Investigations of Analyte-Specific Response Saturation and Dynamic

Sep 30, 2014 - Department of Chemistry and Biochemistry, University of North Carolina at Greensboro, P.O. Box 26170, Greensboro, North. Carolina 27402...
0 downloads 0 Views 960KB Size
Download PDF
Article pubs.acs.org/ac

Investigations of Analyte-Specific Response Saturation and Dynamic Range Limitations in Atmospheric Pressure Ionization Mass Spectrometry Clint M. Alfaro, Agbo-Oma Uwakweh, Daniel A. Todd, Brandie M. Ehrmann, and Nadja B. Cech* Department of Chemistry and Biochemistry, University of North Carolina at Greensboro, P.O. Box 26170, Greensboro, North Carolina 27402, United States S Supporting Information *

ABSTRACT: With this study, we investigated why some small molecules demonstrate narrow dynamic ranges in electrospray ionization-mass spectrometry (ESI-MS) and sought to establish conditions under which the dynamic range could be extended. Working curves were compared for eight flavonoids and two alkaloids using ESI, atmospheric pressure chemical ionization (APCI), and heated electrospray ionization (HESI) sources. Relative to reserpine, the flavonoids exhibited narrower linear dynamic ranges with ESI-MS, primarily due to saturation in response at relatively low concentrations. Saturation was overcome by switching from ESI to APCI, and our experiments utilizing a combination HESI/APCI source suggest that this is due in part to the ability of APCI to protonate neutral quercetin molecules in the gas phase. Thermodynamic equilibrium calculations indicate that quercetin should be fully protonated in solution, and thus, it appears that some factor inherent in the ESI process favors the formation of neutral quercetin at high concentration. The flavonoid saturation concentration was increased with HESI as compared to ESI, suggesting that inefficient transfer of ions to the gas phase can also contribute to saturation in ESI-MS response. In support of this conclusion, increasing auxiliary gas pressure or switching to a more volatile spray solvent also increased the ESI dynamic range. Among the sources investigated herein, the HESI source achieved the best analytical performance (widest linear dynamic range, lowest LOD), but the APCI source was less subject to saturation in response at high concentration.

W

mass analysis,8 or detection.9 This study is primarily concerned with saturation due to inefficient ionization. Several theories have been proposed to explain why production of gas phase ions could be limited at high analyte concentration. Tang and Kebarle explained saturation in ESIMS response as due to a limited rate of evaporation of ions from solution to the gas phase,10,11 whereas Enke described an upper limit as being defined by the limited quantity of excess charge.5 Zook and Bruins proposed that limited space on the surface of electrospray droplets prevents analytes from becoming charged in highly concentrated solutions.3,12 Finally, it has been suggested that droplets with very high analyte concentration form solid particles rather than gas phase ions.13 These theories could explain the ultimate limit that is observed in response with ESI-MS. However, saturation in response appears to be selective: some analytes reach saturation at much lower concentration than others. Studies that examine the sources of variation in dynamic range for small molecules with ESI-MS are lacking in the published literature. With our

ith the advent of electrospray ionization (ESI), mass spectrometry (MS) has become a preferred method for quantitative analysis of trace chemicals in complex nonvolatile samples.1 A significant hurdle in such analyses is ensuring that the response of every analyte falls within the analytically useful portion of the calibration curve (the dynamic range). This dynamic range can vary among analytes, and under some conditions, the range spans less than 1 order of magnitude. To address this, the same set of samples is often analyzed at several dilutions and/or reanalyzed multiple times, a practice that can be problematic given that the magnitude of the matrix effects can change when samples are diluted.2 This problem can be circumvented if the dynamic range is sufficiently wide. Thus, our objective with this study was to provide insights that would be helpful for maximizing dynamic range with atmospheric pressure ionization mass spectrometry of small molecules. In mass spectrometric analyses, an S-shaped calibration curve is obtained when instrument response is plotted versus analyte concentration. Response at low concentration is limited either because analyte signal is too low or chemical noise is too high. Saturation is observed at high analyte concentration and may occur due to limitations in ionization,3−6 ion transmission,7 © 2014 American Chemical Society

Received: July 6, 2014 Accepted: September 30, 2014 Published: September 30, 2014 10639

dx.doi.org/10.1021/ac502984a | Anal. Chem. 2014, 86, 10639−10645

Analytical Chemistry

Article

Table 1. Calibration Curve Data and Physicochemical Properties for the Analytes Studieda analyte quercetin

caffeine

reserpine

apigenin

hesperidin

kaempferol

vitexin

tangeretin

daidzein

naringenin

source ESI HESI APCI ESI HESI APCI ESI HESI APCI ESI HESI APCI ESI HESI APCI ESI HESI APCI ESI HESI APCI ESI HESI APCI ESI HESI APCI ESI HESI APCI

LODb (μM) 0.07 0.03 4.40 0.13 0.13 0.03 0.03 0.03 0.03 0.40 0.40 4.00 0.40 0.40 4.00 0.40 0.40 4.00 4.00 4.00 7.00 0.98 0.98 0.98 0.98 0.98 2.00 0.98 0.98 3.90

LL LDRc (μM) 1.1 0.6 35.0 1.0 2.0 0.5 2.0 0.3 0.03 4.0 3.9 16.0 4.0 8.0 16.0 8.0 16.0 32.0 4.0 4.0 7.0 3.9 3.9 16.0 1.0 7.8 31.0 1.0 3.9 31.0

UL LDRc (μM) 35 70 560 250 500 500 250 500 500 64 250 500 32 130 500 130 1000 1000 15 60 230 250 500 1000 8 250 1000 16 130 1000

OOMc 1.5 2.1 1.2 2.4 2.4 3.0 2.1 3.3 4.2 1.2 1.8 1.5 0.9 1.2 1.5 1.2 1.8 1.5 0.6 1.2 1.5 1.8 2.1 1.8 0.9 1.5 1.5 1.2 1.5 1.5

pKbd 16.6

16

logPe 1.68

10.434

0.063

7.435

5.01

15.716

2.46

−0.55

16.516

2.17

0.52

2.66

2.78

18.716

3.19

a Quercetin, reserpine, and caffeine were analyzed with isocratic conditions of 59.9% water with 0.1% formic acid/ 39.9% acetonitrile with 0.1% formic acid. The flavonoids apigenin, hesperidin, kaempferol, and vitexin were analyzed using a linear acetonitrile/water gradient as described in Methods. bThe LOD corresponds to the analyte concentration where analyte response is 3 times the mass spectral noise. cLL,UL (lower and upper limits of the linear dynamic range, respectively), and OOM (the orders of magnitude of analyte concentration spanned by linear dynamic range). The procedures used for these determinations are described in Methods. dReported pKb values were obtained from the literature where available. e logP, the water-octonal partition coefficient, was calculated using the logP predictor tool in ACD Laboratories Chemsketch (Toronto, Ontario, Canada). Lower values indicate a more polar (greater affinity for water) compound, and larger values indicate more nonpolar (greater affinity for octonal) compound.

equipped with an Acquity UPLC BEH C18 column (1.7 μm particles, 2.1 × 50 mm) was employed with a 3 μL injection volume and a 0.3 mL/min flow rate. Solvent A was 99.9% Optima grade water with 0.1% formic acid, and solvent B was 99.9% Optima acetonitrile with 0.1% formic acid. For the studies comparing dynamic range with caffeine, quercetin, and reserpine, an isocratic mobile phase of 60% A/40% B was employed for the first 3 min, followed by a 1 min column flush with 100% B and a 1.5 min flush of starting mobile phase (5.5 min total runtime). For the studies comparing dynamic range of flavonoids, a linear mobile phase gradient was used with 90% A/10% B at t = 0 min, 50% A/50% B at t = 2 min, 48% A/52% B at t = 3 min, 100% B from t = 4 to t = 4.5 min, and 90% A/ 10% B from t = 4.75 to t = 5.5 min. Mass Spectrometry. Analyses were conducted with a TSQ Quantum Access Triple Quadrupole mass spectrometer (ThermoFisher Scientific, San Jose, CA) fitted with ESI, HESI-II, or APCI probes. Analyses were conducted in the positive ion mode with full scan followed by data-dependent collision induced dissociation (CID). Ionization source parameters in ESI mode were 3800 kV spray voltage, 300 °C capillary temperature, 35 V capillary offset, and sheath gas, ion

investigation, we sought to fill this gap in the knowledge base. In so doing, we explored the hypotheses that inefficient analyte protonation in solution and/or inefficient transfer of protonated analyte into the gas phase dictate analyte-specific saturation in ESI-MS response at high concentration.

■

METHODS Sample Preparation. HPLC Optima grade water, acetonitrile, methanol, formic acid, and Bioreagent grade dimethyl sulfoxide were purchased from Fisher Scientific (Hampton, NH). Quercetin, daidzein, reserpine (purity ≥98%), kaempferol, apigenin, tangeretin (purity ≥97%), vitexin, naringenin (purity ≥95%), and hesperidin (purity ≥80) were purchased from Sigma-Aldrich (St. Louis, MO). Caffeine (purity ≥99%) was purchased from Alfa Aesar (Ward Hill, MA). Flavonoids, caffeine, and reserpine were prepared at 1 mg/mL in solvents of 1:4 DMSO/methanol, water, and 1:4 DMSO/acetonitrile, respectively. These solutions were then diluted with methanol (flavonoids and reserpine) or nanopure water (caffeine) to prepare solutions for analysis. Liquid Chromatography. An ultrahigh performance liquid chromatograph (Waters Acquity UPLC, Milford, MA) 10640

dx.doi.org/10.1021/ac502984a | Anal. Chem. 2014, 86, 10639−10645

Analytical Chemistry

Article

Figure 1. Comparison of linear dynamic ranges with ESI, HESI, and APCI for seven flavonoids.

to be limited due to incomplete desolvation. The experimental [MH+] value for quercetin at each standard concentration (Ci) was obtained by substituting the relevant peak area into the equation for the best fit line of the quercetin calibration curve. To obtain predicted values for [MH+] as a function of Ci, we relied on an equation derived in a previous study (eq 1).15 A derivation of this equation is provided as Supporting Information.

sweep gas, and auxiliary gas pressures of 40, 0, and 40 (arbitrary units), respectively. A vaporizer temperature of 300 °C was employed for HESI and APCI, and the discharge current for APCI was set at 4A. To conduct the dual source experiments, an additional 8 kV cable was spliced in with the 8 kV cable attached to the Ion Max housing,14 enabling the same voltage to be applied to the ESI needle and the corona needle simultaneously. Solvent Effect Studies. A series of caffeine stock solutions were analyzed with flow injection via UPLC. The analysis time was 3 min, flow rate was 0.3 mL/min, and the same solvents described under Liquid Chromatography were used. The mobile phase was varied by changing percentages of solvents A and B, for a total of four different conditions: 100% A, 70% A, 40% A, and 0% A. Detection was accomplished with the triple quadrupole mass spectrometer operated in the positive ion mode with a range of m/z 150−800 and a 0.32 s scan time. Samples were analyzed with ESI using the parameters described under Mass Spectrometry. Interpretation of Calibration Curve Data. For simplicity, we chose with this study to measure and compare only the linear portion of the dynamic range (rather than the full dynamic range). To obtain the upper limit (UL) and lower (LL) of this linear dynamic range, we chose the points at which back calculated analyte concentration (from linear regression of the calibration curve) deviated by more than 20% from the actual concentration of the sample. The OOM (orders of magnitude spanned by the linear dynamic range) was then calculated using eq 2.

⎛ UL ⎞ ⎟ OOM = log⎜ ⎝ LL ⎠

[MH+] =

■

KbC i[H3O+] Kb[H3O+] + K w

(1)

RESULTS AND DISCUSSION Caffeine, reserpine, and a series of flavonoids (Chart S-1) were chosen as test analytes in this study. The alkaloid reserpine was selected because it is an ideal analyte for ESI-MS analysis.6 Caffeine (another alkaloid) and quercetin (a flavonoid) were chosen on the basis of preliminary investigations in our laboratory, which showed that they demonstrated very narrow dynamic ranges under some experimental conditions. Finally, a series of additional flavonoids were included to more broadly evaluate the influence of source conditions on dynamic range. It is apparent from Table 1 that the OOM (orders of magnitude spanned by the linear dynamic range) varies greatly among analytes and between sources. As expected, reserpine demonstrated a very wide linear dynamic range in ESI and HESI modes (OOM = 2.1 and 3.3, respectively). Caffeine performed less optimally, but still provided linear dynamic ranges spanning several orders of magnitude (OOM = 2.4 with both ESI and HESI sources). In contrast, the linear dynamic ranges for quercetin with ESI and HESI sources spanned only 1.5 and 2.1 orders of magnitude, respectively. As shown graphically in Figure S-1, this limitation in the ESI linear dynamic range for quercetin was due to saturation in response at relatively low concentration (35 μM, Table 1). In contrast, the reserpine ESI calibration curve saturated at much higher concentrations (250 μM, Table 1). A selection of additional flavonoids was analyzed using the three different sources (Table 1, Figure 1). Like quercetin,

(2)

OOM, LL, and UL for each flavonoid (n = 8) were compared for different combinations of ESI, HESI, and APCI using a one tailed student’s t test (two sample unequal variance). Fit of Experimental Data. Quercetin was employed to examine the relationship between predicted and experimentally measured [MH+]. HESI data were used for this comparison with the assumption that they would be less likely than ESI data 10641

dx.doi.org/10.1021/ac502984a | Anal. Chem. 2014, 86, 10639−10645

Analytical Chemistry

Article

these flavonoids demonstrated protracted ESI-MS linear dynamic ranges (mean flavonoid OOM = 1.24 ± 0.050, Table 2) as compared to that achievable for reserpine (OOM = Table 2. Averaged Performance Characteristics for the Eight Flavonoids Studied (Quercetin, Apigenin, Hesperidin, Kaempferol, Vitexin, Daidzein, Naringenin, And Tangeretin) ionization source

parameter

mean ± SD

ESI HESI APCI ESI HESI APCI ESI HESI APCI

OOM OOM OOM UL (μM) UL(μM) UL(μM) LL(μM) LL(μM) LL(μM)

1.24 1.76 1.50 68 310 720 3.3 6.0 23

± ± ± ± ± ± ± ± ±

0.50 0.60 0.16 83 310 310 2.5 4.8 11

2.1). Additionally, saturation was observed for the flavonoids at a mean concentration of 68 ± 83 μM (Table 2), which is substantially lower than the 250 μM upper limit achieved with reserpine (Table 1). Across all of the flavonoids, the linear dynamic range improved when switching from the ESI source (OOM = 1.24 ± 0.50) to the HESI source (OOM = 1.76 ± 0.60) (Figure 1, Table 2). This observed increase in linear dynamic range was statistically significant (p = 0.04). Furthermore, switching from the ESI source to the HESI or APCI source gave significantly (p = 0.04 or 0.005, respectively) higher upper limits of the calibration curve (Table 2). A demonstration of this effect for the flavonoid quercetin is shown in Figure 2. On the basis of these observations, it can be concluded that whatever factors contribute to saturation in flavonoid response are at least partially overcome using HESI or APCI sources. In the following sections we describe a series of experiments designed to determine factors that contribute to this advantage. Limits in Ionization Efficiency as a Cause for Saturation. Saturation in the calibration curve with ESI-MS can be related to factors other than limited ionization efficiency (given in the background information); however, the data suggest that limited production of detectable ions is the primary cause of saturation in this study. First, the upper limit in the linear dynamic range was extended to higher concentration for several analytes when switching from ESI to APCI (Table 1, Figure 1); no such extension should be possible if factors involved with ion transmission, mass analysis, or detection were limiting. Additionally, different analytes demonstrated distinct differences in linear dynamic range (Figure 1, Table 1). Finally, the same limited linear dynamic range was observed for quercetin operated with ESI-MS using both an ion trap mass spectrometer and a triple quadrupole mass spectrometer (Figure S-1), despite differences in ion transfer optics and type of mass analyzer. Investigation of Protonation in Solution. Although quercetin is not particularly basic in solution,16 it is detectable with positive ion ESI-MS. We propose the quercetin protonation mechanism shown in Scheme S-1. One explanation for saturation in ESI-MS response would be a shift at high concentration from protonated quercetin to neutral quercetin. To explore this possibility, we compared experimental response of quercetin with predictions based on thermodynamic equilibria (Figure 2).

Figure 2. Investigation of the effects of acid−base equilibria on the HESI-MS dynamic range for quercetin. (A) Comparison between [MH+] and bulk concentration (Ci) determined by experimental measurements and predicted with eq 1. Kb (9.87 × 10−12) was determined by substituting the experimental [MH+] (5.8 × 10−5 M) at a Ci value of 7.0 × 10−5 M into eq S-3. (B) Derived [MH+] (from eq 1) versus Ci for an expanded range of concentrations. Measured [MH+] values are not provided for the higher concentrations due to experimental constraints.

The perfect agreement between theoretical and calculated [MH+] at the fourth highest data point (7.0 × 10−5 M) (Figure 2A) is inherent to the methods used for the fit; this data point was used to determine the experimental value for Kb (eq S-3). For the lower concentrations, the deviations between predicted and measured values of [MH+] are on the order of 10% or less. However, saturation begins to be observed experimentally at ∼1 × 10−4 M, whereas the calibration curve predicted from eq 1 must be extended above 1 × 10−2 M (Figure 2B) before saturation is observed. The data in Figure 2 suggest that saturation occurs in 0.1% formic acid solutions even though quercetin should be protonated in solution. This finding indicates that at high concentration, the charged analyte is either unable to escape the electrospray droplet (is not fully desolvated), is paired by a counterion (effectively neutralizing it), or that some factor inherent in the droplet evaporation process shifts the equilibrium from protonated (charged) analyte to deprotonated (neutral) analyte. It is notable that the Kb value calculated from the experimental data (9.8 × 10−12) is considerably higher than the reported Kb value for quercetin in 50% aqueous ethanol with sulfuric acid (2.51 × 10−17).16 Although our experiments were conducted in 40% aqueous acetonitrile with 0.1% formic acid, it is unlikely that a difference in Kb of such magnitude could be caused by solvent effects alone, particularly given that 10642

dx.doi.org/10.1021/ac502984a | Anal. Chem. 2014, 86, 10639−10645

Analytical Chemistry

Article

Kw (the solvent autoprotolysis equilibrium constant) values for water and 40% aqueous acetonitrile differ by less than an order of magnitude (1.0 × 10−14 versus 1.82 × 10−15, respectively).17 This finding is consistent with a number of previous literature reports that analytes can be detected in positive ion electrospray under conditions that should not favor their protonation in the bulk solution.18 Contributions of Surface Activity. Among charged analytes, it is widely observed that surface activity (the analyte’s affinity for the air-solvent interface at the droplet surface) is associated with responsiveness to analysis with ESI-MS.13 This effect has been exploited to improve analyte sensitivity to analysis with ESI-MS. For example, N-linked glycans can be derivatized with hydrophobic hydrazide reagents to enhance their responsiveness,19 and trace amounts of singly charged anions can be detected in the positive mode by complexing them with hydrophobic dicationic reagents.20 Several theories have been developed to explain the importance of surface activity in dictating ESI-MS response. Kebarle and Tang11,21,22 suggested that selectivity with ESI-MS is governed by the rate of evaporation of ions from ESI droplet surfaces and that surface-active ions would have higher ion currents by virtue of being concentrated at these surfaces.21 An alternate model that has been employed to describe selectivity in ESI-MS response is Enke’s equilibrium partitioning model,5 in which the electrospray droplet is considered biphasic, with a surface phase where the excess charge resides, and a neutral interior phase that contains anion−cation pairs.5 According to the equilibrium-partitioning model, surface-active analytes are expected to be more easily detectable with ESI-MS due to their enhanced ability to compete for the excess charge on droplet surfaces, while less surface-active analytes partition to the droplet interior, where they are paired with counterions and lost as neutrals. In single-analyte solutions, the Enke5 and Kebarle21 models predict that analyte response increases with surface activity, but the saturation concentration will be extended to slightly higher concentrations for poorly surface-active analytes. Our data (Table 1, Figure S-2) only partially agree with this prediction. For our studies, logP, the log partitioning coefficient between octanol to water, was used as an approximate estimation of surface activity, consistent with previous studies.23,15 As expected on the basis of both the Enke and Kebarle models, quercetin, which is presumed to be significantly less surfaceactive than reserpine (logP values of 1.67 and 5.01, respectively), demonstrates lower overall response than reserpine (Figure S-2). However, saturation in signal for quercetin occurs at a much lower concentration than it does for the more surface-active reserpine, rather than the other way around. The findings here suggest some other factors besides those included in the Enke or Kebarle models can cause signal saturation for analytes with poor surface activity. One important consideration is that both Enke and Kebarle worked under the assumption that all of the analyte is charged and, therefore, did not consider the loss of charge by the analyte as a cause for saturation in response. As discussed earlier, our calculations (Figure 4) suggest that quercetin should be completely protonated in bulk solution under the conditions employed for this study. However, it is possible that factors inherent in the droplet formation and evaporation process favor the neutral analyte at high concentration. Electrospray droplets are known to fission unevenly, with the majority of the charge

splitting off into the small offspring droplets, while the majority of the mass remains in the initial “parent” droplet.24 Surfaceactive analytes are likely to follow the charge in this fissioning process, ending up in highly charged offspring droplets that may form ions through ion-evaporation.25 Less surface-active analytes, however, are more likely to end up in less highly charged droplets. It is possible that this uneven fissioning could leave an analyte such as quercetin in droplets that are highly depleted in protons and favor the formation of neutral, rather than protonated, quercetin. This effect would presumably be enhanced at high concentrations, where the concentration of the analyte relative to the concentration of excess charge would be much higher. It is also possible that these charge-depleted droplets may form ions via the charged residue model rather than via ion evaporation.26 This might explain deviations from the behavior predicted by the Enke or Kebarle models, which are both based on the assumption that ionization occurs solely via ion evaporation. Investigations with a Dual HESI/APCI Source. In the preceding sections, we compared the concentration−response behavior of several flavonoids obtained using ESI and APCI. We consistently observed an increase in the upper limit of the linear dynamic range for quercetin and other flavonoids when switching from ESI or HESI to APCI (Table 2), and we hypothesized that this extension could occur because of the ability of APCI to protonate neutral, gas phase analyte. However, the commercial ESI, HESI, and APCI sources employed in this study differ both in the source conditions (temperature, sheath gas pressure) and method of protonation. Thus, differences in performance between the different sources (Table 1) could be due to differences in the efficiency of transfer of analyte into the gas phase or in method of ionization. To enable more direct comparisons between HESI and APCI, the commercial HESI source was modified as described in Methods to create a combination HESI/APCI source or “dual source”. With this source, it is possible for the high voltage to be applied to the corona needle and electrospray needle simultaneously. In essence, the dual source passes an electrosprayed aerosol through a corona discharge at atmospheric pressure (see schematic in Figure S-3). As shown in Figure 3, the dual source configuration enabled an extension of the saturation point in the quercetin calibration

Figure 3. Quercetin calibration curves generated with HESI and the dual HESI/APCI source. All conditions are identical except that the APCI voltage is applied for the dual mode. The fit line represents the linear dynamic range for each calibration curve. 10643

dx.doi.org/10.1021/ac502984a | Anal. Chem. 2014, 86, 10639−10645

Analytical Chemistry

Article

curve to higher concentration as compared to that achieved with HESI. These data suggest that the HESI probe can efficiently transfer neutral quercetin into the gas phase from solution and that the corona discharge can subsequently protonate gas phase neutral quercetin. Thermodynamic equilibrium calculations indicate that quercetin should be protonated in the bulk solution across the entire concentration range evaluated (Figure 2). Thus, the dual source experiments lend further support to the concept that saturation can occur due to a shift from protonated to neutral analyte at high concentration during ESI. Comparison of Limit of Detection (LOD) with APCI, ESI, and the Dual Source. For this study, we observed that the LODs achievable for all of the flavonoids were consistently lower with ESI or HESI than with APCI (Table 1). For the alkaloids (caffeine and reserpine), however, LOD was similar or better with APCI than to that obtained with ESI or HESI. It has been reported that sensitivity with APCI-MS is largely dictated by clustering (solvation) effects and gas phase basicity of the analyte.27,28 The gas phase basicity of a flavonoid similar to quercetin (3-hydroxy-4H-1-benzopyran-4-one) is 204.0 kcal/ mol,29 whereas the gas phase basicity of caffeine is reported as 214.8 kcal/mol.30 The lower gas phase basicity of the flavonoid as compared to caffeine, along with unexplored clustering effects, could explain the observed differences in flavonoid and alkaloid LOD. Interestingly, the analyses conducted using the dual ESI/ APCI source (Figure 3) demonstrated the same higher LOD achieved with APCI (Table 1, ∼4.4 μM). Thus, whatever factor facilitates improved detection at low concentration with HESI is negated in the presence of a corona discharge. These data further support the conclusion that LOD with APCI is compromised as a result of gas phase interactions of the analyte with the charged reagent gas. Optimizing Gas Phase Ion Formation. It has previously been shown that factors that favor the formation of smaller droplets in ESI-MS enhance sensitivity.6,31 Indeed, we observed (Table 2) that the HESI source consistently exhibited wider linear dynamic ranges than the ESI source, presumably because the addition of heat forms smaller droplets that facilitate the efficient transfer of ions into the gas phase.31 These results are not at all surprising in light of the proliferation of commercial ESI sources that reportedly increase ionization efficiency through the use of heated desolvation gas.32 To the best of our knowledge, however, our study is the first to report on the influence of heated gas on linear dynamic range. Two other conditions that can influence droplet size and efficiency of ion transfer to the gas phase are coaxial gas flow rate and solvent composition. To investigate the influence of these parameters on dynamic range, we chose caffeine as a test case. It was our expectation that the low surface activity of this molecule (logP 0.063) should make it susceptible to changes in source parameters. Consistent with this expectation, at the lower coaxial gas flow rate, linear dynamic range for caffeine was quite poor (3.9 × 10−6 M to 6.2 × 10−5 M, OOM = 1.2), although higher coaxial gas flow significantly expanded the caffeine linear dynamic range (1.0 × 10−6 M to 2.5 × 10−4 M, OOM = 2.4, Figure 4A). In addition, increasing the organic solvent (acetonitrile) content in the flow injection analysis resulted in a wider linear dynamic range for caffeine (1.2 OOM with 100% water versus 1.79 OOM with 100% acetonitrile). These findings are in agreement with previous findings that response with ESI-MS is higher in organic solvents than in

Figure 4. Influence of coaxial gas flow rate and solvent content on the caffeine dynamic range with ESI-MS. (A) Influence of coaxial gas flow rate on dynamic range. To normalize, the signal at each concentration was divided by the signal of the highest analyzed concentration. Sheath and auxiliary gas flows were set to 10 and 4 for “low” gas, and 40 and 40 (arbitrary units) for “high” gas, respectively. Low and high gas flow were normalized to 5.0 × 107 and 1.5 × 108 signal counts, respectively. (B) Influence of organic solvent content on dynamic range.

water.33 Overall, the data in Figure 4 indicate that dynamic range can be extended for an analyte with poor surface activity by adjusting the source or solvent conditions. Which Source is Best? Collectively, the data presented herein demonstrate that the various atmospheric pressure ionization sources have trade-offs that warrant careful consideration by the analyst. For some ideal analytes (in this case reserpine), similar performance can be achieved regardless of ionization source. However, for less ideal analytes, such as the small polar flavonoids used as a test case for this study, choice of ionization source must be made more carefully. The use of APCI or HESI resulted in wider linear dynamic ranges than those achievable with ESI for all of the flavonoids analyzed. For these same analytes, HESI had the distinct advantages of lower limits of detection than those achievable with APCI. By employing the dual source, we had hoped to combine the strengths of HESI (better LOQ) with those of APCI (extension of the saturation point to higher concentration) and achieve the widest possible dynamic range. However, due to the aforementioned high LL (Figure 3), the dynamic range for the dual source mirrored that achievable with APCI. Additionally, the dual source demonstrated overall lower signal intensity than the HESI source (Figure 3). If these issues could be resolved, it is possible that extension of the dynamic range with a dual source could be achieved. This would be a worthy topic for future investigation. 10644

dx.doi.org/10.1021/ac502984a | Anal. Chem. 2014, 86, 10639−10645

Analytical Chemistry

■

Article

(6) Tang, K.; Page, J. S.; Smith, R. D. J. Am. Soc. Mass Spectrom. 2004, 15, 1416−1423. (7) Page, J. S.; Marginean, I.; Baker, E. S.; Kelly, R. T.; Tang, K.; Smith, R. D. J. Am. Soc. Mass Spectrom. 2009, 20, 2265−2272. Busman, M.; Sunner, J.; Vogel, C. R. J. Am. Soc. Mass Spectrom. 1991, 2, 1−10. Page, J. S.; Kelly, R. T.; Tang, K.; Smith, R. D. J. Am. Soc. Mass Spectrom. 2007, 18, 1582−1590. (8) Hall, A. B.; Coy, S. L.; Kafle, A.; Glick, J.; Nazarov, E.; Vouros, P. J. Am. Soc. Mass Spectrom. 2013, 24, 1428−1436. Kharchenko, A.; Vladimirov, G.; Heeren, R. M. A.; Nikolaev, E. N. J. Am. Soc. Mass Spectrom. 2012, 23, 977−987. (9) Yuan, L.; Zhang, D.; Jemal, M.; Aubry, A. F. Rapid Commun. Mass Spectrom. 2012, 26, 1465−1474. (10) Kebarle, P.; Tang, L. Anal. Chem. 1993, 65, 972A−986A. (11) Kebarle, P.; Verkerk, U. H. Mass Spectrom. Rev. 2009, 28, 898− 917. (12) Bruins, A. P. J. Chromatogr. A 1998, 794, 345−357. (13) Constantopoulos, T. L.; Jackson, G. S.; Enke, C. G. J. Am. Soc. Mass Spectrom. 1999, 10, 625−634. (14) Ion Max and Ion Max-S API Source, Thermo Fisher Scientific, 2008. https://www.unifr.ch/inph/vclab/assets/files/Software/ IonMax_API_IonSouce_Hardware.pdf. (15) Ehrmann, B. M.; Henriksen, T.; Cech, N. B. J. Am. Soc. Mass Spectrom. 2008, 19, 719−728. (16) Rybachenko, A. I. Soviet progress in chemistry. 1981, 47, 944− 948. (17) Kilic, E.; Aslan, N. Microchim. Acta 2005, 151, 89−92. (18) Zhou, S.; Cook, K. D. J. Am. Soc. Mass Spectrom. 2000, 11, 961− 966. (19) Walker, S. H.; Lilley, L. M.; Enamorado, M. F.; Comins, D. L.; Muddiman, D. C. J. Am. Soc. Mass Spectrom. 2011, 22, 1309−1317. (20) Breitbach, Z. S.; Wanigasekara, E.; Dodbiba, E.; Schug, K. A.; Armstrong, D. W. Anal. Chem. 2010, 82, 9066−9073. (21) Tang, L.; Kebarle, P. Anal. Chem. 1993, 65, 3654−3668. (22) Sunner, J.; Nicol, G.; Kebarle, P. Anal. Chem. 1988, 60, 1300− 1307. (23) Henriksen, T.; Juhler, R. K.; Svensmark, B.; Cech, N. B. J. Am. Soc. Mass Spectrom. 2005, 16, 446−455. (24) Cech, N. B.; Enke, C. G. Anal. Chem. 2001, 73, 4632−4639. (25) Taflin, D. C.; Ward, T. L.; Davis, E. J. Langmuir 1989, 5, 376− 384. Fenn, J. B. J. Am. Soc. Mass Spectrom. 1993, 4, 524−35. (26) Fernandez de la Mora, J. Anal. Chim. Acta 2000, 406, 93−104. Dole, M. J. Chem. Phys. 1968, 49, 2240−2249. (27) Herrera, L. C.; Grossert, J. S.; Ramaley, L. J. Am. Soc. Mass Spectrom. 2008, 19, 1926−1941. (28) Sunner, J.; Nicol, G.; Kebarle, P. Anal. Chem. 1988, 60, 1300− 1307. (29) Aleman, C. J. Mol. Struct.: THEOCHEM 2000, 528, 65−73. (30) H, Y.; Liu, L.; Liu, S. J. Chem. Phys. 2012, 527, 73−78. (31) Ikonornou, M. G.; Kebarle, P. J. Am. Soc. Mass Spectrom. 1994, 5, 791−799. (32) ESCIMulti-Mode Ionization Source, Waters Corporation: U.S.A., 2008. http://www.waters.com/waters/en_US/ESCi-MultiMode-Ionization-Source-/nav.htm?cid=1000392; HESI-IIProbe, ThermoFisher Scientific: U.S.A, September 2008. http://www. thermoscientific.com/content/dam/tfs/ATG/CMD/cmd-support/ tsq-quantum-access-max/manuals/HESI-II-Probe-User.pdf; DuoSprayIon Source, AB Sciex LP: Canada, June 2012. http://www. absciex.com/Documents/Downloads/Literature/duospray-ion-sourceoperator-guide-eng.pdf. (33) Peri-Okonny, U. L. Effects of Eluent pH and Different Types of Acidic Modifiers on the Retention and Electrospray Ionization Efficiency of Basic Analytes in LC-ESI-MS. Seton Hall University Dissertations and Theses (ETDs); 2001, Paper 1254. (34) Martin, A.; Bustamante, P. Physical Pharmacy: Physical Chemical Principles in the Pharmaceutical Sciences, 4th ed.; Lippincott Williams & Wilkins: Philadelphia, PA, 1993. (35) Szasz, G.; Budvari-Barany, Z. Pharmaceutical Chemistry of Antihypertensive Agents; CRC Press: Boca Raton, FL, 1991; Vol. 1.

CONCLUSIONS This study demonstrates that limited ability to convert the analyte to gas phase ions can compromise the dynamic range observed with ESI-MS. A practical implication of this finding is that it is possible to improve dynamic ranges for small polar molecules by operating under conditions that make the formation of gas phase ions more favorable. These include shifting to a higher organic content of the solvent (when not limited by chromatographic conditions), increasing the pressure of the coaxial gas used to enhance desolvation, or using a vaporizer to heat the coaxial gas (as with the HESI source). At least under the conditions employed in this study, our results indicated that inability to protonate the analyte in solution did not contribute to saturation in ESI-MS response at high concentration. Nonetheless, it is still possible that protonation in solution could become limiting for analytes that are less basic than quercetin, or if the solution pH were not sufficiently acidic. This would likely be particularly true for analytes with poor surface-activity that would not be concentrated in the low-pH regions at the surface of electrospray droplets. For such analytes, analysis with APCI could be a good alternative to the application of ESI. Our findings demonstrate that saturation of flavonoid calibration curves with ESI can be overcome using APCI. The practical implications of this observation are not obvious, because although the linear dynamic range was extended to high concentrations by using APCI, the limit of detection was higher for all flavonoids with APCI than with ESI or HESI. However, the findings are interesting in that they suggest that neutral quercetin molecules are present in the gas phase with atmospheric pressure ionization. The mechanisms by which these neutral gas phase molecules are formed, and strategies by which protonation of quercetin in the gas phase could be enhanced for ESI-MS or HESI-MS analyses, would be an interesting area of future study.

■

ASSOCIATED CONTENT

S Supporting Information *

Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org/.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Phone: 336-324-5011. Fax: 336-334-5402. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS We thank Dr. Christie G. Enke for his valuable insights about the data presented herein, and Michael Shelton and Charlie Shepherd for technical assistance.

■

REFERENCES

(1) Cech, N. B.; Enke, C. G. Mass Spectrom. Rev. 2001, 20, 362−387. (2) Kruve, A.; Herodes, K.; Leito, I. Rapid Commun. Mass Spectrom. 2011, 25, 1159−1168. Kruve, A.; Auling, R.; Herodes, K.; Leito, I. Rapid Commun. Mass Spectrom. 2011, 25, 3252−8. (3) Zook, D. R.; Bruins, A. P. Int. J. Mass Spectrom. Ion Processes 1997, 162, 129−147. (4) Cole, R. B. J. Mass Spectrom. 2000, 35, 763−772. (5) Enke, C. G. Anal. Chem. 1997, 69, 4885−4893. 10645

dx.doi.org/10.1021/ac502984a | Anal. Chem. 2014, 86, 10639−10645