A Strategic Approach to the Quantification of Therapeutic Peptides in Biological Fluids

The growing market for biotherapeutic peptides and the development of quantitative methods for those analytes has brought to light the challenges facing the analysis of this broad range of compounds. Market forces and regulatory requirements are encouraging analytical groups to develop methodologies that are time- and cost-effective, while still producing assays that are sensitive enough to cope with biological matrices. This article discusses the development of a robust, fast, and generally applicable assay for a panel of therapeutic peptides that can be used in a real-world setting. Highly targeted, specific assays could be developed individually for each of these peptides, but an approach that looks at a diverse panel serves to examine the multiple factors that need to be considered in detail for assay development. A systematic approach to bioanalytical method development for therapeutic peptides is presented. A single sample preparation screening method and UHPLC starting conditions were identified and assessed using a set of 12 diverse peptide therapeutics. The generic sample preparation method, based on the selectivity of complementary mixed-mode SPE sorbents, resulted in greater than 82% recovery for nine out of the 12 peptides tested, and less than 10% matrix effects where measured. Minor modifications to the generic method resulted in recoveries greater than 82% for all peptides tested. The generic chromatographic method uses a column packed with 1.7-µm particles that have a pore size of 300 Å. The resulting peak widths were 2–4 s at the baseline, which provides a significant increase in signal-to-noise ratio over standard HPLC. The total chromatographic run time was 3.5 min, providing the high throughput required for a bioanalytical laboratory. Lower limits of quantitation (LLOQs) in the 5–10 pg/mL range were obtained for a subset of the test peptides using the combination of multiple reaction monitoring (MRM), selective mixed-mode solid-phase extraction, and a chromatographic screening method using highly efficient columns. Validation data are also presented. Standard curves were linear with 1/x weighting and QC sample concentrations were within ±15% of nominal concentration.

Biomolecules such as proteins and peptides represent a growing number of pharmaceutical drug entities in the development pipelines of major pharmaceutical companies. Therapies based on biomolecules are often more specific than those based on small molecules, because biomolecules typically replace or augment an endogenous molecule known to effect a disease or physiological state (1). Furthermore, because they are analogous to or are endogenous compounds, they tend to be well tolerated by the body, minimizing adverse reactions and maximizing potency. Advances in the various technologies applied to drug discovery and biomolecule characterization (that is, recombinant DNA, fermentation, proteomics, genomics, and informatics) (2) have made it possible for drug manufacturers to develop and characterize biopharmaceuticals successfully. These types of compounds have been used to treat a variety of serious diseases such as diabetes, cancer, arthritis, and hemophilia.

One such example is the synthetic peptide desmopressin, a modified form of the human hormone arginine vasopressin. It is prescribed as a replacement for antidiuretic hormone (vasopressin) and is used to treat bedwetting and diabetes insipidus. Desmopressin provides several treatment benefits over recombinant vasopressin. It degrades more slowly, enabling less frequent dosing, and it does not raise blood pressure as the unmodified peptide does.

With more than 40 peptide drugs already on the market, and more than 400 in late-stage clinical trials (3), there exists an immediate need for an efficient method development workflow for bioanalysis of peptide therapeutics. Quantification of peptides is important not only for synthetic peptide drugs, but also for peptide biomarkers and for quantification of proteins based on measurement of unique or signature peptides. Historically, quantification of biologically based drugs, or biopharmaceuticals, has been performed by ligand binding assays (LBAs). In contrast, liquid chromatography (LC) coupled to triple-quadrupole mass spectrometry (MS) is the platform that dominates most bioanalytical laboratories, and is recognized widely as the technique of choice for drug quantification of small molecules. Recently, there has been a move away from LBAs and toward LC–MS-MS for the quantification of biopharmaceuticals as well as small molecules. This is because LC–MS-MS offers several distinct advantages over LBAs in general:

• Broader linear dynamic range — often 4 or 5 orders of magnitude (triple quadrupole systems) compared to 1.5–2 orders for LBAs.

• Higher degree of accuracy and precision when compared to LBAs.

• Reduced method development times, often measured in weeks for LC–MS-MS, where LBAs may take months to develop. In addition, LBAs are subject to cross-reactivity.

The shift in therapies to a greater proportion of biological therapeutics also means that many traditional "small-molecule" laboratories now find themselves faced with the task of developing bioanalytical methods for peptide therapeutics. An LC–MS-MS approach provides numerous benefits. The development of bioanalytical methods even for the detection of small molecule oral pharmaceuticals in humans and animals is already a challenging process. This same type of method now needs to be developed for protein and peptide pharmaceuticals. Working with complex biological matrices and meeting rigorous regulatory criteria are just a few of the hurdles one must overcome. Naturally, methods also need to be acceptable in terms of linearity, sensitivity, accuracy and precision, selectivity, stability, and carryover.

A review of almost 200 articles in the literature (4) shows that many different combinations of LC, MS, and sample preparation conditions have been used for bioanalysis of peptide therapeutics, making it challenging to identify a common starting point for method development. In addition, many of the published references use nonselective sample preparation methods, such as protein precipitation (PPT) and reversed-phase solid-phase extraction (SPE). However, the use of less-selective sample clean-up may necessitate the use of longer chromatographic runs to eliminate coelutions of endogenous materials with the analyte. In particular, challenges related to peptide handling represent one of the most difficult aspects of method development.

In this article, we will address the three main portions of a bioanalytical method: LC, MS, and sample preparation. To evaluate the feasibility and utility of a systematic approach to peptide bioanalysis, 12 peptide therapeutics were selected for analysis. The peptides chosen were diverse, having a broad range of molecular weights, acidity–basicity, and hydrophobicity. Details of each peptide, along with their chemical properties, are listed in Table I. A high performance liquid chromatography (HPLC) index is used here as a measure of relative hydrophobicity. A low value indicates a more polar peptide, whereas a high value indicates a more hydrophobic peptide.

Table I: Chemical properties of therapeutic test peptides

We present methodically developed LC and sample preparation screening methods and the results obtained to demonstrate the utility of this approach for bioanalytical method development for peptide therapeutics.

Experimental

Reagents: Human plasma was purchased from Lampire Biologicals (Pipersville, Pennsylvania). Water for mobile phase and sample preparation was obtained from a Milli-Q laboratory water system, (Millipore, Billerica, Massachusetts). Acetonitrile, methanol, and ammonium hydroxide (concentrated solution, 28–30%) were purchased from Fisher Scientific (Fair Lawn, New Jersey). Formic acid (>98%) was purchased from Acros Organics (Morris Plains, New Jersey). Enfuvirtide, teriparatide, and bivalirudin were purchased from ChemPep, Inc. (Wellington, Florida). Octreotide, desmopressin, vasopressin, neurotensin, angiotensin I and II, brain natriuretic peptide (BNP), goserelin, and somatostatin were purchased from Sigma-Aldrich (St. Louis, Missouri).

Sample Preparation: Protein precipitation (PPT) was performed using acetonitrile–human plasma in a 3:1 ratio. Samples were vortex-mixed and the supernatant was removed. Liquid–liquid extraction (LLE) was performed using ethyl acetate–plasma in a 5:1 ratio. Samples were vortex-mixed and centrifuged, and the appropriate layer was removed. Reversed-phase SPE was performed with a Waters Oasis HLB 96-well µElution plate (Milford, Massachusetts), using the generic method provided by the manufacturer. Initial mixed-mode SPE studies were performed using Oasis MCX (strong cation exchanger), WCX (weak cation exchanger), MAX (strong anion exchanger), and WAX (weak anion exchanger) in µElution (reduced sorbent bed) format using the manufacturer's generic protocols. For determination of limit of detection (LOD) and lower limit of quantification (LLOQ), stock solutions (1 mg/mL) of desmopressin, angiotensin I, and angiotensin II were prepared by dissolving the appropriate amount of compound in water. Aliquots were stored at –80 °C and freshly thawed each day for preparation of the working solutions. Working solutions of each peptide were then prepared by diluting the appropriate volume of stock solution with water. These solutions were used then to spike human plasma for determination of LOD and LLOQ. Spiked samples were prepared at the following concentrations: 0.001, 0.005, 0.01, 0.02, 0.05, and 0.1 ng/mL (on average, approximately 1 to 100 fmol/mL).

Liquid Chromatography: The peptide separations were performed on an Acquity UHPLC system with an 50 mm × 2.1 mm, 1.7-µm Acquity UHPLC BEH300 C18 column. During method development, a 50 mm × 2.1 mm, 3.5-µm XBridge BEH C18 column and a 50 mm × 2.1 mm, 1.7-µm Acquity UHPLC BEH130 C18 column were used. Mobile phase A was 0.1% formic acid in water, and mobile phase B was acetonitrile. The flow rate was 0.4 mL/min. The linear gradient was 15–75% B over 2 min, followed by a 0.1-min ramp to 98% B, which was held for 1 min before returning to initial conditions. The total cycle time was 3.5 min. Injection volumes ranged from 10 to 35 µL and samples were injected using partial loop mode. Column temperature was 35 °C and samples were maintained at 15 °C in the autosampler. The strong needle wash was 60:40 acetonitrile–isopropanol plus 0.5% formic acid (600 µL) and the weak needle wash was 95:5 water–methanol (400 µL.)

MSMS: Peptides were detected using either a Waters Quattro Premier, Waters Xevo TQ MS, or Waters TQ-S triple-quadrupole MS system operating in positive ion electrospray mode. The desolvation temperature was typically 450 °C and the desolvation gas flow was 800–1000 L/h. The collision cell pressure was approximately 2.6 × 10^-3 mbar.

Results and Discussion

SPE Method Development: In anticipation of developing a single SPE screening method for a wide range of peptides, traditional sample preparation methods (PPT, LLE, and reversed-phase SPE) were performed first using two of the 12 peptides (desmopressin and bivalirudin) to assess analyte recovery and matrix effects. Equation 1 was used to calculate the recovery of each peptide from human plasma:

% SPE recovery = (average peak area in prespiked extracted samples/average peak area in post-spiked samples) × 100 [1]

Matrix effects (ME) were calculated according to equation 2:

Results are shown in Figure 1. As the data show, none of the techniques provided both high recovery and low matrix effects. Mixed-mode SPE has been shown to reduce matrix effects to a greater extent than other sample preparation techniques while still providing high analyte recovery (5). For this reason, it was evaluated for peptide extractions. Because of their zwitterionic nature, the behavior of peptide therapeutics under various sample preparation conditions can be difficult to predict. Therefore, initial proof-of-concept studies were performed on four different mixed-mode sorbents, using a generic set of conditions originally developed for small molecule screening. Each sorbent consists of a moiety that gives reversed-phase behavior as well as an ion exchange group for additional selectivity. Recovery was calculated for both the Elute 1 fraction (100% organic, containing analyte bound by reversed-phase) and the Elute 2 fraction (containing analyte bound by ion-exchange) on all four of the mixed-mode sorbents. Recovery for test peptides was split between the two elutions and was on average <60%. It was clear that new conditions were needed if the advantages of mixed-mode SPE were going to be applied successfully for peptide extraction. Additional experimentation and further examination of the data indicated that better results might be obtained more readily on either strong anion exchange or weak cation exchange than strong cation or weak anion exchange media (data not shown). Many changes were made to the original protocols, including optimization of wash and elution solutions, to generate a protocol developed specifically for peptides.

Figure 1: Extraction recovery and matrix effects for an acidic and basic peptide using traditional sample preparation techniques.

Final Peptide Extraction Screening Protocol: The final extraction protocol is as follows:

Oasis WCX and Oasis MAX µElution 96-well format

Figure 2: SPE recovery for 12 peptide therapeutics using a mixed-mode screening method optimized for peptides and with minor modifications.

1. Condition the wells with 200 µL methanol.

2. Equilibrate with 200 µL water.

3. Load 700 µL of diluted, acidified plasma sample.

4. Wash with 200 µL of 5% ammonium hydroxide in water.

5. Wash with 200 µL 20% acetonitrile in water.

6. Elute with 1 or 2 × 25 µL 1% trifluoroacetic acid in 75:25 acetonitrile–water.

7. Dilute with 25 or 50 µL of water if necessary.

Recovery for the 12 peptides tested using mixed-mode SPE optimized for peptides is summarized in Figure 2. Recovery for nine out of the 12 peptides was acceptable on a first pass using the screening method, clearly indicating that a single SPE platform could be used successfully to initiate method development. Minor modifications to the method resulted in improved recovery for the remaining three peptides. Final SPE recovery and matrix effect values are summarized in Table II.

Table II: Final SPE recovery and matrix effect values using SPE and sub-2-µm LC screening methods optimized for peptides

Liquid Chromatography: Sub-2-µm porous particle LC, specifically ultrahigh performance LC (UHPLC) has been used extensively in bioanalytical applications for the benefits it provides in terms of resolution, sensitivity, and speed (6–15). As valuable as these characteristics are to small-molecule analysis, they can be even more critical to successful peptide analysis. Peptide drugs are often modified versions of substances already in the body, meaning that there almost always will be a closely related interference present in the sample. The drug desmopressin and the endogenous hormone vasopressin upon which it is based are good examples. Desmopressin differs from human vasopressin by the loss of an amino group. Thus, the resolving power of the chromatography system used for peptide separations is critical, and sub-2-µm LC has demonstrated improvements in resolution and detection limits that are significantly better than conventional HPLC.

Figure 3: Van Deemter plot using flow rate for compounds of increasing molecular weight on a 2.1-mm i.d., 3.5-µm particle size column. Assumptions are as follows: A, B, and C terms are equal to 1.5, 0.5, and 0.1666, respectively.

A van Deemter plot was constructed using flow rate instead of linear velocity to demonstrate how plate height for large molecules degrades much more rapidly than for small molecules at higher flow rates (Figure 3). Although small molecules can be chromatographed at much higher flow rates without a significant loss in column efficiency, larger molecules like peptides must be chromatographed at lower flow rates to achieve maximum performance. This is primarily because of the lower diffusion rates of peptides in and out of the pores in the stationary phase. However, for high throughput applications such as bioanalysis, it is not practical to use the very low flow rates needed for peptide separations.

To compensate for the slower diffusion of peptides, smaller particle sizes are used. Figure 4 shows calculated van Deemter plots for a peptide analyzed on 2.1-mm diameter columns packed with 1.7- and 3.5-µm particles. Although the optimum flow rate for this peptide is similar for both particles (~25–50 µL/min), the separation performance of the 3.5-µm particle column degrades more rapidly than the 1.7-µm particle column as the flow rate increases. At 0.4 mL/min, the flow rate used for separation of the 12 therapeutic peptides in this study, the column packed with 3.5-µm particles has a 5X increase in plate height compared to its optimum flow rate, whereas the column packed with 1.7-µm particles only exhibits a 2X increase in plate height. This clearly shows that the use of smaller particles is preferred for high-throughput bioanalysis of peptides.

Figure 4: Van Deemter plot using flow rate for a model 2500 MW peptide on 1.7-µm and 3.5-µm particle size, 2.1-mm i.d. columns. Assumptions are as follows: A, B, and C terms are 1.5, 0.5, and 0.1666, respectively, and the temperature used was 35 °C.

From a practical standpoint, however, peptides are typically analyzed with gradient rather than isocratic methods to reduce analysis times. To illustrate the benefit of small particles for peptide separations, 3.5-µm and 1.7-µm columns packed with the same stationary phase were compared (Figure 5). These columns have the same base particle and differ only in particle size. Both were run on an Acquity UHPLC system using the same flow rate and gradient. The peptides run on the 1.7-µm particle column consistently show sharper, more efficient peaks, which translates into higher signal-to-noise ratios and the ability to achieve lower limits of detection. These data correlate well with previous findings by Gilar and colleagues (16), who demonstrated a marked increase in peak capacity for peptides using 1.8-µm particles compared to either 3.5- or 5-µm particles.

Figure 5: Influence of particle size on S/N: 1.7- versus 3.5-μm columns.

Another parameter of the chromatographic column to consider in peptide separations is the pore size. No concrete rule currently exists as to which pore size to use for peptides of a particular size, which means columns packed with particles of different pore sizes need to be tested to determine the effect on peak shape. Figure 6 shows a comparison of two columns packed with 1.7-µm particles but having two different pore sizes. For a majority of the peptides used in this study, using a column with a larger pore size gives superior peak shape, and results in better signal-to-noise ratios and lower detection limits. These data might also suggest that larger pores may mitigate some of the loss in efficiency observed for peptides at higher flow rates.

Figure 6: The effect of chromatographic pore size on teriparatide, a 4118 MW peptide.

As a result of these data, a C18 column with 300 Å pores (50 mm × 2.1 mm) packed with 1.7-µm particles was chosen for peptide screening. A gradient from 15% to 75% organic was used because it brackets the typical elution window for peptides. The 0.4-mL/min flow rate used correlates well to that of previously published findings (17). A representative separation of five of the therapeutic peptides using this generic LC method is shown in Figure 7. It is important to note that vasopressin (peak 1) and desmopressin (peak 3) are baseline-resolved easily using the screening method.

Figure 7: Representative separation of five therapeutic peptides using the LC screening method optimized for peptide bioanalysis.

MS: The overall MS signal for peptides is often lower than for small molecules for several reasons. First, peptides are multiply charged species whereas small molecules are typically singly charged. There may be several different multiply charged precursors present, diluting the overall ion intensity across several species. Furthermore, peptides tend to form many low abundance fragments rather than one or two intense ones, reducing overall signal for multiple reaction monitoring (MRM) experiments. An even greater loss of signal can be observed for large peptides that are not transferred as easily into the gas phase during ionization as small molecules are. It may be advantageous to sum transitions to either improve signal intensity or reduce variability if the relative abundance of a specific precursor changes during the analysis. Another important consideration for the MS detection of peptides is instrument mass range, both in the first and second quadrupoles. For example, the MS infusion of enfuvirtide showed a dominant 3+ precursor at approximately m/z 1498, requiring an instrument with a mass range of at least 1500. Similarly, MS-MS of the 2+ precursor of bivalirudin at m/z 1091 produced two major singly charged fragments at m/z 650 and m/z 1531, again demonstrating the need for adequate mass range. In this case a mass range of >2000 amu is desirable.

Figure 8: LLOQ for desmopressin extracted from 300 µL of human plasma using the sub-2-µm LC and SPE screening methods.

Limit of Detection, Lower Limit of Quantification, and Partial Validation: To demonstrate the capability of the entire bioanalytical method using generic sample preparation and LC conditions, detection limits in extracted human plasma were determined for a subset of the test peptides. Blank human plasma and samples prepared at 0.001, 0.005, 0.01, 0.02, 0.05, and 0.1 ng/mL (on average, approximately 1–100 fmol/mL) were extracted according to the generic screening method and the chromatographic results were evaluated to determine LOD and LLOQ. The resulting chromatograms are shown in Figures 8–10. Partial validation was performed for desmopressin by preparing standard curves and quality control (QC) samples in human plasma. The assay was run on a higher sensitivity triple-quadrupole MS system, and even lower detection limits were achieved. Results of the partial validation are summarized in Table III. Standard curves were linear with 1/x weighting from 0.001 to 20 ng/mL using the updated MS platform and QC samples easily met the accuracy and precision criteria defined in the FDA Guidance for Industry for Bioanalytical Method Validation (18). The lowest QC sample and standard curve point were within 20% of expected and all other QC samples and curve points were within 15% of expected.

Table III: Partial validation data for the analysis of desmopressin in human plasma

Conclusions

An SPE–LC–MS-MS approach was developed for the analysis of peptide therapeutics in human plasma. A single SPE screening protocol, based on two mixed-mode SPE sorbents, resulted in recoveries that exceeded 82% for nine out of 12 diverse peptides extracted from a complex matrix. Minor modifications to the protocol improved recovery to more than 85% for the remaining three peptides. Matrix effects were used as one indicator of the selectivity of the sample preparation method, and were less than 10% where measured. A simple LC screening method, based on sub-2-µm particle columns with a 300-Å pore size, provided both the resolution and speed required for high throughput bioanalysis of peptides. The combination of the generic SPE and sub-2-µm LC conditions, as well as a high sensitivity triple-quadrupole mass spectrometer, resulted in detection limits in the low picograms-per-milliliter range. A representative method for desmopressin was subjected to partial validation and easily met the regulatory criteria for accuracy and precision.

Figure 9: LLOQ for angiotensin II extracted from 350 µL of human plasma using the sub-2-µm LC and SPE screening methods.

Overall, we have shown that bioanalysis studies for peptide therapeutics are amenable to a platform-based approach to methods development. Such standardized approaches for determining optimal SPE enrichment and MRM-based LC–MS analysis should permit companies to reduce development timelines and shorten time-to-market for peptide drugs.

Figure 10: LLOQ and LOD for angiotensin I extracted from 350 µL of human plasma using the sub-2-µm LC and SPE screening methods.

Erin E. Chambers, Kenneth J. Fountain, and Diane M. Diehl are with Waters Corporation, Milford, Massachusetts.

References

(1) C. Bladon, Pharmaceutical Chemistry: Therapeutic Aspects of Biomolecules (Wiley and Sons, West Sussex, UK, 2002).

(2) R. Ho and M. Gibaldi, Biotechnology and Biopharmaceuticals (Wiley and Sons, New Jersey, 2003).

(3) Bioscience Technology, January (2009).

(4) I. Van den Broek, R. Sparidans, J. Schellens, and J. Beijnen, J. Chromatogr. B 872, 1–22 (2008).

(5) E. Chambers, D.M. Wagrowski-Diehl, Z. Lu, and J. Mazzeo, J. Chromatogr. B 852 (1–2), 22–34 (2007).

(6) M. Jemal, Z. Ouyang, and Y.Q. Xia, Biomaed. Chromatogr. 24, 2–19 (2010).

(7) L. Discenza, C. D'Arienzo, T. Olah, and M. Jemal, J. Chromatogr. B 878, 1583–1589 (2010).

(8) R. Plumb, P. Rainville, W. Potts, K. Johnson, E. Gika, and I. Wilson, J. Proteome Res, 8, 2495–2500 (2009).

(9) S. Pedraglio , M.G. Rozio, P. Misiano, V. Reali, G. Dondio, and C. Bigogno, J Pharm. Biomed. Anal. 44, 665–673 (2007).

(10) R. Kay, B. Gregory, P. Grace, and S. Pleasance, Rapid communications in Mass Spectrometry 21, 2585–2593 (2007).

(11) J. Shen, H. Wang, S. Tadros, and R. Hayes, J. Pharm. Biomed. Anal. 40, 689–706 (2006).

(12) G. Mitulovic et al., Anal. Chem. 81, 5955–5960 (2009).

(13) R. Kay et al., Rapid Communications in Mass Spectrometry 23, 3173–3182 (2009).

(14) R. Denooz, M. Mercerolle, G. Lachatre, and C. Charlier, J. Anal. Toxicol. 35, 280–286 (2010).

(15) P. de Bruijn, S. Sleifer, L. Lam, E. Wiemar, and W. Loos, J. Pharm. Biomed. Anal. 51, 934–941 (2010).

(16) M. Gilar et al., J. Sep. Sci. 28, 1694–1703 (2005).

(17) M. Gilar, A. Daly, M. Kele, U. Neue, and J. Gebler, J. Chromatogr. A 1061, 183–192 (2004).

(18) Guidance for Industry: Bioanalytical Method Validation, FDA CDER and CVM, May (2001).

(19) M. Gilar and U. Neue, Journal of Chromatography A 1069, 139–150 (2007).

(20) U. Neue, HPLC Columns Theory, Technology, and Practice, Wiley-VHC (1997).

(21) I. Van den Broek, R. Sparidans, A. Huitema, J. Schellens, and J. Beijnen, J. Chromatogr. B 837, 49–58 (2006).

(22) Y. Jiang et al., Rapid communications in Mass Spectrometry 21, 3982–3986 (2007).

(23) L. Du, C. Miller-Stein, R. Valesky, A. Martucci, and E. Woolf, J. Chromatogr. B 877, 1465–1471 (2009).