A crying shame: Protein panel in tears classifies eye disorders

Tear film indicators

The proteins and peptides that reside naturally in tears can tell us a lot about the condition of the eye. This premise is based on a series of studies that have found that diseases like dry eye and blepharitis (inflammation of the eyelids) change the proportions of certain proteins in normal tears. A new test to diagnose dry eye would be especially useful because it is currently diagnosed by interview backed up by tests which are not particularly reproducible.

At one time, tears were not thought to contain many proteins and peptides, perhaps 50 in total. Then a report published in 2000 identified more than 500, which attracted more research groups to the area and subjected tears to closer scrutiny. Now, Spanish researchers have looked at the protein composition of tears to see if it can distinguish between dry eye and meibomian gland dysfunction (MGD).

In MGD, the secretory Meibomian glands become obstructed, so that the tear film evaporates more quickly causing evaporative dry eye. Dry eye and MGD display overlapping symptoms, impeding correct diagnosis, so a better way of identifying both conditions would be useful to clinicians and lead to quicker patient treatment.

Tatiana Suarez from Bioftalmik, Derio, and coresearchers from four other ophthalmological organisations in Spain undertook a proteomics study using 2D PAGE to identify any potential protein biomarkers before carrying out a detailed validation study of the candidate species.

Eye to eye

The research team enrolled 144 patients from four centres into the study, comprising 63 with dry eye, 38 with MGD and 43 healthy controls. In the initial discovery stages, tears were collected from 44 patients using surgical sponges, 16 with each disease and 12 controls. The abundant protein immunoglobulin A was removed by immunoaffinity chromatography in case it swamped the less abundant proteins.

The remaining proteins were separated by 2D PAGE and their positions on the gel were identified by staining with a fluorescent dye. A total of 130 protein spots were resolved and image analysis was conducted to pinpoint the proteins which had different abundances between the three groups and these were identified by mass spectrometry.

A group of 15 proteins made good markers for separating the three groups and this was confirmed by a principal components analysis. Six proteins were more abundant in the dry eye group compared with the MGD and control groups. Conversely, nine proteins were less abundant with respect to the controls, five of these being associated with MGD and four with dry eye.

However, before refining the panel of proteins to a more manageable number that retained the predictive capability, they were screened by ELISA to identify the most predictive ones. This was followed by two validations steps in which the proteins were collected by sponge or capillary. The best classification method was achieved with the Random Forest and Naïve Bayes machine learning techniques which were trained using one set of results and confirmed on the others.

Protein panel discriminates diseases

A network analysis of the 15 discriminatory proteins revealed that nine of them were directly or indirectly related. The major biological processes in which they are involved include inflammatory response, stress response, immune response and oxidative stress response.

The final discriminatory panel obtained after refinement of the model consisted of five proteins: S100A6, annexin A1, annexin A11, cystatin-S and phospholipase A2-activating protein. Together they were able to discriminate between dry eye and control patients with high sensitivity and specificity. They could also distinguish between the dry eye, MGD and control groups with good accuracy. The results were independent of the tear collection method.

This panel differed from that discovered in an earlier report in which α-enolase, prolactin-inducible protein, lipocalin-1 and S100A9 diagnosed dry eye with an accuracy of 96% and the levels of α1-acid glycoprotein 1, S100A8 and S100A9 were correlated to the severity of the dry eye condition. However, there is a common feature between the two sets of proteins in that many of them are inflammatory proteins.

The team will now examine the biomarkers in a clinical setting. If the group of proteins survives further scrutiny, it could be developed as a test to give more accurate diagnosis of dry eye and meibomian gland dysfunction and to follow patient response during treatment. It might also be able to distinguish between other diseases of the eye.