Relating Electrospray Ionization Response to Nonpolar Character of

Paul M. Flanigan , IV , Johnny J. Perez , Santosh Karki , and Robert J. Levis ... Santosh G. Valeja , Jeremiah D. Tipton , Mark R. Emmett and Alan G. ...
0 downloads 0 Views 139KB Size
Download PDF
Anal. Chem. 2000, 72, 2717-2723

Relating Electrospray Ionization Response to Nonpolar Character of Small Peptides Nadja B. Cech and Christie G. Enke*

Department of Chemistry, University of New Mexico, 103 Clark Hall, Albuquerque, New Mexico 87131

Nonpolar regions in biological molecules are investigated as a determining factor governing their electrospray ionization (ESI) mass spectrometric response. Response is compared for a series of peptides whose C-terminal residue is varied among amino acids with increasingly nonpolar side chains. Increased ESI response is observed for peptides with more extensive nonpolar regions. The basis for this increase is examined by comparing values of nonpolar surface area and Gibbs free energy of transfer for the different amino acid residues. Comparisons of response with octadecylamine are also made, and this highly surface-active ion is observed to outcompete all other analytes in ESI response. These observations are rationalized on the basis of the equilibrium partitioning model, which is used successfully to fit experimental data throughout the concentration range for several two-analyte systems. This model suggests that because excess charge exists on ESI droplet surfaces, an analyte’s relative affinity for the droplet surface determines its relative ESI response. Increased nonpolar character, which leads to enhanced affinity for the surface phase, results in more successful competition for excess charge and higher ESI response. Electrospray ionization (ESI) is the ionization technique of choice for interfacing liquid chromatography to the mass spectrometer. It owes its popularity to its sensitivity and to the ability to analyze large, nonvolatile, chargeable molecules such as proteins and nucleic acids. It is also widely used for the analysis of drugs, natural products, pesticides, carbohydrates, and other small molecules. Despite its advantages, some challenges exist with ESI that can make its implementation difficult. In equimolar solutions, analytes often have ESI responses that differ by orders of magnitude, creating problems with the implementation of internal standards.1,2 In addition, matrix effects can be severe, such that the presence of other analytes and salts can cause drastic signal suppression of the species of interest.2-5 This can severely * Corresponding author: (e-mail) [email protected]; (fax) (505) 277-2609. (1) Cheng, Z. L.; Siu, K. W.; Guevremont, R.; Bergman, S. S. J. Am. Soc. Mass Spectrom. 1992, 3, 281. (2) Tang, L.; Kebarle, P. Anal. Chem. 1993, 65, 3654-3668. (3) Constantopoulos, T. L.; Jackson, G. S.; Enke, C. G. J. Am. Soc. Mass Spectrom. 1999, 10, 625-634. (4) Xu, N. X.; Lin, Y. H.; Hofstadler, S. A.; Matson, D.; Call, C. J.; Smith, R. D. Anal. Chem. 1998, 70, 3553-3556. (5) Kirby, D.; Greve, K. F.; Foret, F.; Vouros, P.; Karger, B. L. Proceedings of the 42nd ASMS Conference on Mass Spectrometry Chicago, IL, 1994, 1014. 10.1021/ac9914869 CCC: $19.00 Published on Web 05/23/2000

© 2000 American Chemical Society

limit detection limits and create challenges in quantitation. For these reasons, a better understanding of the relationship between ESI response and analyte concentration could be very useful to the large group of scientists employing ESI. It was observed as early as 1983 in Iribarne and co-workers’ studies of atmospheric pressure ionization mass spectrometry (API-MS) that analytes with significant nonpolar portions have higher mass spectral response than highly polar analytes.6 Iribarne suggested that this enhanced response was a result of nonpolar analytes preferring the air-liquid interface at droplet surfaces. According to Iribarne’s ion evaporation theory, ions enter into the gas phase when solvent evaporation increases the charge density on the droplets to the extent that Coulombic repulsion overcomes the interaction between the analyte and the solvent. This causes the ions to “evaporate” from the droplet surfaces.7 On the basis of this theory, Iribarne hypothesized that ions already on the droplet surface would evaporate more readily than those in the droplet interior, resulting in enhanced response.6 Kebarle and co-workers used Iribarne’s ion evaporation theory to develop a comprehensive model relating analyte response to rate of evaporation from droplet surfaces.2,8-10 They developed equations to predict ESI response, relying on coefficients that described the relative abilities of various ions to be converted from the solution phase to the gas phase. Comparisons were made between experimentally measured ESI response and ion solvation energy for metal cations. It was generally observed that higher solvation energy could be correlated with lower ESI response. Later research indicated that differences in evaporation rates alone could not account for response differences observed with molecules that have a high affinity for the air-liquid interface of ESI droplets. When analyzing such analytes (termed “surface active”), Tang and Kebarle attributed response differences at low concentrations to differences in surface affinity and response differences at high concentrations to differences in both surface affinity and evaporation efficiency.2 The surface affinity was assumed to contribute to response differences in a constant ratio throughout the concentration range, while suppression in response was attributed to differences in evaporation rates at high and low analyte concentrations. (6) Iribarne, J. V.; Dziedzic, P. J. Int. J. Mass Spectrom. Ion Phys. 1983, 50, 331-347. (7) Iribarne, J. V.; Thompson, B. A. J Chem. Phys. 1976, 64, 2287-2294. (8) Tang, L.; Kebarle, P. Anal. Chem. 1991, 63, 2709-2715. (9) Tang, L.; Kebarle, P. Anal. Chem. 1993, 65, 972A-985A. (10) Sunner, J.; Nicol, G.; Kebarle, P. Anal. Chem. 1988, 60, 1300-1307.

Analytical Chemistry, Vol. 72, No. 13, July 1, 2000 2717

Figure 1. Predicted partitioning of peptides within an ESI droplet. A significant fraction of the more polar peptides remains in the neutral droplet interior where they are neutralized by counterions. Peptides with nonpolar side chains exist mostly on the droplet surface where these side chains can be desolvated. Peptides on the droplet surface can carry a greater fraction of the excess charge and will therefore have a higher ESI response than peptides in its interior.

The research presented here examines the effectiveness of a model (the equilibrium partitioning model3,12,13) that uses relative affinities for droplet surfaces to describe differences in response at both high and low concentrations. We suggest that analytes that can be most successfully analyzed by ESI-MS have both polar and nonpolar portions. The polar portions are necessary to enable ion formation, either through protonation or the formation of adducts with cations, while the nonpolar portions are responsible for increasing the fraction of the analyte molecules that reside on the surface of ESI droplets. The next section describes how the equilibrium partitioning model explains the correlation between an analyte’s affinity for droplet surfaces and its ESI response. EQUILIBRIUM PARTITIONING MODEL The equilibrium partitioning model of ref 12 postulates that two separate phases exist in an ESI droplet. Excess charge produced in the ESI process resides on the surface of the droplet, while its interior is electrically neutral and consists of solvent molecules, electrolytes, and charged analyte molecules matched by an equal number of counterions (Figure 1). The excess charge is described in terms of the average charge concentration in the initial droplet (eq 1), which can then be directly related to

[Q] ) i/ΓF

(1)

concentration of analyte. Here [Q] is the concentration of excess charge in equivalents per liter, i is the current measured in the needle circuit (in amperes), F is Faraday’s constant (96 485 C/equiv), and Γ is the solution flow rate (in L/s).12 In the analysis of protonated analytes (e.g., proteins) in the positive ion mode, the positive charge must be carried by charged (11) Kostiainen, R.; Bruins, A. P. Rapid Commun. Mass Spectrom. 1994, 8, 549558. (12) Enke, C. G. Anal. Chem. 1997, 69, 4885-4893. (13) Constantopoulose, T. L.; Jackson, G. S.; Enke, C. G. Anal. Chim. Acta 2000, 406, 37-52.

2718

Analytical Chemistry, Vol. 72, No. 13, July 1, 2000

analyte, charged solvent, or electrolyte impurities at the droplet surface. Assuming that the concentration of acid is significantly higher than that of electrolyte impurities (as is typically the case for ESI analysis), the sum of the concentrations of protonated analyte and solvent molecules on the surface of the droplet must be equal to [Q]. Protonated molecules on the droplet surface can exchange with those in the electrically neutral interior. When this happens, the charged molecule from the interior gives up its counterion and part of the excess charge on the droplet surface and the charged molecule from the surface becomes matched by the counterion and moves to the electrically neutral droplet interior. An equilibrium partitioning coefficient (K) is defined for each analyte as the ratio of its concentration on the droplet surface to that in its interior.12 Factors such as polarity, charge density, and basicity are important in determining the magnitude of K for a given ion. Analytes must be basic to carry charge (in the form of protons), and their charge density and the extent of their nonpolar regions will determine how favorably they will exist on the droplet surface. Analytes with very high K’s will exist mostly on the surface of ESI droplets, so that, at high analyte concentration, they will be capable of carrying a large fraction of the excess charge. Analytes with small K’s will exist in the droplet interior and be matched by counterions and thus have less ability to carry the excess charge. This partitioning should occur early in the droplet formation process, and therefore, the factors favoring an analyte’s existence in the surface phase will increase as desolvation proceeds. Only ions that are part of the surface excess charge phase and can carry the excess charge will show up in the mass spectrum. Those that exist in the electrically neutral droplet interior will be lost as components of neutral salts. For this reason, the tendency of a given ion to exist on the surface of a charged droplet should be directly related to its ESI response. The equilibrium partitioning model can be used to derive equations predicting analyte concentration on the droplet surface as a function of analytical concentration. For a two analyte system, this equation is cubic in [AH+]S (eq 2). This equation is a slight

[

(

[AH+]S3 KBH+ - KAH+ + KSH+ 1 -

[

)]

KBH+ KAH+

+

[AH+]S2 CA(2KAH+ - KBH+ - KSH+) +

(

CB KBH+ - KSH+

(

( )) KBH+ KAH+

(

+ CSH+KSH+ 1 -

(

[Q] KAH+ - KBH+ - KSH+ 1 -

KAH+

KAH+

))]

KBH+

)

KBH+

+

-

[AH+]SCA[[Q](2KAH+ - KBH+ - KSH+) + CBKBH+ + CAKAH+ + CSH+KSH+] + CA2[Q]KAH+ ) 0 (2)

modification of the one previously published, which included several typographical errors.12 Here CA is the analytical concentration of analyte A, KAH+ is its partitioning coefficient, and CSH+ is the concentration of protonated solvent.

METHODS All samples were analyzed using a Finnigan MAT triple -stage quadrupole (TSQ) 7000 mass spectrometer with modified Finnigan API source. The modification involved replacing the Finnigan needle assembly with a Taurus-R X-Y-Z micropositioner (World Precision Instruments, Sarasota, FL). The mass spectrometer was calibrated over the mass range of interest (m/z 100-300) using standard peptides of known molecular weight. The mass-dependent transmission of the quadrupoles was not taken into account during the calibration, however, since the mass ranges compared for the various compounds were only 100 mass units wide and the analytes studied were of relatively low mass; this mass dependence should not have been too severe. Since ion transmission is known to decrease with increasing m/z values, the possible effect of ignoring mass-dependent transmission on the experimental data would be a slight suppression of the observed increase in ESI response with increasing nonpolar character. A 5 × 1010 Ω resistor was inserted in series with the ESI power supply. The resistor causes stabilization of response and allows experiments to be performed at constant current regardless of analyte concentration.14 All peptide mixtures were analyzed in the positive ion mode, using a 100-µm-i.d. glass capillary inserted into a 250-µm-i.d. stainless steel needle. Solvent and analyte were injected from a 250-µm gastight syringe (SGE) at a constant flow rate of 5 µL/min using a Harvard Apparatus Syringe pump (South Natick, MA). A voltage of 4.5 kV was applied through direct electrical contact to the metal needle. Spray current for all experiments was monitored via a digital ammeter, which was placed in series with the electrospray needle. Stock solutions of 5 × 10-3 M G-G-G, G-G-A, G-G-V, G-G-L, G-G-Y, and G-G-F (Bachem, Torrance, CA), V-P-L, G-G-H, I-P-I, and T-Y-S (Sigma, St. Louis, MO) were prepared from standards. They were diluted in pure reagent grade methanol (Burdick and Jackson, Mustegon, MI) and Nanopure water (purified with a Barnstead ultrapure water system model 04747). All solvents were acidified with glacial acetic acid (J. T. Baker, Phillipsburg, NJ) to 0.5% v/v. The biological effects of the peptides studied are unknown. Gloves should be worn while handling them to avoid exposure. RESULTS AND DISCUSSION Effect of Nonpolar Regions on ESI Response. A series of simple tripeptides with varying nonpolar character was chosen as a starting point for the study of ESI response. These peptides all contained glycines at residues one and two. Residue three was varied among amino acids with increasing numbers of carbons in their side chains (Figure 2). The effect of changing the number of carbons in a side chain on polarity can be quantitatively measured in several ways. Tables of residue nonpolar surface area and Gibbs free energy of transfer (∆G°transfer) values from water into various solvents have been compiled by biochemists to aid in the prediction of protein folding. A multitude of hydrophobicity scales relying on these values and on empirical data have also been developed.15-19 Values for ∆G°transfer (kcal/mol) relative to (14) Jackson, G. S.; Enke, C. G., Proceedings of the 49th ASMS Conference on Mass Spectrometry, Dallas, TX, 1999. (15) Fauchere, J. L.; Pliska, V. Eur. J. Med. Chem. 1983, 18, 369-375. (16) Karplus, P. A. Protein Sci. 1996, 6, 1302-1307. (17) Kyte, J.; Doolittle, R. F. J. Mol. Biol. 1982, 157, 105-132.

Figure 2. Structures of six tripeptides used in the response comparison shown in Figure 3. The peptides have varying nonpolar character due to differences in their C-terminal residues. Note that the peptide backbone (the polar portion) is unchanged from one peptide to the next. Table 1. Comparison of Nonpolar Surface Area and Gibbs Free Energy of Transfer for Various Peptides C-terminal amino acid glycine (G) alanine (A) valine (V) leucine (L) phenylalanine (F) tyrosine (Y)

∆G°transfer (kcal/mol)a

nonpolar area (Å2)b

log(response)c

0.00 0.42 1.66 2.31 2.43 1.31

47 86 135 164 194 152

5.91 6.04 6.14 6.17 6.30 6.15

a Karplus, P. A. Protein Sci. 1996, 6, 1302-1307. b Fauchere, J. L.; Pliska, V. Eur. J. Med. Chem. 1983, 18, 369-375. c Experimentally determined for an equimolar 10-5 M mixture of peptides.

glycine from water into octanol16 and nonpolar surface areas (in Å2)15 for the N-terminal residue of each peptide are shown in Table 1. Values from hydrophobicity scales are not reported since they vary, depending on how they are calculated.16,17 It is also worth noting that for side chains that can hydrogen bond, significant differences are observed in ∆G°transfer values from scales developed using nonpolar solvents versus those developed using solvents that support hydrogen bonding.16 Figure 3 is a mass spectrum of an equimolar mixture of the six tripeptides whose structures are shown in Figure 2. As expected, increasing nonpolar surface area of the N-terminal residue is correlated with increased response of each peptide (Figure 4a). The correlation between response and ∆G°transfer is almost as good (Figure 4b), with the exception that, based on ∆G°transfer, G-G-Y would be predicted to have a lower response than G-G-V. This anomaly may partly be due to the poor reliability of ∆G°transfer values mentioned above. Tyrosine (Y) is the only amino acid in the series with a side chain that can undergo hydrogen bonding. (18) Wolfenden, R.; Andersson, L.; Cullis, P.; Southgate, C. Biochemistry 1981, 20, 849-855. (19) Rose, G.; Geselowitz, A.; Lesser, G.; Lee, R.; Zehfus, M. Science 1985, 229, 834-838.

Analytical Chemistry, Vol. 72, No. 13, July 1, 2000

2719

Figure 3. A mass spectrum comparing the responses of the six tripeptides whose structures are shown in Figure 2. Response increases with increasing nonpolar residue surface area (see Figure 4). Proton-bound dimers were not observed in the mass spectra of these peptides, possibly because of their small size. The additional small peaks in the mass spectrum represent peptide methanol clusters, whose response is insignificant with respect to that of the bare peptides.

The responses as a function of concentration for the peptides G-G-G, G-G-A, G-G-V, and G-G-L are shown in Figure 5. Equimolar mixtures of the four peptides dissolved in 50:50 methanol-water, 0.5% acetic acid were used to generate the curve. Calibration curves such as those shown in Figure 5 were also generated in different solvent mixtures (with varying percentages of methanol, water, acetonitrile, and chloroform). Similar trends in response were observed. The order of responses in Figure 5 is as predicted, with peptides with the most nonpolar C-terminal residues having the highest ESI response. In the concentration range below 10-4 M, where analyte concentration is significantly less than [Q], analytes compete only with solvent for excess charge. The solvent molecules that are most easily protonated (in this case methanol) must therefore carry the majority of the excess charge. As a result, the solvent response is much higher than the analyte response at low analyte concentrations. At higher concentrations, the analytes consume a large fraction of the excess charge and the solvent response is suppressed. In the high-concentration range (10-4 M and above), the responses of G-G-V and G-G-L level off, while those of G-G-G and G-G-A decrease slightly. This effect can also be explained by competition for excess charge. The use of the series resistor ensures that [Q] is constant throughout the experiment (eq 1). Consequently, as analyte concentration approaches [Q], competition for excess charge should occur among analytes. The suppression in the response of the less surface-active G-G-G and G-G-A is a result of their being less successful in competing for the excess charge than G-G-V and G-G-L. There are several ways to calculate [Q]. One is to use experimental measurements of current and flow rate. For the peptide mixture, this yields [Q] ) 5 × 10-5 M (log[Q] ) -4.3) (eq 1). [Q] can also be determined graphically. Assuming that the analyte with highest response (G-G-L) is completely on the droplet surface, extrapolating the linear portion of the response curve for this analyte to the point where it crosses the saturation 2720 Analytical Chemistry, Vol. 72, No. 13, July 1, 2000

Figure 4. (a) Log response as a function of nonpolar surface area15 for a mixture of the peptides shown in Figure 2. Response was determined for an equimolar 10-5 M solution of peptides in 50:50 methanol-water, 0.5% acetic acid. (b) Log response as a function of ∆G°transfer16 from water into octanol for the same peptide mixture shown in (a). Note that response increases both with increasing nonpolar surface area and with increasing ∆G°transfer.

concentration gives a predicted value of [Q].12 The graphically determined value of [Q], 2.5 × 10-5 M, is approximately equal to the value of 5 × 10-5 calculated from the needle current and flow rate. The difference between these two values may result from other processes that occur during droplet charging, which prevent some of the current from being converted to excess charge. Indeed, we have recently made observations of current pulsation that suggest that the actual current which goes to droplet charging may be significantly lower than that measured with our ammeter. For this reason, we have chosen to rely on the graphically determined [Q] in the calculations presented later. Note that, in Figure 3, saturation begins at ∼4 × 10-4 M (the sum of the analyte concentrations at 10-4 M). This correlates reasonably well to both graphically and experimentally determined values of [Q]. The observed effects of suppression and competition might be explained on the basis of differences in ion evaporation rates rather than competition for excess charge. In this study, however, the polar portions of the six tripeptides are all identical. Provided

Figure 5. ESI response for a mixture of four glycine-containing tripeptides in 50:50 methanol-water with 0.5% v/v acetic acid added as a protonating reagent. The response increases as the nonpolar character of side chain on the N-terminal residue increases. At high concentrations, the response of analytes with more extensive nonpolar regions levels off while that of the more polar analytes is suppressed. This suppression is a result of the peptides with greater surface affinity outcompeting the peptides with less surface affinity for excess charge on droplet surfaces.

Figure 6. Comparison of the responses of four tripeptides with significantly different structures in 50:50 methanol-water, 0.5% acetic acid. The responses of peptides with polar side chains (T-Y-S, G-GH) are lower than that of those with nonpolar side chains (I-P-I, V-PL). This response curve supports the proposed partitioning shown if Figure 1, with the peptides with more nonpolar character residing preferentially on the droplet surface and taking up a greater fraction of the excess charge.

that evaporation takes place from the surface of ESI droplets (and therefore the transition from the bulk to the surface is not involved), the rate of evaporation should depend only on the extent to which the polar portions are anchored in the polar solvent. Consequently, peptides with identical polar portions are expected to have identical evaporation rates. This strongly suggests that differences in ability to compete for excess charge on the droplet surface rather than differences in evaporability are responsible for the competition and suppression in response. Original work by Tang and Kebarle is in agreement with the equilibrium partitioning model in suggesting that the roll-off observed in ESI calibration curves at high analyte concentrations is due to a limit in the amount of charge available.2,9 However, Bruins and Kostainen suggested that this saturation in response may be a result of limited space on droplet surfaces instead.11,20 The latter conclusion is based on experiments that show the rolloff in calibration curves occurring at the same concentration with different applied voltages and electrolyte concentrations. Kostainen and Bruins argued that if competition is limiting, changes in electrolyte concentration and applied voltage should affect the analyte concentration at which saturation in response is observed. On the basis of the equilibrium partitioning model, we suggest that as long as analyte ions have higher partitioning constants than electrolyte ions, they will outcompete electrolytes for response. Consequently, the concentration at which saturation in response is observed should not be greatly effected by electrolyte concentration. The equilibrium partitioning model does, however, predict that increasing the applied voltage (and therefore the ESI current) should increase [Q] (eq 1) and thus cause calibration curves to level off at higher concentrations. The fact that changes in applied voltage had little effect on calibration curves in Bruins and Kostainen’s experiments is compelling evidence for some other basis of saturation. In our experiments, use of the series

resistor limits the ESI current and ensures that, at some point, the analyte concentration must exceed the concentration of excess charge. The question of whether this happens before or after space becomes limiting still remains to be answered. Response Comparisons with More Complex Peptides. The study of the effect of nonpolar character on ESI response was also extended to more complex tripeptides. Figure 1 shows the predicted partitioning within an ESI droplet of four tripeptides whose responses are compared in Figure 6. Two of these peptides, T-Y-S and G-G-H, have primarily polar side chains. The other two, I-P-I and V-P-L, have nonpolar side chains. As expected, the ESI responses of the two nonpolar peptides are much higher than those of the two polar peptides. In addition, the response of the polar peptides is suppressed at high concentrations. These results support the proposed partitioning of the peptides within the ESI droplet (Figure 1) and suggest that the nonpolar peptides are more successful in competing for the excess charge. Comparison of Peptide and Octadecylamine Responses. Even the most nonpolar peptides analyzed have highly polar groups as part of each peptide bond. It is therefore of interest to examine the response of analytes with even higher droplet surface affinity, those without the polar peptide backbone. Surfactants fulfill this requirement and are excellent test compounds since they are well known to prefer the air-liquid interface when dissolved in polar solutions. It is expected that they should also prefer the surface of ESI droplets, where their nonpolar side chains can be desolvated and their chargeable headgroups can carry the excess charge. In their early API-MS experiments, Iribarne and Dziedzic observed that the presence of the surfactant decyl alcohol suppressed the ionization process.6 It was suggested that this suppression could be attributed to the formation of a decyl alcohol film on droplet surfaces, which would prevent ion formation.6,21 However, these experiments were run in the positive ion mode, at pH’s where decyl alcohol is neutral. We predict that basic

(20) Bruins, A. P. In Electrospray Ionization Mass Spectromertry; Cole, R. B., Ed.; John Wiley and Sons: New York, 1997; Vol. 1, pp 107-135.

(21) Fenn, J. B. J. Am. Soc. Mass Spectrom. 1993, 4, 524-535.

Analytical Chemistry, Vol. 72, No. 13, July 1, 2000

2721

Figure 7. Predicted and experimental analyte surface concentration as a function of concentration for a mixture of V-P-L (a nonpolar peptide) and octadecylamine. (oct.). The response of the nonpolar peptide is suppressed by the octadecylamine at high concentrations. This demonstrates that the most surface active analyte in a solution has the highest ESI response. V-P-L suppresses the response of G-G-H and T-Y-S when it is the most surface active analyte (Figure 5) but is suppressed in response when combined with the more surface active octadecylamine. The dashed line represents a fit to the experimental data using eq 2. Parameters were as follows: [Q] ) 3.25 × 10-5 M, KVPL/KSH+ ) 21.1, KODA/KSH+ ) 83.1, and CSH+ ) 2.0 × 10-3 M.

surfactants, which will be charged at low pH, should have a very high ESI response in the positive ion mode due to their high surface activity. The alkylamine octadecylamine (stearylamine), which has an 18-carbon chain, is therefore chosen as a test compound. Figure 7 shows the response curves for a mixture of octadecylamine and the nonpolar peptide V-P-L. To compare predicted response with experimental results, analyte surface concentration was calculated from experimental response. The relationship between response and concentration was calculated by assuming that, in the saturation region, the sum of the analyte responses should equal the response for a total analyte concentration equal to [Q]. To calculate surface concentration as a function of analytical concentration, the sum of the experimentally measured responses at 10-4 M was divided by the graphically determined [Q]. The response of each analyte at each concentration was then divided by this value. The dashed line in Figure 7 represents a fit to the experimental data calculated with eq 2. To apply this equation, relative K values for octadecylamine and V-P-L were calculated by taking the ratio of the responses of the two analytes at 10-4 M (in the saturation region). This method of calculation is valid since, in the saturation region, the analytes compete only with each other for excess charge and thus the ratio of their responses should represent the ratio of their K values. CSH+ was calculated to be ∼2 × 10-3 M based on the concentration of acetic acid in solution and its Ka value. Using these values, the value of KVPL/KSH+ was adjusted (using the Excel solver) until eq 2 yielded the correct experimental surface concentration ([A]S) of octadecylamine at 10-6 M. This value of KVPL/KSH+ was then used to predict [A]s as a function of analytical concentration for both analytes at each concentration. The cubic fit shown in Figure 7 demonstrates that the equilibrium partitioning model’s predictions based on competition 2722 Analytical Chemistry, Vol. 72, No. 13, July 1, 2000

for excess charge successfully model response as a function of concentration throughout the concentration range. This fit was calculated using only one value for each partitioning coefficient at both high and low concentrations. This single constant is related to a single quality of the analyte (its ability to partition to the droplet surface), which is clearly related to a predictable relative surface activity. In comparing Figure 6 with Figure 7, it is apparent that the relative droplet surface affinity of an analyte determines whether its response will be suppressed at high concentrations. V-P-L and I-P-I suppress the response of G-G-H and T-Y-S when they are the most surface active analytes in solution (Figure 6), but the response of V-P-L is suppressed when it is mixed with the still more surface active octadecylamine (Figure 7). This is to be expected if indeed competition for surface area or excess charge determines relative response. The analyte with the highest affinity for the droplet surface will always have the greatest fraction of its concentration on the droplet surface and thus compete most successfully for excess charge. Investigation of Suppression at High Concentration Using Octadecylamine and T-Y-S. Another way to study the relative ability of various analytes to cause suppression in response is to keep the concentration of one analyte constant while increasing the concentration of another. These types of experiments were performed extensively by Tang and Kebarle. They demonstrated that it was far easier to suppress the response of cations such as Cs+ than that of charged organic molecules such as protonated cocaine and morphine2. Since the equilibrium partitioning model predicts response differences as a result of differences in surface affinity rather than evaporation efficiency, it was of interest to study the suppression of peptide response by the highly surface active octadecylamine. Two experiments were performed to demonstrate the relative abilities of octadecylamine and the polar peptide T-Y-S to suppress each other. T-Y-S was chosen for comparison since its high polarity guarantees that its affinity for droplet surfaces is much lower than that of octadecylamine. In the first experiment, the concentration of T-Y-S was kept constant at 10-5 M while the concentration of octadecylamine was increased from 10-7 to 5 × 10-4 M. The resulting experimental response curve and its fit (from the application of eq 2) are shown in Figure 8a. The fit was generated using the same method of calculation discussed for the octadecylamine/V-P-L experiment. A significant suppression in the response of T-Y-S was predicted and observed as the octadecylamine concentration exceeded 10-5 M. Figure 8b shows the reverse case, where the concentration of octadecylamine was kept constant at 10-5 M while the concentration of T-Y-S was increased from 10-7 to 5 × 10-4 M. In this case, almost no suppression of the octadecylamine response was observed, even at very high concentrations of T-Y-S, in both the experimental and predicted response curves. This suggests that even when the T-Y-S concentration is much higher than that of octadecylamine, the octadecylamine occupies the same amount of the droplet surface, and takes up the same fraction of the excess charge, as it did when there was very little T-Y-S present. In the octadecylamine/T-Y-S studies, the results could be rationalized either by differences in ion evaporation or by differences in surface affinity. The T-Y-S has more polar groups than

value order seems to follow that predicted on the basis of the nonpolar character. The K’s of the glycine-containing tripeptides are grouped with increasing nonpolar character of residue three correlating with higher values of K. The two peptides having three nonpolar side chains (V-P-L and I-P-I) have higher K values than the peptides with only one nonpolar residue, and the K value for octadecylamine, by far the most nonpolar analyte, is highest of all. The development of a series of numerical K values predicting relative response for all the analytes studied would be useful for making more quantitative predictions of relative response. The systems that were fit to the model demonstrate that this can be done. However, this would be a lengthy process since the equations governing interanalyte competition are too complex for mixtures of more than two analytes. Consequently, the sequence would need to be developed by comparing analyte pairs. This venture might not be very useful since it is our experience that the values of CSH+ and [Q] are not known with certainty, greatly depend on the solvent matrix used, and may vary from experiment to experiment. This prevents quantitative application of the table of K values that could be generated. Nonetheless, the overall order of K values as predicted above should be generally applicable and the concept of what aspect of the analyte determines its relative K value is beneficial in making qualitative predictions of response.

Figure 8. (a) Observed and predicted [A]S for an experiment in which the concentration of T-Y-S is held constant at 10-5 M while the concentration of octadecylamine is increased. Values used for the fit: KTYS/KSH+ ) 1.8, [Q] ) 6.3 × 10-5, KODA/KTYS ) 48.6, and CSH+ ) 2 × 10-3 M. A significant suppression of T-Y-S response is observed at high concentrations of octadecylamine. (b) The concentration of octadecylamine is held constant at 10-5 M while the concentration of T-Y-S (a polar peptide) is increased. Values for the fit: KTYS/KSH+ ) 4.9, [Q] ) 6.3 × 10-5, KODA/KTYS ) 48.6, and CSH+ ) 2 × 10-3 M. The increasing concentration of the polar peptide has very little effect on the response of octadecylamine in both predicted and observed response curves.

does octadecylamine, and it would therefore be expected to evaporate from the droplets less readily. However, the extreme surface affinity of octadecylamine makes it likely that the surface effect is predominant in this case. The fact that the equilibrium partitioning model can accurately model response curves for these analytes without accounting for differences in evaporation rates supports this hypothesis. Comparisons of Partitioning Coefficients among Multiple Analytes. Using the relative responses observed, it is possible to arrange the analytes studied in order of their K values as follows: KGGG < KTYS < KGGA < KGGH < KGGY < KGGV < KGGL< KGGF < KVPL< KIPI < Koctadecylamine. This order of K values was generated by comparing responses in the saturation range for all of the analytes to the response of G-G-G. An additional experiment comparing the responses of octadecylamine and G-G-G at 10-4 M was required to accomplish this comparison. The empirical K

CONCLUSIONS The data presented here support the hypothesis that nonpolar surface area (and therefore ∆G°transfer) is a major determining factor in the relative ESI response of surface-active analytes. By relying on the concept of competition for excess charge, the equilibrium partitioning model successfully predicts competition and suppression in both the high- and low-concentration regions. Ion evaporation rate has not been ruled out as a contributing factor. However, for the analytes studied, its contribution may be minor in comparison to the contribution of affinity for ESI droplet surfaces. The analytes studied thus far are simple biological molecules, but the profound effect of small changes in side-chain character on response suggests the importance of considering nonpolar character of proteins as well when predicting suitability for ESI analysis. The polar character as well as the nonpolar character will differ considerably from protein to protein. Consequently, it may turn out that the ratio of polar to nonpolar character can be used as a means to predict ESI response for large biological analytes. ACKNOWLEDGMENT The authors gratefully acknowledge the National Institutes of Health (Grant GM 44077) and Pfizer, Inc. for their partial support of this work. Thanks also go to Paul Kebarle for helpful discussions and manuscript review and to Terri Constantopoulos and George Jackson for indispensable support, discussions, and training.

Received for review December 29, 1999. Accepted April 3, 2000. AC9914869

Analytical Chemistry, Vol. 72, No. 13, July 1, 2000

2723