Effect of Polar Protic and Polar Aprotic Solvents on Negative-Ion

Oct 15, 2012 - Department of Chemistry & Biochemistry, James Madison University, MSC 4501, Harrisonburg, Virginia 22807, United States. •S Supportin...
0 downloads 0 Views 1MB Size
Download PDF
Article pubs.acs.org/ac

Effect of Polar Protic and Polar Aprotic Solvents on Negative-Ion Electrospray Ionization and Chromatographic Separation of Small Acidic Molecules Brian A. Huffman, Michael L. Poltash, and Christine A. Hughey* Department of Chemistry & Biochemistry, James Madison University, MSC 4501, Harrisonburg, Virginia 22807, United States S Supporting Information *

ABSTRACT: A comprehensive study investigated the effect of polar protic (methanol and water) and polar aprotic (acetonitrile and acetone) solvents on the chromatographic separation and negative-ion electrospray (ESI) response of 49 diverse small, acidic molecules. Flow injection experiments on a triple quadrupole were used to measure the response in neat solvents after optimization of source conditions and implementation of a rigorous quality control program (the later ensured that changes in analyte response were due to the analyte/solvent measured and not changes in instrument performance over time). In all solvents, compounds with electron-withdrawing groups and extended conjugation ionized best due to resonance and inductive effects. Ionization was greatest in methanol or water for all compounds that elicited a response, thus revealing that enhanced sensitivity and lower limits of detection are achieved with polar protic solvents. Response in acetone was equal to or slightly lower than response in acetonitrile in flow injection experiments; however, the water/ acetonitrile and water/acetone mobile phases produced the better chromatographic separation. Water/methanol produced slightly less satisfactory separation but the greatest overall response. This increase in response was attributed to the protic nature of methanol and the elution of compounds in a higher organic mobile phase composition (retention times were ∼30% later in methanol). This work is intended to facilitate rational liquid chromatography/mass spectrometry method development for small molecule applications, including metabolomics.

N

hydrophobicity) was a better predictor of negative-ion response than pKa. Henriksen et al.,5 however, concluded that a single parameter (log P) did not unequivocally predict negative-ion response. Recent attempts to predict positive-ion ESI response have, therefore, considered multiple physiochemical parameters simultaneously (e.g., polarity, pKa, surface area, molar volume, gas-phase properties).6−9 Ultimately, we aim to use a multivariate approach to quantify factors that affect negative-ion ESI response. In pursuit of this goal, we built upon the foundation of Hendricksen et al. and measured the response of a more diverse set of acidic, small molecules. Compounds selected by Hendricksen et al. spanned pKa and log P ranges of 1.6 (most acidic) to 10.6 (least acidic) and −1.81 (most polar/hydrophilic) to 4.78 (most nonpolar/ hydrophobic), respectively.5 Our test set of 49 compounds spans pKa values from 0.42 to 17 and log P values from −5.1 to 9.62. Compounds selected include n-carboxylic acids, benzoic acids, phenoxyalkanoic acids, dicarboxylic acids, phosphonic acids, phenols, polyphenols, sulfonates, and aniline derivatives. Compounds were selected within a compound class (e.g., ncarboxylic acids, phenols, and benzoic acids) with systematic functional group alterations in order to better understand the role of functional group type and placement on ionization

egative-ion electrospray (ESI) selectively ionizes acidic compounds for mass spectrometric detection by deprotonation. This ionization technique, which is readily coupled to chromatographic separations, enjoys a wide application space that includes small organic molecules and large biomolecules. The work presented here deals primarily with the former, with focus on optimization of mass spectrometric and chromatographic conditions for a diversity of small molecules that form singly charged ions with m/z values less than 1000. Optimization of liquid chromatography/ mass spectrometry (LC/MS) conditions for small molecules is particularly important in untargeted metabolomic studies that aim to maximize the number of metabolites detected in complex biological matrices. Increased ionization efficiency leads to increased response, lower limits of detection and quantitation, and the production of higher quality MS/MS spectra for improved metabolite identification.1 To date, most negative-ion studies have focused on structurally similar compounds, often within a single compound class, in order to better understand the physiochemical factors that affect response.2−4 This led Henriksen et al.5 to measure the response of 31 acidic analytes in neat acetonitrile, neat methanol, and 50:50 mixtures of each organic solvent with water in order to better understand the relative influence of acidity and polarity on negative-ion response. Results indicated that compounds ionized better in methanol than acetonitrile and that log P (the logarithm of the octanol−water coefficient, a measure of © 2012 American Chemical Society

Received: August 19, 2012 Accepted: October 15, 2012 Published: October 15, 2012 9942

dx.doi.org/10.1021/ac302397b | Anal. Chem. 2012, 84, 9942−9950

Analytical Chemistry

Article

flow injection experiments, analytes were injected individually from low to high concentration with two blanks in between analytes to eliminate contamination by carryover. Neat (no modifier) water, methanol, acetonitrile, and acetone were used in the flow injection studies. The grades and suppliers of the solvents were as follows: water (HPLCgrade; Burdick & Jackson, VWR); methanol and acetonitrile (LC/MS-grade, Thermo Fisher Scientific); acetone (certified ACS-grade acetone, Thermo Fisher Scientific). The Agilent negative-ion performance standard (G1946-85005, Agilent Technologies, Santa Clara, CA), available in sealed 1 mL ampules, was used as the quality control standard. Flow Injection Experiments. Flow injection experiments were conducted on an Agilent 6460 triple quadrupole (QqQ) mass spectrometer with a jet stream source (Agilent Technologies, Santa Clara, CA). A thermostated autosampler kept at 7 °C injected 5 μL of each analyte individually into the mobile phase (0.3 mL/min) delivered by one of two Agilent 1200 rapid resolution LC pumps. LCs were configured with equal volume tubing to ensure identical elution profiles. A 10port valve directed the desired solvent through an Agilent 1200 diode array detector (DAD) SL that measured the absorbance of UV-active analytes (see Table S-1 in Supporting Information, bandwidth = 12 nm) prior to quantitation by MS. Analytes were ionized by negative-ion ESI under the following conditions: capillary voltage, −1500 V, nozzle voltage, −500 V; fragmentor voltage, 135 V; drying gas temperature, 350 °C; drying gas flow, 10 L/min; sheath gas flow, 11 L/min; nebulizer, 40 psi. MS data were collected in full-scan mode (500 ms/scan) over the range of 90−600 m/z. The time between consecutive injections was 1 min. Data acquisition was controlled by Mass Hunter software, B.03.01. Quality Control. The quality control (QC) worklist was run at the beginning of each data collection day and periodically between data sets to monitor changes in instrument response over time. Once the total ion chromatogram (TIC) produced a flat baseline, Agilent’s ES negative-ion performance standard (Acid Red 4, 10 μg/mL; CAS 5858-39-9) was injected consecutively 10 times in neat methanol, using the same MS parameters described above. The UV absorbance was measured at 250 nm (bandwidth = 4 nm). Integration of the UV wavelength chromatogram and MS extracted ion chromatogram (EIC, at m/z = 356.6) was automated by use of Agilent’s Mass Hunter Qualitative Analysis B.04.00 software. Data Analysis. All data were collected and analyzed in triplicate. Agilent’s Mass Hunter Quantitative Analysis B.04.00 software was used to integrate peaks. Analyte response was then normalized against that day’s average QC response as follows: (average compound response/average QC response) × 106. Chromatography Experiments. We initially attempted to conduct the chromatographic experiments on the same instrument as the flow injection experiments (under the same flow and source conditions), but we reached the 600 bar limit on the Agilent 1200 pumps, even when a 5 cm column was used. For this reason, the chromatography experiments were conducted on the Agilent Infinity 1290 UHPLC, which tolerated pressures up to 1200 bar and allowed use of a 15 cm column. Maximum pressures obtained for pump A when water/methanol, water/acetonitrile, and water/acetone was used were approximately 800, 550, and 700 bar, respectively. The Infinity 1290 with autosampler was coupled to a 6224 Agilent accurate-mass TOF mass spectrometer (Santa Clara, CA). A mixture of 30 compounds, each at 33 μM, was

efficiency. We also expanded the solvent systems explored in flow injection experiments to include neat water and acetone, in addition to neat methanol and acetonitrile. This expansion allowed the comparison of response in two polar protic solvents (water and methanol) against two polar aprotic solvents (acetone and acetonitrile). Acetone was chosen because it may be used as a widely available, less-expensive substitute for acetonitrile.10−12 At the beginning of the study, the effect of negative-ion ESI voltage (−1000 to −3500 V) as a function of analyte concentration (1−1000 μM) was measured for 30 analytes to determine the optimal voltage for the selected flow rate (300 μL/min). In addition, a rigorous quality control (QC) program was implemented to allow normalization of response over the course of the 1-year study. The QC program ensured that changes in analyte response were the result of the analyte and solvent under investigation, not changes in instrument performance. MS response was also cross-validated by measuring the UV absorption of UV-active compounds. Finally, the quality of chromatographic separation of test analytes was compared when methanol, acetone, or acetonitrile was used as the organic mobile phase with 0.1% formic acid. Analyte response in chromatography experiments was then compared to flow-injection experiments to determine whether gains observed in a neat solvent were realized in a gradient elution. This comprehensive look into the role of source conditions, analyte type/functionality, and solvent composition on the negative-ion ESI response of diverse acidic compounds, coupled to an evaluation of chromatographic separation when different organic mobile phases are used, will undoubtedly facilitate rational method development for metabolomics and many other LC/MS applications.

■

EXPERIMENTAL SECTION Materials and Equipment. The 49 analytes studied were 1-naphthoic acid, 1-pyrenecarboxylic acid, 1,3,5-trihydroxybenzene dihydrate, 17-β-estradiol, 19-norethindrone, 2(methylamino)phenol, 2,4-dichlorophenoxyacetic acid (2,4D), 3-(dimethylamino)phenol, 3-(trifluoromethyl)phenol, 3,4dihydroxybenzoic acid, 4-acetamidophenol, 4-aminophenol, 4chlorophenol, 4-hydroxybenzoic acid, 4-hydroxybenzonitrile, 4nitrophenol, aniline, benzoic acid, bromoxynil, carbazole, catechin, cholesterol, cholic acid, cyanuric acid, D-(+)-raffinose pentahydrate, decanoic acid, dichloromethylenediphosphonic acid disodium salt, dodecanoic acid, estriol, estrone, etidronic acid monohydrate, fumaric acid, gallic acid, hexadecanoic acid (palmitic acid), hexanoic acid, hydroquinone, ibuprofen sodium salt, irgasan, kaempferol, mestranol, octadecanoic acid (stearic acid), octanoic acid, p-cresol, phenol, sodium 1-hexanesulfonate monohydrate, tetradecanoic acid, trichlocarban (3,4,4′-trichlorocarbanilide), undecafluorohexanoic acid, and β-estradiol 17acetate. Most analytes were purchased from Sigma−Aldrich (St. Louis, MO) with the exception of aniline and kaempferol, which were obtained from Fisher Chemical (Thermo Fisher Scientific, Pittsburgh, PA), and Indofine Chemical Company (Hillsborough, NJ), respectively. Table S-1 (Supporting Information) lists all analytes studied, their CAS registry numbers, manufacturer, purity, m/z of ions observed, UV absorption wavelength (if applicable), calculated pKa,13 and calculated log P.13 Structures for all analytes, excluding ncarboxylic acids, are provided in Scheme S-1 (Supporting Information). Stock solutions for all compounds were prepared in LC/MS methanol at 1000 μM and diluted serially to 1.00 μM with half-log dilutions to yield seven calibration levels. In 9943

dx.doi.org/10.1021/ac302397b | Anal. Chem. 2012, 84, 9942−9950

Analytical Chemistry

Article

of a new ion transfer capillary, as evidenced by the constant DAD response between 7/1/11 and 7/22/11. The average MS intraday and interday RSD, post capillary conditioning period, were 3.0% and 24.7%, respectively. A modified exponential with the formula a exp[b/(x + c)] best described the MS QC data. If the MS QC data fell outside the daily residual standard deviation of 1.1 × 107, efforts were made to remedy the problem before data were collected. The use of absolute response data versus data normalized against the QC did not affect the trends observed. Normalized data, however, were used to account for the larger variability in MS response and to enable long-term normalization of data. The work reported here is the first phase of a multiphase study. Negative-Ion ESI Response of Test Compounds. Flow injection experiments were used to measure the negative-ion ESI response of 49 diverse analytes in neat polar protic (methanol and water) and neat polar aprotic (acetonitrile and acetone) solvents. Measurements were made in triplicate over three orders of concentration (1−1000 μM) to yield a data set with 4116 data points, excluding QC measurements. At the beginning of the study, response was measured as a function of negative voltage applied to the ESI source (−1000 to −3500 V at 500 V intervals) and analyte concentration. Two predominant trends emerged: (1) a left-skewed, Gaussian-like response with a maximum response at −1500 V and (2) an exponential decay with a maximum response at −1000 V. Collectively, −1500 V was deemed the best voltage for the majority of the test analytes. This optimal voltage was lower than initially predicted, as most negative-ion ESI experiments are conducted at ∼−2.5 to −3.5 kV. These data, therefore, demonstrate the importance of source optimization prior to a large-scale study in which the highest sensitivity is desired. Recently, Heaton et al.11 utilized signal-to-noise surface plots of cone versus capillary voltage to determine the optimal positiveion ESI conditions for small, basic molecules, while Raji and Schug14 conducted a chemometric experiment to determine the relative contributions of spray voltage, ion transfer capillary temperature, ion transfer capillary voltage, and tube lens voltage. To facilitate discussion of the results, test compounds were divided into three categories: low, moderate, and high responders (Table 1). Classifications were based on the summed normalized response in methanol across all seven calibration levels. Compounds with summed responses below 20k, between 20k and 300k, and above 300k were classified as low, moderate, and high responders, respectively. Compounds in Table 1 are ranked by response, which increases from top to bottom in each category. Compound classes are also designated for acids (A), phenols (P), steroids (S), and aniline derivatives (AD). Raffinose, a trisaccharide, does not fit within these categories. Detection of compounds by UV prior to negative electrospray ionization was used to cross-validate poor ionization. Such cross-validation eliminated concern that poor response was due to compound instability or a bad autosampler injection. UV and MS data for ibuprofen, a representative poor responder, are shown in Figure S-1 (Supporting Information). RSDs for individual MS measurements were high and response did not increase with an increase in concentration. However, ibuprofen exhibited a linear response at 219 nm in acetonitrile (R2 = 0.9815), methanol (R2 = 0.9991), and water (R2 = 0.9996). UV response was not measured in acetone, as the UV cutoff was 331 nm and UV-

separated with an Agilent Zorbax Eclipse Plus C18 rapid resolution HD column that was 150 mm × 2.1 mm i.d. with 1.8 μm particle size (Agilent, Santa Clara, CA). Compounds were separated with a binary mobile phase system of solvent A (water with 0.1% formic acid) and solvent B (acetonitrile, methanol, or acetone with 0.1% formic acid) at 50 °C and a flow rate of 450 μL/min with the following gradient: B was held at 95% for 1 min and then increased to 95% by 25 min and held for 2 min. A 5-min post-run (5% B) equilibrated the column for the next run. Five microliters of each sample was injected onto the column. Analytes were ionized for mass spectrometric detection by negative-ion ESI with the following conditions: spray voltage, −3.5 kV; drying gas temperature, 325 °C; drying gas flow rate, 10 L/min; nebulizer, 30 psi; capillary voltage, 3500 V. MS data were collected in full-scan mode over the mass range 100−1700 m/z at an acquisition rate of 1.43 spectra/s (or 699 ms for one spectrum). Ion optic voltages were as follows: fragmentor, 175 V; skimmer, 65 V; octopole 1 RF Vpp, 750 V. Mass Hunter software version B.04 was used for data acquisition and analysis.

■

RESULTS AND DISCUSSION Quality Control in Fundamental ESI Studies. A QC program was implemented early in the study to monitor changes in DAD and MS response. Such a program is critical in long-term, fundamental ESI studies so that instrument variability is not mistaken for changes in analyte response. Agilent’s negative-ion performance standard, Acid Red 4, was chosen because it is readily available, prepared gravimetrically by ISO9001 quality procedures, has a 2-year shelf life, and is both MS- and UV-active. DAD response was monitored to ensure that the QC was chemically stable throughout its use. Figure 1 (top) demonstrates that the UV QC response varied little from vial to vial and lot to lot (there are five 1 mL vials in a pack). Each new vial is indicated by a different color. The average DAD intraday RSD was 3.4%. The interday RSD over 1 year was 14.0% (n = 890). The drop in MS response (Figure 1, bottom) at the beginning of the study was due to conditioning

Figure 1. (Top) DAD and (bottom) MS response of QC standard, Acid Red 4, over 1 year. Each new QC ampule is represented with a different color. A minimum of 10 QC injections were performed daily. 9944

dx.doi.org/10.1021/ac302397b | Anal. Chem. 2012, 84, 9942−9950

Analytical Chemistry

Article

Table 1. Summed Normalized Negative-Ion Response of Test Analytes in Methanola low respondersb

moderate respondersc

high respondersd

aniline (AD) phenol (P) dichloromethylenediphosphonic acid (A) cholesterol (S) hydroquinone (P) ibuprofen (A) mestranol (P) 4-acetamidophenol (P) irgasan (P) fumaric acid (A) 4-aminophenol (P) norethindrone (S) etidronic acid (A) hexanoic acid (A) 2,4-dichlorophenoxyacetic acid (2,4-D) (A) undecafluorohexanoic acid (A) 3-(dimethylamino)phenol (P) benzoic acid (A) 2-(methylamino)phenol (P) cyanuric acid (P)

p-cresol (P) 4-hydroxybenzoic acid (A) 1,3,5-trihydroxybenzene (P) octanoic acid (A) naphthoic acid (A) 3,4-hydroxybenzoic acid (A) 4-chlorophenol (P) trichlocarban (AD) pyrenecarboxylic acid (A) decanoic acid (A) gallic acid (A) dodecanoic acid (A) estriol (S) β-estradiol 17-acetate (S)

β-estradiol (S) tetradecanoic acid (A) estrone (S) 4-hydroxybenzonitrile (P) carbazole (AD) 4-nitrophenol (P) bromoxynil (P) hexadecanoic acid (A) sodium 1-hexanesulfonate (A) octadecanoic acid (A) cholic acid (A) catechin (P) raffinose (trisaccharide) 3-(trifluoromethyl)phenol (P) kaempferol (P)

a

A = acids; P = phenols; S = steroids; AD = aniline derivatives. Responses increase from the top to the bottom of the table. bSummed normalized response 300k.

active compounds absorbed between 211 and 366 nm. Crossvalidation by UV detection confirmed that aniline, hydroquinone, mestranol, 4-acetomidophenol, norethindrone, 2,4-D, 3-(dimethylamino)phenol, benzoic acid, and 2-(methylamino)phenol ionized poorly. The remaining “low responders” were UV-inactive. Effect of Functional Groups on Negative-Ion ESI Response. Close inspection of the compounds grouped by response in Table 1 revealed the effect of functional group substituents on negative-ion ESI response. Trends were particularly striking when comparisons were made within the n-carboxylic acids and the phenols in neat methanol (Figure 2). For n-carboxylic acids (Figure 2A), response, on average, doubled with each addition of two methylene groups (e.g., decanoic to dodecanoic acid). Since n-carboxylic acids have approximately the same pKa values (Table S-1, Supporting Information), the increase in response with chain length was attributed to the simultaneous increase in nonpolar character and molar volume.8 The equilibrium-partitioning model describes how nonpolar analytes are more surface-active, meaning that they reside in a region of excess charge on the droplet’s surface, and, therefore, are more efficiently ionized.15−17 The second phase of this work, which is currently underway, will use a multivariate approach to determine the factors that best predict negative-ion response (e.g., pKa, polar surface area, nonpolar surface area, molar volume, solvation energy, gas phase proton affinity, etc.).7−9 Therefore, the physiochemical parameters that affect ionization will not be discussed further. The effect of electron-withdrawing versus electron-donating functional groups is clearly seen when negative-ion response is compared among the phenols (Figure 2B) whose structures are provided in Scheme 1. Phenols classified as “high responders” in Table 1 have electron-withdrawing groups (EWGs) para or meta to the hydroxyl that is deprotonated. These functional groups withdraw electrons from the benzene ring through induction and/or resonance, which places a partial positive

Figure 2. Normalized response (n = 3) in neat methanol as a function of concentration (1−1000 μM) for (A) n-carboxylic acids and (B) select phenols.

charge in the ring, thus making it easier to remove a proton to form [M − H]−. EWGs may be classified as weak, moderate, or strong. Weak EWGs include halogens, which withdraw electron density from the ring through induction. Moderate EWGs include carbonyl functional groups [for example, (CO)H, (CO)OH, and (CO)NH2], sulfonates (SO2H), and nitriles (CN), which withdraw electrons through resonance. Resonance and inductive (NO2) or inductive-only effects (CF3) are at work for strong EWGs. The phenol with the highest response (Figure 2B) has a CF3 group meta to the hydroxyl. Bromoxynil, 9945

dx.doi.org/10.1021/ac302397b | Anal. Chem. 2012, 84, 9942−9950

Analytical Chemistry

Article

trihydroxybenzene and 2-(methylamino)phenol] are shown in Figure 2B. All other phenols exhibited responses similar to or lower than 2-(methylamino)phenol. The effects of EWGs and EDGs on negative-ion response have been previously documented in halogenated and amino-substituted benzoic acids3 and aromatic carboxylic acid amides.4 Finally, it is worth noting that negative-ion response increased among benzoic acids as the number of fused rings increased, such that 1pyrenecarboxylic acid (four rings) > 1-naphthoic acid (two rings) > benzoic acid (one ring). As the number of rings increases, the π electron cloud becomes more polarizable and the resulting anion is stabilized through induction. Negative-Ion ESI Response in Neat Polar Protic versus Neat Polar Aprotic Solvents. Examination of functional groups in the previous section revealed that compounds are most effectively deprotonated when electron density is pulled away from the depronatation site. Solvent selection also plays an important role in negative-ion ESI, as previously demonstrated.5,18,19 Henriksen et al.5 showed that response was greater in methanol than acetonitrile for most, but not all, analytes tested. Enhanced response was attributed to methanol’s protic nature and its ability to solvate negative ions and lone pairs. We tested this hypothesis further on an even more diverse set of analytes and measured response in neat acetone and water, as well as neat methanol and acetonitrile. In Figure 3, the 10 μM calibration level was used to compare response in methanol to other solvents by taking the logarithm of response in methanol divided by response in “solvent X” (where X is acetone, acetonitrile, or water). Bars with positive values indicate greater response in methanol at 10 μM, while bars with negative values indicate greater response in acetone, acetonitrile, and/or water. The height of the bar indicates the magnitude of improved response in methanol (for positive bars) or the other solvent (for negative bars). For example, kaempferol has log (methanol/solvent X) values of 0.88, 0.39, and 0.38 when solvent X is acetone, acetonitrile, and methanol, respectively. Therefore, the response of kaempferol in methanol was 7.6 times greater than in acetone and 2.4 times greater than in acetonitrile and water. Thirty-two of the 48 compounds shown in Figure 3 exhibited a greater response in methanol than the other solvents. Thirty-five out of 48 compounds exhibited a greater response in methanol when the same comparison was made at 316 μM (data not shown). Specifically, the increased responses of hydroquinone, 2(methylamino)phenol, and 3,4-hydroxybenzoic acid, which are low to moderate responders, were more discernible at higher concentrations. As a class, the steroids exhibited the greatest improvement in methanol. On average, the improvement was 9.2×, 13.1×, and 6.3× compared to acetone, acetonitrile, and water, respectively. Incidentally, methanol is used as the organic mobile phase in U.S. EPA Method 539 that quantifies hormones in drinking water.20 Finally, the gains observed in methanol for phenols with EDGs were the same as phenols with EWGs. Neat water is normally considered a poor ESI solvent, as its high viscosity and low volatility reduce analyte charging.21 Advances in pneumatically assisted ESI sources, such as Agilent’s JetStream source, facilitate desolvation by use of gases at 350 °C and, therefore, may increase ionization efficiency in water, thus explaining why several compounds, particularly acids, responded better in water than methanol. The trends observed for the n-carboxylic acids are particularly interesting. Hexanoic to dodecanoic acids ionized better in

Scheme 1. Structures of Compounds Discussed in Figure 2Ba

a

Structures are grouped by electron-withdrawing groups (EWGs, red), electron-donating groups (EDGs, blue), and enhanced resonance stabilization (polyphenols). The structures for all 49 compounds are shown in Supporting Information (Table S-1).

4-nitrophenol, and 4-hydroxybenzonitrile have strong or moderate EWGs para to the hydroxyl. Bromoxynil also has two bromine atoms ortho to the hydroxyl. Polyphenols kaempferol and catechin further demonstrate how enhanced resonance stabilization of the resulting anion increases negativeion response. Kaempferol and catechin have the same threering backbone common to flavonoids. Kaempferol, however, has a carbonyl on the second ring that lends extended conjugation and thus greater resonance stabilization. Phenols classified as “moderate responders” in Table 1 contain electrondonating groups (EDGs) that add electron density to the ring. Strong EDGs like OH, NH2, and NRH, where R is an alkyl group, contain a lone pair adjacent to the ring, resulting in resonance structures with a negative charge in the ring. The anion of the conjugate base is thus destabilized and ionization efficiency is reduced. Only two phenols with EDGs [1,3,59946

dx.doi.org/10.1021/ac302397b | Anal. Chem. 2012, 84, 9942−9950

Analytical Chemistry

Article

Figure 3. Normalized log response of each analyte (n = 3) relative to neat methanol (MeOH). Bars with values greater than 0 indicate enhanced response in methanol compared to solvent X (neat acetone, acetonitrile (ACN), or water). Negative values indicate the analyte responded best in acetone (blue), acetonitrile (red), and/or water (green) or that the analyte was not detected at 10 μM.

water, while the tetradecanoic to octadecanoic acids ionized better in methanol. Undecafluorohexanoic acid, with a low overall ionization efficiency, also responded better in water at concentrations ≥316 μM. Cholic acid responded better in water at all concentrations. 4-Hydroxybenzonitrile was the only phenol to ionize better in water than methanol. At concentrations ≥10 μM, responses in water and methanol were approximately equal. Unfortunately, it is difficult to discern why some acids ionized better in water than others. One might conclude from the n-carboxylic acids that more polar acids favored water while less polar acids favored methanol. A broader perspective, however, shows that more polar acids, such as fumaric acid (a dicarboxylic acid) and etidronic acid (a diphosphonic acid), ionized better in methanol. Clearly, acidity (e.g., pKa) and polarity alone do not predict negative-ion response. Regardless, we can conclude that all compounds that exhibited a negative-ion response, which excludes aniline, phenol, and dichloromethylenediphosphonic acid, ionized better in polar protic than in polar aprotic solvents. Generally, responses followed the order methanol > water > acetronitrile ≥ acetone for the majority of test compounds. Solvents with higher polarity and higher dielectric constants stabilize the charge separation in the ESI droplet and allow formation of a compact charge layer at the droplet’s surface.22 As a result, analytes are more effectively ionized. Solvents with a lower dielectric constant, like acetone, produce a more diffuse double layer with excess charge in the droplet interior, thus reducing ionization efficiency. Enhanced response in polar protic solvents led to increased sensitivity and lower limits of detection as shown in Table 2. To demonstrate this improvement, the limit of detection (LOD) is reported for three low-, moderate-, and high-responding

Table 2. Limits of Detection for Select Compounds in Acetone, Acetonitrile, Methanol, and Watera limit of detection (μmol) analyte

acetone

acetonitrile

Low Responders nd 125.397 64.988 43.058

hydroquinone (P) undecafluorohexanoic acid (A) 2-(methylamino)phenol (P) 21.340 7.563 Moderate Responders p-cresol (P) 46.379 18.452 decanoic acid (A) nd 1.095 β-estradiol 17-acetate (S) 47.194 1.398 High Responders estrone (S) 21.699 1.179 cholic acid (A) 4.602 0.847 catechin (P) 0.203 0.116

methanol

water

13.027 69.542

87.626 6.150

4.710

7.578

2.749 0.508 0.174

9.926 0.211 0.331

0.558 0.095 0.078

0.918 0.558 0.985

a

A = acids; P = phenols; S = steroids. Data were collected in full-scan (not MRM) mode.

compounds. LOD was calculated over the linear dynamic range as follows: LOD = 3.3σ /S

(1)

where σ equals the standard deviation of the regression line yintercept and S equals the slope.23 Serial dilutions were continued to 1 nM for moderate and high responders to verify calculated LODs. The lowest LOD value for each compound follows the trends previously discussed. Effect of Polar Aprotic and Polar Protic Solvents on Chromatographic Separation and Negative-Ion ESI 9947

dx.doi.org/10.1021/ac302397b | Anal. Chem. 2012, 84, 9942−9950

Analytical Chemistry

Article

mobile phase composition was ∼10% ± 6% higher. The percent aqueous versus organic mobile phase present during elution has a significant impact on electrospray response, which will be discussed in more detail momentarily. Acetone afforded the fewest coeluting pairs, with only one unresolved doublet of kaempferol (13) and bromoxynil (14) at 11.96 min. In acetonitrile there were two coeluting pairs: 4-acetomidophenol (2) coeluted with 3,4-dihydroxybenzoic acid (3) at 2.50 min, and undecahexanoic acid (11) coeluted with kaempferol (12) at 9.98 min. Methanol yielded three unresolved peaks: 1naphthoic acid (10), undecafluorohexanoic acid (11), and kaempferol (12) coeluted at 14.57 min; pyrenecarboxlic acid (16) and ibuprofen (17) coeluted at 19.69 min; and triclocarban (20) and cyanuric acid (21) coeluted at 21.08 min. Acetone also afforded the most symmetrical peaks (1.32 ± 0.45), followed by acetonitrile (1.57 ± 0.41) and methanol (1.68 ± 0.35). However, the average peak width in acetone was slightly wider at 5.42 ± 0.28 s than methanol (5.31 ± 1.40 s) and acetonitrile (3.76 ± 0.48 s). The large standard deviation for methanol was largely due to octadecanoic acid, the last eluting peak, which did not completely elute off the column even when 95% B was held for 2 min. When octadecanoic acid was excluded, the average peak width for methanol was 4.89 ± 0.63 s. Finally, the overall separation efficiency was evaluated by comparing the effective number of theoretical plates (Neff), which takes into consideration peak width and retention time, as shown in eq 2:

Response. While mass spectrometrists and those interested in ionization fundamentals appreciate the simplicity of flow injection experiments, LC/MS practitioners want to know what conditions will give the best chromatographic separation and highest sensitivity for their analytes of interest. With this practicality in mind, experiments were conducted that measured the response of a subset of test compounds (∼25 at 33 μM) in a chromatographic separation (Figure 4). All

Figure 4. Chromatographic separation when acetone (top), acetonitrile (ACN) (middle), and methanol (MeOH, bottom) was used with 0.1% formic acid as the organic mobile phase. Dashed red line shows the change in organic mobile phase (% B) during gradient elution. Analytes: (1) fumaric acid, (2) 4-acetamidophenol, (3) 3,4dihydroxybenzoic acid, (4) 4-hydroxybenzoic acid, (5) catechin, (6) 1hexanesulfonate, (7) 4-hydroxybenzonitrile, (8) 4-nitrophenol, (9) estriol, (10) 1-naphthoic acid, (11) undecafluorohexanoic acid, (12) kaempferol, (13) bromoxynil, (14) 3-(trifluoromethy)phenol, (15) estrone, (16) 1-pyrenecarboxlic acid, (17) ibuprofen, (18) decanoic acid, (19) mestranol, (20) triclocarban, (21) cyanuric acid, (22) tetradecanoic acid, (23) octadecanoic acid.

Neff = 5.54(t R ′/w1/2)2

(2)

where tR′ is the adjusted retention time and w1/2 is the peak width at half-height. Neff was calculated for the last eluting peak, octadecanoic acid. The tR′ values of octadecanoic acid were 21.91, 24.66, and 25.86 min for acetone, acetonitrile, and methanol, respectively. Therefore, acetonitrile, with the narrowest peaks, yielded the highest Neff of 36 200. Acetone, with wider peaks but shorter retention times, yielded a Neff of 18 200. The high efficiency of acetone was unexpected but is corroborated in the literature.10−12 Fritz et al.10 found the higher elutropic strength of acetone reduced retention times compared to methanol and yielded the same sequence coverage in peptide analysis compared to acetonitrile. Finally, the lowest Neff (12 900) was observed for methanol because octadecanoic acid eluted at the end of the run and was 8.70 s wide. In conclusion, both acetonitrile and acetone provided acceptable separation. Additional optimization of the method would likely yield improved separation with methanol as the organic mobile phase. Chromatograms in Figure 4 are scaled to the peak of highest relative abundance among the three chromatograms. However, the differences in response as a function of organic mobile phase composition are best observed in Figure 5. Compounds are ranked according to their retention time in methanol, with the logarithm of absolute response (peak area) on the x-axis. In the chromatography experiments, response was not normalized against a QC. All data sets, however, were collected within a week’s time to minimize significant variance in instrument response. As expected, the organic mobile phase composition made little to no difference on the response of compounds that eluted during the first 10 min (fumaric acid to hexanesulfonate), as water was the primary mobile phase component (water decreased from 95% to ∼70% in the first 10 min). When the aqueous mobile phase decreased to 70% or less, the organic mobile phase composition exhibited a greater effect on

conditions were kept constant except the organic mobile phase, which used acetone (Figure 4, top), acetonitrile (middle), or methanol (bottom) with 0.1% formic acid. Retention time RSDs averaged 0.037%, 0.109%, and 0.093%, respectively, indicating that the column was adequately equilibrated between chromatography experiments. The quality of the chromatographic separation in Figure 4 was evaluated on the basis of elution order, retention time, peak resolution, peak symmetry, peak width, and effective plate number. The generic elution gradient is overlaid on the methanol chromatogram (Figure 4, bottom) to give an idea of the gradient conditions at the time of elution. Elution order was similar for all mobile phases, with the most similarity between the polar aprotic solvents, acetone and acetonitrile. Analytes separated with water/acetone and water/acetonitrile gradients also exhibited similar retention times. The average retention time difference between water/acetone and water/acetonitrile gradients was 0.24 ± 1.13 min. As a result, analytes eluted under aqueous-to-organic ratios that, on average, differed by only 1% ± 4%. In contrast, analytes eluted much later in the water/methanol gradient due to methanol’s lower eluotropic strength. On average, retention time increased 30% ± 13% compared to water/acetonitrile and 27% ± 17% compared to water/acetone. As a result, analytes eluted when the organic 9948

dx.doi.org/10.1021/ac302397b | Anal. Chem. 2012, 84, 9942−9950

Analytical Chemistry

Article

greater than acetone and acetonitrile in the flow injection experiments and 4.3 ± 3.7 and 28.5 ± 19.5 greater in the chromatography experiments. Previous studies have reported that gains often realized in flow injection experiments are not observed in chromatography experiments due to the presence of water and modifiers in the mobile phase.5,25 Since the ncarboxylic acids used in the chromatography experiments were more nonpolar and eluted when the aqueous percentage was 60% or less, gains with methanol were still observed. Steroids, which also eluted at a lower aqueous percentage, exhibited similar gains. Response was 9.21 ± 1.4 and 5.3 ± 1.7 times higher in methanol compared to acetonitrile in the flow injection and chromatography experiments, respectively. However, the gains for phenols, most of which eluted earlier in the gradient, were not as pronounced (1.4 ± 0.8, acetone; 1.4 ± 0.3, acetonitrile) as gains observed in flow injection experiments (3.5 ± 2.5, acetone; 1.7 ± 1.2, acetonitrile, average excludes 4-hydroxybenzonitrile, whose response improved 120× in methanol). Therefore, analytes that elute at higher organic percentages will experience the greatest increase in response and sensitivity when methanol is used as the organic mobile phase. Of course, sensitivity is not the only factor under consideration during LC/MS method development. Solvent cost, environmental impact, run time, and chromatographic resolution are also factors to consider. Methanol costs less and is greener than acetonitrile, but larger volumes may be consumed due its lower eluotropic strength and, subsequently, longer run times. Figure 4 also demonstrated that methanol yielded less efficient chromatographic separation than acetonitrile as evidenced by lower Neff values. This likely could be improved through optimization of the gradient or use of a different modifier or column. All in all, methanol is a viable solvent choice that should be considered when the highest sensitivity is desired for negative-ion ESI LC/MS applications.

Figure 5. Logarithm of the response (n = 3) for compounds in Figure 4 when acetone (blue), acetonitrile (red), or methanol (black) was used as the organic mobile phase with 0.1% formic acid.

response. Twelve of 15 compounds that eluted after hexanesulfonate responded better when methanol was used as the organic mobile phase. This increase is partially due to later elution in a higher percentage of organic solvent and partially due to the increased ionization efficiency in a protic solvent, as observed in the flow injection experiments. For this reason, increased sensitivity in pesticide quantitation has been reported when methanol was used as the organic mobile phase.24 Bromoxynil responded equally well in all solvent systems, while the fluorinated compounds [3-(trifluoromethyl)phenol and undecafluorohexanoic acid] responded slightly better when acetone (versus methanol) was used as the organic mobile phase. Interestingly, in the chromatography experiments, some analytes responded better with water/acetone than with water/ acetonitrile. This is also apparent when the y-axis values in Figure 4 are compared. The compound that exhibited the greatest response in water/acetonitrile (middle) was about 40% less than the highest responders in water/acetone (top) and water/methanol (bottom). These results are contrary to the flow injection experiments, in which response followed the general trend: methanol > acetonitrile > acetone.

■

ASSOCIATED CONTENT

S Supporting Information *

One table listing relevant data and properties for analytes studied; one scheme showing structures for all analytes, organized by class; and one figure illustrating normalized (A) MS and (B) UV response as a function of ibuprofen concentration. This material is available free of charge via the Internet at http://pubs.acs.org.

■

■

CONCLUSIONS Both flow injection (Figure 3) and chromatography (Figure 5) experiments provide compelling evidence that methanol affords increased sensitivity for quantitation of small, acidic molecules compared to acetonitrile (the solvent most often used in reverse-phase separations) and acetone. In the flow injection experiments (Figure 3), the gains with neat methanol were evident by taking the logarithm of response in methanol divided by response in “solvent X” (where X is acetone, acetonitrile, or water). Positive values indicated that response was greatest in methanol. Calculation of the chromatographic response in the same manner allows gains in flow injection experiments to be compared to gains in chromatography experiments. For instance, four n-carboxylic acids were used in both experiments: undecafluorohexanoic, decanoic, tetradecanoic and octadecanoic acids. Collectively, the response of these compounds in methanol was 2.8 ± 1.4 and 4.5 ± 2.6 times

AUTHOR INFORMATION

Corresponding Author

*E-mail hugheyca@jmu; phone 540-568-6688; fax 540-5687938. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This material is based upon work supported by the National Science Foundation under Grants CHE-0922935 and CHE1046630. Funding from these MRI grants was used to purchase QqQ and TOF MS instruments. Additional support was received from the Research Corporation Development Award 7957 and the James Madison University Department of Chemistry and Biochemistry. A special thanks is extended to Dr. Bruce Wilcox, Director of Instrumentation and Lab Operations at Applied Proteomics, Inc., for assistance with 9949

dx.doi.org/10.1021/ac302397b | Anal. Chem. 2012, 84, 9942−9950

Analytical Chemistry

Article

experimental design and QC implementation; to Dr. Dan Ruderman, Assistant Professor of Research Medicine at the Center for Applied Molecular Medicine within the Keck School of Medicine at the University of Southern California, for help with evaluation of the QC data; and to Dr. Kevin Caran, Associate Professor of Chemistry at James Madison University, for fruitful discussions on effects of EWGs and EDGs.

■

REFERENCES

(1) Yanes, O.; Tautenhahn, R.; Patti, G. J.; Siuzdak, G. Anal. Chem. 2011, 83, 2152−2161. (2) Schug, K.; McNair, H. M. J. Sep. Sci. 2002, 25, 760−766. (3) Schug, K.; McNair, H. M. J. Chromatogr. A 2003, 985, 531−539. (4) Chiu, F. C. K.; Lo, C. M. Y. J. Am. Soc. Mass Spectrom. 2000, 11, 1061−1064. (5) Henriksen, T.; Juhler, R. K.; Svensmark, B.; Cech, N. B. J. Am. Soc. Mass Spectrom. 2005, 16, 446−455. (6) Chalcraft, K. R.; Lee, R.; Mills, C.; Britz-McKibbin, P. Anal. Chem. 2009, 81, 2506−2515. (7) Fusaro, V. A.; Mani, D. R.; Mesirov, J. P.; Carr, S. A. Nat. Biotechnol. 2009, 27, 190−198. (8) Oss, M.; Kruve, A.; Herodes, K.; Leito, I. Anal. Chem. 2010, 82, 2865−2872. (9) Raji, M. A.; Fryčaḱ , P.; Temiyasathit, C.; Kim, S. B.; Mavromaras, G.; Ahn, J.-M.; Schug, K. A. Rapid Commun. Mass Spectrom. 2009, 23, 2221−2232. (10) Fritz, R.; Ruth, W.; Kragl, U. Rapid Commun. Mass Spectrom. 2009, 23, 2139−2145. (11) Heaton, J.; Jones, M. D.; Legido-Quigley, C.; Plumb, R. S.; Smith, N. W. Rapid Commun. Mass Spectrom. 2011, 25, 3666−3674. (12) Keppel, T. R.; Jacques, M. E.; Weis, D. D. Rapid Commun. Mass Spectrom. 2010, 24, 6−10. (13) ACD Labs 12.01, 2009. (14) Raji, M. A.; Schug, K. A. Int. J. Mass Spectrom. 2009, 279, 100− 106. (15) Cech, N. B.; Enke, C. G. Anal. Chem. 2000, 72, 2717−2723. (16) Enke, C. G. Anal. Chem. 1997, 69, 4885−4893. (17) Du, L.; White, R. L. J. Mass Spectrom. 2009, 44, 222−229. (18) Kostiainen, R.; Kauppila, T. J. J. Chromatogr. A 2009, 1216, 685−699. (19) Straub, R. F.; Voyksner, R. D. J. Am. Soc. Mass Spectrom. 1993, 4, 578−587. (20) Smith, G. A.; Zaffiro, A. D.; Zimmerman, M. L.; Munch, D. J. EPA Method 539, 2010. (21) Kostiainen, R.; Bruins, A. P. Rapid Commun. Mass Spectrom. 1996, 10, 1393−1399. (22) Cole, R. B.; Harrata, A. K. J. Am. Soc. Mass Spectrom. 1993, 4, 546−556. (23) ICH. Validation of analytical procedures: text and methodology, Q2 (R1). ICH Harmonised Tripartite Guideline 2005, pp 1−13. (24) Steen, R. J. C. A.; Hogenboom, A. C.; Leonards, P. E. G.; Peerboom, R. A. L.; Cofino, W. P.; Brinkman, U. A. T. J. Chromatogr. A 1999, 857, 157−166. (25) Jemal, M.; Hawthorne, D. J. Rapid Commun. Mass Spectrom. 1999, 13, 61−66.

9950

dx.doi.org/10.1021/ac302397b | Anal. Chem. 2012, 84, 9942−9950