Continuous flow-extractive desorption electrospray ionization: Analysis from “non-electrospray ionization-friendly” solvents and related mechanism

Analytica Chimica Acta 769 (2013) 84–90

Contents lists available at SciVerse ScienceDirect

Analytica Chimica Acta journal homepage: www.elsevier.com/locate/aca

Continuous flow-extractive desorption electrospray ionization: Analysis from “non-electrospray ionization-friendly” solvents and related mechanism Li Li a , Samuel H. Yang a , Karel Lemr b , Vladimir Havlicek c , Kevin A. Schug a,∗ a

Department of Chemistry and Biochemistry, The University of Texas at Arlington, Arlington, TX, USA Regional Centre of Advanced Technologies and Materials, Department of Analytical Chemistry, Faculty of Science, Palacky University, Olomouc, Czech Republic c Institute of Microbiology, v.v.i., Academy of Sciences of the Czech Republic, Prague, Czech Republic b

h i g h l i g h t s

g r a p h i c a l

a b s t r a c t

 We demonstrate a technique for ambient spray ionization from nonpolar solvents.  The technique is mechanistically distinct from extractive electrospray ionization.  Factorial design experiments, among others, elaborate mechanistic details.  CF-EDESI expands the application base of conventional ESI.

a r t i c l e

i n f o

Article history: Received 19 November 2012 Received in revised form 7 January 2013 Accepted 11 January 2013 Available online 20 January 2013 Keywords: Ambient ionization Electrospray Progesterone Mass spectrometry

a b s t r a c t Due to their low polarities and dielectric constants, analytes in solvents such as hexane, chloroform, and ethyl acetate exhibit poor electrospray ionization (ESI) efficiency. These are deemed to be “non-ESIfriendly” solvents. Continuous flow extractive desorption electrospray ionization (CF-EDESI) is a novel ambient ionization technique that was recently developed in our group to manipulate protein charge distributions. Here we demonstrate its potential for ionizing analytes from non-ESI-friendly solvents. This feature makes CF-EDESI attractive to the general analytical community due to its apparent potential in lipidomics, normal phase separations, and hyphenation of mass spectrometry with HPLC-NMR systems. In this context, interest was subsequently initiated to discern mechanistic aspects of CF-EDESI. To achieve this, mechanistic experiments associated with a seemingly similar ambient ionization technique, extractive electrospray ionization (EESI), were emulated to compare CF-EDESI and EESI. Analysis of a series of fatty acids in multiple solvents in the negative ionization mode revealed differences between the two techniques. Whereas EESI has been previously shown to operate via extraction of analytes into the spray solvent, data presented here for CF-EDESI point toward a liquid-liquid mixing process to facilitate ionization. Further, a partial factorial design experiment was performed to evaluate the effects of different experimental variables on signal intensity. Sample flow rate was confirmed to be among the most significant factors to affect sensitivity. As a whole, the work presented provides greater insight into a new ambient ionization process, which exhibits expanded capabilities over conventional ESI; in this case, for direct analysis from non-ESI-friendly solvents. © 2013 Elsevier B.V. All rights reserved.

∗ Corresponding author at: 700 Planetarium Place, Box 19065, Arlington, TX 76019-0065, USA. Tel.: +1 817 272 3541; fax: +1 817 272 3808. E-mail address: [email protected] (K.A. Schug). 0003-2670/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.aca.2013.01.018

L. Li et al. / Analytica Chimica Acta 769 (2013) 84–90

85

1. Introduction Electrospray ionization (ESI) has made a significant impact on the development of mass spectrometry (MS) due primarily to its application as a soft ionization source for analysis of small and large biomolecules from aqueous media [1,2]. ESI can easily be coupled with high performance liquid chromatography (HPLC) and other liquid separation modes to facilitate MS analysis of complex mixtures. Wider application has been achieved for a multitude of analytes in pharmaceutical [3], environmental [4], forensic [5], and clinical [6] analysis, among many others. ESI still has limitations and one is the choice of solvent. Typical ESI solutions are aqueous mixtures of polar organic solvents. They represent a good compromise for efficient ionization as the high surface tension and low volatility of an aqueous solution are balanced by the low surface tension and high vapor pressure of polar organic solvents. Meanwhile, the aqueous mixture maintains good conductivity and a reasonably high dielectric constant. For solvent systems that have low conductivity, low dielectric constant, and high surface tension, low electrospray ionization efficiency results because (a) it is difficult to achieve a stable spray with solvents possessing low surface tension, (b) charge separation is hindered at the spray capillary tip, and (c) nonpolar or low polarity solvents insulate the analyte from efficient charging. Further, nonpolar or low polarity solvents do not dissolve electrolytes that would ensure adequate conductivity of solutions necessary for proper spraying. Consequently, solvents which are widely used in organic reactions (primary solvents in radical reactions), sample extraction, normal phase chromatography, analysis of complex lipids, and NMR analysis, are less amenable for use in ESI-MS. That said, there have been accounts in the literature focusing on the use of non-ESI-friendly solvents for ESI-MS analysis. Such experiments have generally relied on augmenting the solvent through the addition of more ESI-friendly components. Non-polar solvents generally require post-column addition of a polar solvent to attain satisfactory sensitivity [7,8]. Noncovalent complexes and supramolecular assemblies, which are stable in low polarity solvents can be promoted into the gas phase by ESI for interrogation by MS, but success in this regard can be highly system dependent [9,10]. Gas phase ionic metal–ligand complexes can be generated directly from solvents such as chloroform or dichloromethane using ESI, but the vast majority of reports for such systems, some of which may require an anhydrous environment to remain intact, utilize acetonitrile as a dominant component in the spray solvent [11,12]. The use of ionic liquids as an additive to hydrocarbon solvents (such as hexane and toluene) was reported to facilitate ionization of metal complexes by McIndoe and coworkers [13]. Also worthy of note, Van Berkel and coworkers have published a number of papers focused on redox charging of analytes using ESI in combination with halogenated solvents [14–16]. So, while analytes in nonpolar and low polarity solvents have been successfully studied using ESI-MS, these solvents can still be considered to be non-ESI-friendly in the context of the vast majority of routine and highly sensitive applications. Addition of polar solvents as make-up flow to promote ionization dilutes the analyte and can compromise sensitivity (or chromatographic efficiency, where applicable). Recently, a plethora of ambient ionization (AI) techniques have been introduced as alternate means to generate ions for mass spectral analysis [17]. Independent optimization of the ionization and sample introduction processes are generally achievable. Often, samples can be introduced into the ion source in their original state, with minimal sample preparation. Desorption electrospray ionization (DESI) was one of the first and most popular AI techniques introduced [18]. DESI facilitates the sampling of solids and liquids by directing the electrospray plume

Fig. 1. Configuration of continuous-flow extractive desorption electrospray ionization (CF-EDESI). The electrospray probe was aligned on-axis with the MS inlet, with a distance of 8 mm between them. A hypodermic needle for continuous flow sample introduction was set orthogonal to the ESI plume, and about 1.5 mm away from the electrospray source.

to desorb and pick-up analytes from a surface placed between the ionization source and MS inlet in a precise angular arrangement. A multitude of applications have been demonstrated [19–21] and the mechanism has been well established [22,23]. Non-traditional spray solvents have been shown to be beneficial for analysis of hydrophobic analytes [24]. An off-shoot of DESI, developed by the Brodbelt group, is transmission-mode DESI (TM-DESI) [25–28]. In this technique, a sample solution is deposited on a polymeric mesh screen, which is placed between the ESI plume and MS inlet. The source is arranged on-axis with the MS inlet, facilitating transmission of the ESI droplets through the screen and desorption, ionization, and detection of the analyte. Experimental variables and materials have been carefully investigated and characterized to achieve optimum performance for detection of a range of analytes by TMDESI-MS. For liquid samples, especially in the form of a continuous flowing stream, extractive electrospray ionization (EESI) is another AI technique [29]. EESI involves the comingling of the electrospray plume with a pneumatically nebulized sample solution or aerosol in front of the MS inlet to achieve ionization and detection of analytes. EESI has been shown to be viable for analysis of undiluted urine [29], milk [30], perfumes [31], and human breath [32], with minimal or no sample preparation. Non-ESI-friendly solvents such as hexane, cyclohexane, benzene, chloroform were successfully used in a mechanistic investigation of EESI, focused on fatty acid analysis [33]. Inspired by developments in DESI, TM-DESI, and EESI, we have developed a new AI source termed continuous flow-extractive desorption electrospray ionization (CF-EDESI) (Fig. 1). In CF-EDESI, sample solutions are introduced into the ESI spray plume (set onaxis with the MS inlet) by means of a pumped flow through a hypodermic needle set orthogonal to the source. Previously, this AI technique has been demonstrated for the manipulation of protein charge state distributions [34], but little effort was given to characterizing the mechanism associated with its successful performance. In this study, we demonstrated the advantages of CF-EDESI for use in the analysis of analytes dissolved in “non-ESI-friendly” solvents. We have sought to systematically examine the effect of experimental variables on CF-EDESI response in both positive and negative ionization modes. A comparison was also made to EESI in the context of fatty acids analysis [33]. Partial factorial design experiments were able to identify variables that had the greatest impact on ionization efficiency and these were subsequently characterized over a wide range of operational settings. Meanwhile, in

86

L. Li et al. / Analytica Chimica Acta 769 (2013) 84–90

the case of the various analytes targeted in this study, a significantly higher response was obtained by CF-EDESI compared to ESI, indicating the promise of this technique for future applications.

important were also investigated under an extended range of settings. 3. Results and discussion

2. Experimental 2.1. Chemicals and reagents LC–MS grade water (H2 O) and acetonitrile (ACN) were obtained from Burdick & Jackson (Muskeegon, MI, USA). LC–MS grade methanol (MeOH) and glacial acetic acid (HOAc) were obtained from J.T. Baker (Phillipsburg, NJ, USA). Hexanes were purchased from Fisher Scientific (New Jersey, USA). Chloroform was from Pharmco-AAPER (Connecticut, USA). Ethyl acetate, isopropanol, progesterone, hydrocortisone, and menadione (Vitamin K3 ) were obtained from Sigma–Aldrich (St. Louis, MO, USA). Myristic acid, palmitic acid, oleic acid, stearic acid, arachidic acid, and linoleic acid (marketed as GC standards) were purchased from Fluka (St. Louis, MO, USA). 2.2. Instrumental A LCQ Deca XP quadrupole ion trap mass spectrometer equipped with a conventional ESI source (Thermo-Fisher Scientific, Inc., San Jose, CA, USA) was used and modified in-house to construct the CFEDESI arrangement depicted in Fig. 1. A homemade XYZ stage was constructed and used to carefully position the hypodermic needle (22 gauge; SGE Analytical Science, Victoria, Australia) in the spray path. The position of the CF needle was optimized by monitoring the response of a 10 ␮M progesterone solution in hexanes to achieve a stable and high intensity signal. The indicated distances between the electrospray source needle and the CF needle (1.5 mm) and between the electrospray source and the inlet of the mass spectrometer (8.0 mm) are generally consistent with the optimum arrangement. Samples were prepared and introduced through the CF needle by direct infusion with a 500 ␮L SGE syringe and syringe pump (KD scientific, Model KDS-108, Holliston, MA, USA). Typical ESI solvents were introduced into the electrospray source using the syringe pump housed on the LCQ Deca instrument. A 200 ␮L stainless steel injection loop immersed in a hot water bath was integrated into the CF sample solution line when temperature studies were performed. 2.3. Factorial design experiments In order to efficiently obtain information about the impact of experimental variables on measured signal intensity in CF-EDESI, a partial factorial design experiment was devised with the help of a commercial factorial design software package (Design-Ease version 8.0). Five main factors (independent variables), deemed most important for the optimal experimental arrangement, were chosen for evaluation. The factors included sample and electrospray solvent flow rates (5, 50 ␮L min−1 ), nebulizer gas flow rate (40, 80 arbitrary units), electrospray voltage (3, 5 kV) and sample solution temperature (20, 50 ◦ C). Reasonable minimum and maximum settings for each variable (given in parentheses) were chosen according to some preliminary tests. A half-factorial experimental design (25 /2 = 16 randomized variable combinations) was performed in triplicate for a total of 48 analytical determinations under the defined conditions. The monitored output was the mass spectral signal intensity, obtained for analysis of solutions containing 10 ␮M progesterone in either hexane or chloroform. Because the experimental design assumes a linear change in the response variable as a function of its change in setting, additional experiments for variables identified by the factorial study to be the most

Several analytes including progesterone, hydrocortisone and vitamin K3 in different ESI-non-friendly solvents were evaluated for analysis using CF-EDESI. As a bioactive small molecule of significant interest [35–37], we chose progesterone as our primary model compound for extended experiments. While a variety of methodologies have been published for the analysis of progesterone and related chemical compounds [38–41], its hydrophobic nature generally dictates the use of hydrophobic media for its isolation from various sample types. Thus, strategies for analysis of progesterone, and other compounds of similar chemical nature, directly from such media could be considered beneficial; additional sample preparation steps or dilution of the sample (off-line or on-line) to introduce ESI-friendly solvents could be avoided. Based on solubility, some other combinations of analytes and non-ESI-friendly solvents were also investigated. Overall, excellent signal intensities for the combinations tested were obtained by CFEDESI, as shown in Table 1. A solution of 49/49/2 MeOH:H2 O:HOAc was used as an optimal electrospray solvent after some preliminary tests. For comparison, the same sample solutions were directly infused for analysis using conventional ESI-MS (in the standard orthogonal spray geometry on the LCQ instrument). No discernable signal response was observed in conventional ESI for the analytes tested in hexanes and chloroform. Low signal intensity for progesterone was observed in ethyl acetate, but the analysis signal was dwarfed by an interference signal from the solvent (m/z 177 was the observed base peak when EA was used as sample solvent). Significant interference signals were also seen in the CF-EDESI analysis of progesterone in ethyl acetate, but higher signal intensity for the analyte signal of interest was still recorded, relative to ESI. Representative spectra for each condition are given in the on-line Supplementary Information document. An optimal ESI spray solvent (49/49/2 MeOH:H2 O:HOAc) was found to maximize signal response in the CF-EDESI mode. This modified aqueous organic solvent is a good compromise between conductivity, volatility, and charge separation to facilitate ion formation through the ESI mechanism in the positive ionization mode. However, the notion of mixing between the spray and sample solvents, or the possibility of contributions from microextraction processes prompted us to consider the effect of miscibility of the sample and spray phases on signal quality. Ethyl acetate (EA) and chloroform (CHCl3 ) are partially miscible with a 49/49/2 MeOH:H2 O:HOAc spray solvent, whereas hexanes is largely immiscible with it. These assertions were made based simply on mixing the solvents together in a vial, however, they do not account for how charged spray droplets might possibly effect the interaction between the spray and sample solvents. In general, better results were obtained from ethyl acetate and chloroform (Table 1), but a more thorough investigation of phase miscibility was also performed. To investigate the effects of relative miscibilities of sample and spray solutions on analyte response, the analyte progesterone was introduced through continuous flow in hexanes, chloroform, ethyl acetate and methanol, while the composition of the electrospray solvents was varied (0%, 49% and 98% MeOH in aqueous solution with 2% HOAc). According to the data displayed in Fig. 2, when the sample was dissolved in hexanes and ethyl acetate, the use of 49/49/2 MeOH:H2 O:HOAc clearly provided the highest signal response. For other sample solvents, there was no significant difference among performance of spray solvents, within experimental error. A slight preference for an ionization spray with high water content when the sample was present in 100% methanol is notable.

L. Li et al. / Analytica Chimica Acta 769 (2013) 84–90

87

Table 1 Comparison of CF-EDESI and ESI response from non-ESI-friendly solvents for various analytes. Entry

Analyte

Solvent

Concentration (␮M)

ESI response (/105 )a

CF-EDESI response (/105 )a,b

1 2 3 4 5

Progesterone Progesterone Progesterone Vitamin K3 Hydrocortisone

Hexane Chloroform Ethyl acetate Hexanes Chloroform

10 10 10 1000 10

ND ND 1.7 ± 0.3 ND ND

6.5 16.9 14 1.5 21

a b

± ± ± ± ±

0.5 0.9 1 0.3 2

Results presented as average ± standard deviation (n = 3). ‘ND’ denotes not detected. Electrospray solvent was 49/49/2 MeOH:H2 O:acetic acid.

Taken together, these ionizing sprays are either miscible (and facilitate mixing of the phases; e.g. high methanol) or immiscible (and facilitate extraction of the sample solvent by the ionization solvent; e.g. high water) with the sample solvent. Regardless, optimal response appears to be achieved with the use of a spray solvent, which would facilitate ionization in conventional ESI. In other words, the use of a 49/49/2 MeOH:H2 O:HOAc spray solvent provided the optimal response in these experiments. The CF-EDESI moniker was chosen to describe an AI technique that was believed to have similarities with other established techniques, namely DESI (particularly TM-DESI, given the on-axis arrangement of the source and MS inlet) and EESI. A greater similarity is apparent for EESI, which allows the coupling of the source to a continuous flow and has been shown to be a powerful tool for analysis of aerosols, complex solutions, and suspensions without additional sample pretreatment [29,31]. Recent experiments designed to understand the mechanism of EESI have revealed that extraction occurs between the nebulized sample aerosol and charged spray droplets in the intersection region between the two phases, just in front of the MS inlet [33]. In other words, the key factor governing EESI response is that the analyte should be soluble in the charged spray solvent to facilitate extraction from the sample aerosol and subsequent ion formation. This finding was based on the analysis of a series of fatty acids present in a range of sample solvents. During this process, matrix interferences can be reduced if they are extracted to a lesser degree into the electrospray phase compared to the analytes of interest. In order to establish the similarities or differences between EESI and CF-EDESI, an experiment complementary to that reported by Zenobi and coworkers [33] was carried out in our lab. Solutions of

Fig. 2. Effect of electrospray composition (10 ␮L min−1 ) on the signal intensity for progesterone (10 ␮M in 100% hexane, chloroform, ethyl acetate, or methanol) introduced through the continuous flow line (10 ␮L min−1 ). Responses are reported as average ± standard deviation for triplicate analysis.

four fatty acids (1 ␮g mL−1 ), including myristic acid (14:0), palmitic acid (16:0), stearic acid (18:0), and linoleic acid (18:2) in hexane, cyclohexane, chloroform, acetonitrile, and methanol, were examined using CF-EDESI in the negative ionization mode. A reduced analyte set in our work compared to the previous work was necessary due to limits in solubility for some higher chain fatty acids. The spray solvent (80/19.5/0.5 MeOH:H2 O:NH4 OH) was chosen to be identical to that used for EESI. Fig. 3 provides a visual comparison between the results of the two experiments. Our experiments demonstrated a difference between EESI and CF-EDESI for the analysis of fatty acids in the negative ionization mode. Consistent with its moniker, EESI is believed to be largely an extraction process [33]. The solubility of fatty acids has been reported to be higher in methanol compared to acetonitrile [42]. Additionally, it has been shown that the ESI response of fatty acids in protic solvents is much higher than aprotic solvents [43,44]. In EESI, fatty acids originally in acetonitrile provided the highest signal. According to the authors, because of the higher solubility of the fatty acids in methanol, the analytes are efficiently extracted into the methanolic spray phase from their original acetonitrile sample solution. In CF-EDESI, fatty acids displayed the highest response in methanol, a protic solvent; this result is consistent with that expected for conventional ESI, and it is consistent with other experiments presented in this work. Another obvious difference of CF-EDESI and EESI is the droplet-bulk solution interaction (charged solvent spray interacted with continuous flow of sample solution) in CF-EDESI compared to droplet-droplet (solvent aerosol and sample aerosol) interaction in EESI. This difference can be speculated to lead to changes in ionization efficiency under similar operation conditions, and indicates that CF-EDESI likely operates under an independent mechanism (however, further experiments with an extended analyte set would need to be evaluated by both techniques to fully validate this assertion). In an effort to more systematically characterize the importance of experimental variables on measurable signal intensity, a partialfactorial design experiment was performed to predict the effect of individual and multiple factors on the CF-EDESI signal intensity. Commercial software was used to process the results obtained from experiments to evaluate a matrix of five prominent variables in a series of 48 analytical runs, which included triplicate analysis of each condition specified. A half-factorial design provided economy in initial evaluation of each variable’s effect on signal intensity. It allowed those variables which are most important to be identified so that additional experiments over extended variable setting ranges could then be performed. The results of the factorial design experiment are given in the form of a half-normal probability plot, shown in Fig. 4. The points that deviate from the linear region, toward the right of the plot, are considered significant positive and negative effects (depending on the variable specified), whereas the effects close to the “zero region” (on the line) are categorized as insignificant effects relative to experimental error. The positive effects represent a direct relationship of the design factors with the system outcomes (i.e.,

88

L. Li et al. / Analytica Chimica Acta 769 (2013) 84–90

Fig. 3. Comparison of fatty acid ionization response between (A) extractive electrospray ionization (EESI) (Reproduced from reference [33] with permission (© 2010 American Chemical Society)) and (B) CF-EDESI. For CF-EDESI, the electrospray solvent was identical with EESI work [33] (80/19.5/0.5 MeOH:H2 O:NH4 OH) (10 ␮L min−1 ), and the samples were introduced in the indicated solvent through continuous flow at 10 ␮L min−1 .

increasing the setting of these variables aids response), whereas the negative effects represent an inverse relationship (i.e., increasing the setting of these variables hinders response). Additional quality control parameters provided by the software indicated that the data set used to generate these results was of high quality. According to this analysis, a higher setting for sample flow rate (variable “B” in Fig. 4) and spray voltage (D) displayed the most significant positive effect (an order of magnitude increase) on signal intensity. A combined increase of these two variables (BD) provided the next most significant positive effect. An increase in temperature (A) returned the most significant negative effect on signal intensity. Any combinations of variables with temperature (e.g. AB, AC, AD, and AE) produced negative effects. Other factors had lesser effects. An increase in spray solvent flow rate (C) had a positive

Fig. 4. Results of the partial factorial design variable-effect analysis. In this halfnormal plot, standardized values for the magnitude of an effect are given on the x-axis and the significance of the effect are correlated with the random error as a probability on the y-axis. Estimates of errors as triangles for insignificant effects are shown as the “zero region” on the line, whereas significant variables and variable combinations (positive effects are filled and negative effects are open squares) are shown as deviations from the zero region. Positive and negative effects refer to increases and decreases, respectively, in ion response for 10 ␮M progesterone in either hexane and chloroform as a function of increases in the specified variable (A–E). Specific settings for variables are given in Section 2.

effect, whereas an increase in nebulizer gas flow rate (E) gave a negative effect in the context of the selected maximum and minimum variable settings tested. Interestingly, the combination of a high sample flow rate (alone, a strong positive effect) with a high nebulizer gas (alone, a weak but significant negative effect) showed no effect on intensity; this point fell on the zero region line of the plot. Additional experiments were performed to elaborate the effect some identified variables had on signal intensity over a wider range of settings. The expanded evaluation (Fig. 5) largely agreed with the factorial design experiments, with the exception, as expected, that some variables did not show a linear relationship with signal intensity (the half-factorial design, though efficient, only evaluates changes between a minimum and maximum setting for a variable). Sample flow rate was shown to correlate well in a direct manner with intensity, although very high sample flow rates lead to greater variability. Analysis of nebulizer gas flow exhibited a maximum in signal intensity around 40–50 arb units, when the sample and electrospray solvent flow were both maintained at 10 ␮L min−1 . The effect of temperature on signal response was measured several times over extended time periods; the replication was performed due to the high variability observed and the lack of congruence of data with the factorial design experiments. In the more detailed evaluation, an increase in temperature either had no effect, or slightly decreased the measured signal intensity when the standard variable settings for other parameters were held constant. It is possible that temperature effects could be more pronounced as other settings are changed (temperature was found to interact with many of the other variables in the factorial design experiments), but further experiments would be necessary to fully characterize such interactions. Practically speaking, it is much simpler to operate the CF-EDESI source under ambient conditions. The high variability of the data shown for the temperature experiments can be attributed to the fact that long time periods elapsed between the experiments; temporal variation of ESI-based MS instruments is not uncommon and the relative change within a given data set was the most important consideration in these experiments. To summarize, although this is an ambient ionization technique that is still under development, the experiments presented have demonstrated a significant advantage in the use of CF-EDESI for the generation of ions from non-ESI-friendly solvents. The potential for independent optimization of the ionizing spray flow and the sample flow is well apparent in this technique. The best sensitivity is obtained when the electrospray solvent is chosen on the basis of ESI theory for efficient ion generation (e.g., acidified

L. Li et al. / Analytica Chimica Acta 769 (2013) 84–90

89

Fig. 5. Detailed investigation of sample flow rate, sample temperature, and nebulizer gas flow rate on responses in CF-EDESI. (A) Continuous flow sample flow rate vs. electrospray solvent flow rate with 10 ␮M progesterone in hexane. (B) Sample temperature tested from 0 to 69 ◦ C (Hexane b.p.: 69 ◦ C) with 10 ␮M progesterone in hexane on four separate days (replicates 1–4). (C) Sample temperature tested from 0 to 60 ◦ C (Chloroform b.p.: 61 ◦ C) with 10 ␮M progesterone in chloroform on two separate days (replicates 1 and 2). (D) Nebulizer gas flow tested from 15 to 95 arbitrary units with 10 ␮M progesterone in hexane. (E) Nebulizer gas flow tested from 15 to 95 arbitrary units with 10 ␮M progesterone in chloroform. Unless specified, other parameters were held at continuous flow sample flow rate 10 ␮L min−1 , electrospray solvent flow rate 10 ␮L min−1 , spray voltage 5 kV, nebulizer gas flow 40 arb, and ambient sample temperature.

aqueous methanol). Further, where conventional ESI was not able to generate appreciable signal for analytes in nonpolar or low polarity solvents, CF-EDESI could. In cases where partial miscibility of spray and sample solvents was apparent, the signal intensity was higher. This prompts the proposition that some degree of mixing between the phases is optimal for efficient ion generation. Even so, a component of extraction is also possible, given that ions could still be observed using phases that were generally immiscible. Further work, using an expanded set of analytes will be pursued in the future to clarify the relative contribution, limits, and advantages for systems that better align with a total mixing vs. phase extraction mechanism for ion formation. 4. Conclusion CF-EDESI, as a novel ambient ionization technique, was applied to the analysis of analytes from non-ESI-friendly solvents, such as hexanes, chloroform and ethyl acetate. A careful systematic investigation of important variables, along with comparisons to results reported in the literature, has helped solidify the unique nature of this new AI source. Further applications of this technology are underway, and should broaden the scope of analyte systems which can be investigated using a commercial mass spectrometry system modified with a CF-EDESI ion source. Specifically, we are exploring applications, such as tracking organic synthesis product reactions (from the reaction flask or the NMR tube; including reaction kinetics) and normal phase liquid chromatography separations (e.g., lipidomics and chiral separations). Acknowledgements The authors acknowledge support from the National Science Foundation (CHE-0846310), the Ministry of Education, Youth and Sports of the Czech Republic (AMVIS Czech Republic – U.S.

exchange program ME10013), the Operational Program Research and Development for Innovations – European Regional Development Fund (CZ.1.05/2.1.00/03.0058), and the Czech Science Foundation (P206/12/1150). Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.aca.2013.01.018. References [1] M. Yamashita, J.B. Fenn, J. Phys. Chem. 88 (1984) 4451–4459. [2] J.B. Fenn, M. Mann, C.K. Meng, S.F. Wong, C.M. Whitehouse, Science 246 (1989) 64–71. [3] M.S. Lee, E.H. Kerns, Mass Spectrom. Rev. 18 (1999) 187–279. [4] S.D. Richardson, Anal. Chem. 78 (2006) 4021–4045. [5] J.F. Van Bocxlaer, K.M. Clauwaert, W.E. Lambert, D.L. Deforce, E.G. Van den Eeckhout, A.P. De Leenheer, Mass Spectrom. Rev. 19 (2000) 165–214. [6] B. Shushan, Mass Spectrom. Rev. 30 (2011) 491–509. [7] A.P. Zavitsanos, T. Alebic-Kolbah, J. Chromatogr. A 794 (1998) 45–46. [8] L. Charles, F. Laure, P. Raharivelomanana, J.P. Bianchini, J. Mass Spectrom. 40 (2005) 75–82. [9] C.A. Schalley, R.K. Castellano, M.S. Brody, D.M. Rudkevich, G. Siuzdak, J. Rebek Jr., J. Am. Chem. Soc. 121 (1999) 4568–4579. [10] C.A. Schalley, Mass Spectrom. Rev. 20 (2001) 253–309. [11] V. Katta, S.K. Chowdhury, B.T. Chait, J. Am. Chem. Soc. 112 (1990) 5348–5349. [12] V.B. Di Marco, G.G. Bombi, Mass Spectrom. Rev. 25 (2006) 347–379. [13] M.A. Henderson, J.S. McIndoe, Chem. Commun. (2006) 2872–2874. [14] G.J. Van Berkel, F. Zhou, J. Am. Soc. Mass Spectrom 7 (1996) 157–162. [15] G.J. Van Berkel, S.A. McLuckey, G.L. Glish, Anal. Chem. 64 (1992) 1586–1593. [16] G.J. Van Berkel, F. Zhou, Anal. Chem. 67 (1995) 2916–2923. [17] H. Chen, G. Gamez, R. Zenobi, J. Am. Soc. Mass Spectrom. 20 (2009) 1947–1963. [18] Z. Takats, J.M. Wiseman, B. Gologan, R.G. Cooks, Science 306 (2004) 471–473. [19] Z. Takats, I. Cotte-Rodriguez, N. Talaty, H. Chen, R.G. Cooks, Chem. Commun. (2005) 1950–1952. [20] Z. Miao, S. Wu, H. Chen, J. Am. Soc. Mass Spectrom. 21 (2010) 1730–1736. [21] A. Lane, L. Nyadong, A.S. Galhena, T.L. Shearer, E.P. Stout, R.M. Parry, M. Kwasnik, M.D. Wang, M.E. Hay, F.M. Fernandez, J. Kubanek, Proc. Natl. Acad. Sci. U. S. A. 106 (2009) 7314–7319.

90

L. Li et al. / Analytica Chimica Acta 769 (2013) 84–90

[22] M. Nefliu, J.N. Smith, A. Venter, R.G. Cooks, J. Am. Soc. Mass Spectrom. 19 (2008) 420–427. [23] A.B. Costa, R.G. Cooks, Chem. Phys. Lett. 464 (2008) 1–8. [24] A. Badu-Tawiah, C. Bland, D.I. Campbell, R.G. Cooks, J. Am. Soc. Mass Spectrom. 21 (2010) 572–579. [25] J.E. Chipuk, J.S. Brodbelt, J. Am. Soc. Mass Spectrom. 19 (2008) 1612–1620. [26] J.E. Chipuk, M.H. Gelb, J.S. Brodbelt, Anal. Chem. 82 (2010) 16–18. [27] J.E. Chipuk, M.H. Gelb, J.S. Brodbelt, Anal. Chem. 82 (2010) 4130–4139. [28] J.E. Chipuk, J.S. Brodbelt, J. Am. Soc. Mass Spectrom. 20 (2009) 584–592. [29] H. Chen, A. Venter, R.G. Cooks, Chem. Commun. (2006) 2042–2044. [30] L. Zhu, G. Gamez, H. Chen, K. Chingina, R. Zenobi, Chem. Commun. (2009) 559–561. [31] K. Chingin, G. Gamez, H. Chen, L. Zhu, R. Zenobi, Rapid Commun. Mass Spectrom. 22 (2008) 2009–2014. [32] P. Martínez-Lozanoa, J. Fernández de la Mora, Int. J. Mass Spectrom 265 (2007) 68–72. [33] W.S. Law, R. Wang, B. Hu, C. Berchtold, L. Meier, H. Chen, R. Zenobi, Anal. Chem. 82 (2010) 4494–4500.

[34] S.H. Yang, A.B. Wijeratne, L. Li, B.L. Edwards, K.A. Schug, Anal. Chem. 83 (2011) 643–647. [35] C.K. Osborne, M.G. Yochmowitz, W.A. Knight III, W.L. McGuire, Cancer 46 (1980) 2884–2888. [36] C.L. Chen, J. Meites, Endocrinol 86 (1970) 503–505. [37] R.L. Roof, R. Duvdevani, L. Braswell, D.G. Stein, Exp. Neurol. 129 (1994) 64–69. [38] M.S. Diaz-Cruz, M.J. Lopez de Alda, R. Lopez, D. Barcelo, J. Mass Spectrom. 38 (2003) 917–923. [39] M.T.A. Kawaguchi, J. Chromatogr. B 877 (2009) 343–346. [40] P. Regala, S. Anizanb, J.P. Antignacb, B.L. Bizecb, A. Cepedaa, C. Fente, Anal. Chim. Acta 700 (2011) 16–25. [41] S.S.C. Tai, M. Bedner, K.W. Phinney, Anal. Chem. 78 (2006) 6628–6633. [42] C.W. Hoerr, A.W. Balston, J. Org. Chem. 59 (1944) 329–337. [43] S. Cai, J.A. Syage, Anal. Chem. 78 (2006) 1191–1199. [44] P.J. Simms, C.T. Jeffries, X. Zhao, Y. Huang, T. Arrhenius, J. Chromatogr. A 1052 (2004) 69–75.