Avoiding Nonspecific Interactions in Studies of the Plasma Proteome: Practical Solutions to Prevention of Nonspecific Interactions for Label-Free Detection of Low-Abundance Plasma Proteins Joanna L. Richens, Elizabeth A. M. Lunt, Daniel Sanger, Graeme McKenzie, and Paul O’Shea* Cell Biophysics Group, School of Biology, University of Nottingham, Nottingham, England, NG7 2RD Received June 1, 2009
The molecular constitution of blood can be highly representative of the physiological state of an individual and offers an ideal target for studies of biomarkers. High-abundance plasma proteins, particularly albumin, dominate the plasma proteome, but it is the low-abundance proteins (such as cytokines) that are commonly associated with many pathophysiological states. Several detection strategies, and particularly those that involve label-free detection, are available for low-abundance protein detection in plasma, but all can be severely compromised by the high-abundance of serum albumin. In the present study, we examine the effect of albumin interference on accurate label-free detection by protein microarrays. Albumin was found to disrupt specific antigen-antibody binding interactions of low-abundance proteins. In clinical analysis, where it is imperative to preserve the integrity of samples, depletion of albumin may further undermine quantitative measurements. We have optimized procedures that permit accurate analysis to be undertaken without the need for prior treatment of samples. The emphasis is placed on disrupting nonspecific interactions including both electrostatic (i.e., Colulombic) and electrodynamic (hydrophobic and other nonpolar based) interactions. These protocols appear to be generic with potential applications in several areas of analytical biotechnology. Keywords: Plasma proteome • albumin • thiocyanate • microarray • blood • cytokine
Introduction Recent advances in the appreciation of the Systems behavior of many physiological or molecular-cellular networks have also led to less-anecdotal considerations of blood as an organ. The latter adopted by physiologists largely as a convenience for pedagogy is also of relevance, therefore, when considering the total protein makeup of blood, essentially defining the blood proteome. The dynamics of the blood proteome as a route to an understanding of the protein makeup of blood in terms of other related systems such as the immune system, hemodynamics, signaling, and so forth, but particularly from a disease standpoint, therefore, is now much more of a research focus. In particular, variations of the lower molecular weight components such as cytokines under pathophysiological conditions are becoming much more detectable.1 The multifactorial nature of many diseases, however, causes difficulties in the traditional approach to biomarker studies, focused on identification of individual molecular biomarkers as indicators of disease. Similarly, due to inherent variation of some serum proteins, it is often difficult to determine if a disease is present or if a given measurement may represent high ‘normal’ levels of a protein (i.e., this is responsible for many false-positives with the Prostate Specific Antigen (PSA) * To whom correspondence should be addressed. Tel: +44 (0)115 951 3209. Fax: +44 (0)115 970 9259. E-mail:
[email protected]. 10.1021/pr900487y CCC: $40.75
2009 American Chemical Society
test in prostate cancer). Thus, identification of series of molecular markers of disease and other changes of physiological condition are becoming much more prevalent as opposed to detection of elevated levels of a singular molecular species defining a disease (e.g., PSA in prostate cancer2 and HER2 in breast cancer3-5). Accurate diagnosis and evaluation of disease severity by biomarker analysis is critically dependent upon the correct identification and precise quantification of the defined biomolecules. Previous investigations into plasma biomarkers have been hindered by the presence of several highly abundant proteins that dominate the plasma proteome and mask the lower abundance proteins commonly identified as contributing to disease states.6-8 Albumin, present at a concentration of 35-50 mg/mL,9 constitutes over 50% of the total protein milieu in normal human plasma.10 This equates to approximately 10 orders of magnitude higher than the concentration of cytokines such as IL-6 and IL-8 (CXCL8) which are present in normal human plasma at low picogram per milliliter (pg/mL) concentrations.9 Consequently, the presence of albumin in a sample has been shown to be particularly prohibitive to successful identification of low-abundance biomarkers for many proteomic techniques.11,12 In addition to the difficulties caused by high albumin concentrations, studies are also hindered by the rather broad range of properties that predispose a number of different types of nonspecific binding of albumin. Thus, Journal of Proteome Research 2009, 8, 5103–5110 5103 Published on Web 10/09/2009
research articles albumin possesses clusters of different electric charges located over the surface of the molecule. Bovine (BSA) and human (HSA) serum albumin have net charges of -17 and -15, respectively. This charge is distributed unevenly across the three domains of albumin, with charges of -11, -7, +1 (BSA) and -9, -8, +2 (HSA) for domains I, II, and III, respectively.13 Consequently, albumin may interact quite strongly with other types of molecular structure that are either positively or negatively charged. These binding properties are also augmented by the fact that albumin is also equipped with regions on its surface that are distinctly nonpolar. This adds nonspecific hydrophobic interactions and perhaps other ‘nonpolar’ interactions to the overall ability of albumin to interact with many biological structures. These all together seem to underlie the reputation of albumin as a ‘sticky’ molecule and combined with its large concentration in blood mean these are significant problems for any technique that is based on label-free (e.g., SPR) or nonspecific (e.g., silver or coomassie stains) detection. Development of accurate diagnostic analysis requires that interference from high-abundance proteins, of which albumin is the main culprit but there are also several others, be minimized. Dilution of plasma would result in concentrations of many cytokines dropping below the sensitivity levels of current detection technologies. A variety of techniques, based on either physicochemical or affinity capture approaches,14 havebeendevelopedtodepletealbuminfromplasmasamples.15-18 Depletion techniques are reported to remove 70-98% of albumin from blood.14,19 While this can aid in the detection of some proteins15,20 it has also been shown to result in decreased levels of some low-abundance cytokines19 probably due to the fact that albumin is a carrier protein.21 In clinical analysis, where it is imperative to preserve the integrity of samples, therefore, depletion of albumin may skew any information on the concentration of the protein of interest and undermine any quantitative measurement. A majority of investigations into the effect of albumin have been undertaken using 2D electrophoresis12,22 or mass spectrometry.23,24 While these tools are invaluable for the identification of individual biomarkers and disease-specific biomolecular patterns, patient diagnosis will be better facilitated by highthroughput, array-based technologies. Here, we investigate the effect of albumin on detection of low-abundance cytokines using protein microarrays and demonstrate possibilities for conducting analysis without the need for prior treatment of samples. The emphasis is placed on disrupting nonspecific interactions including both electrostatic (i.e., Colulombic) and electrodynamic (hydrophobic and other nonpolar based) interactions. We do this as well as part of a program to develop massive high-throughput arrays based on label-free detection of patterns of protein markers in serum and other biological fluids (see, e.g., Richens et al.25).
Experimental Procedures Protein Array Printing. Capture anti-IL-8 antibody (MAB208; R&D Systems), diluted in a 50:50 mixture of glycerol and either phosphate buffered saline (PBS) or Whatman FAST PAK protein arraying buffer, was printed directly onto Schott Nexterion Slide E slides or Whatman FAST slides using an ArrayJet AJ100 noncontact inkjet printer. Capture antibody concentrations (0.1-0.2 mg/mL) and diluents were selected in accordance with manufacturer’s recommendations. Slides were incubated overnight in a humidified (60-70%) environment at room temperature prior to further processing. 5104
Journal of Proteome Research • Vol. 8, No. 11, 2009
Richens et al. Table 1. Composition of Wash Buffers PBS
total phosphate (mM)
total thiocyanate (mM)
1 2 3 4 5 6 7
5.7 5.7 2.85 5.7 2.85 5.7 2.85
2.85 2.85 10 10 100 100
Array Processing: Epoxide-Coated Slides. Microarrays printed on Nexterion Slide E slides were processed in accordance with the manufacturer’s instructions. Briefly, the slides were washed with PBS-0.1% Tween (3 × 5 min), blocked with 1% BSA in PBS-0.1% Tween for 1 h and washed again (3 × 5 min). For the initial exploratory experiments, slides were then exposed to samples containing IL-8 (250 ng/mL) in the presence or absence of HSA-TRITC (1 mg/mL) for 1 h. Slides were washed and incubated for 1 h with anti-IL-8 antibody (AB-208-NA; R&D Systems) that had been conjugated using the Alexa Fluor 488 Protein Labeling Kit (A-10235; Invitrogen). Finally, slides were washed (3 × 5 min) and spin-dried by centrifugation at 1700 rpm. For all other experiments, slides were exposed to solutions of either HSA (50 mg/mL; Sigma) only, IL-8 (100 pg/mL; Abcam) only or a combination of both HSA (50 mg/mL) and IL-8 (100 pg/mL) for 1 h. Slides were then washed to remove any unbound proteins (3 × 5 min) and incubated with 400 ng/ mL biotinylated detector anti-IL-8 antibody (BAF208; R&D Systems) for 1 h. Finally slides were washed to remove unbound antibody (3 × 5 min), exposed to streptavidin conjugated with alexa-fluor-647 (5 µg/mL; Invitrogen) for 1 h, washed (3 × 5 min) and spin-dried by centrifugation at 1700 rpm. Array Processing: Nitrocellulose-Coated Slides. Microarrays printed on FAST slides were processed using Whatman FAST PAK reagents according to manufacturer’s instructions. Briefly, the slides were blocked with Protein Array Blocking Buffer for 1 h followed immediately by incubation with either HSA (50 mg/mL) only, IL-8 (100 pg/mL) only, or a combination of both HSA (50 mg/mL) and IL-8 (100 pg/mL) for a further 1 h. Slides were then washed with Protein Array Wash Buffer (3 × 5 min), incubated with 400 ng/mL biotinylated detector anti-IL-8 antibody for 2 h, washed (3 × 5 min) and exposed to streptavidin conjugated with alexa-fluor-647 (5 µg/mL) for 1 h. Finally, slides were washed (3 × 5 min) and dried in an oven at 60 °C for 5 min until white in color. Image Analysis. Slides were scanned using a GenePix 4000AL microarray scanner (Axon Instruments). The laser power was 100% and the microarrays were scanned under various PMT gains. PMT gains were kept constant within any one set of experiments. Image analysis was performed with GenePix Pro software (version 6.0). Relative signal intensity was calculated as the difference between signal mean and background mean, followed by adjustment for buffer only controls. Assessing Suitability of Buffer Conditions. Assays were undertaken on Nexterion Slide E slides and processed as described above. Wash buffers, based upon PBS-0.1% Tween, were altered to include varying levels of thiocyanate and phosphate as illustrated in Table 1 with the emphasis on controlling the concentration of the chaotropic anion without large ionic strength changes. Decreases of the ambient ionic strength could lead to a decrease of binding, but any changes of the ionic strength as a result of experimental need were always as the larger concentration and thus would promote
Avoiding Nonspecific Interactions in Plasma Proteome Studies protein binding. As the changes we implemented were small, however, such effects were deemed negligible. The pH buffering capacity was broadly similar but always enough to ensure that no pH changes could take place. Buffers were all adjusted to pH 7.4 and stored at 4 °C for no longer than 1 week. Slides were blocked using 1% BSA diluted in either PBS-0.1% Tween or in the appropriate wash buffer. Standard curves were generated using varying levels of IL-8 (50-350 pg/mL) in the presence/absence of HSA (50 mg/mL). Human plasma (Sigma) was spiked with 100 pg/mL IL-8 and assayed as previously described with unspiked plasma, 100 pg/mL IL-8 and 50 mg/ mL HSA acting as control samples.
research articles
informed consent from Severalls Hospital, Colchester, U.K., and urine was kept frozen at -70 °C until used. The Bence Jones paraprotein, λRG57, was isolated from the urine by a two-step procedure involving protein concentration followed by chromatographic purification as described previously.28 Protein purity was monitored following each procedure using agarose gel electrophoresis. 3. Binding Profile Analysis. Serial additions of λRG57 were made to FPE-labeled erythrocytes, suspended in different media. Experiments were performed in 100 mM KCl, 100 mM
Membrane Preparation and Labeling with Fluoresceinphosphatidylethanolamine (FPE). Egg phosphatidylcholine (PC) was supplied by Lipid Products (U.K.). FPE was synthesized as previously described.26 Membrane vesicles were prepared as previously described.26 Briefly, PC dissolved in chloroform in a round-bottom flask was dried under a stream of oxygen-free nitrogen gas by rotary evaporation until a thin film was formed. The lipid film was rehydrated with 4 mL of PBS pH 7.4. The resulting multilamellar solution was frozen and thawed 5 times and finally extruded 10 times through 25mm diameter polycarbonate filters with pores 1 µm in diameter (Nucleopore Corp.) using an extruder (Lipex Biomembranes, Inc.) according to previously described extrusion procedures.27 This resulted in a monodisperse, unilamellar suspension of phospholipid vesicles which were labeled in the outer bilayer leaflet with FPE as previously described.26 Briefly, the phospholipid vesicles were incubated with ethanolic-FPE (never more than 1% ethanol of the total aqueous volume) at 37 °C for 1 h in the dark. Any remaining unincorporated FPE was removed by gel filtration on a PD10 Sephadex column equilibrated with PBS. Fluorescence Measurements. Fluorescence spectroscopy was conducted on a PerkinElmer Spectrofluorimeter LS55 with emission and excitation wavelengths set at 520 nm and 490 nm, respectively. FPE-labeled PC vesicles (400 µM) were dissolved in either PBS or PBS-SCN (100 mM Cl- replaced with 100 mM SCN-). Serial additions of 50 mg/mL HSA dissolved in the appropriate buffer were added every 100 s, stirring constantly. Appropriate controls whereby HSA additions were replaced with buffer additions were undertaken. Binding Interactions of Paraproteins with Erythrocytes. 1. Preparation of FPE Labeled Erythrocyte. Whole red blood cells from healthy donors were provided by local General Hospitals. Whole blood was stored at 4 °C and was anticoagulated with EDTA (1.5 mg in PBS/mL of blood). Erythrocytes were isolated by centrifugation of blood at 2500g for 10 s at 4 °C with the plasma and leukocyte layers on the surface of the erthrocytes being removed by aspiration. The packed erythrocytes were washed with sucrose solution (280 mM sucrose, 5 mM Tris at pH 7.5) three times, and after the final wash, the cells were left packed (i.e., 100% hematocrit). To label erythrocytes, 30 µg of FPE was resuspended in 15 µL of 95% ethanol and made up to 500 µL with sucrose solution. Packed blood cells (100 µL) were added to the FPE solution and incubated at 37 °C for 1 h. Unbound FPE was removed by repeated centrifugation at 2500g for 10 s at 4 °C until the supernatant appeared clear; three washes were generally sufficient. Finally, the labeled cells were resuspended in the appropriate buffer, stored in the dark at 4 °C, and used within 8 h. 2. Isolation of the Myeloma Paraprotein, λRG57. Urine from a patient with multiple myeloma was obtained with
Figure 1. The difficulties associated with working with albumin in protein microarray experiments. (A) Nonspecific binding of human serum albumin to a covalent slide surface (Nexterion slide E); (B) the effect of nonspecifically bound albumin on the binding intensity of IL-8. Arrays of anti-IL8 antibody (0.1 mg/ mL), printed on Nexterion Slide E slides (Epoxide), were blocked and exposed to (A) HSA-TRITC (1 mg/mL) or buffer only; (B) IL-8 (250 ng/mL) in the presence/absence of HSATRITC. Detection was via alexa fluor-conjugated anti-IL-8 (AB208-NA; 250 µg/mL). Slides were scanned using a GenePix 4000AL microarray scanner and image analysis was performed with GenePix Pro software (version 6.0). Journal of Proteome Research • Vol. 8, No. 11, 2009 5105
research articles
Richens et al.
Figure 2. The interference of albumin in protein microarrays. (A) The effect of albumin on IL-8 binding. (B) Nonspecific binding of albumin to the printed capture antibody. Capture anti-IL8 antibody was printed onto Nexterion Slide E slides (Epoxide) or FAST slides (Nitrocellulose) at (A) 0.1 mg/mL and (B) 0.1-0.2 mg/mL. Arrays were blocked and then exposed to solutions of HSA only (50 mg/mL), IL-8 only (100 pg/mL) or a combination of both HSA and IL-8. Sandwich assay detection was achieved using biotin-conjugated anti-IL-8 antibody (400 ng/mL) exposed to streptavidin-alexa fluor 647. Slides were scanned using a GenePix 4000AL microarray scanner and image analysis was performed with GenePix Pro software (version 6.0). Values represented are mean ( SE for seven repeat experiments.
KSCN and 280 mM sucrose solutions, all buffered with 5 mM Tris at pH 7.5. Fluorescence measurements were performed using a Perkin-Elmer LS-50 spectrofluorimeter (Perkin-Elmer, U.K.) with the excitation and emission wavelengths set at 490 and 518 nm, respectively. Binding curve analysis was performed with Ultrafit software (BIOSOFT, U.K.) and dissociation constants and the relative capacities of binding were determined by fitting curves to simple binding models The fitting used 95% confidence limits and was carried out using the Gauss-NewtonMarquardt algorithm. Robust weighting which rejected outliers with a rejection factor of 6 was employed.
Results Human serum albumin conjugated with a fluorescent label was used to explore the extent of albumin interference during protein microarray procedures directed toward serum protein analysis. IL-8 was chosen as an example of a low-abundance plasma protein but also forms part of a broader research focus of our laboratories (reviewed in Richens et al.25). Figure 1A indicates that even at a concentration of 1 mg/mL, 50 times lower than physiological levels, there is a large amount of nonspecific binding of albumin to a slide surface. Importantly, Figure 1B shows that the presence of albumin also promotes a decrease in the intensity of the signal observed from IL-8 binding. ‘Mock plasma’ samples comprising different combinations of HSA and IL-8, at concentrations chosen to match closely those found physiologically, were used to investigate in further detail the effect of nonspecific albumin binding on detection levels of IL-8. Two different slide surfaces, epoxide (covalent attachment) and nitrocellulose (noncovalent attachment), were examined. The presence of albumin in the mock plasma caused a significant decrease in the IL-8 binding intensity on the epoxide-coated slides (p ) 0.002). While the same pattern was observed on the nitrocellulose-coated slides, Figure 2A indicates that the decrease in IL-8 binding due to the presence of albumin did not reach significance (p ) 0.061). If there was no difference between the intensity of IL-8 binding observed in the presence and absence of HSA, the IL-8/HSA + IL-8 ratio would equate to one. During these experiments, IL-8 only/HSA + IL-8 ratios of 1.27 and 1.11 were observed on epoxide- and nitrocellulose-coated slides, respectively. On both slide surfaces, the background signal caused by nonspecific binding of albumin to the printed spot shown in Figure 2B varied according to the concentration at which the capture antibody was printed (epoxide surface, p ) 0.064; nitrocellulose surface, p ) 0.0006). 5106
Journal of Proteome Research • Vol. 8, No. 11, 2009
Attenuation of Nonspecific Binding Interactions. The ultimate aim of this work is the development of disease-specific microarray chips based on a label-free interrogation assay for use in disease diagnosis.25 The detection technology is based on a sensitive variant of surface plasmon resonance (SPR) that will identify the binding of selective molecules (i.e., the ligands) to an array surface. SPR requires the ligand to be immobilized very close to the surface (ca. 100 nm) to maximize the response seen from antigen binding. This limits the usefulness of nitrocellulose slides in such technologies as the thinnest coating available measures approximately 500 nm. With this is mind, and due to the comparative results obtained on both the epoxide and nitrocellulose surfaces (Figure 2), subsequent studies were undertaken with epoxide-coated slides. A variety of wash buffers containing different concentrations of thiocyanate and phosphate were examined in an attempt to eliminate nonspecific binding of albumin to both the slide surface (Figure 1) and the capture antibody (Figure 2B). These studies were undertaken using blocking solutions diluted in either the corresponding wash buffer or the normal manufacturer’s recommended buffer (PBS-Tween). Inclusion of 2.85 mM thiocyanate was associated with a significant decrease in the intensity of nonspecific HSA binding (BSA in wash buffer, p ) 0.04; BSA in PBS-Tween, p ) 0.03). With addition of 100 mM thiocyanate, however, a total eradication of all signal associated with nonspecific albumin binding was observed (BSA in wash buffer, p ) 0.004; BSA in PBS-Tween, p ) 0.058) (Figure 3). Altering the buffer and the blocking conditions resulted in variations in the IL-8 only/HSA + IL-8 ratio. The most improved ratios were observed using wash buffer 5 (blocking solution diluted in buffer 5) or wash buffer 7 (blocking solution diluted in either wash buffer 7 or PBS-Tween) which varied from a ratio of one by 0.07, 0.07 (both Figure 4A), and 0.03 (Figure 4B), respectively. To ensure the uniformity of the thiocyanate media conditions, those that produced the most improved ratios in Figure 4A,B were used to generate IL-8 standard curves in the presence and absence of albumin. The IL-8 only/HSA + IL-8 ratio at each concentration was determined and an average value for each buffer across the concentration gradient was calculated (Figure 4C). The most consistent conditions were obtained using either buffer 1 or buffer 7 in combination with BSA diluted in PBSTween, but the ratio obtained with buffer 7 (0.002) deviated from zero far less than that obtained with buffer 1 (0.230). With the demonstration of the improvement in measurements in the presence of thiocyanate in an albumin model
Avoiding Nonspecific Interactions in Plasma Proteome Studies
research articles
Figure 3. The effect of thiocyanate as the major anion in the medium on albumin binding. Arrays of anti-IL-8 antibody (0.1 mg/mL), printed on Nexterion Slide E slides, were blocked and exposed to solutions of HSA only (50 mg/mL). Arrays were processed using PBS-Tween containing increasing concentrations of thiocyanate (0-100 mM) and blocking solutions were diluted in either PBS-Tween or the corresponding wash buffer. Sandwich assay detection was achieved using biotin-conjugated anti-IL-8 antibody (400 ng/mL) exposed to streptavidin-alexa fluor 647. Slides were scanned using a GenePix 4000AL microarray scanner and image analysis was performed with GenePix Pro software (version 6.0). Data represented are mean ( SE for three repeat experiments.
solution, the optimized buffer (Buffer 7) was transferred to a more physiologically relevant system and compared to the original wash buffer (PBS-Tween). Pooled plasma was spiked with IL-8 (100 pg/mL) and levels of signal intensity were compared to those obtained for unspiked plasma, IL-8 only and HSA only (Figure 5). Processing the arrays with buffer 7 was found to eradicate nonspecific HSA binding (p ) 0.016), while no corresponding variation in the specific IL-8 binding signal was observed. In the pooled plasma samples, use of buffer 7 resulted in a significant decrease in signal intensity for the unspiked plasma samples (p ) 0.049). A decrease in intensity was also observed for the spiked plasma samples, although this did not reach significance. To confirm that these decreases in nonspecific binding events are due to the thiocyanate anions, membrane binding assays were undertaken (Figure 6). Examination of nonspecific binding interactions between HSA and PC membranes confirmed the efficacy of thiocyanate anions in overcoming nonspecific albumin binding (Figure 6A). Interactions undertaken in the presence of thiocyanate resulted in a significantly higher dissociation constant (Kd ) 16.69 µM) than that obtained in the absence of thiocyanate (Kd ) 0.91 µM) indicating lower binding affinity (Figure 6B; p ) 0.0147). While the above results have focused on the nonspecific binding properties of albumin, there are many other nonspecific binding interactions that may be influenced by the properties of SCN-. The physiological effects of multiple myelomas may be due, in part, to the nonspecific interactions that occur between the Bence-Jones proteins (immunoglobulin free light chains) and cell membranes.28 The ability of SCNto interrupt these nonspecific interactions was investigated using an erythrocyte-based in vitro model. Serial additions of the Bence-Jones protein, λRG57, to FPE-labeled erythrocytes suspended in different media resulted in a change in the fluorescence intensity of the membrane bound FPE, representing binding, in each of the media conditions examined (Figure 7). Titrations undertaken in 100 mM KCl, 100 mM KSCN and 280 mM sucrose solutions all exhibited saturation and displayed similar Kd values, of 14.3 ( 2.1, 15.1 ( 4.1, and 11.9 ( 8.6, respectively. The composition of the buffer did, however, have
Figure 4. The effect of buffer conditions on IL-8 only/IL-8 + HSA ratios. Capture anti-IL8 antibody (0.1 mg/mL) was printed onto Nexterion Slide E slides and arrays were processed using buffers containing varying concentrations of thiocyanate and phosphate (Table 1). Blocking solutions were diluted in either (A) PBS-Tween or (B) the corresponding wash buffer. Arrays were exposed to IL-8 (50 pg/mL) followed by biotin-conjugated anti-IL-8 antibody (400 ng/mL) and streptavidin-alexa fluor 647. IL-8 only/IL-8 + HSA ratios were calculated and values are plotted as the deviance of the ratio from zero. (C) IL-8 only/IL-8 + HSA ratios measured over a range of IL-8 concentrations. IL-8 standard curves (50-350 pg/ mL) were generated in the presence/absence of HSA (50 mg/ mL) using different buffer conditions. Mean ratios were calculated across the concentration gradient and values are plotted as the deviance of the ratio from zero. All slides were scanned using a GenePix 4000AL microarray scanner and image analysis was performed with GenePix Pro software (version 6.0). Data represented are mean ( SE for three repeat experiments.
an effect on the capacity of binding of λRG57 to the erythrocytes. The nonspecific binding capacity in a 100 mM SCNsolution was 5-fold lower than that observed in a KCl solution. In contrast, the binding capacity observed in a sucrose solution was only decreased 2-fold from those observed in the KCl solution.
Discussion The difficulties associated with analysis of low-abundance proteins in plasma samples containing high albumin concentrations are well-documented and present obstacles to proteomic technologies such as 2D-electrophoresis and mass Journal of Proteome Research • Vol. 8, No. 11, 2009 5107
research articles
Figure 5. The effect of buffer conditions on IL-8 measurements within human plasma. Capture anti-IL8 antibody was printed onto Nexterion Slide E slides at 0.1 mg/mL. Arrays were blocked and exposed to solutions of HSA only, IL-8 only, unspiked plasma or plasma spiked with IL-8 (100 pg/mL) followed by biotinconjugated anti-IL-8 antibody (400 ng/mL) and streptavidin-alexa fluor 647. Blocking solutions were diluted in PBS-Tween and arrays were processed using either buffer 1 or buffer 7 (Table 1). Slides were scanned using a GenePix 4000AL microarray scanner and image analysis was performed with GenePix Pro software (version 6.0). Data represented are mean ( SE for three repeat experiments.
spectrometry.6-8,11,12 Problems arise due to the vast dynamic range of proteins found within plasma with at least 10 orders of magnitude separating the most abundant plasma protein (e.g., albumin) from low-abundance protein targets.9 Such difficulties will also be inherent in array-based technologies. The extent of interference by albumin and other high-
Richens et al. abundance proteins on array-based detection systems has not, however, been well-characterized despite the upsurge in their use. Here, model plasma samples, containing human serum albumin and IL-8 (as a representative low-abundance cytokine), were used at physiological levels to investigate the difficulties needing to be overcome to facilitate accurate plasma proteomic studies by array technologies. The problems encountered occur despite highly selective molecular-recognition-based protein identification technologies, that is, antibody-based, being employed. Use of a fluorescently conjugated HSA demonstrated that, in addition to the well-established plasma proteomics obstacles, that is, the difference in magnitude between the high- and lowabundance proteins, array based detection systems are also subject to interference from nonspecific binding of nontarget proteins (Figure 1). Nonspecific binding of HSA was shown to interfere with specific binding of the target protein, in this case IL-8 (Figure 2). This effect was observed on slide surfaces that employed both covalent (epoxide) and absorbance (nitrocellulose) antibody immobilization strategies (Figure 2A). The effect was more apparent on the epoxide surface, indicating that some of the effect is mediated through interactions between HSA and the slide surface. Nonspecific binding interactions were also found to occur between HSA and the capture antibody (Figure 2B). There are several ways in which nonspecific binding of HSA may interfere with the detection of cytokines. Nonspecific binding of HSA to the capture antibody could prevent specific IL-8 binding, while interactions between HSA and bound IL-8
Figure 6. Thiocyanate anions reduce nonspecific binding interactions between HSA and PC membranes. (A) The binding profile was derived from the time course fluorescence variations caused by addition of HSA to FPE-labeled vesicles suspended in either PBS (O) or PBS/SCN (b). (B) Kd values derived from binding profiles. The lipid concentration was 400 µM. Temperature was 37 °C. Data represented are mean ( SE for three repeat experiments.
Figure 7. Thiocyanate anions reduce nonspecific binding of λRG57 to FPE-labeled cells. (A) Changes occur in the fluorescence profile of FPE following nonspecific binding of an antibody to the cell surface. (B) The binding profile of λRG57 to FPE-labeled erythrocytes in various media; 100 mM KCl (b), 100 mM KSCN (O) and 280 mM sucrose (9) were all supplemented with 5 mM Tris at pH 7.5. FPE-labeled erythrocytes were suspended at 0.04% hematocrit and experiments were performed at 20 °C. Fluorescence intensity was recorded using a Perkin-Elmer LS-50 spectrofluorimeter (Perkin-Elmer, U.K.) with the excitation and emission wavelengths set at 490 and 518 nm, respectively. We are grateful to Dr. F. Ayoub for help with some of these studies. 5108
Journal of Proteome Research • Vol. 8, No. 11, 2009
research articles
Avoiding Nonspecific Interactions in Plasma Proteome Studies (acting in its function as a carrier protein) could inhibit detection by a secondary antibody. Additionally, nonspecific binding of HSA to the slide surface and the downstream detection reagents may increase background signal, thus, causing a decrease in the overall signal intensity. Our studies were undertaken using fluorescently labeled protein arrays. Interference from nonspecific binding events would, however, be amplified in label-free detection systems such as surface plasmon resonance (SPR)29 where analytes are detected according to their mass and, by definition, lack the secondary detection step which can provide added specificity. In the case of either protein array format, it is critical that the interference from nonspecific binding is minimized or removed completely to provide accuracy and reliability of results, particular in clinical diagnostic tests. Depletion of albumin from plasma samples should reduce this interference but such is the extent of the excess of albumin over, for example, cytokines that even if a small percentage remains it may still be at levels much greater than the cytokines. Moreover, evidence suggests albumin depletion may also disrupt levels of the target proteins19 and preserving the integrity of clinical samples is imperative in ensuring accurate disease diagnosis. We have demonstrated that interference from nonspecific binding of serum albumin can be overcome completely by the inclusion of thiocyanate in the array buffers (Figure 3). By optimizing the conditions of these solutions, we were able to decrease the ratio of IL-8 signal in the absence and presence of HSA from 1.27 to 1.03 (Figure 4A,B). This ratio could be maintained over an IL-8 concentration gradient (Figure 4C) which is essential for quantitative analysis. Application of the thiocyanate containing buffer to arrays examining pooled plasma samples, as shown in Figure 5, demonstrated the appropriateness of this system to physiologically relevant plasma samples. The signal obtained for unspiked plasma samples is likely due to the IL-8 contained within healthy plasma,30 and thus, the molecular patterns do not appear to be compromised by the presence of thiocyanate. We have concentrated our attention on the problems of detecting low-abundance plasma proteins that may be important disease markers, but we emphasize that our proposed solutions to this may well be generic. Thus, we also indicated how nonspecific binding may be reduced in cellular assays of molecular binding. We are not suggesting, however, that our approach will solve all such problems, but it does appear to represent a promising starting point that may contribute to a broader solution. Thiocyanate is a relatively large anion that possesses a very high entropy of hydration and we have demonstrated in the past that it disrupts the interactions of macromolecules with living cells.28 Thiocyanate achieves this essentially by modulating the structure of water in the vicinity of the interacting surfaces of macromolecules and thus disrupting almost selectivelythenonpolareffectsthatunderlieattractive‘protein-protein interactions’. This is coupled to a reduction in the intensity of the charge-charge interactions by virtue of the elevated ionic strengths associated with these modified salines at the order of 150 mM. In other words, reduction of the total salt concentration would augment charge-charge attraction and repulsion (see, e.g., O’Shea 200428); thus, at the thiocyanate concentrations we employ, these interactions are minimized.31 With the electrolytes normally employed for bioassays, such conditions would allow the nonspecific hydrophobic (etc.) interactions to dominate. Thus, the combined virtue of utilizing
-
SCN at the concentrations we employ is that all nonspecific interactions are effectively removed. Other anions and cations associated with the Hofmeister series could also clearly be utilized to accomplish similar effects, but we suggest that SCNoffers an effective compromise of all the effects that come into play. Thus, we suggest that SCN- is the optimal anion for such studies.
Acknowledgment. This project has been funded by the RCUK Basic Technology Programme. We are grateful to Dr. F. Ayoub for help with some of these studies. We would like to thank Dr. Lucy Fairclough and Dr. Paddy Tighe for helpful discussion. References (1) Drake, R. R.; Cazares, L.; Semmes, O. J. Mining the low molecular weight proteome of blood. Proteomics Clin. Appl. 2007, 1 (8), 758– 768. (2) Crawford, E. D.; Abrahamsson, P. A. PSA-based screening for prostate cancer: how does it compare with other cancer screening tests. Eur. Urol. 2008, 54 (2), 262–273. (3) Pal, S. K.; Pegram, M. HER2 targeted therapy in breast cancer. .beyond Herceptin. Rev. Endocr. Metab. Disord. 2007, 8 (3), 269– 277. (4) Rubin, I.; Yarden, Y. The basic biology of HER2. Ann. Oncol. 2001, 12, S3-8. (5) Yarden, Y. Biology of HER2 and its importance in breast cancer. Oncology 2001, 61, 1–13. (6) Gong, Y.; Li, X.; Yang, B.; Ying, W.; Li, D.; Zhang, Y.; Dai, S.; Cai, Y.; Wang, J.; He, F.; Qian, X. Different immunoaffinity fractionation strategies to characterize the human plasma proteome. J. Proteome Res. 2006, 5 (6), 1379–1387. (7) Fountoulakis, M.; Juranville, J. F.; Jiang, L.; Avila, D.; Roder, D.; Jakob, P.; Berndt, P.; Evers, S.; Langen, H. Depletion of the highabundance plasma proteins. Amino Acids 2004, 27 (3-4), 249– 259. (8) Merrell, K.; Southwick, K.; Graves, S. W.; Esplin, M. S.; Lewis, N. E.; Thulin, C. D. Analysis of low-abundance, low-molecular-weight serum proteins using mass spectrometry. J. Biomol. Tech. 2004, 15 (4), 238–248. (9) Anderson, N. L.; Anderson, N. G. The human plasma proteome: history, character, and diagnostic prospects. Mol. Cell. Proteomics 2002, 1 (11), 845–867. (10) Prinsen, B. H.; de Sain-van der Velden, M. G. Albumin turnover: experimental approach and its application in health and renal diseases. Clin. Chim. Acta 2004, 347 (1-2), 1–14. (11) Rengarajan, K.; de Smet, M. D.; Wiggert, B. Removal of albumin from multiple human serum samples. BioTechniques 1996, 20 (1), 30–32. (12) Seferovic, M. D.; Krughkov, V.; Pinto, D.; Han, V. K.; Gupta, M. B. Quantitative 2-D gel electrophoresis-based expression proteomics of albumin and IgG immunodepleted plasma. J. Chromatogr., B: Anal. Technol. Biomed. Life Sci. 2008, 865 (1-2), 147–152. (13) Peters, T., Jr. All About Albumin: Biochemistry, Genetics, and Medical Applications; Academic Press: San Diego, CA, 1995; pp 9-75. (14) Greenough, C.; Jenkins, R. E.; Kitteringham, N. R.; Pirmohamed, M.; Park, B. K.; Pennington, S. R. A method for the rapid depletion of albumin and immunoglobulin from human plasma. Proteomics 2004, 4 (10), 3107–3111. (15) Rothemund, D. L.; Locke, V. L.; Liew, A.; Thomas, T. M.; Wasinger, V.; Rylatt, D. B. Depletion of the highly abundant protein albumin from human plasma using the Gradiflow. Proteomics 2003, 3 (3), 279–287. (16) Linke, T.; Doraiswamy, S.; Harrison, E. H. Rat plasma proteomics: effects of abundant protein depletion on proteomic analysis. J. Chromatogr., B: Anal. Technol. Biomed. Life Sci. 2007, 849 (12), 273–281. (17) Vestergaard, M.; Tamiya, E. A rapid sample pretreatment protocol: improved sensitivity in the detection of a low-abundant serum biomarker for prostate cancer. Anal. Sci. 2007, 23 (12), 1443–1446. (18) Bjorhall, K.; Miliotis, T.; Davidsson, P. Comparison of different depletion strategies for improved resolution in proteomic analysis of human serum samples. Proteomics 2005, 5 (1), 307–317. (19) Granger, J.; Siddiqui, J.; Copeland, S.; Remick, D. Albumin depletion of human plasma also removes low abundance proteins including the cytokines. Proteomics 2005, 5 (18), 4713–4718.
Journal of Proteome Research • Vol. 8, No. 11, 2009 5109
research articles (20) Ahmed, N.; Barker, G.; Oliva, K.; Garfin, D.; Talmadge, K.; Georgiou, H.; Quinn, M.; Rice, G. An approach to remove albumin for the proteomic analysis of low abundance biomarkers in human serum. Proteomics 2003, 3 (10), 1980–1987. (21) Mehta, A. I.; Ross, S.; Lowenthal, M. S.; Fusaro, V.; Fishman, D. A.; Petricoin, E. F., III; Liotta, L. A. Biomarker amplification by serum carrier protein binding. Dis. Markers 2003, 19 (1), 1–10. (22) Qin, S.; Ferdinand, A. S.; Richie, J. P.; O’Leary, M. P.; Mok, S. C.; Liu, B. C. Chromatofocusing fractionation and two-dimensional difference gel electrophoresis for low abundance serum proteins. Proteomics 2005, 5 (12), 3183–3192. (23) Aresta, A.; Calvano, C. D.; Palmisano, F.; Zambonin, C. G.; Monaco, A.; Tommasi, S.; Pilato, B.; Paradiso, A. Impact of sample preparation in peptide/protein profiling in human serum by MALDI-TOF mass spectrometry. J. Pharm. Biomed. Anal. 2008, 46 (1), 157– 164. (24) Dekker, L. J.; Bosman, J.; Burgers, P. C.; van Rijswijk, A.; Freije, R.; Luider, T.; Bischoff, R. Depletion of high-abundance proteins from serum by immunoaffinity chromatography: a MALDI-FT-MS study. J. Chromatogr., B: Anal. Technol. Biomed. Life Sci. 2007, 847 (1), 65–69. (25) Richens, J. L.; Urbanowicz, R. A.; Lunt, E. A.; Metcalf, R.; Corne, J.; Fairclough, L.; O’Shea, P. Systems biology coupled with label-free
5110
Journal of Proteome Research • Vol. 8, No. 11, 2009
Richens et al.
(26) (27) (28) (29) (30)
(31)
high-throughput detection as a novel approach for diagnosis of chronic obstructive pulmonary disease. Resp. Res. 2009, 10, 29– 46. Wall, J.; Ayoub, F.; O’Shea, P. Interactions of macromolecules with the mammalian cell surface. J. Cell Sci. 1995, 108, 2673–2682. Mayer, L. D.; Hope, M. J.; Cullis, P. R. Vesicles of variable sizes produced by a rapid extrusion procedure. Biochim. Biophys. Acta 1986, 858 (1), 161–168. Wall, J.; Ayoub, F.; O’Shea, P. A study of the interactions of an immunoglobulin light chain with artificial and B-lymphocyte membranes. Front. Biosci. 1996, 1, a46–58. Yu, X.; Xu, D.; Cheng, Q. Label-free detection methods for protein microarrays. Proteomics 2006, 6 (20), 5493–503. Martich, G. D.; Danner, R. L.; Ceska, M.; Suffredini, A. F. Detection of interleukin 8 and tumor necrosis factor in normal humans after intravenous endotoxin: the effect of antiinflammatory agents. J. Exp. Med. 1991, 173 (4), 1021–1024. O’Shea, P. Membrane Potentials; Measurement, Occurrence & Roles in Cellular Function. In Bioelectrochemistry of Membranes; Walz, D. , et al., Eds.; Birkhauser Verlag: Switzerland, 2004; pp 23-59.
PR900487Y
