Article Text

Download PDFPDF

The biology of micrometastases from uveal melanoma
  1. Nicola J Borthwick1,
  2. Jasmine Thombs2,
  3. Marta Polak3,4,
  4. F Guy Gabriel3,
  5. John L Hungerford2,
  6. Bertil Damato5,
  7. Ian G Rennie6,7,
  8. Martine J Jager8,
  9. Ian A Cree1,2,3
  1. 1Institute of Ophthalmology, London, UK
  2. 2Moorfields Eye Hospital, London, UK
  3. 3Cancer Laboratory, Level F—Pathology Centre, Queen Alexandra Hospital, Portsmouth, UK
  4. 4Division of Infection, Inflammation and Immunity, University of Southampton, School of Medicine, Southampton, UK
  5. 5St Paul's Eye Unit, Royal Liverpool University Hospital, Liverpool, UK
  6. 6Department of Ophthalmology, The University of Sheffield, Royal Hallamshire Hospital, Sheffield, UK
  7. 7Department of Orthoptics, The University of Sheffield, Royal Hallamshire Hospital, Sheffield, UK
  8. 8Department of Ophthalmology, Leiden University Medical Center, Leiden, The Netherlands
  1. Correspondence to Professor Ian A Cree, Cancer Laboratory, Level F—Pathology Centre, Queen Alexandra Hospital, Portsmouth PO6 3LY, UK; ian.cree{at}


Background The aim of this study was to investigate the possible causes of tumour latency in uveal melanoma primarily through the analysis of micrometastases in tissue obtained from donors postmortem. Various explanations have been proposed but there is no clear answer from animal studies and few human data. The main hypotheses may be divided into several areas—immunological control of metastatic cells, lack of angiogenesis within micrometastases and reduced cell turnover.

Methods 196 patients were recruited to the study between 2003 and 2007. Patients were invited to take part and their relatives agreed to postmortem examination of their liver and lungs in the event of their death, including tissue sampling to assess the presence of micrometastases and their biology. Metastatic cells were detected by immunohistochemistry using a pan-melanoma antibody reagent, and by quantitative reverse transcriptase (qRT)–PCR for three melanoma-associated genes (tyrosinase Melan-A, and gp100) and a housekeeping gene (HMBS/PBGD) in samples stored in RNAlater or as formalin-fixed paraffin-embedded tissue.

Results 22 deaths were investigated at autopsy as part of the study. Sixteen patients died with large deposits of metastatic melanoma, while six patients died of other causes. In addition, a liver resection for hepatic adenoma provided further tissue from a case without clinical evidence of metastasis. Metastatic melanoma cells were identified by immunohistochemistry of the liver samples in one case and by qRT–PCR in two further cases without macrometastases. There was no evidence of multicellular micrometastases sufficiently large to require angiogenesis and no associated inflammation was observed.

Conclusion The most likely explanation for latency in this setting is the inability of uveal melanoma cells in metastatic sites to grow.

  • Autopsy pathology
  • eye
  • growth
  • immunosuppression
  • latency
  • melanoma
  • metastasis
  • microenvironment
  • PCR

Statistics from

Uveal melanoma provides an excellent model of haematogenous metastasis, as the primary tumour does not have access to lymphatics. It also exhibits latency: clinically evident metastases may only appear many years (even several decades) after removal of the primary tumour. Uveal melanoma is relatively rare with an incidence of 0.5 per 100 000 in most Caucasian populations.1 However, it accounts for 13% of melanoma deaths.2 It has a 50% mortality rate at 15 years and in 85% of these cases, haematogenous liver involvement predominates with lung involvement common in the remainder.3–6 Metastasis of uveal melanoma is therefore a selective process that follows the ‘seed and soil’ hypothesis proposed by Paget in 1889.7 The molecular biology of uveal melanoma is still poorly understood, but we have shown from epidemiological evidence that there are likely to be four biological events required to obtain a metastatic tumour, and that three of these are required for primary tumour formation.1 Activating mutations of the GNAQ gene have recently been identified as an early event in uveal melanoma,8 and a model of genetic progression has been proposed.9 There is evidence that metastasis occurs more readily in tumours with monosomy 3, which have a worse prognosis.10

Latency of metastatic cells has long puzzled cancer researchers. Various explanations have been proposed but there is no clear answer from animal studies and few relevant human data. The main hypotheses may be divided into several areas—immunology, angiogenesis and cell turnover/biology.7 11

Tumour cell quiescence—reversion of malignant metastatic cells to a non-proliferative state, has been a common explanation. However, cell proliferation is permanently switched to the ‘on’ position by mutation of key genes in melanoma and other cancers.12 13 Nevertheless, many tumour cells also remain dependent on autocrine or paracrine growth factor secretion, so this may occur. Limited information from studies of bone marrow aspirates and lymph node resection specimens14 do suggest that metastatic cells may be in a non-proliferative state.15 Kivela and Grambsch16 suggest that the doubling time of uveal melanoma metastases is of the order of 50 days.

The critical role of angiogenesis in tumour biology suggests that metastatic cells do not grow because they lack a necessary angiogenic switch that would permit a small focus of cancer cells to form a tumour mass. Uveal melanoma is held to be an excellent example of this as angiogenesis within the tumour correlates well with prognosis.17–20 The balance of anti-angiogenic and tumour-produced pro-angiogenic signals within a tissue may differ from the primary tumour, leading to an inability of the metastatic focus to grow to detectable size.21

Escape of tumour cells from immune surveillance/control has long been postulated as a mechanism of latency for metastases. In uveal melanoma, Jager and others have produced convincing evidence of the influence of changes in tumour antigenicity, particularly HLA class I expression, on outcome.22–25 We have shown that uveal melanomas contain immature dendritic cells with an immunosuppressive phenotype.26 There is a stronger suspicion that for cutaneous melanoma immunosuppression by the tumour is important.26–28 Both the eye and the liver are immune privileged sites, which may explain the predilection of uveal melanoma to metastasise to the liver.

These models are not mutually exclusive. Cell kinetics must favour growth of a tumour before it will become clinically apparent—ie, the rate of proliferation must exceed the rate of cell loss. This study set out to find latent metastatic cells in uveal melanoma patients who died from other causes and might therefore have latent metastasis. This required an autopsy study. Autopsy-based research has serious ethical considerations, and the authors consulted widely to ensure that current and predicted future standards were met in full from inception of the study.29


In this prospective postmortem study of patients with uveal melanoma, we have used immunohistochemistry and quantitative reverse transcriptase (qRT)–PCR to examine the presence of micrometastases and to explore possible mechanisms of latency.


Patients with uveal melanoma attending each of the three main UK centres (London, Liverpool and Sheffield) were included in the study, which was approved by the Coroners' Society and by a Multicentre Research Ethics Committee. The recruitment procedure involved an initial introduction of the study to patients at routine follow-up appointments following treatment of their primary tumour, at which they were given a copy of the information sheet and consent form by experienced research nurses employed specifically for this project.29 Patients were encouraged to discuss these with their close relatives as both were asked to give their written consent. Patients giving their consent were asked to carry a donor card to carry so that they could inform medical staff of their participation in the study, and letters were sent to their family doctors.

Inclusion and exclusion criteria

All patients with a history of large (>10 mm largest tumour diameter) uveal melanoma treated by enucleation or radiotherapy were eligible for the study. Follow-up followed usual practice within the centres. Patients not attending for follow-up but meeting the eligibility criteria were offered a special visit to discuss the study. All patients and their immediate relatives gave fully informed consent to postmortem for removal and retention of tissue as detailed above. Exclusion criteria were a history of previous non-ocular melanoma or metastatic disease at the time of diagnosis of the primary ocular tumour.

Postmortem examinations

Autopsies were conducted by local pathologists in accordance with normal procedures approved by the Royal College of Pathologists following reporting to the central coordinating office of the death of a patient by their relatives, general practitioner or oncologist. The consent document signed by the patient and relatives was faxed to the relevant pathologist or other physician following telephone contact, and a confirmation sought from the relatives. A decision was sought from the coroner regarding deaths occurring within his or her jurisdiction. Autopsies were conducted as soon as possible after death to reduce the influence of autolysis on tissue procured for the study, and the influence of autolysis was assessed by histology.


Liver samples were taken as 1-cm thick slices from each lobe, part of which was stored in RNAlater (Applied Biosystems Ltd, Foster City, California, USA), and the remainder fixed in 4% buffered formaldehyde, together with further slices from any macroscopically suspicious areas. In some cases, bone marrow samples were also obtained from the sternum or vertebral bodies as preferred by the pathologist. After the liver, the lung is the most common site of metastasis in uveal melanoma. Slices were taken from the left lower and right upper lobes for examination, together with a further slice from any macroscopically suspicious areas. Three millimetre punch samples of the material stored in RNAlater were snap frozen in liquid nitrogen and stored in a −80°C freezer until required for RNA analysis.

Detection of micrometastases

Two methods were used and tested on macrometastatic specimens before moving on to examine samples that may contain micrometastases:

  1. Frozen material and RNA extracted from sections of formalin-fixed paraffin-embedded (FFPE) liver blocks were screened by qRT–PCR. RNA was extracted from frozen material using a RNEasy kit (Qiagen Ltd., Crawley, UK) or a RecoverAll Total Nucleic Acid Isolation kit (no. 1970; Ambion, Huntington, UK) according to the manufacturer's instructions. Extraction of RNA from FFPE sections was performed as described in detail by Polak et al.30 Briefly, two 4-μm sections were taken from both lung and liver, including melanoma involved and grossly uninvolved tissue. The RNA samples were reverse transcribed to cDNA as soon as possible after isolation and stored for further investigation using an ABI High-Capacity cDNA Archive kit (Applied Biosystems, Warrington, UK).30 A two-stage qRT–PCR method was used to examine the presence of the target melanoma-associated genes tyrosinase (TYR), premelanosome protein (gp100) and Melan-A (also known as MART-1) and a housekeeping gene (hydroxymethylbilane synthase, HMBS, also known as porphobilinogen deaminase, PBGD) in RNA from both frozen and FFPE liver samples using Taqman assays obtained from Applied Biosystems for the targets listed as previously described.30 Real-time PCR was performed in an i-cycler PCR machine (Bio-Rad Laboratories Ltd., Hemel Hempstead, UK) for 42 cycles. Results were regarded as positive if they rose above the background threshold set manually using no template control wells within these 42 cycles.

  2. Large paraffin-processed blocks were screened for micrometastases by Avidin-Biotin Complex (ABC)-alkaline phosphatase using a pan melanoma antibody cocktail (Dako Ltd, Ely, UK) and the Vector Red substrate to avoid problems of interpretation due to brown pigments such as melanin or haemosiderin. Further micrometastases not identified in the initial section from each block were sought by cutting three deeper levels at intervals of 100 μm.

Data analysis

The data (linked-anonymised) were collected in a relational database (ACCESS) and extracted to Excel spreadsheets for analysis.



The number of patients recruited to the study and fully consented was 196 out of 561 approached (35%). However, a significant number of patients said that they would like to participate and took forms to discuss with their relatives: many decided to keep these until they became ill and were therefore regarded as verbal consent pending written consent. If these patients are included, then 76% of those approached were prepared to take part in the study (table 1). There was some variation in the percentage recruitment between centres due to local factors. The median age of patients in the study was 65 years (range 22–84).

Table 1

Recruitment to study by centre

Follow-up and autopsy numbers

The study ran for 54 months, with a median follow-up of 69 months (range 24–500 months). A total of 25 deaths, 12.8% of those who gave written consent, occurred during the study and were reported to the organisers. Of these, 22 were investigated at autopsy as part of the study and tissue obtained. Sixteen patients died with large deposits of metastatic melanoma, while six patients died of other causes. In addition, a liver resection for hepatic adenoma provided further tissue from a case without clinical evidence of metastasis, and was grouped with tissue from those dying of other causes for the purposes of analysis. The median age of all 23 patients was 64 years, range 22–89 years. There were 13 male and 10 female patients.

RNA isolation from postmortem tissue

The yield of RNA from RNAlater samples varied, but was normally approximately 50 μg/sample, determined by optical density. RNA from a primary uveal melanoma was used to standardise primers for three melanoma genes (tyrosinase, Melan-A and gp100) and the preferred housekeeping genes (HMBS). The cDNA from postmortem lung and liver tissue was found to be suitable for qRT–PCR and it was possible to detect melanoma-associated transcripts in most cases, despite significant histological evidence of autolysis in liver samples in particular. Comparison of histological versus molecular techniques for the detection of melanoma in this postmortem-derived tissue suggested both techniques to be reliable and consistent.

In a second approach, we standardised the isolation of RNA from FFPE tissue blocks normally used for histology. This is a more economical approach due to the large amount of RNAlater otherwise required. Autolysis is a major problem for postmortem studies and we did not find RNAlater any better than FFPE samples for this purpose. We were able to isolate good quality RNA from FFPE tissue blocks of postmortem lung and liver samples and successfully used these to measure melanoma-associated gene transcripts. There is some evidence that this may be a more successful method of obtaining RNA postmortem, although the current study did not examine this directly.

Histology and qRT–PCR results in those dying without clinical evidence of melanoma

Convincing immunohistochemical evidence of metastatic cells was only found in one patient's liver sample of those dying from other causes (table 2, case 3). In this case (figure 1) a small clump of three cells was identified within the liver parenchyma, positively stained with pan-melanoma antigen. This particular sample showed some autolysis and qRT–PCR (table 3) did not confirm the presence of melanoma cells. In three other samples (cases 1, 5 and 6), equivocal results were obtained—either there was weak staining or the morphology was not convincing. Nevertheless, two of these (cases 1 and 6) showed positive qRT–PCR with more than one melanoma marker. Two further samples were regarded as equivocal by qRT–PCR—in both cases only weak gp100 expression was identified. Three liver samples were not evaluable by qRT–PCR, probably due to autolysis.

Table 2

Cases, cause of death and summary of findings by immunohistochemistry and histology (H&E) or PCR

Figure 1

Immunohistochemical demonstration of melanoma cells (stained red with pan-melanoma antigens) in liver (×200 original magnification) in a patient (case 3) without clinical evidence of metastasis at the time of death.

Table 3

Quantitative RT–PCR results for melanoma markers (a) liver, (b) lung

Combining immunohistochemistry and qRT–PCR data suggest positive results in three of the seven cases studied based on unequivocal positivity by either method. Histologically, the equivocal or convincing cells were widely scattered. Only single cells or in one case a small clump of cells were seen, rather than true micrometastases (ie, 0.2–2 mm clumps). There was no evidence of multicellular micrometastases sufficiently large to require angiogenesis and no inflammatory cells were seen associated with the cells.

Histology and PCR results in those dying with metastatic melanoma

Histology and immunohistochemistry showed large metastases with small satellites around them in the livers of all 16 patients who died of metastatic uveal melanoma, as shown in figure 2 (case 14). Seven of the 13 cases in which lung samples were available also had macrometastases in the lung, while four showed limited metastatic spread (mostly single cells and small clumps of cells, figure 2) and two had no evidence of lung metastasis. In the seven cases with bone marrow samples, three showed macrometastatic spread and four showed no bone marrow involvement (table 2).

Figure 2

Immunohistochemistry (pan-melanoma antigens) of lung with micrometastatic cells in a patient (case 14) dying of metastatic uveal melanoma (×400 original magnification).

The PCR results were evaluable based on HMBS housekeeping gene expression positive for more than one marker in seven of 11 liver samples and in these all seven were positive for at least two melanoma markers, as expected. For lung samples, six out of seven were evaluable based on HMBS expression and all six showed positivity for at least two melanoma markers (table 3).


This study suggests that uveal melanoma latency is due primarily to the inability of cells in metastatic sites to grow. Single cells or a few cells in a single clump were found within the liver of otherwise asymptomatic patients. These are likely to be derived from circulating tumour cells, which despite the failure of early attempts31 to find them, have been demonstrated in uveal melanoma patients using more sophisticated methods.32 33 Such cells must pass through the lung first, yet our data suggest that although lung metastases occur, they are usually smaller and may be microscopic even in the presence of very large liver metastases. Animal experiments by Fidler7 and others suggest that the reason these cells do not grow may be due to differences in the microenvironment at the two sites. The process of metastasis involves circulating cell adherence to liver endothelium and migration into the liver parenchyma. It is unclear whether uveal melanoma cells adhere less well to lung endothelium, are unable to cross it, fail to grow, or are killed by immunological processes. Many molecular differences have been noted between primary and metastatic uveal melanomas.34 Differences in the expression of adherence-associated molecules have been noted between melanoma cell lines,35 and between primary tumours and their metastases.36 Common cytogenetic abnormalities have also been associated with metastasis, and at least one, partial deletion of chromosome 8p, has been associated with a molecule, LZTS1, which alters the motility and invasive potential of uveal melanoma cell lines.37 Loss of chromosome 1p is also associated with metastatic disease, and it has recently been suggested that the loss of a p53-like molecule, p73, may be important in these cases.38 This is particularly interesting, as it would affect the ability of cells to resist growth arrest, apoptosis and possibly differentiation, and represents loss of a differentiation gene, which has previously been postulated in uveal melanoma.1 39 40

This study has found no evidence that latency is due to a failure of the ‘angiogenic switch’, which would be expected to produce growth arrest of cell clumps less than 1 mm in diameter.21 Angiogenesis is in any case required for growth of the primary tumour and it is difficult to see why this would be switched off. Lack of growth, as described above, seems a much more likely explanation for the single cell latency observed.

Equally, we found no evidence of immunological control of micrometastases, although data from other studies, including animal models, suggest that immunosuppressive mechanisms may play an important role in the development of larger metastatic foci.41 They suggest that systemic immune responses to uveal melanoma may modulate the ability of micrometastatic cells to grow following treatment of the primary tumour. The cytokines of most interest appear to be IL-10 and both isoforms of transforming growth factor beta. Natural killer cells are thought to be important mediators of melanoma cell kill and these cytokines are also important regulators of their function.42

Recruitment to this study has been discussed elsewhere, but was impressive given the level of press interest in the topic of postmortem research in the UK over the years preceding this project.29 Of great importance was the ability of individual centres to employ experienced cancer research nurses able to talk openly to patients about the study. Many patients were grateful for the opportunity to talk about their own mortality and used the study to discuss matters with their relatives. We have found no similar autopsy studies of other tumour types, although micrometastasis to bone marrow and to draining lymph nodes is well documented in a number of tumour types, notably cutaneous melanoma,43 44 in which it has prognostic significance but has not been related to latency.


Although the reasons why metastatic single cells fail to grow are unknown, this study perhaps suggests that a dual approach to discourage melanoma cell growth and enhance anti-tumour immunity may be required to improve patient survival. Identification of treatment strategies that target micrometastatic cells, for example within the insulin-like growth factor pathway,45 should perhaps be allied to immunological approaches. It may be possible to examine such hypotheses within animal models of tumour dormancy.46

Take-home messages

  • Animal studies have limited relevance to the study of metastasis in comparison with the analysis of tissue containing micrometases, obtained from donors at autopsy.

  • Tumour latency has been postulated to result from immunological control of metastatic cells, lack of angiogenesis within micrometastases, or reduced cell turnover.

  • In this study, metastatic melanoma cells were identified by immunohistochemistry of the liver samples in one case and by qRT–PCR in two further cases without macrometastases. There was no evidence of multicellular micrometastases sufficiently large to require angiogenesis and no associated inflammation was observed.

  • This study suggests that the most likely explanation for latency is the inability of cells in metastatic sites to grow.


The authors are grateful to the patients and their families who considered participation in this trial at times of great stress or personal loss, and to those pathologists who performed the autopsies. They also wish to thank the many research nurses who contributed to the project.



  • Funding This project was funded by Cancer Research UK, grant no C7498/A2838.

  • Competing interests None declared.

  • Ethics approval This study was conducted with the approval of the Eastern Multi-centre Research Ethics Committee, House No 1, Papworth Hospital NHS Trust, Papworth Everard, Cambridge CB3 8RE. Approval for the study was also given by the Home Office and the Coroners' Society.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.