-xylenesulfonate (K salt; 10% ACN) nortriptyline (0.2 M formate) nortriptyline (0.0079 M TFA) diphenhydramine (0.0079 M TFA) nortriptyline 2-NSA nortriptyline (0.2 M formic) nortriptyline (0.02 M formic) nortriptyline (0.02 M phosphate) 2-NSA (0.02 M phosphate) procainamide (0% ACN)
2.7 2.7 2.7 2.7 2.7 2.7 2.7 2.7 2.7 2.3 2.3 2.7 2.7 2.7 3.8 2.7 2.7 2.7
1.9 1.9 1.9 1.9 1.9 1.9 1.9 1.9 18.3 7.8 7.8 1.9 1.9 18.3 10.5 22.0 22.0 1.9
4.7 2.1 1.6 6.4 1.5 0.5 1.8 8.3 11.1 17.7 6.6 10.3 0.7 10.9 9.9 11.2 0.6 1.3
19000 18200 9900 11900 9900 11800 14300 12000 15500 16900 15500 7200 4500 8500 8600 8600 4600 11400
1.1 1.2 1.3 1.4 1.6 1.3 1.1 1.5 1.2 1.3 1.3 1.4 1.4 1.3 1.3 1.2 1.3 1.2
>200000 >100000 300 400 300 100 >100000 400 3100 2700 2400 400 200 2800 2200 3000 2600 700b
Mobile phase as in Figure 1 except where indicated. k, N, and As are for small sample mass injections. b Normalized to 15-cm length.
Figure 3. Overlaid chromatograms for 5, 2.5, 1, 0.5, 0.25, 0.05, and 0.01 µg (Lf ) 1.3-0.003%) of nortriptyline. Conditions as Figure 1.
(different) low sensitivity of the y axis scales for these high mass injections in Figure 4. The appearance of this peak corresponds approximately to the estimate of ws calculated from a much smaller mass injection (0.5 µg, Lf ) 0.13%) using the Snyder equation. The behavior is equivalent to that found for the same solute and conditions by Gritti and Guiochon using frontal analysis10 and could be explained by the gradual total saturation of a small subset of high-energy sites and the subsequent occupation of the plentiful weakly retaining sites. The latter weak sites are hardly occupied during the filling of the strong sites.9,10 Note that the retention time of the L-shaped peak (Figure 4) is well above the void volume of the column (∼1.5 min.). Even larger masses of nortriptyline (500 and 1000 µg; Lf ) 125 and 250%) gave somewhat split peaks; however, the sample front did not show further reduced retention toward the column void volume, again providing evidence for the multiple site hypothesis. Nevertheless, caution is necessary in the rationalization of these results. The initial concentration of solute in the injected solution in these experiments far exceeds the concentration of the formate anions in the mobile phase. Because
Figure 4. Chromatograms for nortriptyline on XTerra MS for (a) 250 (Lf ) 63%), (b) 100 (25%), 150 (38%), and 200 µg (50%). (c) 12.5 (3.1%), 25 (6.3%), and 50 µg (13%). to indicates the column void time. Other conditions as Figure 1.
the neutral salt nortriptyline hydrochloride was used in both the present and previous studies,10 the pH of the injected solution (made up in the mobile phase) in either case was the same as that of the formic acid (as verified experimentally in the present study). Thus, the initial injection does not cause a pH disturbance. For small mass injections, chloride ions should be unretained; thus, nortriptyline cations are accompanied in some way by formate ions in their passage through the column. Figure 4a shows that the injected highly concentrated solution is considerably diluted by spreading over most of the length of the column. However, it is possible to estimate the average concentration of nortriptyline in the sharp peak from its area compared with that of the whole and the corresponding volume of mobile phase. Such calculations indicate that the solute concentration exceeds the concentration of formate anions in the mobile phase (“buffer overload”). Thus, solute cations may be accompanied unusually Analytical Chemistry, Vol. 78, No. 8, April 15, 2006
2535
Figure 5. Overlaid chromatograms on XTerra MS for (a) propranolol (cationic), 0.05-5 µg (Lf ) 0.017-1.7%) and (b) 2-NSA (anionic), 0.05-5 µg (Lf ) 0.017-1.7%). Conditions as Figure 1.
by injected chloride anions in order to preserve the neutrality of the system, which may be a factor in the appearance of the peak. Clearly the major concern for practical analysts is the very rapid overloading shown in Figure 3 for smaller solute masses. For difficult analytical separations, the broad peaks caused by the low saturation capacity reduce the resolution and the column peak capacity. Analysis should be carried out under conditions of highest detector sensitivity (e.g., λmax for UV detection) to achieve acceptable signal-to-noise ratios with small solute mass, to minimize the effects of overload. The overloading behavior of a second base, propranolol, together with the anionic species 2-naphthalenesulfonic acid (2NSA, injected as the sodium salt) with the same mobile phase as used for nortriptyline, is shown in Figure 5. Propranolol and 2-NSA have similar retention factors. Overloading of either solute is very similar to nortriptyline, despite 2-NSA carrying an opposite charge, with ws only 300-400 µg (see Table 1). Solubility problems were experienced with 2-NSA when attempting to use still higher loads. Instead, the more hydrophilic anionic solute p-xylene-2-sulfonic acid (XSA, as the potassium salt) was substituted. The organic modifier concentration was changed to 10% ACN (from 28% ACN), reflecting the greater hydophilicity of the new solute. The overloading behavior of XSA up to 5 µg (Lf ) 1.3%) mirrored that obtained for 2-NSA with similar ws values. At very high loads (5500 µg; Lf ) 1.3-125%), very similar results were obtained for XSA as for nortriptyline (Figure 6). Somewhat split peaks were obtained for 500 µg of solute; even higher masses (up to 2000 µg of solute) continued to give split peaks but produced only a very small further reduction in retention time, so that it remained considerably above the column void time, as was found for nortriptyline. Thus, at sample loads up to milligram quantities, the overloading behavior of cationic and anionic sites was extremely similar. If small numbers of higher energy sites indeed cause overloading, then it seems that the effect of these sites is the same for positively and negatively charged species, which is difficult to rationalize. Furthermore, similarly low ws values for charged solutes on purely polymeric phases have been demon2536 Analytical Chemistry, Vol. 78, No. 8, April 15, 2006
Figure 6. Overlaid chromatograms for of p-xylene-2-sulfonic acid. Mobile phase 0.02 M formic acid in acetonitrile-aqueous 0.02 M formic acid wwpH 2.7 (10:90, v/v). (a) 500 (Lf ) 125%), (b) 12.5 (3.1%), 25 (6.3%), and 50 µg (13%), and (c) 5 µg (1.3%). Other conditions as Figure 1.
strated2 (e.g., for PRP-1, a poly(styrene-divinylbenzene) copolymer). Perhaps ionized solutes have difficulty in penetrating the hydrophobic structure of an ODS or polymeric layer and are confined to surface sites which are much more limited in number. Alternatively, this rapid loading can be explained by the concept of mutual solute repulsion, which could occur on any surface and between cations or between anions. However, these alternative theories cannot readily explain the existence of a multiple-site mechanism or the similar high total column saturation capacity for ionized and neutral solutes indicated by frontal analysis. Irrespective of its nature, the retention mechanism does not seem to involve ionized silanols sites on the basis of work by the present author or by Gritti and Guiochon. Previously, it was shown that increased mobile-phase ionic strength gave increased saturation capacities;1,19 such a result argues against the supposition that ionized solutes are held in the stationary phase only as classical neutral ion pairs,20 since it is difficult to see why these would be influenced by buffer concentration. Clearly, solute ions cannot exist as separate entities on the phase surface and must be surrounded (at least loosely) by buffer ions of opposite charge. As buffer concentration increases, solute ions could be increasingly screened from one another, lessening the effects of mutual repulsion. Alternatively, increased ion pairing in the stationary phase may lessen this repulsion1,3 or simply improve access to the hydrophobic stationary phase. Overloading was therefore studied for nortriptyline in acetonitrile-0.2 M formic acid, adjusted to the same swpH (2.9) with ammonia. Ionic strength (I) is increased ∼10-fold over 0.02 M formic acid (see Table 1; note calculations do not take into account the effect of the small acetonitrile concentrations in the mobile phase). The ws concurrently increased from 400 to 3100 µg (see Table 1) showing that buffer concentration indeed affects overload. Figure 7 shows peak profiles for very high loads of nortriptyline in the stronger buffer. The L-shaped peak was not (19) Gritti, F.; Guiochon, G, J. Chromatogr., A 2004, 1041, 63-75. (20) Gritti, F.; Guiochon, G. Anal. Chem. 2004, 76, 7310-7322.
Figure 7. Chromatograms on Xterra MS for nortriptyline. Mobile phase, 0.2 M formic acid in acetonitrile-aqueous 0.2 M formic acid 28:72 (v/v) adjusted to swpH 2.9 with ammonia solution. (a) 250 (Lf ) 8.1%), (b) 12.5 (0.40%), 25 (0.81%), and 50 µg (1.6%), and (c) 5 µg (0.16%).
yet observed at 250-µg load, as obtained for 0.02 M formic acid (Figure 4); evidence of this peak was not obtained until loads of >2000 µg (Lf ) 65%) were utilized. This result could be due merely to the increase in ws or to the increased amount of solute necessary to overload the buffer (see above). The k for nortriptyline increased from 6.4 to 11.1 with increased buffer concentration (Table 1). Previously it was suggested that such increases might be due to increased ion pairing as the concentration of formate anions increases.3 However, the ion pair properties of formate are likely to be low. An alternative explanation is the presence of positively charged sites on XTerra, caused by residues of the particular base catalyst used in its preparation.1,21 Screening of these column sites by stronger buffers might produce increased retention times of similarly positively charged solutes. To investigate the possibility that overloading effects were influenced by the presence of catalyst residues, an experimental hybrid column was obtained, prepared using a classical pyridine catalyst instead of the normal XTerra catalyst. With the experimental hybrid, only small differences in k for nortriptyline using the 0.02 and 0.2 M buffers were obtained (k ) 10.3 and 10.9, respectively; see Table 1). These results suggest that stationaryphase ion pair formation with formate may indeed be low and that the observed retention increases are mainly due to the catalyst effect. Carr and co-workers have demonstrated a very low incidence of ion pairing with hydrophilic anions, at least in the mobile phase,22,23 although the situation is different in the stationary phase. However, overloading of the experimental hybrid column was found to be very similar to the regular column both at low buffer concentration (ws for nortriptyline 400 µg on both columns) and at high buffer concentration (2800 and 3100 µg, respectively). The L-shaped peak was noted also on the experimental column at ∼250 µg of nortriptyline using 0.02 M buffer (21) Mendez, A.; Bosch, E.; Rose´s, M.; Neue, U. D. J. Chromatogr., A 2003, 986, 33-44. (22) Dai, J.; Mendonsa, S. D.; Bowser, M. T.; Lucy, C. A.; Carr, P. W. J. Chromatogr., A 2005, 1069, 225-234. (23) Dai, J.; Carr, P. W. J. Chromatogr., A 2005, 1072, 169-184.
Figure 8. Plot of column efficiency vs mass nortriptyline for XTerra MS. Mobile phase, upper plot, 0.0079 M TFA in acetonitrile-aqueous 0.0079 M TFA pH 2.3; lower plot, 0.02 M formic acid in acetonitrileaqueous 0.02 M formic acid.
(Lf ) 63%), but not until considerably higher solute mass with 0.2 M buffer, similar to the regular column. Furthermore, ws of the anionic solute 2-naphthalenesulfonic acid was very similar on both regular and experimental hybrid phases (300 and 200 µg). These results suggest that while the presence of catalyst residues on the regular column influences retention in low ionic strength buffers such as 0.02 M formic acid, these residues do not have an appreciable effect on the column overloading properties. Formic acid (∼0.02 M) is a popular volatile mobile-phase additive for MS. Its low ionic strength leads to smaller values of ws. Adjusting the pH of formic acid to its pKa (3.75) is an alternative means of increasing ionic strength. For aqueous 0.02 M formate at pH 3.75, the ionic strength is ∼10.5 mM, giving ws for nortriptyline of 2200 µg on the experimental hybrid (Table 1). The similar k of nortriptyline at either pH, and the continuing low asymmetry factor (1.3) for nortriptyline at the higher pH, both indicate the absence of silanol ionization. For nortriptyline on the experimental hybrid, ws was 400, 2200, 2800, and 3000 µg in line with the increasing ionic strength of the aqueous buffers 1.9 (0.02 M formic wwpH 2.7), 10.5 (0.02 M formic wwpH 3.75), 18.3 (0.2 M formate, swpH 2.9), and 22 mM (0.02 M phosphate wwpH 2.7). Phosphate is, like formate, a hydrophilic anion with low ion-pairing properties. Phosphate also considerably increases ws for the anionic solute 2-NSA. Trifluoroacetic acid (TFA) is another popular volatile HPLC-MS additive, often used at approximately the same weight percent as formic acid.24 The ws for nortriptyline on XTerra MS using 0.0079 M TFA (ionic strength 7.8 mM) was 2700 µg, similar to that obtained with 0.2 M formate or 0.02 M phosphate of around double the ionic strength. Figure 8 shows the decline in efficiency with mass of nortriptyline using TFA and formic acid at the same weight percent, indicating improved results with TFA. The considerably enhanced retention of nortriptyline in TFA (k ) 17.7) can be attributed to ion pairing in the stationary phase, which reduces the number of charged species and contributes to the reduction in overload caused by its higher ionic strength. Nortriptyline (pKa ) 10.3) is completely ionized in formic acid or TFA, despite the lower pH of the latter. It is unlikely that the lower (24) McCalley, D. V. J. Chromatogr., A 2004, 1038, 77-84.
Analytical Chemistry, Vol. 78, No. 8, April 15, 2006
2537
pH of TFA could influence the ws on a phase with negligible silanol ionization around this pH. Overloading could conceivably be influenced by the low solubility of ionized solutes in the acetonitrile layer known to exist on silica-ODS phases.9,25 If so, elimination of the layer by use of purely aqueous mobile phases should have a profound effect. A problem of ODS phases such as XTerra MS is lack of wetting by water; however, it does not occur with the related XTerra RP18 phase, which contains an embedded polar carbamate group.26,27 Note XTerra RP18 was shown previously to give similar rapid overloading for charged compounds as nonembedded C18 phases.11 Procainamide, a highly hydrophilic base of low retention, was chosen for the study. However, similar rapid overloading was obtained (ws ) 700 µg), suggesting that the layer of acetonitrile is not a prerequisite for this effect. CONCLUSIONS Overloading effects for charged compounds on RP columns at low pH are much greater than for uncharged compounds due to the significantly lower column saturation capacities. On some modern phases such as the hybrid silica used in this study, the low column saturation capacity, rather than interaction with ionized silanols, is the major cause of the broad peaks obtained. (25) Chan, F.; Yeung, L. S.; LoBrutto, R.; Kazakevich, Y. V. J. Chromatogr., A 2005, 1087, 158-165. (26) Walter, T. H.; Iraneta, P.; Caparella, M. J. J. Chromatogr., A 2005, 1075, 177-183. (27) Neue, U. D.; Cheng, Y. F.; Lu., Z.; Alden, B.; Iraneta, P.; Phoebe, C. H.; van Tran, K. Chromatographia 2001, 54, 169-177.
2538
Analytical Chemistry, Vol. 78, No. 8, April 15, 2006
The presence of a layer of acetonitrile on the phase surface is not a prerequisite for this rapid overloading. Large increases in mass give an L-shaped peak, suggesting retention is taking place on at least two kinds of site. However, very high solute concentrations may lead to “buffer overload”, where locally the solute concentration exceeds that of the buffer, which could influence the production of this peak. The first type of site is rapidly overloaded, followed by the filling of a much larger number of weak sites. Alternatively, ionic repulsion could explain the similar overloading of cations and anions, the similar behavior of silica-ODS and purely polymeric phases, and the increase in ws with buffer concentration, which are difficult to explain on the multisite theory. In addition, charged species may have difficulty in penetrating the hydrophobic structure of phases, leading to low saturation capacities. However, these alternative theories cannot readily explain the high total saturation capacities for both ionized and neutral solutes shown by frontal analysis. Ion pairing with buffer anions such as TFA can increase ws; however, the effect appears to be small in formate or phosphate mobile phases. Results at very high loadings are of limited relevance to practitioners who require high column peak capacity (good resolution) and high signal-to-noise ratio (good sensitivity) due to the loss of efficiency, which occurs already at much lower mass of ionized solute.
Received for review November 29, 2005. Accepted January 31, 2006. AC052098B
