Preliminary identification of differentially expressed tear proteins in keratoconus

Subjects

Two groups were included in the study, one consisting of normal subjects (C) who had not been diagnosed with KC and the other subjects with KC. A total of 54 subjects (C=18, KC=36) were enrolled in the study. Demographics of the study subjects are outlined in Table 1. The subjects either gave a history of no or discontinued contact lens wear for at least 1 month before the study began. Subjects with active allergy or history of previous ocular diseases or ocular surgery or who used topical or systemic medications were excluded from the study. The participants with KC gave a history of frequent eye rubbing.

Corneal topography

The corneal curvature was mapped using a Medmont Corneal topographer E300 (Camberwell, Australia) on patients with KC and C subjects. The steepest simulated keratometry reading was recorded. The difference between the mean inferior and superior (I-S) power of the corneal curvature was used to verify the diagnosis of KC [17].

Tear collection

Sterile thin glass micro capillary tubes (BLAUBRAND intraMARK, Wertheim, Germany) were used to collect tears. A minimum of 4–6 µl tears were collected from the inferior cul-de-sac of the eye by the same investigator for every subject. Tear flow rate was monitored to ensure that reflex tears were not collected [18], and special care was taken not to touch the ocular surface to avoid reflex tearing. After the tears were collected, they were centrifuged at 2700 ×g for 10 min at 4 °C and stored at −80 °C until used for analysis [19].

Evaluation of tear proteins

An outline of experimental techniques followed in the present study is shown in Figure 1. The total protein concentration of individual tear samples was quantified with bicinchoninic acid (BCA) protein assay [20]. The bovine serum albumin (BSA) standard curve was obtained by preparing serial dilutions of BSA. The BCA solution (Pierce BCA kit, Thermo Scientific, Scoresby, Australia) was added to the standards or tears, and the absorbance was measured at 562 nm using a spectrophotometer (TecanSpectrofluoro Plus; Tecan Group, Männedorf, Switzerland). The total protein levels in both groups were calculated against the standard BSA curve.

Protein fractionation

The MF10 instrument was used to partition the tear proteins. As this technique requires at least 50 µl sample to be added before fractionation, it was decided to use pooled tears as it would have been impractical to collect 50 µl tears from each patient. This resulted in the use of only one pooled sample for the controls and one pooled sample for KC. Thus, we identified the results as preliminary evidence of differences between the two groups. The C group had 36 µl tears, obtained by pooling 2 µl tears from 18 eyes, and the KC group had 72 µl tears, obtained by pooling 2 µl tears from 36 eyes. Both pooled samples were made up to 80 µl before the MF10 was applied. The total protein µg in the pooled C or KC tears were equivalent before fractionation.

The MF10 instrument consists of specific pore-limiting membranes or polyacrylamide membranes (Figure 2). Samples containing complex peptide mixtures can be separated based on their size and charge using the mass restricted polyacrylamide membranes in MF-10 under native, denaturing, or reducing conditions [14]. In the present study, a six-chamber cartridge was assembled using 2.7×2.2 cm polyacrylamide membranes with varying pore sizes of 5, 25, 50, 75, and 150 kDa (NuSep, French forest, Sydney, Australia) and a 1 kDa regenerated cellulose membrane (Millipore, Billerica, MA), arranged in order from anode to cathode (Figure 2). Both ends of the cathode and anode were fitted with 5 kDa membranes to remove the charged contaminants from the sample during the fractionation process. Before use, the membranes were washed twice with ultrapure water (H₂O) and then twice with 1X Tris-ε-aminocaproic acid (EACA)/urea buffer (90 mM Tris (hydroxymethyl) aminomethane, 10 mM EACA, 2 M urea). The buffer lines of the MF10 were washed with ultrapure H₂O followed with a 10 min wash with 1X Tris/EACA/urea buffer.

The resulting assembly generated five chambers (F1 to F5) for fractionation (Figure 2). All cartridge assemblies had two lanes of chambers that allowed fractionation of two samples in one run. The chamber cartridge assembly was placed in the MF10 separation unit (Figure 2). A solution containing 80 ml of 90 mM Tris/10 mM EACA/1 M urea (pH 10.2) circulated around the electrodes.

Under native conditions, 35 µL C tears made up to 140 µl sample buffer (90 mM Tris/10 mM EACA/1 M urea, pH 10.2) were loaded into one chamber, and 70 µl KC tears in 280 µl sample buffer were loaded on a separate run. All the remaining chambers of both lanes were filled with an equivalent volume of sample buffer. Fractionation was performed at 50 V for 30 min, 250 V for 2 h (0.5 KvH) at 15 °C. After separation, fractions were collected from the chambers using gel-loading tips (Interpath Services, Heidelberg, Australia).

Sample preparation for liquid chromatography and mass spectrometry

Proteins were recovered for digestion with the C18 Stage tip (Thermo Fisher Scientific, Scoresby, Australia) for the 1 to 5 kDa and 5 to 25 kDa fractions and by acetone precipitation for the remaining fractions to remove salt and buffer contaminants. The stage tips and acetone precipitations were used to concentrate the lower and higher molecular weight proteins, respectively [21,22].

C18 Stage tip

The C18 Stage tip was equilibrated with 25 μl 50% methanol, 5% formic acid (FA) in distilled water (dH₂O; v/v) and washed with 25 μl 5% FA in dH₂O. Samples were slowly passed through the Stage Tip in 25 μl aliquots with elution and reequilibration in between. The tip was washed twice with 25 μl 5% FA in dH₂O (v/v). Peptides were eluted with 10 μl 80% acetone and 5% FA in dH₂O (v/v) into a clean 1.5 ml polypropylene tube. The samples were dried in a vacuum centrifuge and frozen at −20 °C until further use. Lyophilized protein samples were reconstituted with 50 μl 50 mM ammonium bicarbonate buffer (pH 8.50) before trypsin digestion.

Acetone precipitation

Ice-cold acetone was added to the fractions (>25 kDa) in the ratio of 4:1. The mixture was incubated at −20 °C for 1 h. The samples were centrifuged in a 5804R centrifuge (Eppendorf AG, Hamburg, Germany) at 15,300 ×g for 12 min at 10 °C. After centrifugation, the supernatant was discarded, and the pellet was air dried for 20 min and stored at −20 °C. Protein samples were reconstituted with 50 μl of 50 mM ammonium bicarbonate buffer (pH 8.50) before trypsin digestion.

Trypsin digestion

The individual fractions were treated with 1 µg trypsin, and the mixture was incubated for 18 h at 37 °C. Reduction and alkylation before trypsin digestion were not performed as has been reported previously [23]. The reaction was stopped by adding 5 µl neat FA. Samples were then dried in a vacuum centrifuge. The lyophilized samples were treated with 5 µl 2% acetic acid and 0.1% FA. Equivalent amounts of fractionated samples were analyzed for each group (C or KC). Fractions were examined with liquid chromatography and mass spectrometry (LC-MS/MS). One to 5 kDa fractions were analyzed at 1 µl from 10 µl, and all the other individual fractions were analyzed at 0.5 µl from 10 µl.

To allow for quantitative comparison, an equal amount of starting material was analyzed. The 1 to 5 kDa fractions were stage tipped to remove buffer components and to concentrate the proteins. The proteins were resuspended into 10 µl buffer A (2% acetonitrile (ACN)/0.1% FA/98% ultrapure H₂O), and equivalent concentrations of total protein were subjected to trypsin digestion and mass spectrometry. This equated to 1 µl of the 1–5 kDa fraction and 0.5 µl of the other molecular weight fractions. These volumes were correctly assessed such that the instrument would be able to provide MS/MS information for up to 10 ions at any given scan cycle rather than being overwhelmed and missing important ions. This also limited saturation of the detector and provided better quality LC separation.

Liquid chromatography and mass spectrometry using linear ion trap quadrupole Fourier transform

The digested MF10 fractions were evaluated using a LTQ-FT Ultra mass spectrometer (Thermo Electron, Bremen, Germany). Peptides were separated with nano-LC using an Ultimate 3000 HPLC and auto-sampler system (Dionex, Amsterdam, Netherlands). Samples were concentrated and desalted onto a micro C18 precolumn (500 µm × 2 mm) with 0.05% (v/v) heptafluorobutyric acid (HFBA) 20 µl/min. After a 4 min wash, the precolumn was switched (Valco 10 port valve, Dionex, Amsterdam, Netherlands) into line with an in-house built fritless nano C18 column (75 µm × 10 cm) packed with the C18 Magic stationary phase (5 µm, 200 Å pore size; Michrom Bioresources, Auburn, CA). The LC mobile phase comprised buffer A, 2% acetonitrile (ACN)/0.1% FA/98% ultrapure H₂O, and buffer B, 80% ACN/0.1% FA/20% ultrapure H₂O. Peptides were eluted using a linear gradient of 2% to 50% buffer B over 30 min followed by a 100% buffer B wash over 1 min at a flow rate of approximately 300 nl/min. High voltage was applied to low volume tee and a column tip position at approximately 0.5 cm from the heated capillary (T=200 °C) of the LTQ-FT. Positive ions were generated by electrospray, and the LTQ-FT operated in the data-dependent acquisition (DDA) mode. An MS survey scan was acquired (m/z 350–1700) in the Fourier transform ion cyclotron resonance cell (resolution=100,000 at m/z 400) up to seven of the most abundant ions (>2,000 counts) with charge states of +2 or +3 were sequentially isolated and fragmented within the linear ion trap using collision-induced dissociation (CID). Mass to charge ratios selected for MS/MS were dynamically excluded for 45 s. Peak lists were generated using MASCOT Daemon/extract_msn software, version 2.2 (Matrix Science, London, England) using default parameters and submitted to the database search program MASCOT.

Identification of peptides

All MS/MS spectra were searched using the MASCOT search engine. A minimum MASCOT score of 50 was used to identify peptides. MF10 protein/peptide fractions were searched against the Swiss-Prot database release 15 with the following criteria: precursor tolerance and the product ion tolerances were at 6 ppm and ± 0.6 Da, respectively, variable modification of methionine oxidation, and no enzyme.

Although trypsin was used, MASCOT was searched using the “no enzyme” setting. While trypsin was used to cleave the larger proteins and peptides, we decided to look for other possible peptides. This allowed for identification of peptides that had undergone proteolysis with other endogenous enzymes allowing the potential to identify proteins that were degraded in the tears. Thus, selecting “no enzyme” within MASCOT increased the potential number of peptides that could be identified as it was not reliant on the peptides having an arginine or lysine terminus. Because there were some proteins with low mass, restricting the search to only trypsin-derived peptides may have resulted in fewer peptides identified for each protein as there may not have been sufficient tryptic cleavage sites in smaller proteins.

Analysis was performed using Scaffold proteome software version 3 (Proteome Software, Portland, OR). Scaffold is a bioinformatics tool designed to compare the spectral counts obtained between multiple samples for the purposes of relative quantitation. A peptide probability threshold of 95% with at least two peptides required for confirmation. This gave a 1.50% false positive rate for identifications.

We chose not to merge the peak files. However, we combined the analysis within Scaffold. Our aim was first to define and enrich for the low mass component within the tears and then to compare the data sets. Thus, determining which peptides partition to which mass fraction before quantitative analysis was essential.

Statistical methods

Significant differences in total tear protein levels between the groups were analyzed using the Student t test. The differential expression of proteins across the samples was compared between the C and KC groups based on peptide spectral counts using the Scaffold software. The relative abundance of individual proteins across the groups was calculated with spectral counting. Spectral counting compares the MS/MS spectra assigned to each protein and is considered sensitive for detecting the proteins that undergo changes in abundance [24]. The spectral count correlates with the protein abundance [25]. Standard deviations were automatically calculated across replicates, and the spectral counts were normalized between the groups to remove statistical errors created from repeated measurements [26]. The fold change in the Scaffold analytical software was used to test the significance of the ratio of spectral counts between the C and KC groups.

Corneal topography

The mean steepest keratometry reading (K2) in the C and KC groups was 43.7±1.2 D and 57.3±10.5 D, respectively. The KC subjects recruited had mild (<45 D), moderate (45–52 D), and severe (>52 D) stages of the disease. The proportion of the KC subjects with mild, moderate, and severe stages of KC were 13.9%, 38.8%, and 47.2%, respectively. However, for the analysis, the tears of all KC subjects were pooled.

Total tear protein level

The differences in age or gender were not statistically significant between the C and KC groups. The patients with KC had a significantly lower level of total tear protein (3.24±0.70 µg/µL) compared to the C (non-keratoconic) group (6.35±1.24 µg/µl; p<0.0001), and thus, twice as much tears were analyzed for the KC group to normalize this difference.

Tear protein profile

Pooled C (36 µl) and KC group (72 µl) tear samples had equal total protein amounts before protein fractionation. A total of 75 tear proteins in the C and KC groups (Appendix 1) were identified confidently. The proteins were also identified in the individual tear fractions of the C and KC groups, i.e., 1–5 kDa (Appendix 2), 5–25 kDa (Appendix 3), 25–50 kDa (Appendix 4), 50–75 kDa (Appendix 5), and 75–150 kDa (Appendix 6).

Differentially expressed proteins in keratoconus compared to controls

Based on the spectral counts, cathepsin B was increased in abundance in the KC group compared to the C group (Figure 3A). Polymeric immunoglobulin receptor (PIGR), fibrinogen alpha chain or α-fibrinogen, cystatin SN, and cystatin S were decreased in the KC group compared to the C group (Figure 3B–E). The tear proteins found only in the tears from the KC group were keratin type-1 cytoskelatal-14 (K1C14) and keratin type-2 cytoskeletal-5 (K2C5).

Fold change is the ratio of spectral counts in the KC group compared to the spectral counts in the C group of a particular protein. Larger fold change values imply larger differences in a particular protein between the groups. The corresponding molecular weight fractions and the fold change values of the differentially expressed proteins in the KC group are shown in Table 2. The MS/MS spectra of the altered proteins in the C group compared to the KC group are illustrated in Appendix 7.

Ontology analysis

The proteins listed in Table 2 were grouped based on location (Figure 4A), biologic process (Figure 4B), and functions performed (Figure 4C) by using the gene ontology (GO) term analysis in Scaffold. Several proteins belonged to more than one category within each group. The altered proteins identified in the KC group are mainly intra- and extracellular proteins (Figure 4A), implicated in cellular or developmental processes (Figure 4B), and carry out important enzyme regulatory or molecular functions (Figure 4C).