To establish optimal conditions for determination of the tear fluid proteome we performed both in-gel and in-solution digestion, as summarized in Figure 1. Tear fluid was subjected to SDS-PAGE, the gel band cut into 13 slices, in-gel digested with trypsin and the resulting peptide mixtures analyzed by LC MS³ in the LTQ-FT. Alternatively, proteins were digested in solution with Lys-C, or alternatively by Lys-C followed by trypsin, prior to MS analysis. Our results showed a total of 320 proteins identified for in gel-based analysis whereas only 59 proteins were identified using 1 μl of tear digested in solution and 63 proteins for 4 μL of in solution tear digestion.

Figure 2 illustrates an example of MS acquisition and identification for in-gel digestion. In the inset of Figure 2a, the total ion chromatogram (TIC) is represented, and the spectrum shows the ions detected in selected ion monitoring (SIM) mode. Figure 2b shows the fragmentation pattern of the most intense ion in Figure 2a (m/z = 494.2906, peptide mass = 987.5812 Da). The identification is initially done on the basis of the data obtained using the Mascot algorithm 25. Figure 2c shows the MS³ of the most intense ion observed in Figure 2b, which is used to support or discard the identification made on the basis of MS² spectra 2026.

As shown in Figure 1, in situ digestion of 4 μl of tear sample was also analyzed by the LTQ-Orbitrap mass spectrometer. In this case, we were able to identify 368 proteins in the sample. Since MS³ analyses were not performed in the LTQ-Orbitrap, the criteria used for protein identification required at least two peptides with statistical significance (see Materials and methods). When the LTQ-FT protein list was overlaid with the Orbitrap data, we observed that approximately one-third of the proteins identified in the LTQ-FT analysis were not detected on the Orbitrap (112 of 320 proteins). Interestingly, most of the proteins that were exclusive to the LTQ-FT analysis (86 hits) are the ones that were validated due to improvements in the Mascot score resulting from MS³ data, and even though most of these 86 hits are present in the Orbitrap data (61 hits), they were discarded due to statistical reasons. On the other hand, the most abundant proteins in the sample had better sequence coverage in the Orbitrap data than in the LTQ-FT data (Figure 3). Discarding single peptide hits with Orbitrap tandem MS data may be overly conservative, since these spectra have very high resolution and low ppm mass errors, making false positives extremely unlikely. If such single peptide hits had been admitted, more than 100 additional proteins could be reported (data not shown). Either way, the presence of a substantial number of single peptide hits suggests that many more proteins are present in this proteome than we report here.

The complete list of proteins identified is summarized in Additional data file 1; this table lists the number of peptides observed for each identified protein, the Mascot score and the MS³ score for each peptide and protein (if available, that is, LTQ-FT data). Protein identification criteria were extremely stringent, requiring fully tryptic peptides with a mass error less than 3 ppm for the LTQ-FT or less than 5 ppm for the Orbitrap. For the FT data, the criteria needed were two matching peptides with a Mascot score of 27 or one matching peptide 'rescued' by an additional MS³ score, adding to a total probability score of at least 54. For the Orbitrap data, the criteria were two matching peptides with minimal score of 21. These criteria ensure an error rate in protein identification of less than 0.1%, so there should be no false positive protein identifications in our data set. In-gel analysis fully covered in-solution identifications. Therefore, all subsequent discussion is based on the in-gel data set.

The 491 proteins identified in the in-gel analysis were functionally classified using the Protein Center Tool (Proxeon Biosystems, Odense, Denmark) and statistical analysis was done using the BiNGO tool 27, based on cellular localization, molecular function and biological process. It should be kept in mind that Gene Ontology (GO) classification and tools that build on those annotations often comprise very broad and overlapping functional categories. Nevertheless, they provide a useful method of initial classification of a large proteome in terms of origin and molecular processes. Figure 4 illustrates an example of group over-representation determined by BiNGO. Tables 1 and 2 list the two main groups of molecular functions identified in this work, and also indicate the biological process that it is involved. Extracellular proteins are indicated by a dagger and proteins already identified in tear samples by an asterisk. Note that the hydrolase GO classification group is very broad, involving several processes, such as signal transduction (phosphatases), energy-driven reactions (ATPases), and glycolysis. We selected from this group proteins that possibly are directly functional in the tear environment, and not only present as a result of cellular degradation in the epithelia, for example. Thus, our 'hydrolase' group listed in Table 1 considers only extracellular proteins, proteins already described as components of the tear fluid, or proteins that participate in biological processes that are known to occur in the fluid that covers the eye. In this way we identified 32 proteins with hydrolase activity, and 32 proteins classified as protease inhibitors, mainly serine protease inhibitors. From these 64 proteins, only 24 proteins had been previously identified as components of tear fluid in other studies.

Figure 5a shows the cellular localization pattern of all proteins identified. Approximately 199 proteins were not classified by the ontology database. Interestingly, our data show that 41% (200 proteins) belong to the intracellular compartment, mainly present in cytoplasm (136 proteins), and, to a lesser extent, to compartments such as the nucleus (20 proteins), the Golgi apparatus (12 proteins) and the lysosome (11 proteins). On the other hand, only 68 proteins were classified as extracellular proteins, in addition to 2 that were classified as components of the extracellular matrix. When the not mapped protein group is eliminated from the chart, the intracellular proteins represent approximately 68% of the total identification (Figure 5b).

In addition, the classification of the identified proteins based on biological processes (Figure 6) revealed that at least 37 proteins belong to the immune system, 50 proteins are involved in immune response, such as antibodies and proteins from the complement system, 15 proteins are involved in inflammatory response, and 7 proteins are responsible for defense against pathogens. We also identified 31 proteins that are associated with response to wounding and blood coagulation. Finally, we identified 18 proteins that are involved in the metabolism of reactive oxygen species, such as peroxiredoxins and catalase, which may be functioning in the tear film in the defense against toxic oxygen compounds.

Samples of closed-eye tear were collected from one of us (GAS) using a 5 μl calibrated glass microcapillary tube (Blauband intraMARK, Brand GMBH, Werthein, Germany) without touching the eye globe or lids, in the course of one week and at different times of the day to avoid diurnal variation 3435. One sample typically contained 2 μl. After collection, the tears were centrifuged at 14,000 g for 1 minute at 4°C (Eppendorf model 5417C, Eppendorf, Hamburg, Germany) to remove cellular debris, and stored at -20°C until analysis.

A tear sample (4 μL) was added to electrophoretic sample buffer (NuPAGE kit, Invitrogen, Karlsruht, Germany) and tear protein content was resolved by SDS-PAGE using a homogeneous 12% gel (NuPAGE gel, Invitrogen) under reducing conditions for 50 minutes with a constant voltage of 200 V. The gel was stained with Coomassie staining kit (NuPAGE, Invitrogen), as instructed by the manufacturer. After staining, two lanes of the gel were combined and then sliced in 13 pieces as indicated in Figure 1. The pieces were then subjected to in-gel reduction, alkylation and tryptic digestion. To reduce disulfide bonds, 100 mM DTT was added to a final concentration of 10 mM in the protein solutions and incubated for 1 h at 56°C in the dark. Free thiol (-SH) groups were subsequently alkylated with iodoacetamide (50 mM final concentration) for 45 minutes at room temperature. The reduced and alkylated protein mixtures were digested with sequence grade-modified trypsin (wt:wt 1:50; Promega, Madison, WI, USA) for 16 h at 37°C in 50 mM NH₄HCO₃, pH 8.0. Proteolysis was quenched by acidification of the reaction mixtures with 2% trifluoroacetic acid (Fluka, Buchs, Switzerland). Finally, the resulting peptide mixtures were desalted on RP-C₁₈ STAGE tips as described 43 and diluted in 0.1% trifluoroacetic acid for nano-HPLC-MS analysis.

Samples of 1 and 4 μl of tear fluid were resuspended in 20 μl of 6 M urea and 2 M thiourea (Invitrogen) and submitted to reduction and alkylation as described above. For enzymatic digestion, Lys-C (wt:wt 1:50; Wako, Japan) was added to the solution for 16 h at room temperature, and the resulting peptides were desalted on RP-C₁₈ STAGE tips. The same experiment was repeated using Lys-C for 16 h, followed by trypsin (1:50) for 24 h at room temperature.

All nano-HPLC-MS² experiments were performed on an Agilent 1100 nanoflow system connected to a 7-Tesla Finnigan linear quadrupole ion trap-Fourier transform (LTQ-FT) mass spectrometer (ThermoElectron, Bremen, Germany), or connected to a LTQ-Orbitrap mass spectrometer (ThermoElectron), both equipped with a nanoelectrospray ion source (Proxeon Biosystems, Odense, Denmark).

Briefly, for in-gel samples, the mass spectrometer was operated in the data-dependent mode to automatically switch between MS, MS², and MS³ acquisition. Survey full-scan MS spectra (m/z 300 to 1,500) were acquired in the Fourier transform ion cyclotron resonance (FT ICR) with resolution R = 25,000 at m/z 400 (after accumulation to a target value of 10,000,000 in the linear ion trap). The three most intense ions were sequentially isolated for accurate mass measurements by an ICR-FT SIM scan with 10 Da mass range, R = 50,000 and target accumulation value of 50,000. They were then fragmented in the linear ion trap by collisionally induced dissociation at a target value of 5,000. For MS³, up to three ions in each MS² spectra (the most intense ions with m/z > 300) were further isolated and fragmented. Former target ions selected for MS² were dynamically excluded for 30 s. Total cycle time was approximately 3 s. The general mass spectrometric conditions were: spray voltage, 2.4 kV; no sheath and auxiliary gas flow; ion transfer tube temperature, 100°C; collision gas pressure, 1.3 mTorr; and normalized collision energy, 30% for MS² and 28% for MS³. Ion selection thresholds were: 500 counts for MS² and 50 counts for MS³. An activation q-value of 0.25 and an activation time of 30 ms was applied in both MS² and MS³ fragmentation 20. However, due to the expected higher complexity of in solution digestion samples, the acquisition method was adjusted to not perform SIM scan or MS³, but to sequence the five most intense peaks for obtaining MS² data.

Stringent criteria were applied for protein identification, which was performed by searching the data against the International Protein Index database (IPI_human) by MASCOT (Matrix Science) and MSQuant (an in-house developed, open source software program). These criteria comprised: for LTQ-FT-ICR data, a mass accuracy within 3 ppm (in-gel digestion; average absolute peptide mass accuracy was 1.03 ppm) or 25 ppm (in-solution digestion; average absolute accuracy was 8.3 ppm); for LTQ-Orbitrap data, a mass accuracy of 5 ppm (average absolute accuracy of 1.01 ppm); at least two, fully tryptic, matching peptides per protein with a Mascot score for individual peptides (MS²) better than 27 (p ≤ 0.01), or one peptide with MS² + MS³ score better that 54 (p ≤ 0.0001), when MS³ was performed. For in-solution digestion, proteins were considered identified if they had at least two peptides with score higher than 35. For Orbitrap data, the criteria were a mass accuracy within 3 ppm (average absolute peptide mass accuracy was 1.22 ppm) and at least 2 fully tryptic peptides per protein with a Mascot score better then 21 (p ≤ 0.01) for individual peptides (MS²). Differences in mass accuracy between in-solution samples (more complex compared to in-gel samples due to lack of pre-fractionation) was observed because the measurement of ion masses was not performed with the SIM method, leading to higher sequencing speed of the method at the cost of lower mass accuracy.

Experiments with a reversed database were performed as described in 44. The number of statistically significant peptides identified in the IPI database was 1,935, while the reverse database identified 12 peptides with statistical significance (0.6%) for the LTQ-FT data. However, these 12 peptides were not sufficient to identify a single protein (that is, none of the proteins had at least 2 peptides with score higher than 27 or one peptide with MS³ score higher than 54). The Orbitrap data included no peptides within statistical significance in the reverse database. Identified proteins were combined in a larger data set and initial GO characterization was done using the Protein Center tool (v0.62, Proxeon Biosystems).

Our data are freely available at the proteome database of the department of proteomics and signal transduction of the Max-Planck-Institut for Biochemistry 45.