Uveal melanoma (UM) is the most common primary intraocular tumor in adults . Approximately 50% of patients with UM develop metastatic disease and up to 85% of these patients succumb to their visceral metastases. Currently, patients are screened with serial body imaging, which typically presents 2–4 years after primary diagnosis. However, it is suspected that micro-metastatic disease may develop 1–5 years prior to detection. Determining which patients are likely to metastasize has been the focus of recent molecular testing. To determine patients at higher risk for developing metastatic disease, clinical or histopathologic risk factors are considered . More recently, gene expression analysis (i.e., gene expression profiling [GEP] and preferentially expressed antigen in melanoma [PRAME] status) after direct surgical tumor sampling has enhanced prognostic accuracy [3,4,5]. While tumor biopsies are beneficial for assessing mortality risk, they are still subject to certain risks (e.g., risk of retinal detachment and tumor dissemination) and not amenable to repeat testing [3,4,5]. Furthermore, they do not identify protein targets to direct new adjuvant treatments. In contrast, liquid biopsies offer a minimally invasive method for real-time molecular assessment of the primary tumor. Diagnostic vitrectomies are reproducible, repeatable, and carry a lower risk of adverse outcomes since they do not require invasion of the primary tumor. Serial fluid biopsies from the eye might allow prospective metastatic surveillance.
Global protein changes may provide an independent biomarker that reflects the total sum of the tumor effects or a supplemental biomarker that could enhance diagnostic sensitivity and specificity. Our research has demonstrated that retinal proteins leak into the vitreous, and that proteomic analysis of vitreous biopsies can detect molecular changes in the adjacent retina, uncover biomarkers, and identify protein targets for adjuvant therapy . These biomarkers are concentrated in fluid compartments adjacent to the primary tumor (Fig. 1a). There have been proteomic studies on UM tumor cells which have suggested differentially expressed proteins (DEPs), but liquid protein markers have yet to be validated (Table S1). Few studies utilized quantitative platforms and none, to date, measure protein biomarkers in the vitreous. In this study, we use proteomics to detect candidate biomarkers for UM and identified targets for drug repositioning.
Targeted proteomic analysis distinguishes molecular classes of UM
A total of eight patients were diagnosed with UM and used for the discovery phase analysis. The clinical and demographic information are described in Table S2 [4, 7, 8]. Vitreous biopsies were collected in the operating room (see Methods). For patients undergoing I-125 plaque brachytherapy, vitreous biopsies were collected prior to placement of the active plaque and before tumor biopsy. No biopsy complications were observed. In enucleated eyes, vitreous collection was performed following trans-scleral tumor biopsy (without vitreous violation) prior to formalin fixation. There were 77 proteins DEPs among control and UM vitreous (62 upregulated and 15 downregulated proteins; p < 0.05). Among the proteins upregulated in UM vitreous were SIGL6, c-Myc, and SCFR/c-Kit. FABP1 and KLK7 were highly expressed in control vitreous. When comparing protein expression by GEP class, there were 46 DEPs (p < 0.01; Fig. 1b; Table S3). Among the upregulated proteins in GEP Class-2 patients were HGFR/c-MET, SIRT1, DNMT3A, common β-chain (βc), ARGI1, and FasL. When comparing by PRAME status, there were 32 DEPs (p < 0.01; Fig. 1c-d; Table S4). Among the upregulated proteins in PRAME-positive patients were DSG3, ENPP-2, LEG9, and HGF. The top-represented pathway detected in UM vitreous was the STAT3 pathway (p-value = 0.004; activation z-score = 2.7; Fig. 1e). Forty-two downstream STAT3 signaling effectors were differentially expressed. Recent transcriptomic analysis of UM liver metastases detected STAT3 upregulation . Interestingly, vitreous from patients with GEP Class-2 and PRAME-positive tumors displayed higher representation of STAT3 signaling proteins. Taken together, these results identified potential vitreous proteins that correlate with gene expression patterns for UM metastatic risk.
Biomarker validation study
To prospectively validate the discovery dataset, we examined an independent cohort of patients. Twenty proteins were selected for further study based on their statistical significance and biological function (Table S5). We generated a custom ELISA that measured these 20 proteins in control (n = 11) and UM (n = 11) vitreous, as well as in control (n = 5) and UM plasma (n = 8; Fig. 2a). The clinical and demographic information are described in Table S6. KLK7, a negative control marker, was decreased in UM vitreous compared to controls, as predicted (p = 0.001; Fig. 2b). We also confirmed the presence of elevated SCFR (p = 0.011) and HGF (p = 0.044) in UM vitreous (Fig. 2c-d). When analyzed by genetic class, we observed an increase in c-Myc levels in GEP Class-2 vitreous compared to GEP Class-1A (p = 0.009; Figs. S1, S2 and S3). We observed elevated HGFR in AJCC Stage-III and -IV UM vitreous (p = 0.046; Fig. 2e). Interestingly, ENPP-2 levels were increased in UM vitreous in the training dataset but decreased in the validation cohort compared to controls (p = 0.02; Table S7). These results support the verification of HGF, HGFR, and SCFR expression in UM vitreous. ENPP-2 and ARGI1 levels were significantly elevated in UM plasma (p = 0.044 and 0.042, respectively; Fig. 2f-h; Table S8). Despite levels of βc being elevated in UM vitreous in the training dataset, βc levels were decreased in the UM validation cohort (Fig. 2f). These results suggest that ENPP-2 and ARGI1 could serve as independent plasma biomarkers for UM. Prospective, serial studies are needed to further validate their expression in the UM plasma proteome.