Peptide bioanalytical method development poses special problems for conventional liquid chromatography/tandem mass spectrometry (LC/MS/MS) systems.1 The total ion current is distributed among several charge states owing to multiple basic ionization sites. Therefore, many multiply charged species are distributed in an envelope of charge states. Even before method development begins, the scientist has less absolute signal than is frequently observed with singly charged small drug molecules.2
Tandem mass spectrometry (MS/MS) offers increased selectivity over single MS, but, because peptide bonds are nearly equal in energy to each other, the ion current becomes distributed among many equally low abundance fragments. To prevent this low response, selected ion monitoring (SIM) is a frequently used option.3 SIM is a sensitive technique due to the high ion current observed, but it is not very selective and hence the popularity of triple quadrupole mass spectrometry. Complex matrices frequently contribute interferences at the m/z for the analyte during SIM.4 These interferences appear as high chemical background and co-eluting chromatographic peaks.5
For compounds that have low ion current due to inefficient fragmentation, one way to take advantage of the selectivity of a triple quadrupole mass spectrometer is to use pseudo-selected reaction monitoring (pseudo-SRM). This experiment involves both mass-resolving quadrupoles at the same m/z with only a small amount of collision energy. The objective is to reduce the response of the interferences preferentially while keeping the peptide intact. In this study, a method was developed for the analysis of a peptide drug by LC/pseudo-SRM. During validation experiments, matrix interferences prevented the successful implementation of the method for analysis of pre-clinical study samples. Thus additional selectivity was required.
FAIMS (high-Field Asymmetric waveform Ion Mobility Spectrometry) increases the selectivity of an MS assay based on the trajectory of an ion in an atmospheric pressure gas.6 This technology has application in small molecule analysis for quantitation as well as qualitative discovery.4, 7, 8 Thus, FAIMS was used in combination with pseudo-SRM at unit resolution to validate a quantitative method for a peptide in rat serum according to good laboratory practices bioanalytical guidelines.9 As this method had already been validated without FAIMS, the novel FAIMS-pseudo-SRM approach was compared against the established assay with respect to linearity, accuracy, precision, lower limit of quantification and selectivity.
Chemicals and supplies
Peptide 1 and a 15N-labeled internal standard (IS) were synthesized in house by standard amino acid synthesis methods. Rat serum was purchased from Harlan-Winkelmann (Borchen, Germany). HPLC-grade acetonitrile (ACN), formic acid, trichloroacetic acid (TCA) and trifluoroacetic acid (TFA) were obtained from commercial sources and used without further purification. Industrial grade gases and deionized water were used throughout.
Stock solutions of peptide 1 and IS were prepared by accurately weighing the materials and dissolving in TCA 0.02% (m/m). Calibration (standard) and validation (quality control) samples were prepared by adding appropriate volumes of peptide 1 stock solution into control rat serum. All solutions were divided into aliquots and stored at or below −18°C until use. To each aliquot (100 µL) was added 20 µL IS working solution. Samples were processed by the addition of 120 µL of TCA (3.5%, m/m), by shaking on a laboratory shaker for 60 s, and by centrifugation at 12 000 rpm for 10 min. Finally, the supernatant was transferred into sample vials for injection onto the LC system.
At the onset of each measuring sequence a set of eight calibration samples at six non-zero concentrations was measured. The lower level of quantitation (LLOQ) and the upper level of quantitation (ULOQ) were processed in duplicate. The calibration standards were adequately distributed over the whole calibration range (10.0, 30.0, 100, 249, 1000 and 1998 ng/mL). Additionally, one matrix sample (blank) and one matrix sample fortified with IS (zero) was analyzed within every batch.
Validation samples were prepared at the following concentrations: the LLOQ (10.1 ng/mL), a low concentration (between the LLOQ and 2–3 times that of the LLOQ, 29.6 ng/mL), the mid-range concentration (between 30% and 60% of the ULOQ, 499 ng/mL) and the ULOQ (1995 ng/mL).
The LC system consisted of an 1100 series binary pump and degasser (Agilent Technologies, Santa Clara, CA, USA) and a HTC-PAL autosampler (CTC Analytics, Zwingen, Switzerland). The mobile phases were composed of A: 0.1% TFA in water, and B: 0.1% TFA in ACN, at a total flow rate of 0.2 mL/min. The mobile phase gradient proceeded according to the following scheme. After injection, mobile phase B was held at 5% for 1 min. Then, it was linearly increased to 60% over 6 min. After 7.5 min mobile phase B was increased to 80% and held for 2.5 min. Mobile phase B was then immediately reduced to 5% for column re-equilibration. The total run time was 12 min. Twenty microlitres of sample were injected. The column used was an Atlantis dC18 (30 × 2.1 mm, 3 µm; Waters Corporation, Milford, MA, USA).
The FAIMS system (Thermo Fisher Scientific, San Jose, CA, USA) used the recommended standard conditions of dispersion voltage (DV) −5 kV, 4 L/min equimolar nitrogen and helium, inner electrode temperature 70°C and outer electrode temperature 90°C. A reference solution of peptide 1 was infused at 5 µL/min and combined via a mixing tee with 50% mobile phase B flowing at 0.2 mL/min. The compensation voltage (CV) was ramped and the voltage at which the maximum signal appeared was the experimentally determined CV. The resulting CV for peptide 1 and IS was −35 V. Optimum response was achieved by ensuring the probe position was set to the furthest back possible (standard conditions). This provided for sufficient desolvation which in turn translated to a stable maximum signal.
The triple stage quadrupole (TSQ) mass spectrometer was a TSQ Quantum Ultra controlled by LCquan 2.5 software (Thermo Fisher Scientific, San Jose, CA, USA). Positive mode heated electrospray ionization (H-ESI) was used with the following parameters: ion spray voltage, 4000 V; vaporizer temperature, 400°C; tube lens offset, 228 V. Optimized pseudo-SRM conditions for peptide 1 included m/z 1277.3 and 1277.3 for Q1/Q3 masses and 25 eV collision energy. For the IS, the Q1/Q3 masses were 1292.7 and 1292.7, with a collision energy of 25 eV. Both mass-resolving quadrupoles were set to unit resolution (0.7 u FWHM). The m/z values correspond to the quintuply charged [M–H]5+ ions of peptide 1 and the IS. Collision gas pressure was set to 3.0 mTorr.
Integration of the chromatographic peaks was performed by LCquan 2.5. Concentration was determined by plotting peak area ratio vs. nominal concentration and back-calculating area ratios from the regression line using a weighting factor of 1/x2. All statistical calculations were performed using Microsoft Excel Office 2003 (Microsoft Corporation, Redmond, WA, USA). All concentrations are reported with three significant figures except for integer values and numbers greater or equal to 1000. Precision was assessed by the coefficient of variation (CV%) calculated from the measured concentrations. Accuracy (bias %) was calculated as the relative difference between the mean measured and the nominal concentrations.
Method validation criteria
Calibration samples were rejected from the calculation of the response function if the deviation from the nominal concentration exceeded 15% (20% at the LLOQ). At least one sample at the LLOQ and ULOQ had to meet this specification and could not be rejected. If more than 25% of the calibration samples failed then the analysis batch had to be rejected. The precision and accuracy of the method was evaluated after replicate analysis (n = 6 at each concentration level) of the validation samples. The precision had to be less than or equal to 15% (20% at the LLOQ) and the accuracy had to be within ±15% (±20% at the LLOQ).
Selectivity at the LLOQ
For the confirmation of the LLOQ the mean peak area of six zero samples (control rat serum of six different origins fortified with IS) was compared to the mean peak area found in six LLOQ validation samples prepared from the same source serum as the zero samples. The mean peak area of the zero samples should not exceed 20% of the mean LLOQ peak area.
RESULTS AND DISCUSSION
Attempts to fragment this multiply charged peptide and use an SRM transition with different m/z ratios, and thus fully utilize the selectivity of the triple quadrupole mass spectrometer, resulted in no significant single fragment being formed. Thus, LC/pseudo-SRM analysis resulted in chromatograms with abundant signal but numerous background peaks at the retention time of the peptide. The use of highly selective reaction monitoring (H-SRM), which is a high-resolution MS experiment, resulted in no improvement to the signal-to-noise (S/N) ratio. Figure 1 shows a representative chromatogram at the LLOQ. The LLOQ may seem rather high at 10 ng/mL, but for this peptide drug there was no need to develop the method to lower range. The unknown sample concentrations (therapeutic range) were expected to be within the limits used.
A peak was detected and integrated at approximately 6.6 min, but there were closely co-eluting interferences that adversely affected the quality of the integration. Likewise, the internal standard (IS) demonstrated a raised baseline indicating chemical background. The interferences appeared to be multiply charged species.
Despite these observations, the assay passed the initial validation criteria using this method as shown in Tables 1–3 (Without FAIMS results). However, during confirmation of selectivity at the LLOQ, serum demonstrated many interferences, as shown in Fig. 2 and quantified in Table 4 (Without FAIMS results). Blank samples demonstrated chromatographic peaks at the same retention time as the analyte, despite no analyte being added to the samples. According to bioanalytical guideline9 criteria, blank matrix samples should have less than 20% of the LLOQ analyte signal. By LC/MS alone, control serum appeared to have as much as 45% and on average approximately 35% of the LLOQ signal. The interference signal was not due to endogenous analyte because of the synthetic, exogenous nature of the peptide. It was for these reasons that FAIMS was explored as a solution to the interference problem.
Table 1. Calibration standard samples without and with FAIMS. There is essentially no difference between the %-deviation from theoretical
Nominal concentration (ng/mL)
Table 2. Calibration regression statistics without and with FAIMS. The slope of the regression line with FAIMS appeared higher than without FAIMS, suggesting greater detector sensitivity to changing concentrations. With FAIMS, a lower y-intercept confirms that fewer interferences obstructed accurate and precise quantitation. The coefficients of determination were similar for both analyses
Coefficient of determination (r2)
Table 3. Validation levels for quality control. Precision was comparable for both methods, but accuracy was better using FAIMS
Nominal concentration (ng/mL)
Table 4. Confirmation of LLOQ (10 ng/mL) experiment. Without FAIMS blank serum extracts showed greater than 20% of the LLOQ signal. Because FAIMS could remove the interferences, the remaining signal is only 7% of the LLOQ. This remaining signal could be attributed to unlabeled material present in the synthetically prepared IS or carryover
% of LLOQ
Figure 3 shows a representation of the FAIMS electrodes. The electrodes are shown in cut-away form to display the inner electrode and the analysis gap between the inner and outer electrodes. Ions generated by the H-ESI spray emitter enter the FAIMS device through the ion entrance and are analyzed in the mixture of helium and nitrogen between the electrodes. After analysis, ions with a stable trajectory leave the FAIMS device through the ion exit and enter the high-vacuum region of the mass spectrometer through the MS inlet. The FAIMS device acts as a filter because interfering ions do not have a stable trajectory through the device and are removed from the ion beam.10
Standard conditions were used as a starting point for investigating the behavior of peptide 1 and the IS in FAIMS. After 1 h of experimentation, the standard conditions were demonstrated to be the best and so were used without further change. In Fig. 4, representative LC/FAIMS-MS chromatograms demonstrated that the chemical background is greatly reduced versus the without FAIMS chromatograms (Fig. 1). Fewer co-eluting peaks appear in both the analyte and IS transitions. The S/N ratio was improved by 18 times at the LLOQ.
A visual indication that control blank chromatograms with FAIMS contained fewer interference peaks is shown in Fig. 5. Analyte traces are normalized to each other (Figs. 5(A) and 5(C)) as are IS traces (Figs. 5(B) and 5(D)). Figures 5(B) and 5(D) confirm that the IS response is essentially identical for the with and without FAIMS experiments, respectively. The interferences in Fig. 5(C) are removed when using FAIMS (Fig. 5(A)).
Despite the selectivity that FAIMS offers, there is still a trace response in the analyte channel (Fig. 5(A)). The resulting interference level was 7% of the LLOQ. This remaining response is likely due either to residual unlabeled peptide present in the synthetic preparation of the IS or due to autosampler carryover. Further reduction of this contribution might be achieved either by improved peptide synthesis and purification of the IS or by enhanced autosampler washing techniques.
Apart from this dramatic improvement in S/N the performance of the FAIMS assay was comparable or even better than that of LC/pseudo-SRM. The back-calculated concentrations of calibration samples with and without FAIMS resulted in essentially no difference in the %-deviation from the theoretical values, as shown in Table 1. Linear regression in Table 2 revealed a significantly higher slope with FAIMS (0.0109 with and 0.00699) than without FAIMS, respectively. This suggests that changing concentrations of analytes results in greater detector sensitivity when using FAIMS. With FAIMS, a lower y-intercept was obtained (0.0248 with and 0.0370 without FAIMS, respectively) confirming that fewer interferences obstructed accurate and precise quantitation. The coefficient of determination was approximately identical for both techniques.
Validation samples give a measure of assurance that the calibration samples have been prepared and extracted correctly. Table 3 shows that accuracy of these validation samples was better with FAIMS (7.03% difference) than without FAIMS (11.3% difference). The precision of the assay, as defined by the relative standard deviation (RSD) of these validation samples, was comparable for both with FAIMS (9.73% RSD) and without FAIMS (12.3% RSD).
One active area of interest in FAIMS research is the tendency to lower response during analysis.11 By comparing the IS area response with and without FAIMS, the modulation of response could be quantified. Within each experiment type the IS response was consistent. Table 5 demonstrates that FAIMS reduces the chromatographic peak area to approximately 90% of the non-FAIMS experiment similar to previously published results.4 This represents a change within the experimental error of the among-sample variation in the IS response.
Table 5. Chromatographic peak areas for peptide 1 calibration standards. At low concentration fewer interferences are transmitted into the mass spectrometer. At higher concentrations the more important effect is the focusing of ions which results in an increased response relative to the without FAIMS experiment
Area response of peptide 1
Area response of IS
A second interesting observation on the sensitivity of analysis with FAIMS is described in Table 5. The chromatographic peak area at the LLOQ (sample K10, 10 ng/mL) in the with FAIMS experiment is enhanced 12% relative to the without FAIMS experiment. At higher concentrations the chromatographic peak area further increases relative to non-FAIMS. This observation results in the 56% higher calibration line slope with FAIMS as calculated in Table 2.12
The IS response was consistent within each experiment type. A comparison of IS peak areas as performed in Table 5 demonstrates that FAIMS reduces the signal to approximately 90% of the non-FAIMS experiment. This result seems to contradict the response-enhancing effect of FAIMS for peptide 1. It could, however, be explained by the removal of co-eluting interfering signals lying underneath the internal standard peak by FAIMS.
Peptide quantitation according to GLP guidelines was achieved using LC/FAIMS-MS. Endogenous interferences which prevented validation in a traditional LC/MS assay were successfully removed. The resulting chromatographic cleanup was characterized by an 18-fold increase in chromatographic S/N ratio and an acceptable confirmation of the LLOQ experiment. FAIMS is a useful bioanalytical tool and an investment in developing robust, reproducible methods.