|
|
||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Proteomics and Protein Markers |
1 Critical Care Medicine Department and 2 Department of Laboratory Medicine, Clinical Center, National Institutes of Health, Bethesda, MD.
aAddress correspondence to this author at: National Institutes of Health, Department of Critical Care Medicine, Bldg. 10, Rm. 2C145, 10 Center Dr., Bethesda, MD 20892-1662. Fax (301) 402-1213; e-mail nseam{at}cc.nih.gov.
| Abstract |
|---|
|
|
|---|
Methods: Depletion of albumin from pooled serum, performed using IgY immunoaffinity spin columns, was assessed for 21 runs on each of 4 columns. We measured albumin concentrations, after albumin depletion, by use of an immunoturbidimetric assay on the Beckman LX 20 analyzer and assessed mass spectra of albumin-depleted samples by use of SELDI-TOF mass spectrometry.
Results: There was substantial run-to-run variation in efficiency of albumin depletion, with systematic decline in efficiency after multiple uses of the columns. Mean depletion efficiency was >95% for 15 of the 1st 17 runs and <90% for runs 18 to 21. We evaluated the 10 highest-intensity peaks present in all spectra from runs 1, 8, 17, and 21 and assessed the effect of albumin depletion on SELDI-TOF mass spectrometry reproducibility. Comparing the %CV of relative intensities for low and high m/z measurements revealed a significant difference of run 21 compared with runs 1, 8, and 17 (P <0.0001). Six-fold more peaks were found in albumin-depleted than unfractionated serum at both high and low m/z.
Conclusions: Sporadic and systematic variation in efficiency of albumin depletion by spin columns may contribute significant preanalytical bias to proteomic approaches of biomarker discovery. This variation requires ongoing QC of the albumin depletion process by quantification of albumin concentration to assess depletion efficiency.
| Introduction |
|---|
|
|
|---|
Removal of major proteins to enhance proteomic analysis has been widely studied. Methods to remove highly abundant proteins include those based on immobilized dye (10)(11) and immunoaffinity-based depletions (12)(13). Immunoaffinity-based depletion techniques appear to provide the most specific method for removal of highly abundant proteins (14)(15). One immunoaffinity-based method uses IgY antibodies, derived from the egg yolks of chickens, linked to microbeads to separate plasma and serum proteins (16)(17)(18). Chicken IgY antibodies may be superior to other immunoaffinity-based albumin depletion methods because of their high affinity and specificity for albumin from multiple mammalian species (19), so we used them in this experiment.
An appropriate assessment of the reproducibility of albumin depletion data derived from SELDI-TOF MS is important because early studies using this technology were hampered by lack of standardization and reproducibility of protein profiling (20)(21). The latter may be related to both preanalytical and analytical biases (22). With the use of rigorous calibration and standardization techniques, reproducible protein profiling of unfractionated serum was achievable within a single laboratory and between 6 different laboratories during different time periods (23). Prefractionation steps such as albumin depletion have become commonplace in protein profiling techniques, but the preanalytical bias that could be introduced with variability in depletion of albumin by manual spin columns has not been adequately assessed.
We sought to prospectively evaluate potential preanalytical bias, and thus the reproducibility of prefractionation by IgY albumin depletion, by SELDI-TOF MS. We describe the efficiency of immunoaffinity-based albumin depletion using spin columns and compare the SELDI-TOF mass spectra of unfractionated, albumin-bound, and albumin-depleted serum.
| Materials and Methods |
|---|
|
|
|---|
albumin depletion
We packed chicken IgY-antialbumin microbeads (GenWay Biosciences) into 4 spin columns (packed volume of 1 mL). We diluted serum aliquots (20 µL) with 480 µL Tris (hydroxymethyl) aminomethane (Tris)-buffered saline (TBS), pH 7.4, and mixed them with the solid-phase antialbumin beads. Unbound proteins were eluted from the spin columns, and columns were washed with TBS before elution of bound albumin with 100 mmol/L glycine (pH 2.5), followed by neutralization of the eluate with 100 mmol/L Tris, pH 8.0. We pooled 2 successive elutions with the pH 2.5 buffer. This process was performed concomitantly across all 4 spin columns for each run, and 2 runs were performed each day. Although the manufacturer suggested buffer-only cleaning runs when column capacity decreased, we omitted this step so that the natural lifespan of the spin columns could be characterized. All proteins were stored at –80 °C before analysis. We measured albumin concentration by use of a previously described immunoturbidimetric method on an LX20 analyzer (Beckman Coulter) (24). Ineffective albumin depletion was defined as depletion efficiency of <90%, and calculated as follows:
[(concentration in unfractionated serum – concentration in depleted serum) x 100]/concentration in unfractionated serum (25).
This definition was based on the assertion by the column manufacturer that a functioning column depletes >95% of albumin (16), allowing for minimal variation in the assay measurement of albumin concentration. When the mean depletion efficiency was <90% for 3 runs, no further depletions were performed.
albumin depletion reproducibility analysis
We assessed column reproducibility by use of 1-way ANOVA (JMP; SAS Institute), examining the runs as the factor of interest to determine the cutoff where albumin depletion efficiency remained >90%. We measured column-to-column variability by use of 1-way ANOVA of the percentage albumin depletion across all 4 columns for all 21 runs. The experiment concluded after run 21 because the mean depletion efficiency was <90% for 3 consecutive runs.
analysis with seldi-tof ms
Based on the albumin depletion reproducibility analysis, we evaluated the samples from runs 1, 8, 17 (the last effective albumin depletion), and 21 [ineffective depletion (efficiency 87%)] with SELDI-TOF MS. Unfractionated serum and albumin-depleted and albumin-bound aliquots were thawed and centrifuged at 13 000g for 5 min to remove insoluble material. Samples (10 µL) were mixed with 90 µL binding buffer (100 mmol/L Tris, pH 9.0) and added to preequilibrated Q10 ProteinChips (Ciphergen Biosystems) using the Biomek 2000 automated workstation (Beckman-Coulter). After 1-h incubation with horizontal shaking, unbound proteins were removed by 2 washes with binding buffer (100 mmol/L Tris, pH 9.0) followed by 2 final rinses with doubly distilled water. Each wash step was performed with 5-min horizontal shaking. The ProteinChips were removed from the bioprocessor and air dried. A 100% saturated solution of sinapinic acid (3,5-dimethoxy-4-hydroxycinnamic acid) (Fluka) prepared in 0.5% trifluoroacetic acid and 50% acetonitrile comprised the matrix, which was applied robotically as 2 1-µL aliquots.
We analyzed ProteinChips by use of SELDI-TOF MS on the PBS IIc (Ciphergen Biosystems) at low and high intervals to optimize reading of the spectra. Low m/z reads were performed at unitless laser intensity 190, sensitivity 7, and collection interval 2 to 20 kDa; high m/z measurements were performed at unitless laser intensity 195, sensitivity 7, and collection interval 7 to 200 kDa. Mass spectra were generated by using a mean of measurements across each spot with a spot protocol programmed to read 10 times at every 5% interval from 20% to 80%. The PBS-IIc Reader was calibrated daily with external standards (All-in-1 Protein Standard II; Ciphergen Biosystems).
data analysis of seldi-tof ms spectra
We analyzed the low and high m/z spectra separately. The spectra were baseline subtracted, normalized by total ion current, and then mass aligned (CiphergenExpress Client 3.0). Supervised hierarchical clustering was performed with CiphergenExpress 3.0. Peaks were picked with criteria of >10 times signal-to-noise ratio on the 1st pass and 5 times on the 2nd pass for cluster completion. We limited cluster bin width to 0.3% of the m/z and excluded estimated peaks from clusters. All processed spectra were exported as peak cluster data and imported to JMP for analysis. We assessed peak cluster data by use of a 3-factor ANOVA for replicate, column number, and depletion run.
| Results |
|---|
|
|
|---|
|
effect of albumin removal efficiency on seldi-tof ms
We compared SELDI-TOF MS data from run 21 to runs 1, 8, and 17 by use of an ANOVA examining the %CV of the relative intensities of the top 10 peaks present in all spectra of these 4 runs. This analysis allowed for a comparison of the downstream effect of ineffective vs effective albumin depletion on peak data. The data from run 21 were significantly different from those of the other 3 runs (P <0.0001) for both the high and low m/z measurements (even with poor-performing column 4 excluded), showing that when albumin depletion efficiency was <90%, SELDI-TOF MS data were not reproducible. We assessed the effect of IgY albumin depletions on SELDI-TOF MS reproducibility by evaluating the 10 highest-intensity peaks present in all spectra for all 3 columns in runs 1, 8, and 17. There were no significant differences between columns 1, 2, and 3 in the %CV of relative intensities by 3-way ANOVA for column, run, and replicate at both low and high m/z measurements (all not significant).
Fig. 2
contrasts the high m/z read SELDI-TOF MS from column 1, which efficiently depleted albumin through run 17, and column 4, which was ineffective throughout the experiment. The relative intensity of the albumin peak (m/z 66 431) was higher for all runs with column 4 and run 21 for column 1 (mean intensity 22.01) than the 3 runs in which column 1 efficiently depleted albumin (mean intensity 2.43). This result is consistent with the poor efficiency of albumin depletion with column 4 throughout the experiment and with the other 3 columns beyond the cutoff point of 17 runs.
|
With efficient depletion of albumin, SELDI-TOF MS revealed more protein peaks than were found in unfractionated serum. This effect is apparent in Table 1
, in which the mass spectral peak data are quantified and separated based on the fractions in which the peaks were found. Table 1
shows the distribution of peaks for unfractionated, albumin-bound, and depleted serum for runs 1, 8, and 17 for columns 1, 2, and 3 at both high and low m/z measurements. These 3 runs spanned the entire range of effective albumin depletion for these columns, defined a priori as >90% albumin depletion. Table 1
illustrates the complementary peak information provided by the different fractions. For the high m/z measurements, the depleted fraction provided 13 unique peaks (26% of total peaks) and the albumin-bound fraction provided 11 unique peaks (22% of total peaks), compared with 2 unique peaks (4%) in unfractionated serum. For the low m/z measurements, the depleted fraction provided 8 unique peaks (31% of total peaks) and the bound fraction and unfractionated serum each provided 1 unique peak (4% of total peaks). The absolute number of peaks described above is relatively low owing to the stringency of the criteria we used to pick peaks (see Materials and Methods). We used these criteria to evaluate unambiguous peaks that should be readily reproducible.
|
Other preanalytical processes such as sample and matrix manipulation by robotic handling are potential sources of bias in MS data; therefore, we evaluated the effects of these processes on the data set by examining the reproducibility of the relative intensity of common SELDI-TOF MS peaks (26)(27). We evaluated chip-to-chip variability by measuring the %CV for the 10 highest-intensity peaks of unfractionated serum samples that had been simultaneously spotted within the same bioprocessors as the albumin-depleted specimens described above. For bioprocessor 1, the mean CV was 12.92% for the low read and 7.57% for the high read. For bioprocessor 2, the mean CV was 25.36% for the low read and 12.30% for the high read. Similarly, we assessed bioprocessor-to-bioprocessor variability by comparing replicates of albumin-depleted samples across 2 bioprocessors, again using the %CV of the top 10 peak intensities. The mean CV for specimens from all 3 columns and runs 1, 8, 17, and 21 was 10.70% for the low read and 16.04% for the high read. These data suggest that neither chip-to-chip nor bioprocessor-to-bioprocessor variability contributed significantly to preanalytical biases, and most likely, inefficient depletion was the source of the lack of reproducibility of SELDI-TOF MS data from run 21 compared with the other 3 runs.
| Discussion |
|---|
|
|
|---|
Other studies have addressed aspects of QC in the depletion of highly abundant proteins. Huang et al. (29) studied the specificity and reproducibility of an immunoaffinity-based IgY spin column packed with microbeads to deplete the 12 most abundant plasma proteins. They found this method to be specific and reproducible based on analysis with protein assays, 2-dimensional gel electrophoresis, and liquid chromatography-tandem MS of proteins excised from gels. They also found that 1 spin column was recyclable for up to 20 different column runs.
Roche et al. (30) showed that IgY-based immunoaffinity spin column depletion of serum allowed detection of unique low-abundance protein peaks from SELDI-TOF MS that were not present in the bound serum. These investigators, however, assessed the repeatability of SELDI-TOF spectra by analysis of the same serum sample 8 times in 1 day, and reproducibility by analysis of the same sera on 4 consecutive days. Thus, their findings may not mirror real-world proteomic laboratory conditions, in which albumin depletion and proteomic profiling methods are often performed over a series of experiments spanning several weeks or months.
Although automated HPLC-based immunodepletion is likely more reproducible, we used spin columns, which are widely used for immunodepletion because of their ease of use and setup. Whereas the absolute number of times spin columns can be used to deplete albumin is variable, in part because of differences in operator techniques, we found that immunoaffinity-based IgY spin columns packed with microbeads were recyclable for 17 uses before the efficiency of albumin depletion decreased to <90%. Potential problems that could contribute to variable and decreased spin column lifespan include variations in sample and bead mixing or sample pipetting, or variations in the beads themselves. Visible loss of microbeads into the eluate occurred near the point at which 3 of the 4 columns stopped efficiently depleting albumin. This result suggests that when visible column breakdown occurs during manual depletion of albumin through a spin column, these columns should be discarded and new columns used. We also describe several instances, however, in which there was no visible evidence of column failure but the depletion efficiency was <90%.
We have shown that inefficient depletion leads to mass spectra that differ from those obtained with efficient depletion and that failure of effective albumin depletion cannot always be predicted without calculation of efficiency of the depleted protein. We used the relative intensity of the top 10 peaks that were picked by stringent criteria to examine reproducibility, because these peaks are most likely to be reproducible. Lack of reproducibility of these peaks suggests that low-abundance biomarkers, which would presumably be more susceptible to small changes in sample preparation, are more affected by variability in albumin depletion.
Thus, QC processes are necessary for manual immunodepletion before proteomic methods because the manual spin-column lifespan is variable and downstream data are less reproducible after inefficient depletion. We describe quantitative albumin concentration calculation as a reasonable method to evaluate column breakdown. Such a quality assurance measure is not time-consuming or cost prohibitive when 1 highly abundant protein is depleted. The use of IgY columns to deplete multiple highly abundant proteins is widespread, however, and although reproducibility of depletion should be verified in experiments using such columns, QC is much more difficult. In such cases, quantitative albumin concentration may still be a useful QC measure, because changes in albumin concentration would suggest lack of reproducibility in the depletion of the other highly abundant proteins.
We verified that when the columns were effectively depleting albumin, the process of albumin depletion followed by SELDI-TOF MS was reproducible and that the process adds unique information not found in SELDI-TOF MS analysis on unfractionated serum alone (30). Therefore, albumin depletion is a useful prefractionation step before SELDI-TOF MS that does not introduce further preanalytical bias to the data obtained from SELDI-TOF MS when the depleting columns function properly. It is unknown how much influence moderate (90%) compared with higher (95%) depletion efficiency affects downstream proteomic techniques, but it is notable that albumin would still be 1 of the 10 most abundant proteins in serum depleted of 90% to 95% of its albumin content. Therefore, variation in depletion efficiency has a substantial impact on the protein composition of immunodepleted serum.
Our finding of peaks unique to the albumin-bound fraction suggests that depletion of albumin changes the serum proteome such that the albumin-bound fraction should not be discarded, but rather further evaluated for unique peaks. Although albumin binds peptides such as insulin and bradykinin (31), the extent of this peptide binding is unknown (32). Exploration of the albuminome is an evolving area of research interest (33).
| Acknowledgments |
|---|
Financial disclosures: None declared.
Acknowledgments: We thank Dr. Ashaunta Tumblin for help with the experiments and Debra Reda for technical support in figure production.
| References |
|---|
|
|
|---|
The following articles in journals at HighWire Press have cited this article:
![]() |
G. L. Hortin, D. Sviridov, and N. L. Anderson High-Abundance Polypeptides of the Human Plasma Proteome Comprising the Top 4 Logs of Polypeptide Abundance Clin. Chem., October 1, 2008; 54(10): 1608 - 1616. [Abstract] [Full Text] [PDF] |
||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |