Abstract
Purpose.:
To establish the use of selected reaction monitoring (SRM) mass spectrometry for quantification of tear proteins.
Methods.:
Tear samples were collected on multiple occasions (7–10 days) from healthy subjects with contact lens wear (CL = 3) and without contact lens wear (NCL = 4). Tear proteins were denatured using 8M urea, reduced with iodoacetamide, precipitated by acetone, and digested using trypsin. Internal standards were included by adding isotopically-labelled standards of known concentrations to the samples. Lactoferrin, lysozyme, prolactin-induced protein, lipocalin 1, and proline-rich protein 4 were quantified using liquid chromatography-triple quadruple mass spectrometry in conjunction with selected reaction monitoring.
Results.:
The limits of quantification for the selected peptides were below 50 pg/μL. The recovery of peptides from spiked digested tears was greater than or equal to 56% and the coefficient of variation values were less than or equal to 16%. The concentration of lactoferrin (1.20 ± 0.77 μg/μL), lysozyme (2.11 ± 1.50 μg/μL), and lipocalin-1 (1.75 ± 0.99 μg/μL) were consistent with previous ELISA studies. Tear levels of prolactin-induced protein (0.09 ± 0.06 μg/μL) and proline-rich 4 (0.80 ± 0.50 μg/μL) are reported here for the first time.
Conclusions.:
The SRM method can be used for simultaneous detection and quantification of selected proteins in low volumes of human tear samples (2.5 μL per sample) without prior purification of each protein component or need for antibodies.
Ocular comfort is affected by the quality of the pre-ocular tear film and its stability.
1 Contact lenses can function as superficial foreign bodies and can alter the tear film,
2,3 and thus activate mechanisms leading to contact lens–related adverse events,
4 resulting in ocular discomfort and dryness.
5,6 The mechanisms underlying these events are unknown, but might involve changes in the levels of tear proteins similar to those documented in dry eye conditions.
7
A variety of methods for measuring tear proteins have been described, including ELISA
8,9 and two-dimensional SDS-PAGE analysis (2-DE).
10 The results of such studies depend on antibody cross-reactivity and gel mobilities, and frequently need to be verified by alternative methods such as protein sequence analysis; thus, relatively large quantities of protein are required to achieve reliable results.
11 Other methods, aiming at multidimensional protein identification, have also been employed.
12,13 Whether these alternative methods can be reliably applied to clinical samples is still not clear. In view of the increasing interest for measuring protein levels in tear fluids, an improved method is needed that resolves the drawbacks of previous techniques and allows reliable and repeatable quantification of proteins in complex mixtures such as tears.
Recent studies have described mass spectrometry (MS) as a general method for quantification of proteins in human fluids.
14–19 Proteomics and transcriptional profiling methods have been implemented by other researchers to examine the tear proteome
6,7,20 and have resulted in identification of a number of candidate protein markers for detection of diseases, such as cancer
21,22 or dry eye.
20 Complex humeral fluids other than the tear fluid, including serum and urine, have been subjected to quantification by selective reaction monitoring (SRM).
23,24 By using samples with high and intermediate protein abundance, it has been shown that the SRM method is highly sensitive and can simultaneously quantify up to 40 proteins,
25 while maintaining high mass accuracy in protein profiling. Additionally, using the SRM method, quantitative mass spectrometric analysis of endogenous human peptides can be performed against internal standards. Synthetic peptides labelled with stable isotopes are used for this purpose. These isotopes are identical in chemical structure to the corresponding endogenous peptides despite their difference in mass, rendering the preparations of standards reliable. Thus, the SRM methodology can be proposed as a potentially useful experimental tool for studies aimed at identification of proteins in human tears and their changes during ocular disease, such as dry eye or during contact lens wear.
Previous SRM quantification of complex humeral fluids involved derivatization and depletion of proteins and peptides, requiring an additional purification of proteins.
26,27 In view of the limited availability of tear samples and taking into consideration that only 5 to 10 μL of tears are available for collection at every attempt, the main aim of this study was to establish whether the SRM method could be used for simultaneous quantification of proteins in very small sample volumes of human tears without requiring prior purification of each individual protein component. In order to successfully establish the assays, a limited number of proteins described in the literature as potential markers of dry eye
20 were selected to evaluate accuracy, precision, linearity, limit of detection, and quantification of the SRM method.
Dithiothreitol, iodoacetamide, recombinant human lysozyme and lactoferrin (≥90% purity as indicated by suppliers), and Trizma Base were obtained from Sigma Aldrich (St. Louis, MO). Recombinant human lipocalin-1 was obtained from R&D systems (Minneapolis, MN). Sequencing grade modified trypsin was purchased from Promega (San Luis Obispo, CA). Urea was obtained from VWR International Ltd. (Poole, UK). Acetonitrile high performance liquid chromatography (HPLC) gradient grade, ammonium bicarbonate, heptafluorobanzile, acetone, and formic acid were purchased from Merck (Darmstadt, Germany). Deionized water was prepared on a cartridge-deionizer from Milli-Q (Millipore, Billerica, MA).
All studies were approved by human ethics committee of the University of New South Wales (NSW), Australia and followed the Declaration of Helsinki. All subjects signed written informed consent before commencing the study. All experiments were performed on aliquots of tear samples collected from seven healthy volunteers (male = 1, female = 6). Subjects enrolled for tear sample donation had no ocular disease and were either contact lens wearers (CL; N = 3) or noncontact lens wearers (NCL; N = 4).
Subjects were trained to use capillary tubes for collection of tear samples in the morning and evening for 7 to 10 days using a noninvasive method (minimal or no eye irritation) described elsewhere.
28 Briefly, each subject tilted his or her head to one side and gently applied the tip of a 10-μL glass microcapillary tube (BLAUBRAND; intra MARK, Wertheim, Germany) with rounded edges to the corner of the eye (lateral canthus). Samples were collected from both eyes and were expelled into 0.5-mL Eppendorf tubes (Scientific Specialties, Inc., Lodi, CA) using a small manual rubber pump. Subjects were asked to store samples in a −20°C home freezer and transport them on ice to the university laboratory facility, where the samples were stored at −80°C.
The total volume of tear samples collected by each subject was expected to reach 35 to 100 μL. No minimum volumes were designated for sample volume. The first 10 μL of samples donated by each individual were stored separately for comparison of results in later stages of the study. The remaining samples were pooled together to obtain one large sample for each individual for the consistency of results in all method development experiments. The tear samples were centrifuged at 1000
g for 10 minutes at 4°C to remove debris or cells. The supernatants were aliquoted and stored at −80°C for subsequent use. Tears were processed as described elsewhere for plasma with modifications.
29 Briefly, for each experiment 2.5 μL of the pooled sample was diluted 1:1 (vol/vol) with deionized water and 5 μL of 8 M urea, 300 mM Tris, pH 7.8 were added. Urea was used as a denaturant and proteins were reduced with 0.5 μL Dithiothreitol (100 mM), kept for 30 minutes at 37°C. Sufficient reagents were prepared for the complete study so as to minimize the variance produced by such a procedure. Cysteine residues were alkylated by 0.5 μL iodoacetamide (200 mM) in water to obtain 20 mM concentration in the samples and the samples were incubated for 30 minutes at 37°C in the dark. Ice cold acetone was added to the sample to the ratio of 4:1 (vol/vol). The mixture was incubated at −20°C for 1 hour. The samples were centrifuged in a 5840R centrifuge (Eppendorf AG, Hamburg, Germany) at 4000
g for 15 minutes at 4°C. After centrifugation, the supernatant was discarded. The pellet was collected by decanting; the acetone was removed by air drying for 10 minutes and nonfractionated tear proteins were reconstituted in 2.5 μL of 100 mM ammonium bicarbonate buffer pH 8.50. Trypsin (200 ng/μL) was added to achieve a 1:50 enzyme-to-substrate ratio (wt/wt) on the assumption that approximately 10 μg/μL total proteins were present in tears
30,31 and the mixture was incubated at 37°C for 17 hours to produce a mixture of tryptic peptides. Samples were acidified to pH 2.0 by adding 0.05% heptafluorobanzile in 0.1% formic acid to stop digestion. For quantitation of proteins in the samples, a concentration-balanced mixture of 10 isotopically-labelled internal standard peptides (50% acetonitrile in 0.1% [vol/vol] formic acid) was added to each digested tear sample at a ratio of 1 volume of peptides mixture to 8 volumes of acidified tear digest.
The five tear proteins described in the literature as potential markers of dry eye disease (lactoferrin, lysozyme, lipocalin 1, proline rich protein 4, and prolactin-induced protein)
6,7,20 were studied using purely in silico methods. The FASTA
32 formatted sequences of target proteins were directly pasted into the Skyline program v1.5 (MacCoss Lab, University of Washington, Seattle, WA), which generated a full list of peptides and corresponding product ions (Y-ions) and the results were compared with the human National Institute of Standards and Technology (NIST) spectral library. Tear samples were run on a nano-flow liquid chromatography mass spectrometer (LTQFT Ultra; Thermo Electron, Bremen, Germany); the results were compared with the Mascot search engine to identify the more frequently observed tryptic peptides in the dual mass spectrometry (MS/MS) analysis.
33–35 By using the combined information of the previous two steps, a list of peptides ending in C-terminal lysine (K) or arginine (R), containing 5 to 15 amino acids were selected for analysis. Selection of each peptide was based on the signal strength, intensity of electrospray ionization process, and chromatographic peak shape (symmetric peak with a narrow width). Peptides containing methionine and cysteine or those with two basic amino acids (KK, RR, or KR) at the N or C terminus were excluded from experiments.
36 To select appropriate transitions, the list of transitions was saved as a file (in the csv format) and exported into the 4000 quadruple-linear ion trap mass spectrometer (Qtrap; Applied Biosystems, Foster, CA). The double charged and singly charged fragment ions (first and third quadruples) were selected based on their signal intensity and the ability to discriminate target peptides from the rest of the peptides present in the sample. Transitions with lowest coefficient of variation (CV) were chosen as the best transitions. These were mostly the ones with highest abundance and were used to quantify proteins in tear samples. To ensure that the peptides selected using this procedure were sequence-unique to each of the proteins of interest, the Basic Local Alignment Search Tool (BLAST) was used to perform a homology search against the National Center for Biotechnology Information (NCBI) database. Skyline v1.5 (an open source program) was used to analyze the files generated by SRM, which automatically generated the chromatogram, retention time and peak area for each peptide, and its corresponding product ions.
Instrumentation used for chromatography included a nano-LC Ultimate 3000 HPLC (Dionex, Amsterdam, Netherlands) and Switchos and Famos autosampler (LC-Packings, Amsterdam, The Netherlands). One microliter of tear digests were concentrated and desalted onto a micro C18 pre-column (500 μm × 2 mm; Michrom Bioresources, Auburn, CA) with H
2O:CH
3CN (98:2, 0.05% trifluoroacetic acid) at 15 μL/min for 4 minutes. The precolumn (10 port valve; Valco, Houston, TX), which was connected via a fused silica capillary (Upchurch Scientific, Oak Harbor, WA), was automatically switched into a nano-flow liquid chromatography (LC) system.
37 Solvents used in chromatography were H
2O:CH
3CN (98:2, 0.1% formic acid; Solvent A) and H
2O:CH
3CN (36:64, 0.1% formic acid; Solvent B). Samples were eluted at approximately 300 nL/min over 30 minutes. Selection reaction monitoring analyses were performed on a 4000 Qtrap (Applied Biosystems). The MS/MS system consisted of an ion spray (voltage 2.4 kV), curtain gas flow of 12 μL/min and nebulizing gas flow of 5 μL/min with positive ion mode, and was controlled by Analyst 1.5 software (Applied Biosystems) and interface heater temperature set at 150°C. Collision energies (CE) were optimized as described before (CE = 0.043 m/z + 4.756)
38 for maximum transmission of individual “signature peptides.”
39 A total of 60 SRM transitions (3 per peptide) were monitored during an individual sample analysis. Instrument parameters remained unchanged for each unlabelled/labelled pair. To ensure that maximum specificity was achieved with samples as complex as human tears, SRM transitions were acquired at unit resolution in the first and third quadruples. Dwell times of 25 or 50 ms were used for selected transitions and cycle times did not exceed 1 second.
Following selection of the signature peptides (the dominant ions in collision-induced dissociation fragmentations)
40 uniformly isotope-labelled stable peptides containing C-terminal [
13C,
15N]-lysine or [
13C,
15N]-arginine and corresponding normal (nonisotope labelled peptides) were purchased from Sigma Aldrich. Individual, synthetic unlabelled peptide stocks of 1 nmol/μL and labelled peptide stocks of 10 pmol/μL were prepared in 0.1% formic acid and/or 30% to 50% acetonitrile in 0.1% formic acid (vol/vol; based on hydrophobicity of the peptide). To investigate the purity of the labelled and unlabelled peptides and to confirm their homology with the corresponding endogenous forms of the tear peptides, all purchased peptides were analyzed by amino acid analysis (AAA). Amino acid analysis was performed by the manufacturer for the labelled peptides and by the Australian Proteome Analysis Facility (Sydney, Australia) for the unlabelled peptides. To reflect the AAA values, stock dilutions were performed to 0.1 nmol/μL with 30% acetonitrile in 0.1% formic acid. The synthetic peptides also were used to optimize the MS analysis parameters.
Serial dilutions of synthetic [
12C/
14N] peptides were generated by diluting unlabelled standard peptides to final concentrations of 1, 5, 25, 125, 250, 500, 1000 femtomole per microliter (fmol/μL) in the presence of 50 fmol/μL of the labelled peptides. The ratio of the peak areas of the unlabelled to labelled peptides was plotted against the concentration of the corresponding unlabelled peptides. Limits of detection (LOD) and quantification (LOQ) were determined using a previously reported technique.
29,41 Optimal LOD and LOQ were achieved for each peptide by performing three independent identical experiments. The 95th percentile of the blank samples was used as the limit of blank (LOB) and was calculated as mean
b +
t1-α × SD
b, where mean
b and SD
b are the mean and SD of the blank sample. The LOD was calculated using the following formula (LOD = LOB + C
β × SD
s), where SD
s is the SD of the lowest level spike.
29 The LOQ was assumed to be three times the LOD.
27 To determine the effect of matrix on the linearity of the peptide dilutions, the same procedure was carried out in the presence of digested tears. Human tear samples were diluted (1:50 in 25% acetonitrile in 0.1% [vol/vol]) before spiking the unlabelled standard peptides to final concentrations of 1 to 1000 fmol/μL adding 50 fmol/μL signature peptide (
13C/
15N) into the tears. Blank runs of digested tears with labelled peptides provided estimates of chemical background levels in the absence of signature peptide peaks. Furthermore, an estimate of carryover was determined by running a series of gradient HPLC washout runs. The relative intensities of all the three transitions were identified and background correction was applied uniformly throughout the study. A seven-point standard curve (a plot of response versus known concentration) was produced for each of the three transitions of each peptide to evaluate the linearity of the SRM measurement across the range of spiked peptides for each experiment. All experiments were repeated three times. Experimentally measured concentrations of the spiked peptides were plotted against the theoretical concentration values (all in molar units) to find the rate of recovery of each peptide. To ensure the method covered a sufficient range for measurement of peptide levels in unknown samples, the combination of linear regression with response factor ([peak area ratio-y intercept]/concentration) plots were tested to identify analyte concentrations that did not respond in a linear manner for each peptide. To ensure that the sample dilution matched the concentration range of the standards, dilutions of a tryptic digest of the pooled tear samples were prepared to a range of 2- to 18-fold and were spiked with 50 fmol/μL of all internal standards. Three replicate LC-SRM/MS analyses of each sample dilution were performed. The concentration of each of the 10 peptides in the pooled sample was calculated by dividing the concentration of each peptide by that of the corresponding internal standard, averaged for three identical, but independently performed experiments.
To assess the general applicability of the labelled peptide as an internal standard for quantification of selected tear proteins, aliquots of pooled tear samples were processed for analysis (as described above). The concentration of endogenous proteins (μg/μL) in tear samples was deduced using the peak area ratio between the endogenous and labelled peptides (50 fmol/μL). Labelled internal standard peptides were added postdigestion. Replicate analyses of the spiked tear samples provided estimates of assay precision.
Human tears and equal concentrations of commercially available lactoferrin, lysozyme, and lipocalin proteins (10, 5, 2.5, 1.25 μg/μL) were digested separately with trypsin by the procedure described above (proline-rich 4 and prolactin-induced protein were not commercially available). To simulate biological duplicates, each concentration point was digested in duplicate prior to adding the protein digest mixture to diluted tear digests. Internal standards for each peptide were prepared in a similar manner. All LC-SRM/MS runs were performed in duplicate (technical replicates). Digestion of recombinant proteins served as a control for digestion efficiency across diminishing concentration points. To assess the concentration of purchased recombinant proteins, LavaPep total protein assay (Star Scientific, Sydney, Australia) was performed according to the manufacturer's instructions. Comparison of measured versus the nominal spiked protein concentration values was recorded as protein recovery rate.
Accuracy of experimentally-prepared protein concentrations can be affected by acetone precipitation rates,
26,42 which may vary between samples. The potential effect of this confounding factor was examined with pooled tear samples and calculation of the CV of all 10 peptides. Three independent in vitro pooled tears were measured on 3 consecutive days and each sample was run in duplicate.
Comparison of Protein Concentration Determined by SRM-MS and Immunoassay in Tears
The concentration of lactoferrin in the tears of participants was confirmed by ELISA (Hycult Biotech, Uden, The Netherlands) according to the manufacturer's instructions.
The list of target proteins, their signature peptides, three sets of mass-to-charge ratio (m/z) of transitions (first and third quadruple) per peptide and their retention times in the SRM assay are presented in
Table 1.
Excellent linear responses (
R 2 > 0.99) were obtained for 7 of the 10 peptides (
Fig. 1). The peptides that showed lower linear responses (TYLISSIPLQGAFNYK, FYTIEILK, and FPSVSLQEASSFFR) still showed acceptable
R 2 values (0.9425–0.9881). Addition of light, synthetic peptides to digested tears showed high-intensity and peak area ratios for the range of concentrations used indicating that these peptides responded well and could be used in this assay.
Figure 2 shows the chromatographs of NNLEALEDFEK from lipocalin 1 as an example. Also, the relative ratios of the transitions in both solutions agreed closely indicating that there was little interference from the matrix in the MS channels.
Figure 3 shows the extracted ion chromatograms of FQLFGSPSGQK peptide of lactoferrin as an example.
The LOD for the 10 selected peptides was between 0.53 to 8.96 pg/mL and LOQ was 1.59 to 26.88 pg/μL, when peptides were added to the diluted tears (
Table 2). Of note is the fact that each of two signature peptides of each protein could yield a different LOQ for their respective proteins.
Figure 4 shows a compilation of response curves for the 10 peptides plotted on a linear–linear scale of experimentally determined concentrations versus theoretical concentrations of the target analyte and provides an overview of recovery of the peptides. The decrease/increase in the slopes of the response curves might result from the complexity of protein digestion. The LOQ estimations across peptides (
Table 2) and CVs of 1.3% to 15.1% indicated good reproducibility. Protein recovery rates during protein digestion and sample preparation, determined in trypsinised digests, were 75.0%, 60.4%, and 79.8% for lactoferrin, lysozyme, and lipocalin-1, respectively. The results of CVs of the selected peptides after acetone precipitation are presented in
Table 3. For 8 of the 10 peptides the CVs were less than or equal to 18%, whereas only two peptides had a CV between 21% and 24%.
The concentrations of selected tear proteins using SRM-MS were as follows: lactoferrin at 1.20 ± 0.77 μg/μL, lysozyme at 2.11 ± 1.50 μg/μL, lipocalin-1 at 1.751 ± 0.99 μg/μL, prolactin-induced protein at 0.09 ± 0.06 μg/μL, and proline-rich protein 4 at 0.80 ± 0.50 μg/μL.
Table 4 shows the comparison of the results of the concentration of lactoferrin by the SRM and ELISA methods for seven tear samples. Using a paired
t-test, there were no differences between ELISA (1.18 ± 0.46 μg/μL) and SRM (1.22 ± 0.63 μg/μL) results (
P < 0.05).
The tear film is a unique fluid that contains a wide range of proteins. Using nano–HPLC-MS/MS, de Souza et al.
43 identified 491 proteins in the tear film of one individual. Zhou et al.
44 using various fractions of tears and nano-reverse phase HPLC-MS/MS, identified 1543 proteins in tears collected from four healthy non–contact lens wearers, with 714 proteins being present in all samples. A large number of these proteins have an established role in physiological maintenance of the ocular surface,
31,45 but some might be involved in various pathologic conditions and tear film abnormalities. The tear film composition is known to change in response to various pathologic conditions
45,46 ; hence, individual tear film components have been proposed as suitable biomarkers for diseases.
6,7,20,47 However, in view of the limited availability of tear samples, new methods need to be developed in order to facilitate quantification of multiple tear components using small volume tear samples.
This study examined selected reaction monitoring as a technique for quantification of proteins in small-volume human tear samples. A major part of the study was focused on developing the tear processing method directed at increasing digestion efficiency and protein recovery rate. Particular attention was paid to simplification of the tear processing method. The experimental method described here detected and quantified the concentration levels for five proteins. Our results for lactoferrin, lysozyme, and lipocalin are consistent with findings from previous ELISA studies,
8,30,48,49 and we confirmed this for lactoferrin in the current study. Tear concentrations of proline-rich protein 4 (0.80 ± 0.50 μg/μL) and prolactin-induced protein (0.09 ± 0.06 μg/μL) are reported here for the first time.
Quantification of proteins in this study was performed in tear digests using two signature peptides per selected protein as quantitative factors and stable isotope-labelled variant of the same peptides as internal standards. Labelled peptides were also used for normalization of the natural peptide peaks using a modification of earlier described methods.
50 In this method we used precipitation to separate tear proteins from the remaining components of the matrix. An attempt was made to reduce false-positive responses in the MS and chromatography methods by comparing signals for each individual peptide in tear samples with those obtained from pure standards. The comparisons yielded highly similar results suggesting accuracy of the findings. Optimized sample dilution rates were used in these studies to avoid system saturation. Still, in general, poor chromatography technique and inevitable attenuation of MS signals, as well as software-related errors in detection of signal peaks might influence accuracy of quantification in peptide SRM-MS.
51
The results of this study confirm earlier findings,
39 suggesting that equimolar concentration of peptides chosen for quantification of each protein (paired signature peptides) have different MS detectability. The lack of consistency in performance of paired signature peptides could be caused by several factors, including incomplete trypsin digestion,
25 loss of peptides,
12 and unequal ionization rate of individual peptides.
29 It is recommended that at least two signature peptides be used for quantification of each protein so that an average of the two can be used for quantification.
In conclusion, these experiments used 10 stable internal standard peptides to develop rapid, specific, and sensitive SRM assays to quantify the levels of five human tear proteins. This method requires some sample preparation but no protein purification or use of antibodies, and has linearity across a broad range of protein/peptide concentrations. Further work will be required to expand the established panel of isotopically-labelled peptides to include additional proteins. The expanded assay would potentially serve as a valuable investigative tool for exploration of tear proteins in various tear-related studies and the effect of diseases on the tear matrix. This method can potentially be used to target relevant proteins in a range of applications, making this a very specific investigational tool.
Supported in part by the Australian Government through the Vision Cooperative Research Centre and the Tuition Fee Remission Scholarship through the University of New South Wales.
Disclosure: S. Masoudi, None; L. Zhong, None; M.J. Raftery, None; F.J. Stapleton, None; M.D. Willcox, None