The Molecular Classification of Uveal Melanocytic Lesions

Abstract

This Commentary discusses a multi-gene expression assay for predicting metastasis in uveal melanoma.

In 2004, Onken et al¹ reported the strong association between a gene expression profile and outcome in uveal melanoma. In this issue of the Journal of Molecular Diagnostics, Onken et al² describe the migration of their technology (Table 1)² to a new platform, the ability to detect the revised profile from formalin-fixed paraffin-embedded tissues as well as from fine-needle aspiration biopsies of intraocular tumors in patients, and preliminary outcomes data. What has emerged is a clinically useful and accurate molecular prognostic classification for uveal melanocytic neoplasms. This significant advance appears nearly 80 years after Callender's discovery that the morphology of uveal melanoma cells is associated with outcome.³ At that time, and for many decades thereafter, most patients with intraocular melanocytic tumors were treated by removal of the affected eye. Pathologists received the specimen and recorded a number of morphological observations. The Callender classification of uveal melanomas was eventually refined and modified⁴ and in many studies was shown to be a leading independent prognostic indicator for survival.⁵ Sadly, to this day, there is no effective treatment for metastatic uveal melanoma. Nevertheless, the ophthalmic oncology community has remained hopeful that an effective treatment is “around the corner,” and to be prepared to stratify patients in treatment groups, there is a need to assign prognosis to patients.

Table 1.

Genetic Profile for Predicting Metastasis in Uveal Melanoma2

Genes studied	Biological function
Control genes
MRPS22	Mitochondrial 28S ribosomal protein S22, involved in protein synthesis
RBM23	Probable RNA-binding protein 23
SAP130	Histone deacetylase complex subunit
Discriminating genes
HTR2B	Serotinin receptor that signals through GNAQ, which is mutated in half of uveal melanomas
ECM1	Extracellular matrix protein that plays a role in angiogenesis and is overexpressed in some tumors
RAB31	May play a role in mitogenic signaling through vesicle transport machinery
CDH1	Cadherin-1 loss of function results in proliferation, migration, and invasion
FXR1	Nucleolar targeting factor and possible tumor suppressor
LTA4H	Involved in leukotriene synthesis, potential modulating immune function and tumor infiltration
EIF1B	Translation initiation factor that is down-regulated in some cancers
ID2	bHLH inhibitor and tumor suppressor
ROBO1	Axonal cue receptor implicated in angiogenesis and tumor migration
LMCD1	Gata transcription factor inhibitor that regulates migration and differentiation of mesenchymal cells
SATB1	Regulates higher level chromatin organization
MTUS1	Tumor suppressor

For many decades, ophthalmologists struggled to be certain that an intraocular tumor was indeed a melanoma and not a lesion that simulated melanoma, such as choroidal hemangiomas, hemorrhages, or metastases to the eye. By the 1990s, the clinical accuracy of discriminating uveal melanomas from simulating lesions had improved significantly.⁶ Nevertheless, some ophthalmic oncologists began to confirm the diagnosis of uveal melanoma with fine-needle aspiration biopsies.⁷ After the sampling technique was refined, ophthalmic oncologists wondered whether fine-needle aspiration biopsy could be used to obtain prognostic information—to extract tumor cells and apply the Callender classification to assign prognosis. The question was important, because unlike the application of fine-needle aspiration biopsy techniques in other diagnostic settings where the goal of the procedure is to establish a diagnosis, the new goal of the ophthalmic fine-needle aspiration biopsy was changed to obtain prognostic information. Ophthalmic oncologists typically take a very limited number of passes into a tumor within the eye, unlike the approach of taking multiple samples at varying angles and for masses elsewhere in the body to ensure a representative sampling. In using fine-needle aspiration biopsy to establish prognosis based on the Callender classification, the underlying assumption was that the cells that conferred the most adverse prognosis would be captured by a few random samples into a tumor. The assumption was incorrect: epithelioid cells—those cells associated with the worst prognosis under the Callender classification—were not distributed randomly,⁷ and the use of fine-needle aspiration biopsy to establish a Callender-based prognosis never matured into widespread clinical practice.

Meanwhile, the clinical approach to treating patients with intraocular melanomas shifted radically from removal of the eye recommended for almost every patient with uveal melanoma to treatments such as various forms of radiation therapy to eradicate the tumor while hopefully preserving vision whenever possible. Eventually, it was shown that there was no significant difference in survival between patients whose affected eyes were removed and patients with medium-sized intraocular tumors that were treated by radiation therapy.⁸ Therefore, whenever possible, patients were treated by modalities that did not involve the removal of any tissue. Uveal melanomas are among the only forms of cancer treated without a pathologist ever examining tissue to either confirm the diagnosis or assign a prognosis based on the best information available.

The discovery of a cytogenetic profile associated strongly with outcome in 1990⁹ was confirmed widely and was initially applied to tumor tissue from eyes that still required removal. Many ophthalmic oncologists attempted to use fine-needle aspiration biopsy to extract cells for cytogenetic analyses and to establish prognosis for patients with uveal melanomas based on the detection of monosomy 3 and other cytogenetic markers associated with an unfavorable outcome.¹⁰ Once again, the question was raised: given the tendency for many cancers to be heterogeneous in the expression of tumor-related markers, are the cytogenetic markers of interest distributed uniformly so that a limited number of passes into the tumor with a fine needle will capture the marker of interest? Recently, investigators have reported a heterogeneous distribution of cytogenetic markers of interest in uveal melanoma,^11,12 thus raising doubt about the sensitivity and specificity of this test for establishing prognosis in uveal melanoma.

The technical achievements now described by Onken et al²—migrating their platform for gene expression and providing evidence that this test can be applied successfully to material submitted from multiple contributors—are remarkable. Most remarkable, however, are two findings that must be considered within the context of the historical background described above: the degree of separation between patients who have a “class 1” gene expression profile (the survivors), and the observation that heterogeneity of expression is not a factor in this test as it has been for previous attempts to apply prognostic criteria from very limited material extracted by randomly placed fine-needle aspiration biopsies.

Three elements of caution are in order. First, the onset of metastatic uveal melanomas can be delayed by as many as two decades after treatment of the primary tumor, so the molecular classification of patients with uveal melanoma may not hold as extended follow-up becomes available. However, the assay described by Onken et al² can be applied to formalin-fixed paraffin-embedded tissues, perhaps allowing investigators to study archival tissues from patients whose eyes were removed many years ago and infer the long-term reliability of the prognostic test from retrospective analyses. Second, it is not known whether class 1 lesions ever evolve into class 2 lesions. This issue is important clinically because ophthalmologists may choose not to treat patients with clinically asymptomatic class 1 lesions and simply observe the lesion. Third, as Onken et al² admit, additional work must be done to ensure that a needle randomly placed into a uveal melanoma will yield material that is representative and will result in an accurate molecular classification.

Let's assume an optimistic posture that with time, the molecular classification described by Onken et al² holds and that the authors and others provide convincing evidence that the classification is valid in the context of random sampling within the tumor. Now, let's look carefully at the data and probe deeper. Onken et al's Figure 2, a Kaplan–Meier curve of survival with outcome recorded at the percentage of patients developing metastasis over time, indicates that not one patient whose tumor was assigned to class 1 on the basis of the authors' gene expression profile developed metastatic melanoma. When combined with the authors' claim that if tumor cells with a class 2 expression profile would have been detected had they been present in class 1 tumors, one could draw the conclusion that class 1 tumors are not melanomas—they are benign melanocytic lesions.

This conclusion is not at all radical. There are only three melanocytic lesions described in the uvea: nevi, melanocytoma (magnocellular nevus), and melanomas. There are no uveal counterparts to Spitz nevi, pigmented spindle cell tumors, and a variety of other benign melanocytic lesions that are described in the skin.¹³ It should not be at all surprising that benign melanocytic lesions that simulate uveal melanoma exist in the uvea just as benign melanocytic lesions that mimic cutaneous melanomas exist in the skin. It would therefore be helpful for the authors—or other investigators—to look carefully at tumors classified as “class 1”—those patients who are fortunate to have developed benign uveal melanocytic lesions—for either clinical characteristics or histological characteristics that could allow the ophthalmologist to predict the molecular classification by noninvasive means, including imaging studies.

Attention can then be focused on refining the molecular classification of patients whose tumors are assigned as “class 2” by the current gene expression profiling strategy. Are there lesions that are low-grade uveal melanomas or even uveal melanoma precursors that have some risk of metastasis? For example, although the concept of “atypical uveal melanocytic hyperplasia” was advanced as a clinical hypothesis,¹⁴ this theoretical construct has never been described clinically or histologically. Might further refinements of a molecular classification identify such lesions in the uvea?

The molecular characterization of the primary uveal melanocytic lesions may represent only the beginning of an emerging diagnostic strategy for these patients. Uveal melanoma spreads preferentially to the liver,¹⁵ where micrometastases may be difficult to detect through the currently available modalities of imaging studies and elevations in liver enzymes. Uveal melanoma cells secrete a number of proteins, and early studies suggest that elevations in selected serum biomarkers may precede the clinical detection of uveal melanoma hepatic metastases with conventional liver imaging studies by six months.¹⁶ For these reasons, oncologists are now discussing the feasibility of developing large banks of serum from patients with uveal melanomas drawn over an extended time course, from diagnosis through years of follow-up, as well as serum from matched controls. Just as the precision of the molecular classification of primary uveal melanocytic lesions improved with the addition of multiple genes to the profile,^1,2 it is hoped that a proteomic approach to the detection of multiple serum biomarkers may help to identify patients who have subclinical hepatic metastases, thus providing opportunities for therapeutic approaches that can be offered to patients before the metastases become clinically overt.

The Callender classification of uveal melanomas,³ based on morphology alone, was a major advance when it was published nearly 80 years ago. The molecular characterization of these lesions provides yet another platform for continued discovery to the benefit of patients whose disease—uveal melanoma—is not only life threatening but potentially blinding.