Methods xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

Methods journal homepage: www.elsevier.com/locate/ymeth

Review Article

Clinical protein mass spectrometry Alexander Scherl ⇑ Department of Human Protein Science, Faculty of Medicine, University of Geneva, Geneva, Switzerland Department of Genetic and Laboratory Medicine, Geneva University Hospitals, Geneva, Switzerland

a r t i c l e

i n f o

Article history: Received 26 November 2014 Received in revised form 24 February 2015 Accepted 25 February 2015 Available online xxxx Keywords: Mass spectrometry Clinical diagnostics Proteins Top-down proteomics

a b s t r a c t Quantitative protein analysis is routinely performed in clinical chemistry laboratories for diagnosis, therapeutic monitoring, and prognosis. Today, protein assays are mostly performed either with nonspecific detection methods or immunoassays. Mass spectrometry (MS) is a very specific analytical method potentially very well suited for clinical laboratories. Its unique advantage relies in the high specificity of the detection. Any protein sequence variant, the presence of a post-translational modification or degradation will differ in mass and structure, and these differences will appear in the mass spectrum of the protein. On the other hand, protein MS is a relatively young technique, demanding specialized personnel and expensive instrumentation. Many scientists and opinion leaders predict MS to replace immunoassays for routine protein analysis, but there are only few protein MS applications routinely used in clinical chemistry laboratories today. The present review consists of a didactical introduction summarizing the pros and cons of MS assays compared to immunoassays, the different instrumentations, and various MS protein assays that have been proposed and/or are used in clinical laboratories. An important distinction is made between full length protein analysis (top-down method) and peptide analysis after enzymatic digestion of the proteins (bottom-up method) and its implication for the protein assay. The document ends with an outlook on what type of analyses could be used in the future, and for what type of applications MS has a clear advantage compared to immunoassays. Ó 2015 Elsevier Inc. All rights reserved.

1. Introduction Proteins and peptides are essential biomolecules, composed of individual amino acid residues linked together via peptidic bonds. Their biological functions are multiple: structural and mechanical roles, responses to stimuli, gene replication, chemical transporters, catalyzing metabolic reactions, and many others. Proteins are also used as drugs for the treatment of diseases. Typical examples are insulin, growth hormones, and therapeutic antibodies. Protein analysis is thus an important activity in clinical laboratories. Nonspecific detection methods such as chromatographic or electrophoretic separation followed by UV detection or staining are nowadays replaced by specific detection methods. These methods include indirect antibody-based immunoassays such as enzymelinked immunosorbent assay (ELISA) and, more recently, mass spectrometry (MS). For many years, specific quantitative protein assays were performed almost exclusively using immunoassays. But rapid progress of protein MS has taken this method into the spotlight. There is much discussion in the field about the advantage

of MS compared to immunoassays, including reagent cost, specificity, throughput, and the possibility to multiplex assays [1–7]. The following report will first discuss the principles, advantages and inconveniences of mass spectrometric assays compared to immunoassays from the point of view of a clinical laboratory. In opposition to other recently published reviews [7–10], an important distinction is made between top-down and bottom-up MS assays, where respectively the full length protein or only a short specific signature peptide after it’s enzymatic digestion is analyzed. Second, the few most used or most promising clinical MS assays for peptides and proteins will be reviewed. The primary focus of this review is to highlight their limits and advantages compared to immunoassays, not to provide an exhaustive list of what has be proposed by the community. The two last sections will focus on the challenges to be overcome to implement more MS assays in clinical laboratories, and an outlook on what type of assays might be used in the future.

2. Immunoassays ⇑ Address: Department of Human Protein Science, Faculty of Medicine, University of Geneva, Geneva, Switzerland. E-mail address: [email protected]

Immunoassays use the specificity of an antibody to detect an antigen, for example a peptide or a protein. The concentration of

http://dx.doi.org/10.1016/j.ymeth.2015.02.015 1046-2023/Ó 2015 Elsevier Inc. All rights reserved.

Please cite this article in press as: A. Scherl, Methods (2015), http://dx.doi.org/10.1016/j.ymeth.2015.02.015

2

A. Scherl / Methods xxx (2015) xxx–xxx

antigens is then deduced from the number of antibodies bound to antigens, via a signal amplification technique such as a radiolabel, a fluorescent label, or an enzymatic conversion of a reagent to a detectable product. A typical immunoassay assay is summarized in Fig. 1. Immunoassays are today clearly the most common quantitative protein assays in clinical laboratories. They are very sensitive, and protein concentrations in the ng/ml range can be detected with good specificity and linearity, if an antibody with the desired performances is available. The assays are performed in solution, making automation easy with the help of liquid handling robots. At the department of genetics and laboratory medicine of our institution, a university hospital serving a population base of one million habitants, about 320,000 protein immunoassays are performed each years, to measure 72 different parameters. However, it should be noted that the most common protein assays are enzymatic tests, with 470,000 analyses/year and 60 different parameters. Far behind are colorimetric assays, with 77,000 tests on 13 different parameters. No protein mass spectrometry assay is currently used on a routine basis at our department. Although very common, immunoassays are not perfect, and some of them suffer from limitations. Antibodies recognize a polypeptide epitope of typically eight to seventeen amino acids in length. Thus, an antibody might recognize a common portion of a protein present in different proteoforms1 [11] without being able to differentiate between them. In some cases, a distinction of different proteoforms is impossible or simply not known. Antibody specificity for a particular isoform is suited for clinical data interpretation, but is not always possible. Such interferences are for example described for antibodies against the glycated hemoglobin form HBA1c, interacting also with the carbamylated form of hemoglobin [12]. Autoantibodies, i.e. antibodies expressed by a subject against the analyte can also result in false reporting. When such autoantibodies are present, the analyte of interest is captured by them and does not bind to the assay’s antibodies. The measured analyte concentration is thus below the real physiological concentration. The presence of autoantibodies is known against thyroglobulin, a protein routinely measured for the therapeutic follow-up of thyroid cancer [13]. Other examples where autoantibodies interact with proteins are described, and includes mucin-1, PSA, TSH, CRP, troponin I and insulin [4]. Nonspecific interactions between antibodies and other compounds, typically other antibodies present in the patient’s serum, could also be a source of errors. In addition, saturation and limited dynamic range is also a limiting factor for immunoassay. The titer of antibodies linked on the solid surface for the test has to be in the same order of magnitude as the number of analyte molecules binding to it. If the concentration of analyte molecules is higher, the excess will be washed away. This will, in turn, result with an incorrect concentration value. This saturation effect is also called the ‘‘hook’’ effect due to its characteristic shape when the detection signal is plotted against the analyte concentration. The different causes resulting in incorrect test results for immunoassays are summarized in Fig. 2.

3. Mass spectrometry Mass spectrometry (MS) is an analytical method that detects analyte ions in the gas phase. The detection is specific to the mass over charge ratio (m/z) of the ions. In an MS experiment, the analyte is first ionized, i.e. converted into gas-phase ions. Many different ionization methods have been described, including electron impact, fast atom bombardment, chemical ionization, 1 Different proteoforms could be for example different cleavage products, different splice variants, different isoforms or the presence or absence of post-translational modification(s) [11].

photoionization, matrix-assisted laser desorption/ionization (MALDI) and electrospray ionization (ESI). The latter two are particularly soft ionization techniques, i.e. fragmentation of the analyte during the ionization process is very limited. They are thus particularly useful to convert large biomolecules such as DNA and proteins into molecular ions. The inventors of these soft ionization techniques were recognized by the 2002 Nobel Prize [http://www.nobelprize.org/nobel_prizes/chemistry/laureates/2002/]. With MALDI, the analyte is co-crystalized in a matrix, usually an organic acid, and introduced in solid form into the mass spectrometer. A laser pulse is then used to desorb and ionize the sample. The energy of the laser is absorbed by the matrix, resulting in a rapid gas-phase expansion. During this expansion, charges are transferred from the matrix to the analyte molecules, resulting with gas-phase ions (Fig. 3A). With ESI, the liquid analyte solution is directly transformed from the liquid phase into the gas phase. For this, a high potential difference is applied between the tip from which the analyte solution elutes and the mass spectrometer inlet. ESI can be hyphenated to a liquid chromatograph in a process called LC–MS. ESI is widely used in routine and research laboratories to analyze various analytes in complex matrixes, such as biofluids or complex mixtures (Fig. 3B). The exact mechanisms of MALDI and ESI ionization is still under debate, and the subject is extensively reviewed [14,15]. Once into the gas phase, the ions are directed toward a mass selective analyzer and detector, and their m/z value is precisely determined. A mass spectrum represents the ion abundance as function of the m/z value. The relative molecular mass (m) is expressed in unified atomic mass units [u] or Daltons [Da], defined as 1/12th of a 12C atom. The charge (z) is the number of elementary charges. Sometimes, the Thompson [Th] is used as mass to charge ratio unit [16]. If the number of charges of an ion is known, for example by measuring the m/z distance between different isotopes in isotopic clusters, the m/z value can be directly converted into mass units. Different types of mass analyzers can be used. Quadrupoles and ion traps are mostly used today in clinical laboratories. Also popular are time-of-flight, ion cyclotron and orbitrap analyzers, but these high-end instruments are mostly used for research applications. Quadrupoles (Q) consist of four parallel rods (electrodes), diagonally electrically connected. A radiofrequency (RF) is applied between them, and a direct current is superimposed on it (RF + DC). The electric field generated focuses the ions into the axial center of the quadrupoles. For each electric potential and radiofrequency amplitude pair, only ions at unique m/z ratio have a stable trajectory and pass across the quadrupole. Ions at other m/ z values hit the electrodes [17]. The quadrupole can thus be used as mass filter to selectively transmit ions at a particular m/z value, or be used as scanning device to acquire a mass spectrum, i.e. a plot of m/z values and their relative abundances. Three-dimensional ion trapping devices (3D-IT) are operating on the same principle, but consist of a ring electrode and two endcap electrodes [17]. They can be used as ion storage and ion manipulating devices, and a mass spectrum is acquired when they are ejected out of the ion trap according to their m/z value [18]. Quadrupoles can also operate as ion trapping devices if an electrode at each end is used to create a potential well in the quadrupole. Such devices are called linear ion traps [19,20]. Time-of-flight (TOF) analyzers are fieldfree tubes that ions travel through after an initial acceleration using an electrical field pulse. All ions are accelerated with the same kinetic energy, and their flight time toward the detector is measured and converted into an m/z value. Ions with low m/z value will travel at higher velocity and reach the detector first [21,22]. Ion cyclotrons use a high magnetic field. The ions are accelerated orthogonally to the field, which induces a cyclotronic motion. The image current of this cyclotronic motion is recorded on

Please cite this article in press as: A. Scherl, Methods (2015), http://dx.doi.org/10.1016/j.ymeth.2015.02.015

A. Scherl / Methods xxx (2015) xxx–xxx

3

Fig. 1. Immunoassay. A modern antibody-based assay consists of a surface coated with a specific capture antibody (A). As the sample is added, the antigens present binds specifically to the antibody (B). Free antibodies are added, binding again to the antigens (C). These antibodies are then detected via a secondary antibody directed to the Fc domain of the specific antigen-bound antibodies (D). The secondary antibodies are linked to an enzyme converting a substrate to a detectable product, or linked to any other type of detectable label such as a fluorophore or a metal nanoparticle.

Fig. 2. Sources of errors in immunoassays. A similar protein epitope can bind to the antibody (A). The presence of autoantibodies in the sample against the analyte can prevent binding (B). Non-specific interactions, typically with other antibodies present in the sample, can result with binding of the secondary antibody (C). In case of high analyte concentration, the antibodies present on the surface can be saturated (D).

Fig. 3. MALDI and ESI ionization. For MALDI, the analyte is co-crystalized with solid matrix molecules. A laser pulse hits the co-crystal, resulting with the rapid expansion of the analyte and matrix into gas phase. During this expansion, charges are transferred from the matrix molecules to the analyte molecules (A). For ESI, a high potential difference is applied between the analyte solution and the mass spectrometer inlet. As the analyte solution exits from the emitter tip, charged droplets are created. As the solvent evaporates, analyte ions are ejected from the droplets (ion evaporation model) or charges stay on the analyte (charge-residue model). This results with gas-phase analyte ions. ESI can be hyphenated with a liquid chromatograph (LC–MS). The eluting analyte molecules are ionized and analyzed by MS (B).

detector plates, and transformed into frequency values using Fourier-transformation, and subsequently converted into m/z values. These instruments are thus called Fourier-transform-ion cyclotron resonance (FT-ICR) mass spectrometers [23]. Orbitap (OT) instruments are operating on a similar principle of image current detection and Fourier-transformation. However, the oscillating axial frequency of ions spinning around a central electrode in a quadro-logarithmic field is recorded [24,25]. The different mass analyzers are summarized in Fig. 4.

All mass analyzers have different performance characteristics. Their price difference can be very important as well. Main performance metrics are resolution and mass accuracy. The resolution is the ability to distinguish two peaks very close to each other. Resolution is expressed as the ratio of the m/z value of a compound divided by the full width at half maximum (FWHM) of the peak. Resolution is thus a ratio, without units. Mass accuracy is expressed as the difference between the theoretical mass of a compound and the measured mass, expressed either in absolute m/z

Please cite this article in press as: A. Scherl, Methods (2015), http://dx.doi.org/10.1016/j.ymeth.2015.02.015

4

A. Scherl / Methods xxx (2015) xxx–xxx

Fig. 4. Mass analyzers. In a quadrupole, only ions with a particular m/z ratio value have a stable trajectory and hit the detector (A). Similarly to the quadrupole, ions can be stored in a 3D-ion trap or ejected in an m/z specific manner toward the detector (B). In a linear ion trap, an axial potential well is formed to keep ions within the quadrupole. Similarly to the ion trap, they are ejected in an m/z specific manner (C). A pulsed electric field is applied at the beginning of the flight tube in a time of flight analyzer. All ions receive the same kinetic energy, and low m/z ions hit the detector first. Their m/z value is calculated from the time of flight value (D). In an ion cyclotron resonance cell, the ions are placed in a magnetic field. A potential is added to the exciting plates and induces a cyclotronic motion. This ion motion induces a current between the detection electrodes. The frequency of the current is converted into m/z values via Fourier-transformation (E). In an orbitrap analyzer, the ions spin around the center electrodes. The shape of the electrodes induces a quadro-logarithmic potential resulting with an oscillating motion along the z axis. This motion induces a current between the detector electrodes, converted into m/z values via Fourier-transformation (F).

units (Da or Th) or in relative units, typically parts per million [ppm]. In summary, quadrupoles and ion trapping devices have relatively low mass accuracy and resolution, but are robust and relatively economical. On these instruments, mass accuracy and resolution varies according to the scan speed. Typical values are mass accuracies of 0.1 Th and resolution of 3000 at usual operating conditions for an ion at m/z = 1000. Time-of-flight analyzers have intermediate resolution and accuracy performances, but are more expensive and very sensitive to temperature variations and other operating conditions. They thus demand high maintenance and advanced user knowledge. Typical mass accuracy values of modern instruments are 0.005 Th and resolution of 25,000 for an ion at m/z = 1000. Finally Fourier-transform instruments (FT-ICR and OT) have highest resolution and mass accuracy, but are very expensive and their use is generally reserved to specialized laboratories with expert personnel. Such instruments have sub-ppm mass accuracies, meaning that the third decimal of a

compound at m/z = 1000 is accurate. Resolution can reach 400,000 or more. Two stages of mass analysis can be combined in a tandem MS experiment (MS2 or MS/MS). Typically, two mass analyzers are combined, separated by a collision cell, in a tandem mass spectrometer. During the first stage of MS, ions of one particular m/z values are isolated and transferred into the collision cell. There, the so-called precursor ions are activated into product ions (also called fragment ions) and neutrals. The product ions are then subjected to a second stage of MS analysis. Such an experiment is called product ion scan [26]. In ion trapping devices, activation and subsequent m/z analysis is performed in the same analyzer after ejection of the undesired precursor ions. This process can be repeated n times after re-isolation of product ions (MSn). Structural information about the molecular ion of interest can be obtained from product ion scans. This includes polypeptide sequence determination and glycan analysis. Collisions with

Please cite this article in press as: A. Scherl, Methods (2015), http://dx.doi.org/10.1016/j.ymeth.2015.02.015

A. Scherl / Methods xxx (2015) xxx–xxx

neutral gas molecules, or collision induced dissociation (CID) is typically used as activation method to fragment molecular ions. For larger molecular ions such as proteins, alternative activation methods have been developed. Among them, electron transfer dissociation (ETD), resulting from the ion-ion reaction between a multiply charged cation and a radical anion [27], and electron capture dissociation (ECD), the result of an electron capture by a multiply-charged cation [28] is of particular interest. These two activation methods are indeed particularly useful to dissociate large polypeptides and proteins for their structural determination. Tandem MS can also be used as a very specific and selective assay to identify and quantify an analyte of interest in a complex matrix. During the first stage of MS, only ions with the exact m/z value of the analyte ion to be measured are selected and activated. Second, only product ions at m/z ratio specific from the analyte are selected and detected in the second stage of MS. This results in a very specific assay, called selected reaction monitoring (SRM). Traditionally, such SRM assays are performed in triple quadrupole mass spectrometers (Q-q-Q). The first quadrupole is used to select the precursor ion of interest, the second quadrupole is the collision cell to activate and dissociate the precursor ion, and the third quadrupole is used to select the specific product ion. The matrix containing the analyte is infused into the mass spectrometer via a liquid chromatograph, where a first separation of all analyte molecules is already performed. A list of precursor ion and product ion m/z with their respective retention time and collision energy is programmed into the mass spectrometer, and these analytes are subsequently measured. An SRM assay is summarized in Fig. 5. Triple quadrupole instruments have been used routinely for more than three decades for quantification of all sorts of molecules in analytical laboratories of all types, including the pharmaceutical industry, food industry, environmental analysis, forensics, and clinical laboratories [29]. Other tandem mass spectrometers, for example ion traps or hybrid instruments combining two distinct mass analyzers separated by a collision cell can be used to acquire product ion scans and to perform assays similar to SRM. Instead of selecting a specific product ion in the second mass analyzer, all product ions are collected and subsequently separated according to their m/z value, as in any product ion scan. An extracted ion chromatogram (XIC) representing only the abundance of the specific product ion of interest is then generated and used for identification and quantification purpose. Such assays are termed ‘‘pseudo-SRM’’ [30,31], or ‘‘multiple reaction monitoring’’ by some instrument vendors. Although scan speed, sensitivity and resolution can vary between such pseudo-SRM assays and ‘‘traditional’’ SRM assays in a triple quadruple instrument, they are conceptually identical. A specific product ion mass resulting from the dissociation of a specific precursor ion mass is measured in a complex matrix, resulting in a very selective assay. If a high accuracy and resolution detector is used, the specificity of the assay is increased. Selected reaction monitoring in quadrupole instruments is used for decades in clinical laboratories to quantify various analytes. Typically, this includes therapeutic drug monitoring, drugs and drug of abuse detection and quantification, toxicology screenings, and metabolite measurements. In recent years, steroid hormone analysis is also increasingly performed by mass spectrometry, and typically with selected reaction monitoring [32]. 3.1. Protein analysis by mass spectrometry The invention of soft ionization methods such as MALDI and ESI made the routine analysis of full length proteins and peptides possible, mainly because molecular ions could be obtained from such large biopolymers without extensive fragmentation during the ionization step. In tandem MS experiments using CID, dissociation

5

occurs primarily along the peptide backbone. Extensive fragmentation is observed for shorter peptides, typically up to 15 or 20 amino acids long, permitting in most cases de novo sequencing of the peptide from a product ion scan acquired at optimal collision energy and mass range [33]. The peptide fragmentation pattern depends on the primary structure, the amount of energy introduced to activate the peptide, the nature of the activation, and many other factors such as molecular mass, charge state, nature of the collision gas, etc. In particular, the number and position of amino acids with a strong affinity to protons in the gas phase (for positively charged ions) have a strong influence. For longer peptides and proteins, the number of observed fragments to the protein length ratio becomes lower, and mainly the presence of acidic residues drives the fragmentation process. Peptide and protein activation mechanisms were extensively studied and reviewed, for example by Shukla [34] or by Sleno [35]. As CID is less efficient for full length proteins due to the high mass of the precursor ions, alternative activation techniques such as ECD and ETD were introduced [27,28]. However, these techniques are only available for high end instruments and mostly used in specialized laboratories. In summary, CID remains the most popular way to activate peptides in routine tandem MS experiments. However, alternative activation techniques such as ETD are very promising and open new perspectives for the routine analysis of full length proteins. 3.2. Top-down versus bottom-up A common strategy for analyzing proteins is to digest them first with an enzyme into peptide components. The endopeptidase trypsin is very often used, due to its high specificity. Trypsin cleaves polypeptide chains almost exclusively on the c-terminal side of lysine and arginine residues, if not followed by a proline residue. For identification and quantification purposes, one or more tryptic peptides specific to the protein to be analyzed, i.e. with a sequence corresponding to one unique gene product, are selected. This process is termed ‘‘bottom-up’’ analysis. Despite undeniable successes using the bottom-up approach for protein analysis, digesting all proteins from a complex mixture into many more peptides has also many drawbacks. First, the complexity of the mixture increases dramatically during the digestion. If thousands of proteins are present in a sample such as a biofluid or a cellular extract, the digestion step will result in a mixture of tens of thousands of peptides, assuming perfect enzyme specificity. Second, endopeptidases such as trypsin follow first-order reaction kinetics [36]. This means that the digestion rate is proportional to protein concentration. In other words, a low-abundance protein will have a slow digestion rate compared to an abundant one. Third, the digestion step will always result in peptides either too short or too long to be optimally analyzed by one single LC–MS condition. Full protein sequence recovery is thus impossible. Similarly, digesting a full length protein will result with peptides with different physicochemical proprieties such as isoelectric point and hydrophobicity, again difficult to be analyzed using a single analytical system. Last but not least, enzymatic digestion often introduces variability in the sample preparation. In an interlaboratory study, coefficients of variations of up to 60% have been recorded for quantitative measurements when digestion was performed in different laboratories. In contrast, this value dropped to about 15% when digestion was performed in a reference laboratory before inter-laboratory measurement [37]. In another report, using direct analysis of peptides from high abundance proteins and immunoenrichment of low-abundance peptides after digestion, reported interlaboratory coefficients of variations up to 30% after choosing best performing peptides [38]. Another major issue for clinical laboratories relies in the time constraints coming

Please cite this article in press as: A. Scherl, Methods (2015), http://dx.doi.org/10.1016/j.ymeth.2015.02.015

6

A. Scherl / Methods xxx (2015) xxx–xxx

Fig. 5. Selected reaction monitoring (SRM). The analyte solution, typically a complex mixture, is separated by liquid chromatography and infused into a triple-quadrupole mass spectrometer. Several analyte molecules can co-elute at a given chromatographic retention time. The first quadrupole mass filter (Q1) selects ion with a particular m/z ratio value. This ion is subsequently activated into product ions in the collision cell of the mass spectrometer (q2). A product ion specific to the analyte of interest is then selected in the third quadrupole (Q3) and detected (A). A chromatogram showing the ion abundance of this specific product ion from a specific precursor ion can then be plotted (B).

from the digestion. Typically, the digestion duration varies from several hours up to 48 h, making prompt reporting of results difficult. An alternative approach to ‘‘bottom-up’’ analysis is to analyze the full-length protein, without previous enzymatic digestion. This approach is termed ‘‘top-down’’. The advantage of the top-down method is that the full length protein is selected. The measured mass corresponds therefore to the specific proteoform, including post-translational modifications. For example, the glycated, carbamylated and native form of hemoglobin beta chain is unambiguously differentiated if the full length protein is analyzed. In contrast, after enzymatic digestion, all peptides but the N-terminal one are identical and do not allow differentiation between the three different proteoforms. The difference between the top-down and bottom-up approach is illustrated in Fig. 6. Small proteins and native peptides such as peptide hormones, typically of 15–30 amino acids in length, are typically analyzed without enzymatic digestion as well. This approach is often called ‘‘peptidomics’’ [39]. The important difference to the bottom-up strategy is that the full length peptide is analyzed, and that different proteoforms can be distinguished, as in any other top-down approach. Of course, the top-down strategy also has several drawbacks compared to the bottom-up approach. With electrospray ionization, the protein may appear in many different charge states. For example, on a typical electrospray mass spectrum, hemoglobin is visible at all charge states from 14+ to 24+ (Fig. 7). This results in charge-dilution, and overall sensitivity will decrease. In addition, as proteins appear in several different charge states, many of them

can appear at the same m/z ratio window, potentially decreasing selectivity of the assay. Once ionized and in the gas-phase, activating full length proteins can also be a limiting factor. CID is not very effective for this purpose, and ETD or ECD are mostly available on high-end instruments, usually only in specialized laboratories. Protein separation/enrichment via liquid chromatography or other methods is also challenging. However, if the different physicochemical proprieties such as hydrophobicity are taken into account, a specific chromatography can also be used to enrich a particular protein from the matrix or to simplify the mixture, for example by keeping only the most hydrophilic proteins in solution. 3.3. Introduction of an internal reference standard (IRS) No matter if protein quantification will be performed using the top-down or the bottom-up strategy, an internal reference standard (IRS) is needed. This is due to the fact that the response of a mass spectrometer is not always linear with the analyte concentration, and ionization suppression and matrix effects are observed, especially in the presence of complex matrixes such as serum, plasma or urine. Due to these effects, the signal of one analyte may vary with the presence of co-eluting species and the composition of the matrix. An IRS is thus spiked into the sample at a known concentration. The relative concentration of the analyte according to one of the IRS is then calculated. If the absolute concentration of the IRS is known, absolute concentration of the analyte can be determined as well. The ideal IRS is a compound having the exact same ionization response as the analyte. A molecule of the same structure, but with

Please cite this article in press as: A. Scherl, Methods (2015), http://dx.doi.org/10.1016/j.ymeth.2015.02.015

A. Scherl / Methods xxx (2015) xxx–xxx

7

Fig. 6. Top-down versus bottom-up analysis. With the top-down approach, the full-length protein is ionized and analyzed. The total mass, including post-translational modifications (PTMs) is observed. With the bottom-up approach, the protein is first digested into peptides. The individual peptides are then analyzed by LC–MS/MS. Some peptides may be too small to be observed by MS, or may be lost during the sample preparation or sample analysis.

Fig. 7. Typical mass spectrum of hemoglobin. The multiply-charged alpha chain (blue labels) and beta-chain (red labels) is visible. The heme group is observed at m/z = 616.2.

a different isotopic composition is thus the golden standard. Quantification using an IRS with heavy isotopes is called isotopedilution MS. In this situation, the IRS is also called stable-isotope standard (SIS) or stable isotope labeled (SIL) standard. Typically, 13 C, 15N or 18O is introduced during (bio)synthesis of the IRS. Deuterium should generally be avoided, as H/D exchange can be observed and its incorporation can result in a retention time shift with reverse-phase chromatography. As the natural compound contains also approximately 1% of 13C and traces of 15N, an isotopic distribution is observed in the pure analyte spectrum. The IRS should therefore contain sufficient heavy isotopes to appear at an m/z value sufficiently away from the analyte to allow resolution. For small molecules (mass range from 200 to 600 Da) the mass difference should typically be between 4 and 6 Da. For peptides (mass range 600–1800 Da), a mass difference of 6–8 Da between the

natural compound and the IRS is generally considered sufficient. In the case of a full length protein, this number should be much higher. In a typical Hemoglobin spectrum, about 20 different isotopic compositions are observed. In this case, a mass difference of 40–50 Da between the natural compound and the IRS would be necessary to differentiate them clearly. In addition, the isotopic incorporation should be as close as possible to 100% to ensure a correct isotopic pattern and a reliable signal measurement. For top-down protein analysis, obtaining a good IRS with heavy isotopes, namely with a sufficient mass difference and correct isotopic purity might be difficult in practice. Thus, alternative IRS strategies are used. Typical examples are proteins of a similar but not identical aminoacid sequence. For example, the hemoglobin delta chain can be accurately quantified by using the alpha or beta chain as IRS [40]. Bovine insulin is also used as IRS for human

Please cite this article in press as: A. Scherl, Methods (2015), http://dx.doi.org/10.1016/j.ymeth.2015.02.015

8

A. Scherl / Methods xxx (2015) xxx–xxx

insulin [41]. Similarly, derivatization of amino-acid side chains with alkylating compounds such as methylations result in an important mass shift, and they can be used as an IRS. For bottom-up protein quantification, there are different issues. As described above, the highest source of variability is in the enzymatic digestion of the protein. Ideally, the IRS should anyway be introduced as early as possible in the quantification workflow, to compensate for all experimental variations, not only digestion. The ideal IRS would be the heavy isotope analog of the full length protein, which would account for variations during the enzymatic digestion [42]. However, introducing a full length protein with sufficient mass difference and acceptable isotopic purity is often difficult. In addition, the same post-translational modifications occurring in the natural protein should also be present. Therefore, the most common approach is to introduce an isotopically labeled specific peptide, corresponding to one of the digestion products. This approach is known as AQUA, for absolute quantification [43]. However, due to incomplete digestion, different digestion kinetics, and variability in trypsin accessibility, the AQUA method often lacks the precision required for clinical decision-taking [44]. Indeed, imprecision of up to 30% or even 60% were observed in quantitative measurements if the same analyte was digested using the same digestion protocol in different laboratories [37,38]. To minimize imprecision, specific digestion protocols should be developed and best performing peptides for quantification should be selected [45]. Alternative approaches include the synthesis of shorter peptides specific to the analyte protein sequence, but including one or two adjacent trypsin cleavage site, also called ‘‘winged peptides’’ [46–48]. Other solutions to at least partially compensate for digestion variability is the biosynthesis in isotopically enriched media of a protein construct of concatenated tryptic peptides specific to several proteins to be measured. Enzymatic digestion of this protein will result in the generation of specific tryptic peptides. This approach is known as QconCAT [49].

4. Proteins quantified by bottom-up mass spectrometry The idea of quantifying proteins by quantifying their specific peptides after enzymatic digestion and isotope-dilution MS was already presented during the late 80s and early 90s, when analytical chemists used fast atom bombardment (FAB) ionization. During these very early days of peptide MS, Dass and co-workers presented a method for the semi-quantitative analysis of the neuropeptide beta-endorphin. The hormone was enriched from pituitary glands and digested with trypsin. A specific tryptic fragment was then selected for quantitative analysis. The detection limit was fairly high (90 fmol on column) as compared to today’s instrument sensitivity, but the feasibility of protein quantification by MS was demonstrated [50]. FAB-MS was also proposed as a reference method for the quantification of apolipoprotein A1. The international federation of clinical chemistry (IFCC) introduced reference material for this protein in 1983, and quantification was done by amino acid analysis (AAA) after full hydrolysis. Reference quantification was thus dependent on the protein purity, since this method is not specific to the protein sequence. Specific peptides to apolipoprotein A1 were later synthesized with heavy isotopes and used as internal reference standards for the quantification of protein digests by MS. The method allowed the quantification of the reference material (1 mg tubes) with a CV of 3.95% for 96 measurements [51]. In comparison, the precision reported with AAA was slightly above 5% [52]. Although the sensitivity and limits of detection were not very good, the proof of principle that proteins could be quantified with high precision by MS was made. Improvements in instrumentation and use of electrospray ionization allowed rapid progress of the isotope-

dilution methods for protein digests. With the growing popularity of the omics sciences, and especially proteomics, MS assays were developed and proposed for many protein biomarkers. Selected reaction monitoring (SRM) in triple quadrupole instruments, used during decades for small molecule analysis, was re-discovered for peptide quantification. Assays were proposed for many proteins, including for example the C-reactive protein [53] and urinary albumin [54]. However, the complexity of the instrumentation compared to the simplicity of automated immunoassays and the necessary fast turnaround time between sample drawing and the reporting of results prevented its use in routine clinical laboratories. Also, the sensitivity of MS instruments is below immunoassays, and only highly abundant proteins could be directly measured with this method. Precision is good for high abundance proteins, but insufficient for lower abundance ones, and digestion protocols are not robust enough for inter-laboratory agreement if compared to immunoassays [37]. Typically, direct MS could be used as reference methods for calibration material. This is for example the case for glycated hemoglobin (HbA1c). HbA1c is of clinical importance for the management and follow-up of diabetic patients. Here, a synthetic N-terminal of the glycated and nonglycated peptide, labeled with 7 atoms of deuterium instead of hydrogen, is used as internal reference standard. The ratio of glycated to non-glycated hemoglobin is then calculated according to the ratio of glycated and non-glycated n-terminal peptide after digestion with endopeptidase Glu-C [55]. MS methods represented also an interesting alternative for relatively high abundance proteins where immunoassays do not perform well or cannot differentiate protein isoforms. This is the case for the Apolipoprotein E isoforms. Three isoforms are described: ApoE2, with cysteine residue at positions 112 and 158, ApoE3, with a cysteine residue at position 112 and an arginine residue at position 158, and ApoE4, with arginine residues at positions 112 and 158. To differentiate between them, synthetic tryptic peptides specific to the different sequences with stable heavy isotopes are used as internal reference standards. To reach the required sensitivity, sample preparation includes an enrichment step of lipoproteins using an absorbent prior to trypsin digestion [56]. The detection is based on a presence/absence base of each isoform, and precise quantification is thus not an issue. Apolipoproteins A1 and B and their ratio is used to predict cardiovascular risk. Although immunoassays exist for these proteins, LC–MS/MS was suggested as alternative due to the high structural variability of apolipoproteins in artherogenic particles. A method relying on trypsin digestion and using synthetic tryptic peptides labeled with stable heavy isotopes used as internal reference standard was thus proposed. Good agreement was observed between the immunoassay and the LC–MS/MS method, with r2 correlations of 0.86 and 0.82 for ApoA-1 and ApoB respectively [57]. As these apolipoproteins are present in the g/l concentration range, previous enrichment or purification is not necessary. Another high abundance protein relevant for clinical applications is hemoglobin (Hb). Sequence variants are responsible for various diseases, like HbS for sickle-cell anemia. More than one thousand variants are described today (http://globin.cse.psu.edu/ hbvar/menu.html). Identification of variants for diagnosis relies mostly on indirect protein analysis such as electrophoresis and chromatography, and comparison with reference standards. MS represents thus an attractive and specific detection method. Also, for the detection of variants, only presence/absence needs to be determined, and precise quantification is generally not an issue. Many MS methods for the identification of numerous variants were described, as reviewed by Edwards and co-workers [58]. Among them, Daniel and colleagues proposed an SRM screening method for the detection of peptides specific to the most common clinically relevant variants, namely HbS, C, E, O and D. Due to the specificity

Please cite this article in press as: A. Scherl, Methods (2015), http://dx.doi.org/10.1016/j.ymeth.2015.02.015

A. Scherl / Methods xxx (2015) xxx–xxx

of tandem MS and the high concentration of hemoglobin in blood, only lysis and digestion of the sample is performed prior to dilution and direct MS/MS analysis, without chromatographic separation. The throughput is thus potentially very high [59]. Recently, the same principle was applied to analyze hemoglobin peptides directly from dried blood spots, and refined for the detection of only disease states (homozygote) and not carrier state (heterozygote) of Hb C, S, D, E, O, and Lepore. More than 13,000 newborns were screened, and a direct comparison with a routine HPLC method was used. The study concluded that no false negatives were detected and that all clinically relevant cases were correctly identified [60]. For the quantification of proteins present at lower concentrations, strategies to increase the detection limit have been proposed. For example, Arsene and co-workers used cysteinecontaining peptide conjugation with N-(3-iodopropyl)-N,N, N-dimethyloctylammonium (IPDOA) iodide to increase the detection limit and the limit of quantification for human growth hormone (GH, 22 kDa isoform) [61]. The labeled full length protein is used as internal reference standard. After alkylation with IPDOA iodide, the serum sample is digested with trypsin and the peptide to be quantified is separated using a combination of strong cation exchange and reverse-phase chromatography prior to MS. The limit of quantification for GH using this method is 0.4 lg/l. A reference method for the measurement for the C-reactive protein (CRP) was also developed, combining immunoenrichment with isotope-dilution mass spectrometry. This method demonstrated that the combined approach could reach the requirements in precision and limits of quantification [62]. Since, many other bottomup methods to increase sensitivity based on immunoenrichment were developed. This is for example the case for thyroglobulin, one of the best studied protein for the comparison of immunoassays and MS for clinical diagnosis. Thyroglobulin is a marker for recurrence or persistence of thyroid cancer after initial treatment [63]. The presence of antithyroglobulin autoantibodies in 10–30% of patients can cause wrong results [13]. MS assays were thus developed and are now proposed routinely for the measurement of thyroglobulin when the presence of autoantibodies against the protein is detected. The most recent assays are based on addition of labeled winged peptides (comprising trypsin cleavage sites) into the sample, digestion, immunoenrichment of specific peptides with magnetic beads coated with antibodies and LC–MS/MS analysis [46]. One of the advantages of MS relies in its multiplexing capabilities. Indeed, modern instruments operate so fast that tens of different transitions specific to different peptides can be performed over a very short chromatographic elution peak. This multiplexing capability was nicely demonstrated in an application where 45 serum proteins were directly quantified in a single assay. Among other proteins, antithrombin, apolipoproteins, ceruloplasmin, transthyretin and vitamin D binding protein were quantified. Labeled synthetic peptides for each protein were added and used as internal reference standard after digestion [64]. Serum proteins present in the lM range could be quantified with this method, which would correspond to the high mg/l range for a 30 kDa protein. This approach was later refined using immunoenrichment of the tryptic peptides and their corresponding internal reference standard after enzymatic digestion, a method known as SISCAPA (stable isotope standards and capture by anti-peptide antibodies). Low lg/l level sensitivity can be achieved this way [65]. In a proofof principle demonstration, the method was applied to troponin I, a routinely tested biomarker of cardiac injury, and interleukin 33 (IL-33), a potential marker for cardiovascular disease. The limit of quantification is 1.5 lg/l for both proteins [66]. However, as very rapid, robust and fully automated immunoassays exists for them proteins, this multiplexed MS method is mostly used for discovery

9

experiments in research environments rather as in routine clinical laboratories. Nevertheless, the SISCAPA method demonstrates that immunoenrichment combined with bottom-up proteomics techniques can reach sufficient sensitivity and precision to be used in clinical laboratories. The concept of using antibodies for immunoenrichment of peptides prior to multiplex LC–MS/MS analysis is nowadays well accepted, and numerous technical solutions for potential clinical assays have been proposed since [38,67]. Finally, MS assays are also used to measure therapeutic monoclonal antibodies (mABs). As mABs are increasingly used drugs, specific detection methods are needed for therapeutic monitoring. These compounds are typically administered at concentrations between the lg/kg and mg/kg level. Due to the relatively high sensitivity necessary to quantify them, all published methods use a combination of immunoenrichment and MS, as reviewed by van den Broek and co-workers [9].

5. Proteins and peptides quantified without enzymatic digestion (top-down) As already detailed above, enzymatic digestion is one of the main source of variability and imprecision for protein quantification. It’s probably one of the main limiting factors preventing the wide use of MS in clinical laboratories for protein analysis. An attractive alternative consists thus in analyzing the full length protein without previous digestion. The term ‘‘protein’’ refers to a native polypeptide in this context, i.e. anything from a small peptide of several amino acids up to a large protein of several hundreds of amino acids, but not previously cleaved by an enzyme. Mass spectrometry analysis is indeed fairly straightforward for small peptides and proteins, but challenging for longer sequences. ESI-MS instruments provide best sensitivity in the 300–1200 m/z range. A small peptide hormone, as for example angiotensin II, with a mass of 1046 DA, will appear similarly to a peptide produced by enzymatic digestion of a protein, and be visible as singly, doubly and triply charged molecular ions. However, higher molecular weight proteins will appear in several multiple charge states (Fig. 7) in this range, ‘‘diluting’’ the total signal. In parallel, the multiple natural isotopic compositions also contribute to this ‘‘dilution’’ effect. Last but not least, solubilization, chromatography and liquid handling in general can be very difficult for full length proteins. As a rule of thumb, a high molecular weight protein will be more difficult to analyze than a short polypeptide. But the important distinction between bottom-up and top-down analysis is the fact that no enzymatic digestion is used prior to the mass spectrometry analysis, to ensure that a single proteoform is selected and analyzed. Short proteins or polypeptides, such as peptide hormones, are thus particularly attractive due to their solubility and ideal mass for direct analysis by MS. Plasma renin activity is measured to test the capacity of renin to generate angiotensin I from angiotensinogen. Angiotensin I is further converted to angiotensin II by the converting enzyme. Direct immunoassays cannot be used due to the high cross-reactivity between the three species. The test is thus generally done via a radioimmunoassay (RIA) measuring the competitive binding between radioactive labeled angiotensin already present on the antibodies and natural angiotensin I from the sample. As radioassays generally tends to be avoided nowadays, an isotope-dilution LC–MS/MS assay was developed for the measurement of angiotensin I and angiotensin II. After addition of the internal reference standard into the sample, peptide preconcentration is performed via preparative HPLC prior to LC–MS/ MS analysis. In general, the results correlated very well with the RIA results. In addition, the MS method permitted to identify 2–5% of samples in which endogenous enzymes degraded

Please cite this article in press as: A. Scherl, Methods (2015), http://dx.doi.org/10.1016/j.ymeth.2015.02.015

10

A. Scherl / Methods xxx (2015) xxx–xxx

angiotensin I very rapidly, and where the RIA assay thus reported falsely low results [68]. The method was later adapted for solid-phase extraction separation, and the reported lower limit of quantification for renin activity is 0.08 nmol/l/h after a 24 h incubation period [69]. Another short signaling peptide hormone perfectly suited for MS analysis is ghrelin. The active peptide is acetylated on the serine-3 residue, and the modification, critical for activity, is labile. As antibodies cross-react with the different proteoforms, the precision of traditional immunoassays is much debated [70]. MS methodologies based on the direct detection of the active, acetylated form were thus developed. After addition of an isotopic labeled internal reference standard in blood samples, larger proteins are precipitated and all ghrelin isoforms are immunoprecipitated prior to MALDI-TOF analysis. Peptide concentrations in the pg/ml range can be measured with this method [71]. Hepcidin is a 25 amino acid cysteine rich peptide that regulates iron metabolism. Other isoforms, namely the 20, 22 and 24 amino acid long peptides are also present. Immunoassays cross-react with the different isoforms, making MS the ideal method for hepcidin quantification. In addition, the relatively low molecular weight of the peptide is perfectly suited for analysis by MS. A method for the hepcidin quantification in serum was proposed, based on precipitation of abundant proteins and solid-phase extraction of all hepcidin isoforms prior to LC–MS/MS quantification. A short protein, calcitonine-gene related peptide (CGRP) was used as internal reference standard. The assay had a reported lower limit of quantification of 1 ng/ml [72]. Later, a fast and direct assay using chip-based nanoLC chromatography and a labeled synthetic hepcidin as internal reference standard was developed. The assay has limit of quantification of 6 ng/ml and the correlation with ELISA test is high (r2 = 0.96) [73]. Immunoassays also cross-react between different synthetic insulins, porcine insulin and the natural human form. Insulin is a 51 amino acid long polypeptide. Its mass is thus well suited for mass spectrometry analysis. Specific MS methods for the different proteoforms were developed for doping control and forensic science applications after suspicion of insulin overdose or poisoning. Immunoaffinity is used for the enrichment of all potential insulin isoforms from blood or urine, prior to specific LC–MS/MS detection and quantification of all different isoforms and their metabolites. Detection limits relevant for physiological concentrations (high pg/ml level) can be obtained with such methods [41]. Antibody-based detection has also limitations for the distinction of amyloid beta isoforms (Ab1–40 and Ab1–42) in cerebrospinal fluid, due to sequence similarity and protein self-aggregation. However, immunoaffinity can be used for the enrichment of all Ab isoforms prior to MS detection. Different methods using MALDI-TOF MS or LC–MS/MS have been published, with quantification limits around 200 pg/ml [74]. MALDI-TOF MS combined with immunoenrichment was also used by Trenchevska and co-workers for the detection and quantification of Transthyretin. This protein is indeed present in several isoforms (glycated, oxidized, sulfonated, cysteinylated, cysteinglycin and native form, as well as common point-mutations) rendering immunodetection non-optimal. The assay compared well with conventional ELISA methods, and limits of quantification (low mg/l range) are within the desired range for clinical applications [75]. A very similar method was published for the detection of Apolipoprotein C-I, C-II, C-III and their isoforms [76]. Similarly, top-down analysis of the full-length protein was also developed for many clinically relevant proteins, including parathyroid hormone [77], chemokines [78] and Macrophage Inhibitory Factor [79]. Insulin-like growth factor 1 (IGF1) is also a 70 amino acid long protein that can be measured by top-down MS. High-resolution MS was combined with off-line IGF1 enrichment by precipitation of

abundant serum proteins to quantify the peptide [80]. The internal reference standard was oxidized rat IGF1. The lower limit of quantification was about 15 lg/l for the assay. The method was then further optimized and transposed to low-resolution triplequadrupole instruments for an inter-laboratory study, using full-length labeled IGF1 internal reference standard. Reported inter-laboratory imprecision was below 16% for a study involving 5 laboratories and 152 samples [81]. In parallel, a method using solid-phase extraction of tryptic IGF peptides rather than precipitation of abundant proteins was also developed and achieved similar sensitivity [82]. A high-throughput assay combining immunoenrichment and MS analysis was also proposed for IGF [83]. Finally, a very interesting way of using MS was used by Murray’s group for the detection of Ig M-protein monoclonal light chains in patients with multiple myelomas. The light chain is released after disulfide bond reduction and enriched using a gel-affinity procedure. The sample is than directly infused into the mass spectrometer, and presence of clonal M-proteins is characterized by an abundant signal at a precise mass, unique for each patient. After recording of the accurate mass of the antibody’s light chain, treatment efficiency and disease relapse could be monitored via the MS assay [84]. Top-down protein analysis could also be used for the diagnosis of hemoglobin-related diseases. Our laboratory developed an assay based on isolation of the different globin chains in the mass spectrometer followed by activation via electron-transfer dissociation (ETD). Protein product ions specific to the most common variants such as Hemoglobin S (HbS, responsible for sickle-cell disease), HbC, HbE, HbO and HbD can be detected [85]. In parallel, precise quantification of the hemoglobin delta chain can be performed [40]. Such an increase of the delta chain ratio is pathognomonic for beta-thalassemia trait. Rapid detection of clinically significant variants and diagnosis of beta-thalassemia can thus be performed on a single MS platform.

6. Current state and challenges Today, it appears that a majority of the proposed protein MS assays rely on bottom-up analysis, with prior digestion of the clinical sample and detection of a specific peptide from the protein of interest. Historically, this is certainly related to the popularity of the bottom-up approach in discovery proteomics. It is indeed certainly the most sensitive way to analyze proteins by MS, and throughput is adapted for current requirements. However, by digesting the protein into several dozen peptides, the fundamental problem of immunoassays, i.e. recognizing only a single epitope, might be repeated. Not all peptides will be detected by MS, resulting only in partial sequence coverage. Also, some of the identified peptides will inevitably not be specific and could be attributed to several different proteins. And enzyme specificity will never be perfect. In other words, the pitfalls of immunoassays, namely non-specific detection, are somewhat repeated with the bottomup approach. Nevertheless, the problems of autoantibodies and saturation effects can still be overcome using the bottom-up approach. This is very nicely demonstrated with the thyroglobulin mass spectrometry-immunoassay [46]. Top-down MS is not common practice for clinical applications yet. There are indeed only few laboratories equipped with high resolution instruments, knowledge and experience with full-length protein analysis. In addition, there is a lack of bioinformatics tools to reach the required robustness for clinical applications. But the fact that a unique proteoform can be distinguished by MS makes the top-down approach particularly attractive for clinical applications. Sensitivity is a problem today and will probably still be a

Please cite this article in press as: A. Scherl, Methods (2015), http://dx.doi.org/10.1016/j.ymeth.2015.02.015

A. Scherl / Methods xxx (2015) xxx–xxx

problem in the future, because of lack of good liquid-phase full-length protein separation technologies hyphenated to mass spectrometry, charge-state dilution and overlapping isotopic patterns. Measuring abundant proteins like hemoglobin is obviously possible with current methods and instrumentation. To meet the sensitivity requirements for other clinically relevant proteins, immunoenrichment prior to MS analysis might be a valid solution. With immunoenrichment, all proteoforms could be enriched to a sufficient level, and the mass spectrometer will then detect the isoform of interest in a specific manner. The pitfalls of immunoassays described above could thus be avoided. In addition, top-down analysis of the full length protein satisfies the clinical requirements for sample preparation time and avoids variability and imprecision due to enzymatic digestion. Another challenge of protein MS assays is method development and method validation. This time-consuming work has to be properly done for each single assay, and repeated for every new one. Finding an appropriate internal standard is the first difficulty. Finding appropriate matrixes to test for interferences, linearity and limits of quantifications is the next one. If immunoenrichment is used, the antibodies and enrichment procedures need to be developed and validated as well. But once this work is done, method transfer to other laboratories and analytical platforms can be relatively straightforward. In summary, an MS method could potentially be developed, validated and transferred to other laboratories in a very short time. It is thus well suited to solve ‘‘acute’’ analytical problems in a fast way. MS is particularly well adapted for clinical studies, biomarker validation on large cohorts and other small- and mid-size studies were development and validation of a fully automated immunoassay might be time consuming and expensive. The regulatory context has also to be considered for the implementation of protein MS assays. In vitro diagnostic (IVD) certification in Europe and FDA approval in the US is common practice for immunoassay kits and instruments. In the field of protein mass spectrometry, such certifications are certainly not common practice today. In fact, only very few instrument vendors propose such certification on a limited choice of instruments, mostly triple quadrupoles, and no protein assay is yet FDA or CE certified for diagnostic applications. This does not mean that MS assays cannot be used for diagnostic purposes, but signifies that the methods have to be developed and validated according to relatively strict guidelines, and that a strict quality control has to be applied. The fact that the rules are different in each country does not simplify the procedure. For example in Switzerland, university hospitals can develop and use clinical mass spectrometry assays if the methods are developed and validated according to the ISO15189 norms and the QUALAB regulations, if the laboratory is certified and part of an internal and external quality control program (http://qualab.ch/). While this practice is well-established and relatively straightforward, specialized laboratory personnel with a high level training are needed to implement them. In contrast, if a certified assay is implemented on a certified instrument, the burden is much lower for the laboratory. In this case, only calibration and quality controls have to be performed according to the guidelines. Assay validation, traceability, and corrective measures are the responsibility of the instrument/assay vendors in this case. Availability and large use of certified instrumentation and reagents might thus change the paradigm that MS assays are complicated to establish in the future.

7. Conclusions and outlook MS has a demonstrated capability to detect peptides and proteins in a specific manner. Combined with appropriate

11

sample preparation and/or enrichment, sensitivity is high enough to quantify peptides and proteins for clinical applications. Also, the multiplexing capabilities are well demonstrated. It is clear today that MS can overcome some of the limitations observed with antibody-based assays. In particular, the detection of isoforms and different post-translational modifications is of highest interest for clinical applications. These are typical cases where antibody cross-reactivity is observed and where immunoassays have their limitations. Concrete examples are the renin activity assay with the measurement of angiotensinogen, angiotensin I and angiotensin II, as well as the peptide hormone ghrelin. Also, MS is a good alternative to overcome the problem of autoantibodies. This is well illustrated with the thyroglobulin assay. On the other hand it is clear that the instrumentation is complex, there is a need for well trained and very specialized personnel to use mass spectrometers, and fully automated data analysis is not available yet. In other words, the manual input is still very high. The few examples discussed above summarize well the current state of protein MS analysis: It’s an expensive, specialized technique, and not completely mature for fully automated routine analysis yet. However, the potential added value makes it a viable alternative to solve complex problems were antibodies simply cannot provide the easy answer. It’s very likely that that common routine assay, validated within a well-defined frame and routinely used on fully automated clinical chemistry liquid handling robots such as troponin and CRP will not be replaced soon with MS. MS assays will probably be first reserved for niche applications with high added value, to measure parameters where fast turnaround time is less critical, but fine interpretation of results is clinically important. Future routine applications of mass spectrometrybased peptide and protein assays in clinical laboratories might include peptide hormones, ideal because of their limited mass and close structure, proteins related to iron-metabolism like hepcidin, due to the presence of various isoforms, characterization of sequence variants with close structural similarity, and generally all proteins were autoantibodies create analytical interferences. Concrete issues to be resolved are enzymatic digestion protocols, appropriate internal standards of sufficient length and with sufficiently pure isotopic composition, and automation of the instrument interface. This last point includes a ‘‘one button’’ fully automatic instrument calibration and performance monitoring. This automated step should include the interfacing with HPLC devices, as well as the automatic generation of results and reports. To overcome sensitivity issues, immunoenrichment can be performed. Antibodies can be used to enrich several proteoforms, including interfering species for immunoassays, and the relevant one is finally measured by MS. For this, antibodies with lower specificity can be used (compared to a classical specific immunoassay), which are less costly and easier to produce. Last not but least, robust and easy to automate immunoaffinity enrichment procedures are necessary for this essential step. To conclude, a unique advantage of MS is the fact that the analysis is specific to the entire analyte molecule, which is total mass and structure. In comparison, other analytical techniques use non-specific physicochemical proprieties such as retention time or isoelectric point. Similarly, antibodies recognize only an epitope, not the entire molecule. MS is thus in a unique position to distinguish between proteoforms with subtle differences. But to simplify distinction between these subtle differences, enzymatic digestion prior to MS analysis should be avoided. Enzymatic digestion creates orders of magnitude more complexity, certainly not suitable for specific detection of a single proteoform. In all cases where such a single proteoform is relevant for the clinical context, top-down MS analysis has a great potential.

Please cite this article in press as: A. Scherl, Methods (2015), http://dx.doi.org/10.1016/j.ymeth.2015.02.015

12

A. Scherl / Methods xxx (2015) xxx–xxx

References [1] A. Di Meo, E.P. Diamandis, H. Rodriguez, A.N. Hoofnagle, et al., Clin. Chem. 60 (2014) 1258–1266. [2] P.J. Monaghan, B.G. Keevil, P.J. Trainer, Clin. Endocrinol. (Oxf.) 78 (2013) 344– 346. [3] J.E. Hale, Int. J. Proteomics 2013 (2013) 1–6. [4] A.N. Hoofnagle, M.H. Wener, J. Immunol. Methods 347 (2009) 3–11. [5] R.P. Grant, A.N. Hoofnagle, Clin. Chem. 60 (2014) 941–944. [6] J.O. Becker, A.N. Hoofnagle, Bioanalysis 4 (2012) 281–290. [7] F.G. Strathmann, A.N. Hoofnagle, Am. J. Clin. Pathol. 136 (2011) 609–616. [8] S. Lehmann, A. Hoofnagle, D. Hochstrasser, C. Brede, et al., Clin. Chem. Lab. Med. (2013) 51. [9] I. Van den Broek, W.M.A. Niessen, W.D. van Dongen, J. Chromatogr. B 929 (2013) 161–179. [10] V. Thongboonkerd, J. LaBaer, G.B. Domont, J. Proteome Res. 13 (2014) 4493– 4496. [11] L.M. Smith, N.L. Kelleher, M. Linial, D. Goodlett, et al., Nat. Methods 10 (2013) 186–187. [12] R.R. Little, W.L. Roberts, J. Diabetes Sci. Technol. 3 (2009) 446–451. [13] C.A. Spencer, M. Takeuchi, M. Kazarosyan, C.C. Wang, et al., J. Clin. Endocrinol. Metab. 83 (1998) 1121–1127. [14] P. Kebarle, J. Mass Spectrom. 35 (2000) 804–817. [15] R. Zenobi, R. Knochenmuss, Mass Spectrom. Rev. 17 (1998) 337–366. [16] R. Cooks, A. Rockwood, Rapid Commun. Mass Spectrom. 5 (1991) 93. [17] W. Paul, Rev. Mod. Phys. 62 (1990) 531–540. [18] K.R. Jonscher, J.R. Yates, Anal. Biochem. 244 (1997) 1–15. [19] J.W. Hager, Rapid Commun. Mass Spectrom. 16 (2002) 512–526. [20] J.C. Schwartz, M.W. Senko, J.E.P. Syka, J. Am. Soc. Mass Spectrom. 13 (2002) 659–669. [21] R.J. Cotter, Biomed. Environ. Mass Spectrom. 18 (1989) 513–532. [22] M. Guilhaus, V. Mlynski, D. Selby, Rapid Commun. Mass Spectrom. 11 (1997) 951–962. [23] A.G. Marshall, C.L. Hendrickson, G.S. Jackson, Mass Spectrom. Rev. 17 (1998) 1– 35. [24] Q. Hu, R.J. Noll, H. Li, A. Makarov, et al., J. Mass Spectrom. JMS 40 (2005) 430– 443. [25] A. Makarov, Anal. Chem. 72 (2000) 1156–1162. [26] J.C. Schwartz, A.P. Wade, C.G. Enke, R.G. Cooks, Anal. Chem. 62 (1990) 1809– 1818. [27] R.A. Zubarev, D.M. Horn, E.K. Fridriksson, N.L. Kelleher, et al., Anal. Chem. 72 (2000) 563–573. [28] J.E.P. Syka, J.J. Coon, M.J. Schroeder, J. Shabanowitz, D.F. Hunt, Proc. Natl. Acad. Sci. 101 (2004) 9528–9533. [29] R.W. Kondrat, G.A. McClusky, R.G. Cooks, Anal. Chem. 50 (1978) 2017–2021. [30] V. Lange, P. Picotti, B. Domon, R. Aebersold, Mol. Syst. Biol. 4 (2008) 222. [31] A. Scherl, S.A. Shaffer, G.K. Taylor, H.D. Kulasekara, et al., Anal. Chem. 80 (2008) 1182–1191. [32] G. Hopfgartner, E. Varesio, TrAC, Trends Anal. Chem. 24 (2005) 583–589. [33] K.G. Standing, Curr. Opin. Struct. Biol. 13 (2003) 595–601. [34] A.K. Shukla, J.H. Futrell, J. Mass Spectrom. JMS 35 (2000) 1069–1090. [35] L. Sleno, D.A. Volmer, J. Mass Spectrom. JMS 39 (2004) 1091–1112. [36] E. Mihalyi, J. Biol. Chem. 201 (1953) 197–209. [37] T.A. Addona, X. Shi, H. Keshishian, D.R. Mani, et al., Nat. Biotechnol. 29 (2011) 635–643. [38] A. Prakash, T. Rezai, B. Krastins, D. Sarracino, et al., J. Proteome Res. 11 (2012) 3986–3995. [39] P. Schulz-Knappe, M. Schrader, H.-D. Zucht, Comb. Chem. High Throughput Screening 8 (2005) 697–704. [40] A.E. Acosta-Martin, D.C. Graça, P. Antinori, L. Clerici, et al., Anal. Chem. 85 (2013) 7971–7975. [41] M. Thevis, A. Thomas, W. Schänzer, Forensic Sci. Int. 213 (2011) 35–41. [42] V. Brun, A. Dupuis, A. Adrait, M. Marcellin, et al., Mol. Cell. Proteomics 6 (2007) 2139–2149. [43] S.A. Gerber, J. Rush, O. Stemman, M.W. Kirschner, S.P. Gygi, Proc. Natl. Acad. Sci. 100 (2003) 6940–6945. [44] C.G. Arsene, R. Ohlendorf, W. Burkitt, C. Pritchard, et al., Anal. Chem. 80 (2008) 4154–4160. [45] I. Van den Broek, N.P.M. Smit, F.P.H.T.M. Romijn, A. van der Laarse, et al., J. Proteome Res. 12 (2013) 5760–5774. [46] M.M. Kushnir, A.L. Rockwood, W.L. Roberts, D. Abraham, et al., Clin. Chem. 59 (2013) 982–990. [47] H. Neubert, D. Muirhead, M. Kabir, C. Grace, et al., Anal. Chem. 85 (2013) 1719– 1726.

[48] M. Razavi, L.D.S. Johnson, J.J. Lum, G. Kruppa, et al., Clin. Chem. 59 (2013) 1514–1522. [49] R.J. Beynon, M.K. Doherty, J.M. Pratt, S.J. Gaskell, Nat. Methods 2 (2005) 587– 589. [50] C. Dass, G.H. Fridland, P.W. Tinsley, J.T. Killmar, D.M. Desiderio, Int. J. Pept. Protein Res. 34 (1989) 81–87. [51] J.R. Barr, V.L. Maggio, D.G. Patterson, G.R. Cooper, et al., Clin. Chem. 42 (1996) 1676–1682. [52] J. Shepherd, Commission of the European Communities, Directorate-General for Science, R., and Development, Purification and certification of human apolipoprotein AI and AII reference materials: CRM 393 and 394 : final report, Commission of the European Communities, Directorate-General Science, Research and Development, Luxembourg 1991. [53] E. Kuhn, J. Wu, J. Karl, H. Liao, et al., Proteomics 4 (2004) 1175–1186. [54] J.C. Seegmiller, D.R. Barnidge, B.E. Burns, T.S. Larson, et al., Clin. Chem. 55 (2009) 1100–1107. [55] P. Kaiser, T. Akerboom, R. Ohlendorf, H. Reinauer, Clin. Chem. 56 (2010) 750– 754. [56] K.R. Wildsmith, B. Han, R.J. Bateman, Anal. Biochem. 395 (2009) 116–118. [57] S.A. Agger, L.C. Marney, A.N. Hoofnagle, Clin. Chem. 56 (2010) 1804–1813. [58] R.L. Edwards, N.J. Martin, H.J. Cooper, Bioanalysis 5 (2013) 2043–2052. [59] Y.A. Daniel, C. Turner, R.M. Haynes, B.J. Hunt, Br. J. Haematol. 130 (2005) 635– 643. [60] S.J. Moat, D. Rees, L. King, A. Ifederu, et al., Clin. Chem. 60 (2014) 373–380. [61] C.G. Arsene, D. Schulze, J. Kratzsch, A. Henrion, J. Mass Spectrom. 47 (2012) 1554–1560. [62] E.L. Kilpatrick, D.M. Bunk, Anal. Chem. 81 (2009) 8610–8616. [63] L.M. Demers, C.A. Spencer, Clin. Endocrinol. (Oxf.) 58 (2003) 138–140. [64] M.A. Kuzyk, D. Smith, J. Yang, T.J. Cross, et al., Mol. Cell. Proteomics 8 (2009) 1860–1877. [65] N.L. Anderson, A. Jackson, D. Smith, D. Hardie, et al., Mol. Cell. Proteomics 8 (2009) 995–1005. [66] E. Kuhn, T. Addona, H. Keshishian, M. Burgess, et al., Clin. Chem. 55 (2009) 1108–1117. [67] B. Krastins, A. Prakash, D.A. Sarracino, D. Nedelkov, et al., Clin. Biochem. 46 (2013) 399–410. [68] C.E. Bystrom, W. Salameh, R. Reitz, N.J. Clarke, Clin. Chem. 56 (2010) 1561– 1569. [69] S. Carter, L.J. Owen, M.N. Kerstens, R.P.F. Dullaart, B.G. Keevil, Ann. Clin. Biochem. 49 (2012) 570–579. [70] C. Prudom, J. Liu, J. Patrie, B.D. Gaylinn, et al., J. Clin. Endocrinol. Metab. 95 (2010) 2351–2358. [71] J.A. Gutierrez, J.A. Willency, M.D. Knierman, T. Coskun, et al., Methods Enzymol, vol. 514, Elsevier, 2012, pp. 129–146. [72] A.T. Murphy, D.R. Witcher, P. Luan, V.J. Wroblewski, Blood 110 (2007) 1048– 1054. [73] C. Delaby, J. Vialaret, P. Bros, A. Gabelle, et al., EuPA Open Proteomics 3 (2014) 60–67. [74] T. Oe, B.L. Ackermann, K. Inoue, M.J. Berna, et al., Rapid Commun. Mass Spectrom. 20 (2006) 3723–3735. [75] O. Trenchevska, E. Kamcheva, D. Nedelkov, Proteomics 11 (2011) 3633–3641. [76] O. Trenchevska, M.R. Schaab, C.R. Borges, D. Nedelkov, R.W. Nelson, Multiplexed mass spectrometry immunoassay for quantitative determination of apolipoprotein C-I, C-II and C-III and their isoforms. MSACL 2014 US 6th Annu. Conf. 2014, San Diego, Tu poster 7A. [77] M.F. Lopez, T. Rezai, D.A. Sarracino, A. Prakash, et al., Clin. Chem. 56 (2010) 281–290. [78] O. Trenchevska, N.D. Sherma, P.E. Oran, P.D. Reaven, et al., J. Proteomics 116 (2015) 15–23. [79] N.D. Sherma, C.R. Borges, O. Trenchevska, J.W. Jarvis, et al., Proteome Sci. 12 (2014) 52. [80] C.E. Bystrom, S. Sheng, N.J. Clarke, Anal. Chem. 83 (2011) 9005–9010. [81] H.D. Cox, F. Lopes, G.A. Woldemariam, J.O. Becker, et al., Clin. Chem. 60 (2014) 541–548. [82] R. Kay, D.J. Halsall, A.K. Annamalai, N. Kandasamy, et al., Clin. Endocrinol. (Oxf.) 78 (2013) 424–430. [83] P.E. Oran, O. Trenchevska, D. Nedelkov, C.R. Borges, et al., PLoS One 9 (2014) e92801. [84] J.R. Mills, D. Murray, D.R. Barnidge, A. Dispenzieri, et al., Using the accurate molecular mass of monoclonal light chains to detect residual disease in patients with multiple myeloma. MSACL 2014 US 6th Annu. Conf. 2014, San Diego. [85] D. Coelho Graça, P. Lescuyer, L. Clerici, Y.O. Tsybin, et al., J. Am. Soc. Mass Spectrom. 23 (2012) 1750–1756.

Please cite this article in press as: A. Scherl, Methods (2015), http://dx.doi.org/10.1016/j.ymeth.2015.02.015