The proteome of perilymph in patients with vestibular schwannoma. A possibility to identify biomarkers for tumor associated hearing loss?

Abstract

Background: Due to the surrounding bone, the human inner ear is relatively inaccessible and difficult to reach for cellular and molecular analyses. However, these types of investigations are needed to better understand the etiology, pathophysiology and progression of several inner ear disorders. Moreover, the fluid from the inner ear cannot be sampled for micro-chemical analyses from healthy individuals in vivo. Therefore, in the present paper, we studied patients with vestibular schwannoma (VS) undergoing trans-labyrinthine surgery (TLS). Our primary aim was to identify perilymph proteins in patients with VS on an individual level. Our second aim was to investigate the proteins identified at a functional level and our final aim was to search for biological markers for tumor-associated hearing loss and tumor diameter.

Methods and findings: Sixteen patients underwent TLS for sporadic VS. Perilymph was aspirated through the round window before opening the labyrinth. One sample was contaminated and excluded resulting in 15 usable samples. Perilymph samples were analyzed with an online tandem LTQ-Orbitrap mass spectrometer. Data were analyzed with MaxQuant software to identify the total number of proteins and to quantify proteins in individual samples. Protein function was analyzed using the PANTHER Overrepresentation tool. Associations between perilymph protein content, clinical parameters, tumor-associated hearing loss and tumor diameter were assessed using Random Forest and Boruta. In total, 314 proteins were identified; 60 in all 15 patients and 130 proteins only once in 15 patients. Ninety-one proteins were detected in at least 12 out of 15 patients. Random Forest followed by Boruta analysis confirmed that alpha-2-HS-glycoprotein (P02765) was an independent variable for tumor-associated hearing loss. In addition, functional analysis showed that numerous processes were significantly increased in the perilymph. The top three enriched biological processes were: 1) secondary metabolic processes; 2) complement activation and 3) cell recognition.

Conclusions: The proteome of perilymph in patients with vestibular schwannoma has an inter-individual stable section. However, even in a cohort with homogenous disease, the variation between individuals represented the majority of the detected proteins. Alpha-2-HS-glycoprotein, P02765, was shown to be an independent variable for tumor-associated hearing loss, a finding that needs to be verified in other studies. In pathway analysis perilymph had highly enriched functions, particularly in terms of increased immune and metabolic processes.

Fig 1. The number of proteins identified in declining order of patients.

This figure shows the distribution of 314 proteins identified in perilymph samples from fifteen patients. Sixty proteins were identified in all samples, and 91 in 12 or more samples. Note that 130 proteins were only identified once in 15 patients. Mass spectrometer data was analyzed using MaxQuant software.

Fig 2. Boruta analysis for the outcome tumor-associated hearing loss.

In brief Boruta analysis identifies variables performing better than random variables based on repeated Random Forest analysis. The shadow value represents random data. Boruta analysis showed that Alpha-2-HS-glycoprotein (P02765) was stronger than random data, while the four proteins identified as significant by univariate linear regression (P01857, P01834, P01024 and P01860) were rejected.

Fig 3. PANTHER overrepresentation test of biological processes.

Orange bars represent the proteins detected in the perilymph while gray bars represent the expected number of proteins in each biological process group. The scale is logarithmic. In total, perilymph proteins were involved in 38 up-regulated biological processes while involved in two down-regulated biological processes.

Fig 4. PANTHER overrepresentation test of molecular function.

Radar diagram, with a logarithmic scale, showing a comparison between the numbers of expected and detected proteins in each molecular function group. Orange boxes represent the detected number while gray triangles show the expected number. In total, 12 enriched molecular functions and one down-regulated.

Fig 5. PANTHER overrepresentation test of cellular component.

Radar graph with a logarithmic scale showing a comparison between the numbers of expected and detected proteins in each cellular component group. Within the perilymph (orange boxes), nine functions were up-regulated compared to expected levels (gray triangles) while two were down-regulated.

Fig 6. PANTHER overrepresentation test of PANTHER pathways.

Image shows a logarithmic radar graph comparison between the detected (orange boxes) and expected (gray triangles) numbers of proteins in each group. Seven significantly up-regulated PANTHER pathways were found.

Fig 7. PANTHER overrepresentation test of PANTHER protein class.

Image shows a logarithmic radar graph of the comparison between the number of detected (orange boxes) and expected (gray triangles) proteins in each group. Twenty up-regulated protein classes and one down-regulated were identified.