Identification and Relative Quantification of Proteins
This invention relates to the identification and relative quantification of proteins, especially through the use of matrix-assisted laser desorption ionization mass spectrometry (MALDI MS).
Immuno affinity methods and mass spectrometry are established techniques found in many bio-analytical laboratories. Their combination in the form of affinity capture mass spectrometry has the potential to detect, isolate and identify a vast range of target analytes in a very simple manner. For protein analysis, various types of immurioaffmity separations are combined with MALDI MS; with antibody- antigen interactions being the most often employed. Affinity capture MALDI MS has also been applied to study peptide - metal ion affinity (Qian et al., Anal. Biochem., 274, 174-180, 1999, Papac et al, Anal. Chem., 66, 2609-2613, 1994), lectin and carbohydrate affinity of microorganisms (Bundy et al, Anal. Chem., 73, 751-757, 2001) and genetic polymorphisms (Griffin et al, Nat. Biotechnol., j[5, 1368-1372, 1997, Tang et al, Nucleic Acids Res., 23, 3126-3131, 1995). In addition, peptide and protein digestion has been performed on enzymatically active MALDI probes, yielding an increased sensitivity and fewer enzymatic autolysis fragments (Dogruel el al, Anal. Chem., 67, 4343-4348, 1995, Krogh et al, Anal. Chem, 274, 153-162, 1999). Under the MALDI conditions, the non-covalent biospecific interactions between binding molecule and analyte are disrupted and the captured biomolecule can then be analyzed in the mass spectrometer (Zhao et al, Anal. Chem., 66_, 3723-3726, 1994). Frequently used 'indirect' methods imply immobilization of a binding molecule on a solid matrix, i.e. on a solid support for immunoaffinity chromatography or on magnetic beads (Papac et al, Anal. Chem., 66, 2609-2613, 1994, Rϋdiger et al, Anal. Biochem., 275, 162-170, 1999, Schriemer et al, Anal. Chem., 68, 3382-3387, 1996, Hurst et al, Anal. Chem., 71, 4727-4733, 1999), which are then placed on a conventional MALDI target for mass analysis. A practical development is the use of a 'direct' methodology where a binding molecule is immobilized on a MALDI target for affinity capture of a protein (Hutchens et al, Rapid Commun. Mass Spectrom., 7, 576-580, 1993, Brockman et al, Anal.
Chem., 67, 4581-4585, 1995, Brockman et al, Rapid Com un. Mass Spectrom., 10, 1688-1692, 1996, Liang et al, Anal. Chem., 70, 498-503, 1998) i.e. an 'immuno ffinity target' surface which employs immunoaffinity by target surface immobilised immunoglobulins for the capture of analyte(s) for analysis by MALDI. The 'direct' MALDI MS approach seems to be superior to the 'indirect' method because the latter has been shown to degrade the MALDI performance (Papac et al, Anal. Chem., 66, 2609-2613, 1994).
Different strategies have been used to immobilize antibodies on a MALDI target. The affinity of proteins to hydrophobic polymer films, i.e. nitrocellulose, has been exploited to produce MALDI probes with immobilized antibody (Liang et al, Anal. Chem., 70, 498-503, 1998) or streptavidin (Schriemer et al, Anal. Chem., 68, 3382-3387, 1996). Major drawbacks of that approach are the random immobilization of antibody and the predisposition to non-specific binding of other proteins to the nitrocellulose surface when complex solutions are analysed (Liang et al, Anal. Chem., 70, 498-503, 1998). Alternatively, methods for covalent activation of MALDI targets have been investigated by many researchers. Activation can be achieved by using the homobifunctional linker dithiobis(succinimidyl propionate) (DTSP) on gold coated MALDI surfaces (Dogruel et al, Anal. Chem., 67, 4343-4348, 1995, Brockman et al, Anal. Chem., 67, 4581-4585, 1995, Zhang et al, Anal. Chem., 7 4753-4757, 1999). Anchoring of the DTSP linker molecule onto MALDI targets is based on the dissociative chemisorption of its disulphide group on gold surfaces as first reported by Katz et al for the modification of gold electrodes (Katz et al, Electroanal. Chem., 291, 257-260, 1990). The N-hydroxysuccinimide active ester functionalities of the monolayer are used to immobilize biomolecules via their primary amino groups. Self-assembled monolayers (Ulman et al, Chem. Rev., 96, 1533-1554, 1996) (SAM) organized from DTSP have also been successfully employed in biosensor applications for covalent immobilization of biomolecules (Darder et al, Anal. Chem., 5530-5537, 1999). For MALDI purposes it was soon realized that the loading capacity of the antibody surfaces on DTSP SAM was limited, because the relatively short length of the linker sterically hinders antibody immobilization and accessibility of the antigen to the binding sites (Brockman et al, Rapid Commun. Mass Spectrom., 10, 1688-1692, 1996). To
overcome this problem Brockman and Orlando proposed a new antibody immobilization scheme which is not restricted by monolayer formation, however, the assembly of their affinity probes takes several days.
The number and accessibility of available antibody binding sites on the surface of MALDI MS targets is now recognized as a crucial factor for improving detection capabilities.
Fc receptors (such as Protein A, G and A/G) are used in various immunoassays and biosensors to orientate antibodies (Lu et al, Analyst, 121, 29R-32R, 1996, Spitznagel et al, Bio/Technology, JJL 825-829, 1993, Schramm et al, Anal. Biochem., 205, 47-56, 1992, Klonisch et al, Immunology, 89, 165-171, 1996). Protein A/G is a sythesised protein, described in Eliasson et al, J. Biol. Chem., 263, 4323-4327, 1988. Biosensors are dynamic devices used for protein detection revealing kinetic information of bioprocesses such as affinity constants etc. Biomolecular interaction analysis (BIA) which utilises surface plasmon resonance detection is an example (biacore instrument). The specificity of such an assay system to a great extent is determined by the antibody, but it cannot be easily verified which protein was actually bound to the antibody as the associated end-point detection with these techniques does not give molecular mass information. This lack of information on molecular mass leaves doubt as to what is actually being measured, i.e. what is actually bound to the antibody binding sites. BIA has been coupled to MALDI where the real-time capabilities of biosensors were combined with the qualitative specificity of mass spectrometry directly (Krone et al, Anal. Biochem, 244, 124-132, 1997, Nelson et al, Anal. Chem, 69, 4363-4368, 1997) or indirectly (Sonksen et al, Anal. Chem, 70, 2731-2736, 1998); however, to the best of our knowledge, neither Protein A or G has been used in such a combination.
Protein G has been used in conjunction with MALDI by immobilising it on beads to capture antibody/antigen complexes, the beads then being placed on MALDI targets for analysis (Zhao et al, Anal. Chem., 66, 3723-3726, 1994, Zhao et al, Proc. Natl. Acad. Sci. U.S.A., 93, 4020-4024, 1996). As capture is not directly on the MALDI target this can be considered as an 'indirect' approach, albeit the analyte is not dissociated from
the Protein G antibody complex prior to laser desorption. However, the chief disadvantage of this application of beads is that it has been shown to degrade MALDI performance compared to conventional MALDI MS experiments (Papac et al, Anal. Chem., 66, 2609-2613, 1994), i.e. a lower intensity of analyte is measured. The explanation given is attributed primarily to surface inhomogeneities (Papac et al, Anal. Chem., 66, 2609-2613, 1994). The present invention does not use beads but instead orientates antibodies directly on the MALDI target thus allowing homogeneity of the matrix with the analyte on the target surface.
The orientation of antibodies using Fc receptors on solid supports, such as Protein A, Protein G or recombinant or synthesised Protein A G, has already been reported for imimmoassays, immunosensors and immuno affinity chromatography but not for MALDI MS.
In accordance with the present invention, a strategy was devised for improving the loading capacity of a MALDI surface. The present invention provides a matrix-assisted laser desorption/ionisation mass spectrometry (MALDI MS) target for the immobilisation of immunoglobulins, the target having a support surface to which an Fc receptor is linlced, the Fc receptor being oriented with the "receptor portion" positioned away from the support surface. In particular the target may be used in Matrix-assisted desorption ionisation time of flight mass spectrometry (MALDI TOF MS).
The present invention allows mass identification of analytes captured on orientated antibodies. In the inventors' technology, an F0 receptor is orientated and immobilised directly on a target to bind and orientate antibodies. The subsequent mass spectrometric analysis of the captured antigen gives qualitative information.
The target according to the invention achieves improved antibody immobilization by allowing correct orientation and optimised surface density of antibodies on the target. The antibodies are oriented such that the antigen binding site or sites are positioned away from the support surface to increase antigen recognition. The surface density of the antibodies is controlled because the distance between the support surface and the
bound antibodies influences antigen-binding due to steric hindrance. The correct orientation and optimum surface density are both achieved by the correct positioning of the Fo receptor on the support surface.
The term "support" means a structure, usually solid which forms the basis of the target. "Support surface" is used to mean the surface of the support on to which the Fc receptors are mounted. Any support suitable for use in MALDI MS may be used to form the basis of the target of the invention. In particular gold coated MALDI surfaces are preferred, however other surfaces may be used as indicated below.
The Fc receptor is preferably immobilised on the support surface by a chemical linker. The chemical linker and support may be formed as a self-assembled monolayer (SAM). An Fc receptor may be immobilised on the (SAM). Examples of this technology include:
• monolayers of fatty acids on surface metal cation (e.g. aluminium oxide Al2O3, silver oxide AgO, copper oxide CuO);
• monolayers of organosilicon derivatives by the formation of a polysiloxane SAM from alkylsilanes on hydroxylated substrate surfaces;
• organosulfur adsorbates from e.g. alkanethiol, dialkyl disulfide, dialkyl sulfide, alkyl xanthate, dialkylthiocarbamate on metal surface (such as gold, copper, silver, platinum, mercury, iron);
• selenium compounds on transition metal surfaces; and
• synthetic polymers carry suitable functional groups for modification
Preferably the homobifunctional linker dithiobis(succinimidyl propionate) (DTSP) is used as the chemical linker.
Once the chemical linker is surface-immobilised using one of the above approaches the chemical moiety of the linker that is oriented away from the SAM is used to covalently link to the Fc receptor. Such orientating chemistries include:
• Primary Amine Selective o Amide bond formation with N-hydroxysuccinimde ester (such as
DTSP); o Imidoester linker to form amidine bonds; o Carbodiimide activated carboxylic acids.
• Sulfhydryl Reactive o Maleimides, alkyl and aiyl halides and alpha-haloacyls are thiol reactive and target cysteine residues.
• Arginine Specific o Glyoxals for targeting the guanidinyl moiety of arginine residues.
Carbonyl Specific o Carbonyl (aldehyde and ketones) react with amines. Carbonyls can be formed by mild oxidation of vicinal hydroxyls using NaIO4.
The Fc receptor may also be tagged with a chemical marker (not an amino acid) that uniquely identifies a position for specific cross-linking.
Particularly preferred is the use of an immobilisation chemistiy that was first reported for the modification of gold electrodes (Katz et al, Electroanal. Chem., 291, 257-260, 1990) and has been previously applied on gold coated MALDI surfaces (Dogruel et al, Anal. Chem., 67, 4343-4348, 1995, Brockman et al, Anal. Chem., 67, 4581-4585, 1995, Zhang et al, Anal. Chem., 71, 4753-4757, 1999) which uses the homobifunctional linker dithiobis(succinimidyl propionate) (DTSP). Anchoring of the DTSP linker molecule onto MALDI targets is based on the dissociative chemisorption of its disulphide group on gold surfaces. The N-hydroxysuccinimide active ester functionalities of the monolayer are used to immobilise biomolecules via their primary amino groups. This chemistry has not been used for immobilisation and orientation of Fc receptors onto MALDI targets, prior to this invention.
The term "Fc receptor" means a molecule that selectively binds to the Fc region of an antibody, leaving the antigen binding region of the antibody free for antigen binding. In particular, Fc receptors are often proteins, but other molecules such as protein mimetics may also be used. The term "Fc receptor" also encompasses molecules which bind Fc, but which also comprise domains other than that used for Fc binding. The portion of the receptor which binds Fc is herein referred to as the Fc binding region.
Particularly preferred Fc receptors include, but are not limited to Protein A, Protein G, synthesised Protein A/G, or functional fragments or variants thereof. A functional fragment or variant of Proteins A, G or A/G is a fragment or variant of one of those proteins which is capable of functioning as an Fc binding region.
Proteins A and G are naturally found on the surface (cell walls) of a variety of staphylococci and streptococci species. The first Fc receptor immobilised on a solid support was Protein A (from cell wall of Streptococcus aiirens) which has been used successfully to bind the F0 portion of IgG from many mammalian species. However, Protein A does not react with IgG from several species such as goat, sheep, cow and horse. A more versatile and efficient alternative for binding IgG isotypes is Protein G (from cell walls of Streptococcus) (Lu et al, Analyst, 121, 29R-32R, 1996). However, it does not bind strongly to several IgGs with which Protein A reacts well (Godfrey et al, A Practical Approach; Matejtschuk, pp 141-195, 1997). The protein from group G streptococci (Protein G) shows a broader range of binding to IgG subclasses compared to Protein A from staphylococci (Bjorck et al, Immunol., 133, 969-974, 1984). Using Protein G for antibody immobilisation is particularly attractive because a wide range of γ-immunoglobulins can be bound with affinity constants ranging from 109 to 1010 M"1 (Akerstrδm et al, J. Biol. Chem., 261, 10240-10247, 1986). For all examined immunoglobulins, the determined affinity constants were found higher for Protein G than for Protein A).
A schematic representation of the structure of Protein G is given in Figure 1. The protein is constructed largely of three sets of repeated amino acid sequences. The two A-repeats of unknown function are followed by two B-repeats for binding to
immuno globulins. Therefore, each Protein G molecule can bind 2 molecules of IgG, allowing the formation of a precipitate. The highly charged C-repeats are involved in the attachment to the cell wall and consist of the pentapeptide sequence AspAspAlaLysLys, repeated five times. Protein G does not contain any cysteine residues which could be used for site-directed immobilisation.
Recombinant or synthesised variants of Protein A and G are also available. These variants are tailored for covalent immobilisation onto solid supports and may have sequences removed i.e. for membrane anchoring, albumin binding etc., or sequences inserted i.e. for cross linking. Variants are available that contain a C-terminal cysteine residue for site-directed immobilisation onto gold surfaces using the thiol chemistiy or have additional lysine residues added for the purpose of covalent immobilisation using active esters. In addition, Protein G also binds to albumin at several sites, which are structurally separate from the IgG-binding sites (Sjobring et al, J. Biol. Chem, 261, 10240-10247, 1986). Those sites may be deleted in recombinant material to enhance the separating power of the Fc receptor-containing solid supports.
Preferred functional variants or fragments according to the present invention are those which retain the Fc binding function. Preferably the functional variants or fragments include a region which has at least 90% amino acid sequence homology with the Fc binding region of Protein G, A or A/G. More preferably the fragments or variants have at least 90% amino acid sequence homology with Protein G, A or A/G. Preferably, the fragments or variants differ from Protein G, A or A/G by only 1 to 25 amino acids. More preferably the fragments or variants differ within the binding site from one of Protein G, A or A/G by only 10% of the total number of amino acids in the binding site, more preferably by 1 to 5 amino acids, most preferably by 1 to 2 amino acids.
Sequence homology is preferably measured using 2 sequence BLAST analysis Tatiana et al, FEMS Microbiol. Lett., 174, 247-250, 1999; (www.ncbi.nlm.nih.gov/BLAST/)
It is further preferred that any amino acid changes are conservative. Conservative changes are those that replace one amino acid with one from the family of amino acids
which are related in their side chains. For example, it is reasonable to expect that an isolated replacement of a leucine with an isoleucine or valine, an aspartate with a glutamate, a threonine with a serine, or a similar conservative replacement of an amino acid with a structurally related amino acid will not have a major effect on the biological activity of the protein. Mutations which increase the number of amino acids which are capable of forming disulphide bonds with other amino acids in the protein are particularly preferred in order to increase the stability of the protein. Other mutations which increase the function of the protein except the cell adhesion destabilising function can also be made.
Preferably, functional fragments of Fc receptors are at least 10 amino acids in length. It is further preferred that the functional fragments comprise at least one B-repeat region.
The Fc receptor and chemical linker are selected so that the Fc receptor will selectively link to the chemical linker in such a position that when antibody binds to the Fc receptor the antibody is oriented with its antigen binding region or regions positioned away from the support surface, that is to say, the Fc receptor is oriented with its receptor portion positioned away from the support surface. The Fc receptor or chemical linker may be modified to improve site-directed binding between the two. Any linker and modification which allows correct orientation of the F0 receptor may be used. Examples of linkers are given above.
Preferred modifications include attaching one or more amino acid residues to the Fc receptor. The one or more residues may be attached to any part of the receptor provided that the positioning allows correct orientation of the Fc receptor on the support surface, and that the residues do not significantly obstruct the Fc binding region.
Any amino acid residue which allows linkage to a chemical linker may be used. Particularly preferred amino acid residues are cysteine and/or lysine, more preferably lysine.
A particularly preferred modification is the attachment of between 5 and 15, more preferably 10, lysine residues to the C repeat region of one of Protein G, A or A/G, most preferably to Protein G. The C repeat is the repeated pentapeptide AspAspAlaLysLys in the cell wall attachment region of the protein. Preferably between 5 and 15 lysine residues are attached to at least one of the C repeats enabling easy attachment to a support surface via a chemical linker (e.g. DTSP).
The target according to the invention may further comprise an antibody molecule bound to the Fc receptor. Such a target may additionally comprise an antigen molecule bound to the antibody.
Antibodies can be directly immobilised on targets by covalent coupling (Brockman et al, Anal. Chem., 67, 4581-4585, 1995) or adsorption (e.g. Schriemer et al, Anal. Chem., 68, 3382-3387, 1996, Liang et al, Anal. Chem., 70, 498-503, 1998) but this results in random orientation. Orientating antibodies through orientated Fc receptors on MALDI targets in accordance with this invention allows the antigen binding regions to be oriented away from the target surface thus giving the advantage of increased loading capacity onto a MALDI target, as evident by the increase in MALDI signal intensity compared to randomly immobilised antibodies, as shown in Figure 6. An additional advantage is direct chemical coupling or passive adsorption of antibody onto MALDI targets causes partial antibody denaturation and results in considerable loss of biological activity (Olson et al, Immunol., 26, 129-136, 1989). The inventors' use of Fc receptors avoids direct chemical coupling or passive adsorption of antibodies and preserves biological activity thus avoiding wasting antibody.
The term antibody is used to mean a molecule comprising at least one Fc region and at least one antigen binding region. The antigen binding region may be, for example an Fv region or an Fab region. The antibody molecule may be single stranded such as llama or camel antibodies or double stranded. The antibody molecule may be a recombinant antibody molecule, such as a chimeric antibody molecule. Methods for producing such antibodies are well known in the art.
Quantification of MALDI signals has been demonstrated on conventional MALDI targets (Tang et al, Anal. Chem, 65, 2164-2166, 1993, Nelson et al, Anal. Chem, 66, 1408-1415, 1994) but has never been reported on analyte signals from on-target captured MALDI signals. The present invention shows for the first time that relative quantification of MALDI signal from on-target affinity captured protein is possible because the signal of the analyte to that of the external standard increases linearily with increasing concentrations of analyte present. This enables optimisation and standardisation of an affinity capture MALDI system (Figure 5).
The availability of immobilised orientated F0 receptors on MALDI targets could have widespread use amongst those who do not wish to perform surface chemistry or have the laboratory facilities to linlc antibodies to MALDI targets. In addition, such pre-prepared targets could allow antibodies to be linlced very rapidly, which could have many advantages, e.g. in method development or screening of proteins.
Prepared targets offer certain advantages, for example, they are easy to use and are standardised requiring no optimisation. No method development is required. Such targets may be universal allowing any antibody to be loaded. Further, the use of such targets can save the user significant amounts of time.
Also provided by the invention is apparatus for MALDI MS comprising a target according to the invention.
Such apparatus generally comprises a MALDI source, a laser being the core element, a mass analyser and a detector. Commonly used lasers include N2 lasers. Examples of mass analysers include Time-of-flight, ion traps and Fourier Transform Ion Cyclotron Resonance analysers. The use of MALDI in protein analysis is described in US Patents 5,118,937 (Hillenlcamp) and 5,045,694 (Beavis & Chait).
Further provided is a method of maximising the amount of an antibody that may be immobilised on a target comprising the steps of: a) determining the surface area of the antibody to be bound;
b) determining the surface area of the target; and c) comparing the two surface areas to ensure that the surface area of the total number of antibodies to be immobilised is not greater than the surface area of the target.
The theorectical maximum amount of antibody that may be immobilised on a target may be calculated by determining the surface area+ of an antibody, by for example, x-ray diffraction (Sarma, Silvertori et al, 1971) and comparing the results with the surface area of the target surface.
Using higher concentrations (Spitznagel et al, Bio/Technology, JT, 825-829, 1993, Sada et al, Biotechnol. Bioeng., 28, 1497-1502, 1986) of antibody for random or oriented immobilisation on MALDI targets results in steric hindrance between surface-bound antibodies at higher concentrations. This is because accessibility of antigen binding sites on immobilised antibodies is restrained by neighbouring antibodies, thus resulting in a diminished loading capacity of the affinity capture MALDI target. Our invention optimises the spatial spread of immobilised antibodies on the surface of the MALDI targets by varying the antibody/ Fc receptor concentrations for immobilisation to afford a maximum antigen loading.
Preferably the total surface area of the antibodies to be immobilised is between 1% and 10%, more preferably about 5%, less than the total surface area of the target. By ensuring that the total surface area of the antibodies being immobilised is less than the surface area of the target, the likelihood of steric hindrance will be reduced.
Further provided is the use of a target or apparatus according to the invention in MALDI MS.
The targets according to the invention may also be useful in other types of mass spectrometry such as electrospray or nanospray mass spectrometry (Femi, JB., Mann,
M., et al, Science 246 (4926) 64-71 (1989), and Gaskell, S. J., J.Mass Spectrom. 32 (32) 677^688 (1997).
The invention will now be described in detail with reference to the drawings in which:
Figure 1 shows the sequence of Protein G; adapted from Fahnestock, S.R., Trends Biotechnology, 5, 79-83, 1987. Two homology groups A are followed by two B-repeats for immunoglobulin-binding and by a C-region for cell wall attachment.
Figure 2 shows anchoring of dithiobis(succinimidyl propionate) DTSP onto a MALDI target.
Figure 3 shows a schematic representation of the preparation of the immunoaffmity discs (left: random immobilization of immunoglobulin; right: oriented immobilization of immunoglobulin using recombinant Protein G).
Figure 4 shows the effect of washing with different concentrations of the surfactant Triton X-100 following incubation with hCGβcf (10 pmol/μl) on the signal area ratio. These experiments were carried out in duplicates. (Untreated gold targets - D; gold targets without linlcer but with antibody - •; and gold targets with covalently immobilized antibody - r).
Figure 5 shows the immobilization of different concentrations hCG MAb on either DTSP activated gold discs (Q), random covalent immobilization of MAb) or discs where recombinant Protein G was covalently attached to a DTSP self-assembled monolayer on gold for biospecific immobilization of hCG MAb. Protein G concentration ranged from 0.5 μg/ml ( • ), 1.0 μg/ml (Δ) to 5.0 μg/ml (T) (oriented immobilization). Immunoaffmity discs were incubated with 20 ml of 1 pmol/ml hCGβcf (n = 3) in PBS by shaking for 1.5 hours at 20 °C. The MALDI signal area ratios are expressed as mean +/- SEM.
Figure 6 shows MALDI signal area ratios obtained from random and oriented immunoaffmity discs incubated for 1.5 hours at room temperature in 20 μl hCGβcf (concentrations ranging from 100 fmol/μl to 2 pmol/μl). Antibody orientation increased
loading capacity of immunoaffmity discs around threefold. The MALDI signal area ratios are expressed as mean +/- SEM (n = 3).
Figure 7 is an AC-MALDI TOF mass spectra of hCGβcf (with cyto chrome C as an external standard for signal area comparison) using sinapinic acid matrix. 3 -dimensional overlay of two spectra of hCGβcf captured from 20 μl of a 1 pmol/μl solution on a surface where hCG MAb (5 μg/ml) was randomly immobilized (foreground) and on a surface where the same antibody was oriented using 0.5 μg/ml Protein G (background). The typical broad MALDI pattern of hCGβcf is due to its diversity of glycoforms and was previously deconvoluted by Jacoby et al.
Figure 8 shows conventional MALDI (i.e. using a stainless steel target) and affinity-capture MALDI analysis of a peptide and protein mixture containing 1 pmol/ml hCGβcf. (A) Conventional MALDI analysis of the protein mixture, (B) hCGβcf capture from the peptide and protein mixture on random (B) and oriented (C) immunoaffmity discs. To compare amounts of captured hCGβcf the external standard cytochrome C was added (B) and (C). (Remark: Annotated signals are oxidized insulin chain B (peak 1), for bovine α-lactalbumin (peaks 2 and 6 respectively), an unknown contaminant (peak 3), human growth hormone (pealcs 4 and 9 respectively), two polypeptides present in the from pregnancy urine purified hCGβcf (pealcs 5 and 5 respectively), bovine carbonic anhydrase (peaks 7, 10 and 12) and bovine serum albumin (pealcs 11 and 13 respectively). For details see text.).
Example 1
The development of an optimized affinity capture MALDI MS system using a recombinant Protein G as our model Fc receptor is described. The chosen model test substance is the core fragment of the beta sub unit of human chorionic gonadotropin (hCGβcf), a major urinary metabolite of the glycoprotein human chorionic gonadotropin (hCG).
In order to capture hCGβcf, DTSP self-assembled monolayer technology was used to covalently immobilize and orientate recombinant Protein G for antibody capture and orientation onto a gold-coated surface of a MALDI target. The effects of antibody
orientation and antibody density on the differences in MALDI signal intensity after incubation with the antigen followed by extensive washing were measured. This was achieved by adding an external standard (cytochrome C) on the target together with the MALDI matrix to compare relative signal areas. To determine the effectiveness of antibody orientation, data was compared to that obtained using the same monoclonal antibody randomly immobilized to the SAM on the gold surface.
Washing of surface boimd antibody-antigen complexes with detergents is routinely employed in immunoassays and biosensors systems. For washing affinity capture MALDI support detergents such as Triton X-100 (Nelson et al, Anal. Chem., 67, 1153-1158, 1995), n-octylglucoside (Rϋdiger et al, Anal. Biochem., 275, 162-170, 1999, Wang et al, J. Biol. Chem., 271, 31894-31902, 1996) or Tween 20, respectively; are employed whereas the majority of MALDI studies used only detergent-free washing buffers (Schriemer et al, Anal. Chem., 68, 3382-3387, 1996, Brockman et al, Anal. Chem., 67, 4581-4585, 1995, Brockman et al, Rapid Commun. Mass Spectrom., 10, 1688-1692, 1996, Liang et al, Anal. Chem., 70, 498-503, 1998). In this invention the inventors show the effect of washing with different concentrations of the detergent Triton X-100 on the non-specific binding of a protein to the surface of an affinity capture MALDI target (Figure 4). From that an optimum detergent concentration can be elucidated for maximum removal of non-specific binding without causing disruption of the antibody-antigen binding.
Example 2
Calculating optimum amount of bound antibody
The maximum amount of antibody that can be immobilised on a target can be calculated theoretically, assuming no steric hindrance and that all the antibodies are orientated with the Fab portions pointing upwards, i.e. bound the gold surface via the Fc domain. The surface area of a Gl -immunoglobulin is 70nm2 [calculation: 14 nm x 5 n ; dimensions determined by X-ray diffraction (Sarma, Silverton et al, 1971)] and thus a maximum of 5.4 x 1010 immunoglobulins can be immobilised on our MALDI target of 3.8 mm2, equating to 100 fmol. In practice, the surface of the immunoaffmity discs is roughened and hence it is reasonable to assume that the area may be several
times greater. In addition, IgG antibodies can bind simultaneously two antigens assuming no negative allosteric effects. Nonetheless, the binding capacity may be attenuated to some degree because of "incorrect" antibody orientation, steric crowding, diminished binding affinities of immobilised antibodies (Schramm and Paek, 1992).
Experimental Section
Materials
Dithiobis(succinimidyl propionate) (DTSP) was purchased from Sigma (Poole, UK). Anhydrous dimethylsulfoxide (DMSO) and anhydrous ethanol, both over molecular sieves, were obtained from Fluka (Buchs, Switzerland). Trifluoroacetic acid (TFA) and acetonitrile were of analytical grade and from BDH (Poole, UK) and Rathburn (Walkerburn, Scotland), respectively. Sinapinic acid was purchased from Aldrich (Gillingham, UK) and was recrystallized from hot ethanol prior to use. Water was purified on an Elgastat UHQ system (High Wycombe, UK). Monoclonal antibody directed against human chorionic gonadotropin (hCG MAb 147-PA) was obtained from Akzo Nobel (Boxtel, Netherlands). The affinity constants of this antibody for hCG, beta sub unit of hCG and hCGβcf are quoted as being 1.8 x 109, 2.1 x 109 and 5.6 x 109 1 mol"1 respectively. Recombinantly engineered Protein G was obtained from Calbiochem (Nottingham, UK) and contained an extra 10 lysine residues in the highly charged C-repeats of the protein for cross-linking purposes (distinct from the B-repeats for antibody binding). The hCGβcf was a kind gift from Professor Laurence A Cole (University of New Mexico, USA), having been purified from normal pregnancy urine. Oxidized bovine insulin chain B, horse heart cytochrome C, bovine alpha-lactalbumin, bovine carbonic anhydrase and bovine serum albumin were purchased from Sigma (Poole, UK). Recombinant human growth hormone was obtained from Serono
Pharmaceuticals Ltd. (Feltham, UK). MALDI autosampler strips consisting of 48 targets, these being stainless steel discs with an anodised surface diameter of 2.2 mm, were supplied by Thermo Bioanalysis Ltd. (Hemel Hempstead, UK).
Preparation of Immunoaffmity Discs
MALDI autosampler strips were located in a customized jig and the stainless steel discs were punched out. A gold layer of approximately 30 nm thickness was sputtered onto the scratched sides of the discs using a SEM Coating Unit E5100 (Polaron Equipment Ltd., Watford, UK). The preparation of the immunoaffmity discs is schematically presented in Figure 3. A self-assembled monolayer (SAM) was prepared by dissociative chemisorption of the dialkyl disulphide DTSP onto the gold surface, the likely mechanism of anchoring being by disulphide bond cleavage and subsequent formation of gold thiolate species. To this purpose, each sputtered disc was transferred into a polypropylene tube (39 mm height x 10 mm diameter), 200 ml of anhydrous DMSO containing 10 mM of DTSP was added and incubation was performed at 20°C for 1.5 hours. The discs were then washed twice with 200 ml anhydrous ethanol.
For producing immunoaffmity discs that have only antibody bound to the surface, i.e. randomly immobilized antibody the SAM was reacted with 100 ml hCG MAb in phosphate buffered saline (PBS), at concentrations ranging from 100 ng/ml to 100 μg/ml, at 20°C for 16 hours. For producing immunoaffmity discs that accommodate oriented antibody on a Protein G monolayer, i.e. oriented antibody, the SAM was reacted with 100 ml recombinant Protein G in PBS (5.0, 1.0 or 0.5 mg/ml) at 20°C for 16 hours. Following aspiration, 100 ml of hCG MAb in PBS in concentrations ranging from 100 ng/ml to 100 μg/ml was added and incubated at 20°C for 3 hours. After random or oriented immobilization of hCG MAb, the discs were washed with 200 ml 0.5 % (w/v) Triton X-100 in PBS and twice with 200ml PBS to ensure that all non-specifically bound proteins were removed. The immunoaffmity discs were not allowed to dry in the tubes in order to prevent the possibility of denaturation of the immobilized Protein G and/or antibody. Following preparation, the discs were left in the tubes for subsequent immunoaffmity capture and washing steps, which were performed on the same day.
Development of Detergent Washing Conditions
The effect of washing with different concentrations of the non-ionic detergent Triton X-100 was investigated on random immobilized antibody discs (prepared using 5 \mg/ml hCG MAb; as described in the previous section). Gold discs without linlcer but having been incubated with the same concentration of antibody, and untreated gold discs, underwent the same washing process for control purposes. The discs were incubated with the antigen hCGβcf (10 pmol/ml; 20 μl) in PBS by shaking for 1.5 hours at 20°C.
Immunoaffmity capture of antigen
The immunoaffmity discs were either incubated in 20 ml solution of hCGβcf in PBS by shaking for 1.5 hours at 20°C (the antigen concentration ranging from 2 pmol/ml to 100 fmol/ml), or in 20 ml of a test solution containing several peptides and proteins. The test solution contained oxidized chain B of insulin (200 fmol/ml), hCGβcf (1000 fmol/ml), bovine α-lactalbumin (400 fmol/ml), recombinant human growth hormone (800 fmol/ml), bovine carbonic anhydrase (800 fmol/ml) and bovine serum albumin (2000 fmol/ml) in PBS. The discs were then washed once with 200 ml 0.5% (w/v) Triton X-100 in PBS, twice with 200 ml PBS and finally with 200 ml water to remove buffer salts which are likely to interfere with the MALDI process and then left at room temperature until dry.
MALDI Mass Spectrometry
The immunoaffmity discs were affixed to the mass spectrometer autosampler strip using 1 mm2 squares of double-sided tape. A solution of 20 mg/ml sinapinic acid in 60:40 acetonitrile/0.05 % (v/v) TFA-water was prepared containing cytochrome C (125 fmol/ml) for standardizing MALDI signal areas. Cytochrome C was selected because of its molecular weight of around 12.2 kDa (same range as hCGβcf) and because it gives rise to intense MALDI signals even at low concentrations, therefore ion suppression was not expected. 0.8 ml of this mixture was added onto each disc by pumping the micropipette three times to promote mixing. In this way, the molar ratios between cytochrome C and sinapinic acid were kept constant. The sample surfaces were
crystallized on air before analysis using a LASERMAT 2000 MALDI-TOF mass spectrometer (Thermo Bioanalysis Ltd., Hemel Hempstead, UK). Pulsed light from a nitrogen laser (λmax = 337 nm) was used to desorb ions from the samples, which were then accelerated by a 20 keV electric field down a linear 0.5 m drift tube and detected by a micro-channel plate detector. The detector signal was digitized at a sampling rate of 500 MHz and transferred to a PC for data analysis. To account for heterogeneity within a MALDI sample, spectra were generated across the target from 20 adjacent laser aims, firing 15 shots per aim equating to 300 shots per spectrum.
Data Processing
Saved data files were exported from the LASERMAT Data Review software as ASCII files and imported into GRAMS Spectral Notebase. After smoothing and baseline subtraction, the signals for the singly charged molecular species of hCGβcf and CytC were integrated. To compare relative signal intensities the ratio of the signal areas was calculated.
Results and Discussion
Development of Affinity Capture MALDI system
Our affinity capture MALDI system was constructed in a stepwise manner, each disc being treated individually, thus simplifying coating, washing and incubation steps and also preventing cross contamination between targets, e.g. on a large MALDI sample holder. In addition, the dimensions of the targets made it possible to use very small incubation volumes.
The formation of the SAM was not monitored directly, but confirmed through final antigen capture. The acylation reaction of the DTSP linker with primary amino groups of proteins competes with the hydrolysis of the active ester (as the acylation is in an aqueous medium). The rate of hydrolysis increases with pH and temperature and decreases with the concentration of primary amine. However, hydrolysis eliminates the need for quenching the active moiety, as any ester group not reacted with protein is
converted into carboxylic acid. A drawback is that these surface-immobilized carboxylic acid groups may perform as cation exchangers that are capable of retaining highly basic analytes. In addition, adsoiption of proteins onto metal surfaces, and gold coatings in particular, is a well-studied phenomenon. Although this problem was not specifically addressed here, the need for washing conditions strong enough to remove non-specifically bound compounds to the gold surface and the SAM (linlcer) was taken into consideration. Indeed, MALDI has been used previously as a tool to analyze proteins adsorbed onto target surfaces. A strategy of keeping non-specific binding to a minimum is important when comparing relative signals in the development of immunoaffmity discs, i.e. optimizing the analysis of just antibody bound compounds. A washing protocol was developed that involved the use of a detergent. Currently, only a very limited number of studies have employed detergents for washing affinity capture MALDI supports, i.e. Triton X-100, n-octylglucoside or Tween 20, respectively; whereas the majority of other MALDI studies used only detergent-free washing buffers. At the beginning of our investigation we examined the influence of different concentrations of the non-ionic surfactant Triton X-100 on our affinity capture MALDI surface incubated in 10 pmol/μl hCGβcf (Figure 4). A high concentration of hCGβcf was used in order to observe large MALDI signal area changes upon washing with detergent, especially for non-specifically bound hCGβcf. These experiments were carried out in duplicates. A Triton X-100 concentration of 0.4% (w/v) washed away all non-specifically bound hCGβcf from untreated gold targets (open triangles) or gold targets without linlcer but with antibody (closed circles), i.e. the signal area ratio (hCGβcf/CytC) was zero. Around this concentration (~0.3 to 0.5 % w/v), only hCGβcf specifically bound to the surface activated discs (with antibody) was measured (open squares). At a detergent concentration of 1.0 % (w/v), the signal area ratio decreased greatly, most probably because of disrupted antibody antigen interaction. From this data, 0.5 % (w/v) Triton X-100 was chosen as the most suitable strength for our washing protocol. The concentration of Triton X-100 chosen in this study was five times higher than what was used by others who have incorporated a detergent washing technique in their methodology.
In coating "surface activated discs" with antibody, it is desirable to use a quantity of the antibody that will yield a maximized mass spectrometric response. Too little antibody chosen for surface coating will result in a diminished response, as will too much, paradoxically, because an excessive surface density of antibody can result in diminished antigen-binding efficiency. In order to find an optimal antibody concentration for immobilization, immunoaffinity discs were coated with varying concentrations of monoclonal antibody. This resulted in immunoaffinity discs exhibiting different antibody densities on their surfaces. After incubation with antigen, the resulting standardized MALDI signal areas for hCGβcf were plotted against the antibody concentration used for immobilization (Figure 5). Signals from hCGβcf trapped on surfaces with randomly immobilized antibody are shown in the bottom trace. Expectedly, signals increased with higher loading of antibody. However, the MALDI response decreased considerably for higher antibody densities resulting in a bell-shaped curve. We were able to identify the hCG antibody coating concentration of 5 μg/ml to yield a maximized MALDI response. This phenomenon has also been observed with preparation of immunoaffinity chromatography matrices, where it was demonstrated that the antigen binding capacity decreases with increased antibody coupling efficiencies and may be explained by steric hindrance of antibody active sites at higher surface densities preventing interaction with antigen. Indeed, Cooper et al. found that surface crowding of a self-assembled monolayer on gold presenting D-alanine resulted in lower affinity binding of the antibiotic chloroeremomycin as monitored with a surface plasmon resonance biosensor.
The binding capacity of immunoaffinity targets may be attenuated because of inappropriate antibody orientation and steric crowding. Our antibody orientation approach using recombinantly engineered orientated Protein G yielded higher MALDI MS read-outs for all antibody concentrations analyzed compared to random immunoaffinity discs (Figure 5). All experiments were carried out in triplicates. For the three employed Protein G concentrations, 5.0, 1.0 and 0.5 μg/ml, standardized signals intensified with increasing antibody concentrations from 100 ng/ml, reaching a maximum signal at 5 μg/ml, albeit there being very little difference in hCGβcf signals between the Protein G concentrations used. The signal amplification for Protein G discs
in the maximum was around three-fold compared to those having antibody randomly immobilized. Generally, oriented coupling techniques offer a antigen-binding capacity which is of a factor two to eight higher than efficiencies obtained with random coupling methods. For antibody coating concentrations above 5 μg/ml the following situation was observed: The trace corresponding to the lowest Protein G concentration leveled off for the higher MAb concentrations suggesting free accessibility to all antigen-binding sites on the surface. Here, the Protein G was spatially spread out on the surface so that steric crowding was prevented, even for higher antibody concentrations. Hindered accessibility of antigen binding sites was not observed for this Protein G concentration. In fact, the plateau is explained by saturation of surface immobilized antibodies with hCGβcf. Upon increasing the Protein G concentration, the traces descended for the higher antibody concentrations and showed bell shapes. They followed the same pattern as the trace for surfaces with randomly immobilized antibody. As demonstrated here, crowding on solid supports can also be observed with oriented antibodies. Moreover, the binding capacity seems to decrease as the loading of Protein G and therefore oriented antibody increases. Generally, a combination of low antibody density and oriented coupling may represent the best approach to prepare a high capacity affinity-capture surface.
The adsorption equilibrium between immobilized antibody and antigen and therefore the performance of the immunoaffmity disc depends upon the loading capacity (a function of antibody surface density), antigen concentration, incubation time, pH, temperature, ionic strength, etc. To demonstrate the difference in loading efficiency between random and oriented immobilization, the discs were incubated with increasing concentrations of hCGβcf. Activated discs with or without Protein G were coated with 5 μg/ml antibody to give maximized MALDI signals. For oriented immobilization a Protein G concentration of 0.5 μg/ml was employed to minimize steric crowding. Incubation of hCGβcf for 1.5 hours at room temperature resulted in curves that plateau at 500 or 1000 fmol/μl hCGβcf incubation concentration for random (closed squares) or oriented (closed circles) immobilization respectively (Figure 6). MALDI signals in the plateau region of the curve were almost 3 times higher for oriented antibody surfaces. This plateau region of both curves represents surface saturation with antigen.
In keeping with this observation, the maximum binding capacity for the oriented immobilized antibody, surface was ~3 times higher than for the random immobilized discs (point of inflexion of -900 μmol/1 vs -300 fmol/μl). MALDI surface area ratios for hCGβcf analyzed from 10 pmol/μl using 0.5% (w/v) Triton X-100 on a random MAb surface (Figure 7) are the same as for the plateau region of the bottom trace in Figure 7; which again demonstrates antigen surface saturation. Under the chosen incubation conditions, 100 fmol/μl was the smallest hCGβcf concentration detectable for both types of surfaces. As an adjunct, lower concentrations such as 25 fmol/μl (equates to 250 ng/ml) were routinely observed with both types of surfaces after incubation for 24 hours at +4°C but further investigations are required to establish the optimum conditions to achieve maximum sensitivity (limit of detection).
The final washing step with purified water removed remnant salts from the surface of the immunoaffinity disc, because salts are likely to diminish the spectral quality and deteriorate resolution. Salt removal leads to increased spectral resolution that is demonstrated by the fact that even for small amounts of captured hCGβcf glycoforms started to resolve. For comparison, only very low spectral resolution was observed for hCGβcf in a previous study on conventional stainless steel targets. On-target sample desalting to improve spectral quality was previously achieved by employing SAM technology to manufacture a Cl 8 surface on a MALDI probe.
To illustrate the difference in MALDI signal intensity, two exemplary spectra generated from a 1 pmol/μl solution of hCGβcf in PBS trapped on a random (5 μg/ml hCG MAb) and oriented surface (0.5 μg/ml Protein G; 5 μg/ml hCG MAb) were overlaid (Figure 7). An intensified signal and enhanced spectral resolution can be clearly observed for oriented antibody discs.
Immunoaffinity capture of hCGβcf from a test solution
The performance of the random and oriented immunoaffinity discs prepared was examined on a test mixture containing various peptides and proteins. The conventional MALDI spectrum (on a stainless steel target) of this mixture was recorded from a solution containing no salts as the PBS buffer ions suppress analyte signals (Figure 8A).
Apart from the typical hCGβcf doublet signal, the spectrum shows the [M+H]+ for oxidized insulin chain B (peak 1), the [M+2H]2+ and [M+H]+ for bovine α-lactalbumin (pealcs 2 and 6 respectively), an unknown contaminant (peak 3), the [M+2H]2+ and [M+H]+ for human growth hormone (pealcs 4 and 9 respectively), two polypeptides present in the from pregnancy urine purified hCGβcf (peaks 5 and 8 respectively), the [M+2H]2+, [M+H]+ and [2M+H]÷ for bovine carbonic anhydrase (pealcs 7, 10 and 12) and finally the [M+2H]2+ and [M+H]+ for bovine serum albumin (peaks 11 and 75 respectively). The signal intensity for hCGβcf was. diminished in the presence of the other proteins and peptides. This mixture was incubated on random (Figure 8B) and oriented (Figure 8C) immunoaffinity discs to capture hCGβcf and following washing, the external standard cytochrome C was added with the MALDI matrix in order to compare the amounts captured by both surfaces. For cytochrome C, signals for the single and double charged molecular ions were observed, the peak area of [M+H]+ being used for signal area standardization. Both immunoaffmity surfaces were able to capture and enrich hCGβcf on-target and to isolate it from the deliberately added contaminant proteins and peptides, which were either entirely or almost completely removed by washing. An unidentified component was retained (peak 8), suggesting a polypeptide molecule that may be structurally related to hCG beta. Relative MALDI signals for hCGβcf on the oriented antibody surface were again approximately 3 times higher compared to the random antibody surfaces.
All documents referred to above are hereby incorporated herein by reference.