Acta Ophthalmologica 2014
MicroRNA expression analysis and Multiplex ligation-dependent probe amplification in metastatic and non-metastatic uveal melanoma Ann-Cathrine Larsen,1 Line Holst,2 Bogumil Kaczkowski,3 Morten T. Andersen,4 Valentina Manfe,2 Volkert D. Siersma,5 Miriam Kolko,6,7 Jens F. Kiilgaard,7 Ole Winther,3,8 Jan U. Prause,1 Robert Gniadecki2 and Steffen Heegaard1,7 1
Eye Pathology Institute, University of Copenhagen, Copenhagen, Denmark Department of Dermatology, Bispebjerg Hospital, University of Copenhagen, Copenhagen, Denmark 3 Department of Biology, Biotech Research and Innovation Centre, University of Copenhagen, Copenhagen, Denmark 4 Department of Clinical Genetics, Copenhagen University Hospital, Copenhagen, Denmark 5 The Research Unit and Section of General Practice, Institute of Public Health, University of Copenhagen, Copenhagen, Denmark 6 Department of Neuroscience and Pharmacology, University of Copenhagen, Copenhagen, Denmark 7 Department of Ophthalmology, Glostrup Hospital, University of Copenhagen, Copenhagen, Denmark 8 DTU Informatics, Technical University of Denmark, Lyngby, Denmark 2
ABSTRACT. Purpose: To determine the association of microRNA expression and chromosomal changes with metastasis and survival in uveal melanoma (UM). Methods: Thirty-six patients with UM were selected based on the metastatic status, and clinicopathological data were collected. Multiplex ligation-dependent probe amplification (MLPA) was used to identify chromosomal changes. Chromosomal changes and clinicopathological data were correlated with survival and metastasis. The microRNA expression was analysed in 26 of the 36 archived UM samples. Unsupervised analysis, differential expression analysis and Cox regression analysis were performed to determine the association with metastasis and survival. Results: Metastasis and metastatic death occurred in 20 patients, two patients died of other causes and one patient of unknown causes. A significant association between increasing size category (p = 0.002, log-rank), extraocular extension (p = 0.001), chromosome 3 loss (p = 0.033) and 1p loss (p = 0.030) and development of metastases was observed. Tumour, node, metastasis (TNM) staging showed a significant association with survival (p < 0.0001, log-rank). Adjusting for gender and age TNM size category T4 (p = 0.016, Cox regression analysis), mixed (p = 0.029) and epithelioid (p = 0.0058) cell types, chromosome 3 loss (p = 0.014) and 8q gain (p = 0.010) were significant prognosticators for a poor survival. Hierarchical clustering divided the UM into three groups based on microRNA expression. The clusters showed no association with clinical or histopathological features, TNM staging, metastasis or survival. Differential expression analysis did not reveal microRNAs related to metastasis or survival. Conclusions: The prognostic significance of chromosome 3 loss and 8q gain identified by MLPA analysis was confirmed in archived UM samples. The value of microRNA expression as a predictor of metastasis and survival in UM could not be confirmed. Key words: chromosomal aberrations – expression – genetic – metastasis – microRNA – Multiplex ligation-dependent probe amplification – prognostic factors – survival – uveal melanoma
Acta Ophthalmol. 2014: 92: 541–549 ª 2013 Acta Ophthalmologica Scandinavica Foundation. Published by John Wiley & Sons Ltd
doi: 10.1111/aos.12322
Introduction Uveal melanoma (UM) is the most common primary intraocular tumour in adults with an incidence of approximately 0.7 per 100.000 in Denmark (Isager et al. 2005). The course of UM is highly aggressive due to frequent metastases, and the 5-year survival rates range from 66% to 69% in Denmark (Jensen 1982; Isager et al. 2006). Despite advances in surgery, radiotherapy and chemotherapy, the survival rate has not changed. MicroRNAs (miRNA) are small non-coding RNAs that regulate gene expression through inhibition of translation or destabilization of targeted messenger RNAs. Research has proposed that miRNAs function as oncogenes or tumour-suppressor genes in various cancer types (Esquela-Kerscher & Slack 2006). Recent findings support the use of targeted therapy as a novel method to counteract the tumourpromoting functions of miRNAs (Kasinski & Slack 2010). MicroRNA expression analysis has identified expression variations in miRNAs in UM compared with normal tissue (Yang & Wei 2011). Additionally, miRNAs specific for metastasizing melanoma and miRNA acting as
541
Acta Ophthalmologica 2014
oncogenes have been detected (Radhakrishnan et al. 2009). Studies on UM cell lines have revealed miRNA-34a, b, c, miRNA-182, miRNA-137 and miRNA-9 as important tumour suppressors in UM functioning through several target genes (Yan et al. 2009, 2012; Chen et al. 2011; Liu et al. 2012). The investigation of the gene expression profile in UM has led to a classification of patients with UM into two prognostically significant classes (Onken et al. 2004). The utility of miRNAs for predicting metastatic risk has only been assessed in one study which showed a clustering into the same two prognostically significant classes (Worley et al. 2008). Multiplex ligation-dependent probe amplification (MLPA) is a PCR-based technique applied for genomic tumour typing, and a commercial kit is available. Multiplex ligation-dependent probe amplification analysis has been used to predict metastatic death by evaluation of chromosome 3 losses and 8q gains (Damato & Coupland 2009; Damato et al. 2009; Vaarwater et al. 2012). Recently, a study has shown MLPA to be efficient in classifying chromosome 3 loss in formalin-fixed, paraffin-embedded (FFPE) UM samples (Lake et al. 2012). The purpose of the present study was to identify differentially expressed miRNAs in metastatic UM and investigate the association between miRNA expression and survival. For comparison, the prognostic value of chromosomal changes identified by MLPA, and clinicopathological features were evaluated.
Material and Methods Approval for this study was obtained from the Committee on Biomedical Research Ethics (no. 23179), the Danish Data Protection Agency (no. 200941-3749) and the Registry of Human Tissue Utilisation. The study followed the tenets of the Declaration of Helsinki. Tumour samples and clinicopathological data
The pathology reports of patients with UM that had their eye enucleated were retrieved at the Institute of Eye Pathology, University of Copenhagen. Patients were selected between 1986 and 2009, with the aim of achieving a
542
metastatic group and an age- and gender-matched non-metastatic group. Patients treated with radiation therapy, cases of iris melanoma, cases with incomplete patient records, incomplete pathology reports or without available tumour blocks were excluded. Thirteen patients with metastatic disease (preferably present at the time of diagnosis) and 13 age- and gendermatched control patients without metastatic disease were selected for miRNA analysis. These patients and additionally five more patients with metastatic disease (preferably present at the time of diagnosis) and five ageand gender-matched control patients without metastatic disease were added for MLPA analysis. During the study period (2009–2012), one patient without metastatic disease left the country and died, and we were not able to assess any development of metastatic disease or cause of death in this patient. Furthermore, two patients in the non-metastatic group developed metastasis in the study period. This resulted in 15 patients in the metastatic group, seven of these with metastatic disease present at the time of diagnosis and 10 patients in the nonmetastatic group for inclusion in the miRNA analysis. For the DNA analysis, this resulted in the inclusion of 20 patients in the metastatic group, eight of these with metastasis at the time of diagnosis and 15 patients in the non-metastatic group. The patient with unknown metastatic status was included in analysis of all-cause mortality. Clinical files, pathology reports and histopathological samples were reviewed to collect the following: Age at diagnosis, gender, cell type, tumour height and thickness and extraocular growth. Presence or absence of closed extravascular matrix patterns (closed loops) was evaluated from H&E sections. The number of mitoses were evaluated in H&E sections in 40 highpower fields (HPF) using the 409 objective and characterized as low mitotic activity (0–1 per 40 HPF) and medium to high mitotic activity (≥2 per 40 HPF) as described by McLean et al. (1977). The presence of metastatic disease was identified using clinical patient records, abdominal ultrasound examination and chest X-ray records and searching the Danish Registry of Pathology. Tumour size categories were established based on thickness and
diameter, and the Tumour, Node, Metastasis (TNM)-based Staging System 7th edition was applied (Edge et al. 2010; Kujala et al. 2013). Follow-up time was either defined as the time from diagnosis to the date of death or from the time of diagnosis to the last followup date (17th August 2012). MicroRNA and DNA extraction
For DNA and miRNA extraction, four sections each of 20 lm thickness were cut from the FFPE eyes on a RM2255 microtome (Leica, Bannockburn, IL, USA). Sections were transferred to glass slides, and tumour tissue was isolated by hand with a scalpel. The tissue was melted in 1 ml 100% xylene (Ambion, Foster City, CA, USA) at 50°C for 3 min. The protocol of RecoverAll Total Nucleic Acid Isolation Kit (Ambion) was followed according to manufacturer’s instructions. MicroRNA array
The concentration of total RNA was measured spectrophotometrically using a NanoDrop ND-1000 (Thermo Scientific, Wilmington, DE, USA). Extraction resulted in a total yield in the range of 20–757 ng/ll. The OD A260/A280 ratio was in the range of 1.50–2.79 indicating high sample purity. The quality of total RNA was measured with an Agilent 2100 Bioanalyzer using the Agilent Nano RNA 6000 kit (Agilent Technologies, Santa Clara, CA, USA). The concentration of total RNA was adjusted to 500 ng/ll, and labelling was performed with the miRCURYTM LNA microRNA Hy3/Hy5 Power labelling kit (Exiqon, Vedbæk, Denmark) according to the manufacturer’s instructions. A dual colour system, in which the samples were labelled using fluorescent dyes: Hy-3 (532 nm) for the tumour samples and Hy-5 (635 nm) for the common reference (Ambion), was used. The samples were hybridized overnight with HS400 pro (Tecan, Gr€ odig, Austria) to the preprinted miRCURYTM LNA microRNA arrays, v. 11.0, hsa, mmu, rno arrays (Exiqon), containing optimized probes targeting known human mature miRNAs in mirBase release 11.0 (http://microrna. sanger.ac.uk, accessed October 2009) along with 4168 miRNAs discovered through deep sequencing. After hybridization, the arrays were scanned on
Acta Ophthalmologica 2014
two-colour mode, ratio 635/532, default scaling 5 lm/pixel with the DNA Microarray Scanner (Agilent Technologies, Waldbronn, Germany). Multiplex ligation-dependent amplification
probe
Multiplex ligation-dependent probe amplification was performed with the SALSA MLPA KIT P027-B1 UM (MRC-Holland, Amsterdam, The Netherlands) followed by capillary electrophoresis on a Genetic Analyzer 3130xl (Applied Biosystems, Life Technologies, Carlsbad, CA, USA). Data were quantified using GeneMapper (Applied Biosystems) and subsequently subjected to normalization and statistical analysis using an in-house developed software (Gerdes et al. 2005). In brief, normalized probe signals were used to calculate mean ratios between patient and reference samples for each of the chromosomes 1p, 3, 6 and 8. Chromosomes with mean ratios below 0.87 or above 1.13 and with a corresponding p-value of 0.025 or less (one-tailed t-test) were considered abnormal with significant losses or gains, respectively. Reference samples were either DNA extracted from normal choroidal archival tissue samples or from blood drawn from healthy donors. Bioinformatic analyses
analysis
and
statistical
For the analysis of microRNA expression data, images of hybridized arrays were analysed by GenePix Pro 6.0.1.25 package and converted to text files (gpr.) (Molecular Devices, Sunnyvale, CA, USA), which were imported into R environment (R Development Core Team 2013). LIMMA package (Smyth 2004) was used for importing and preprocessing of the data. The mean pixel intensities were taken to estimate spot intensities (foreground) and the median pixel intensities to estimate background intensity. Spots with quality flags less than 49 were excluded from the analysis. The spot intensities were normalized by quantile normalization implemented in LIMMA package (Smyth 2004), log2 transformed, and three intra-array replicates of each probe were averaged. The probes with a variance of more than 0.5 were used for further analysis. To identify subgroups based on miRNA expression, a principal component
analysis (PCA) and a hierarchical clustering were performed, and a heatmap was generated. To identify miRNAs that were differentially expressed between metastatic and non-metastatic UM, a supervised analysis of differential expression was performed using a moderated t-test as implemented in LIMMA. Additionally, to identify miRNA associated with poor survival, we performed Cox regression analysis for each miRNA as implemented in survival package (Therneau & Grambsch 2000; Therneau 2013). The p-values from differential expression and Cox regression analyses were adjusted for multiple testing by the Benjamini and Hochberg method. The patient with unknown metastatic status was not included in differential expression and survival analyses. MicroRNA clusters identified from heatmap analysis, chromosomal aberrations identified by MLPA, and clinicopathological features were compared with metastasis and survival using Kaplan–Meier analysis and logrank test or log-rank test for trend. The primary end-point for survival was defined as the time from the date of diagnosis to the date of death from all causes or metastatic death. The influence of clinicopathological characteristics and chromosomal aberrations in all-cause mortality was also assessed in Cox regression models (with age as underlying time scale; Korn et al. 1997). These analyses were conducted both unadjusted and adjusted for possible confounding by gender and age dichotomized at 65 years at diagnosis through inclusion of these as variables in multivariable Cox regression models. Statistical analyses were performed using SPSS (SPSS version 20; SPSS Science, Chicago, IL, USA) and SAS (SAS statistical package ver. 9.1, Cary, NC, USA), and p-values 65 years 15 (42) TNM size category T1 (small) 3 (8) T2 (medium) 10 (28) T3 (large) 14 (39) T4 (very large) 9 (25) Ciliary body involvement Present 12 (33) Absent 24 (67) Extraocular extension Present 4 (11) Absent 32 (89) Cell type Spindle 15 (42) Mixed 9 (25) Epithelioid 12 (33) Closed loops Present 23 (64) Absent 13 (36) Mitotic count§ 0–1 13 (36) ≥2 23 (64) TNM stage at diagnosis I 2 (6) IIA 6 (17) IIB 11 (30) IIIA 8 (22) IIIB 1 (3) IV 8 (22)
Survival status
M
NM
p value
Dead
Alive
p value
9 11
6 9
0.87†
10 13
5 8
0.97†
10 10
10 5
0.28†
13 10
8 5
0.56†
1 4 6 9
2 6 7 0
0.002‡
2 5 7 9
1 5 7 0
0.004‡
5 15
6 9
0.62†
8 15
4 9
0.81†
4 16
0 15
0.001†
4 19
0 13
0.003†
7 5 8
8 3 4
0.07‡
7 6 10
8 3 2
0.007‡
14 6
8 7
0.33†
16 7
7 6
0.30†
5 15
8 7
0.08†
6 17
7 6
0.07†
1 2 3 5 1 8
1 4 8 2 0 0
MicroRNA expression analysis and Multiplex ligation-dependent probe amplification in metastatic and non-metastatic uveal melanoma Ann-Cathrine Larsen,1 Line Holst,2 Bogumil Kaczkowski,3 Morten T. Andersen,4 Valentina Manfe,2 Volkert D. Siersma,5 Miriam Kolko,6,7 Jens F. Kiilgaard,7 Ole Winther,3,8 Jan U. Prause,1 Robert Gniadecki2 and Steffen Heegaard1,7 1
Eye Pathology Institute, University of Copenhagen, Copenhagen, Denmark Department of Dermatology, Bispebjerg Hospital, University of Copenhagen, Copenhagen, Denmark 3 Department of Biology, Biotech Research and Innovation Centre, University of Copenhagen, Copenhagen, Denmark 4 Department of Clinical Genetics, Copenhagen University Hospital, Copenhagen, Denmark 5 The Research Unit and Section of General Practice, Institute of Public Health, University of Copenhagen, Copenhagen, Denmark 6 Department of Neuroscience and Pharmacology, University of Copenhagen, Copenhagen, Denmark 7 Department of Ophthalmology, Glostrup Hospital, University of Copenhagen, Copenhagen, Denmark 8 DTU Informatics, Technical University of Denmark, Lyngby, Denmark 2
ABSTRACT. Purpose: To determine the association of microRNA expression and chromosomal changes with metastasis and survival in uveal melanoma (UM). Methods: Thirty-six patients with UM were selected based on the metastatic status, and clinicopathological data were collected. Multiplex ligation-dependent probe amplification (MLPA) was used to identify chromosomal changes. Chromosomal changes and clinicopathological data were correlated with survival and metastasis. The microRNA expression was analysed in 26 of the 36 archived UM samples. Unsupervised analysis, differential expression analysis and Cox regression analysis were performed to determine the association with metastasis and survival. Results: Metastasis and metastatic death occurred in 20 patients, two patients died of other causes and one patient of unknown causes. A significant association between increasing size category (p = 0.002, log-rank), extraocular extension (p = 0.001), chromosome 3 loss (p = 0.033) and 1p loss (p = 0.030) and development of metastases was observed. Tumour, node, metastasis (TNM) staging showed a significant association with survival (p < 0.0001, log-rank). Adjusting for gender and age TNM size category T4 (p = 0.016, Cox regression analysis), mixed (p = 0.029) and epithelioid (p = 0.0058) cell types, chromosome 3 loss (p = 0.014) and 8q gain (p = 0.010) were significant prognosticators for a poor survival. Hierarchical clustering divided the UM into three groups based on microRNA expression. The clusters showed no association with clinical or histopathological features, TNM staging, metastasis or survival. Differential expression analysis did not reveal microRNAs related to metastasis or survival. Conclusions: The prognostic significance of chromosome 3 loss and 8q gain identified by MLPA analysis was confirmed in archived UM samples. The value of microRNA expression as a predictor of metastasis and survival in UM could not be confirmed. Key words: chromosomal aberrations – expression – genetic – metastasis – microRNA – Multiplex ligation-dependent probe amplification – prognostic factors – survival – uveal melanoma
Acta Ophthalmol. 2014: 92: 541–549 ª 2013 Acta Ophthalmologica Scandinavica Foundation. Published by John Wiley & Sons Ltd
doi: 10.1111/aos.12322
Introduction Uveal melanoma (UM) is the most common primary intraocular tumour in adults with an incidence of approximately 0.7 per 100.000 in Denmark (Isager et al. 2005). The course of UM is highly aggressive due to frequent metastases, and the 5-year survival rates range from 66% to 69% in Denmark (Jensen 1982; Isager et al. 2006). Despite advances in surgery, radiotherapy and chemotherapy, the survival rate has not changed. MicroRNAs (miRNA) are small non-coding RNAs that regulate gene expression through inhibition of translation or destabilization of targeted messenger RNAs. Research has proposed that miRNAs function as oncogenes or tumour-suppressor genes in various cancer types (Esquela-Kerscher & Slack 2006). Recent findings support the use of targeted therapy as a novel method to counteract the tumourpromoting functions of miRNAs (Kasinski & Slack 2010). MicroRNA expression analysis has identified expression variations in miRNAs in UM compared with normal tissue (Yang & Wei 2011). Additionally, miRNAs specific for metastasizing melanoma and miRNA acting as
541
Acta Ophthalmologica 2014
oncogenes have been detected (Radhakrishnan et al. 2009). Studies on UM cell lines have revealed miRNA-34a, b, c, miRNA-182, miRNA-137 and miRNA-9 as important tumour suppressors in UM functioning through several target genes (Yan et al. 2009, 2012; Chen et al. 2011; Liu et al. 2012). The investigation of the gene expression profile in UM has led to a classification of patients with UM into two prognostically significant classes (Onken et al. 2004). The utility of miRNAs for predicting metastatic risk has only been assessed in one study which showed a clustering into the same two prognostically significant classes (Worley et al. 2008). Multiplex ligation-dependent probe amplification (MLPA) is a PCR-based technique applied for genomic tumour typing, and a commercial kit is available. Multiplex ligation-dependent probe amplification analysis has been used to predict metastatic death by evaluation of chromosome 3 losses and 8q gains (Damato & Coupland 2009; Damato et al. 2009; Vaarwater et al. 2012). Recently, a study has shown MLPA to be efficient in classifying chromosome 3 loss in formalin-fixed, paraffin-embedded (FFPE) UM samples (Lake et al. 2012). The purpose of the present study was to identify differentially expressed miRNAs in metastatic UM and investigate the association between miRNA expression and survival. For comparison, the prognostic value of chromosomal changes identified by MLPA, and clinicopathological features were evaluated.
Material and Methods Approval for this study was obtained from the Committee on Biomedical Research Ethics (no. 23179), the Danish Data Protection Agency (no. 200941-3749) and the Registry of Human Tissue Utilisation. The study followed the tenets of the Declaration of Helsinki. Tumour samples and clinicopathological data
The pathology reports of patients with UM that had their eye enucleated were retrieved at the Institute of Eye Pathology, University of Copenhagen. Patients were selected between 1986 and 2009, with the aim of achieving a
542
metastatic group and an age- and gender-matched non-metastatic group. Patients treated with radiation therapy, cases of iris melanoma, cases with incomplete patient records, incomplete pathology reports or without available tumour blocks were excluded. Thirteen patients with metastatic disease (preferably present at the time of diagnosis) and 13 age- and gendermatched control patients without metastatic disease were selected for miRNA analysis. These patients and additionally five more patients with metastatic disease (preferably present at the time of diagnosis) and five ageand gender-matched control patients without metastatic disease were added for MLPA analysis. During the study period (2009–2012), one patient without metastatic disease left the country and died, and we were not able to assess any development of metastatic disease or cause of death in this patient. Furthermore, two patients in the non-metastatic group developed metastasis in the study period. This resulted in 15 patients in the metastatic group, seven of these with metastatic disease present at the time of diagnosis and 10 patients in the nonmetastatic group for inclusion in the miRNA analysis. For the DNA analysis, this resulted in the inclusion of 20 patients in the metastatic group, eight of these with metastasis at the time of diagnosis and 15 patients in the non-metastatic group. The patient with unknown metastatic status was included in analysis of all-cause mortality. Clinical files, pathology reports and histopathological samples were reviewed to collect the following: Age at diagnosis, gender, cell type, tumour height and thickness and extraocular growth. Presence or absence of closed extravascular matrix patterns (closed loops) was evaluated from H&E sections. The number of mitoses were evaluated in H&E sections in 40 highpower fields (HPF) using the 409 objective and characterized as low mitotic activity (0–1 per 40 HPF) and medium to high mitotic activity (≥2 per 40 HPF) as described by McLean et al. (1977). The presence of metastatic disease was identified using clinical patient records, abdominal ultrasound examination and chest X-ray records and searching the Danish Registry of Pathology. Tumour size categories were established based on thickness and
diameter, and the Tumour, Node, Metastasis (TNM)-based Staging System 7th edition was applied (Edge et al. 2010; Kujala et al. 2013). Follow-up time was either defined as the time from diagnosis to the date of death or from the time of diagnosis to the last followup date (17th August 2012). MicroRNA and DNA extraction
For DNA and miRNA extraction, four sections each of 20 lm thickness were cut from the FFPE eyes on a RM2255 microtome (Leica, Bannockburn, IL, USA). Sections were transferred to glass slides, and tumour tissue was isolated by hand with a scalpel. The tissue was melted in 1 ml 100% xylene (Ambion, Foster City, CA, USA) at 50°C for 3 min. The protocol of RecoverAll Total Nucleic Acid Isolation Kit (Ambion) was followed according to manufacturer’s instructions. MicroRNA array
The concentration of total RNA was measured spectrophotometrically using a NanoDrop ND-1000 (Thermo Scientific, Wilmington, DE, USA). Extraction resulted in a total yield in the range of 20–757 ng/ll. The OD A260/A280 ratio was in the range of 1.50–2.79 indicating high sample purity. The quality of total RNA was measured with an Agilent 2100 Bioanalyzer using the Agilent Nano RNA 6000 kit (Agilent Technologies, Santa Clara, CA, USA). The concentration of total RNA was adjusted to 500 ng/ll, and labelling was performed with the miRCURYTM LNA microRNA Hy3/Hy5 Power labelling kit (Exiqon, Vedbæk, Denmark) according to the manufacturer’s instructions. A dual colour system, in which the samples were labelled using fluorescent dyes: Hy-3 (532 nm) for the tumour samples and Hy-5 (635 nm) for the common reference (Ambion), was used. The samples were hybridized overnight with HS400 pro (Tecan, Gr€ odig, Austria) to the preprinted miRCURYTM LNA microRNA arrays, v. 11.0, hsa, mmu, rno arrays (Exiqon), containing optimized probes targeting known human mature miRNAs in mirBase release 11.0 (http://microrna. sanger.ac.uk, accessed October 2009) along with 4168 miRNAs discovered through deep sequencing. After hybridization, the arrays were scanned on
Acta Ophthalmologica 2014
two-colour mode, ratio 635/532, default scaling 5 lm/pixel with the DNA Microarray Scanner (Agilent Technologies, Waldbronn, Germany). Multiplex ligation-dependent amplification
probe
Multiplex ligation-dependent probe amplification was performed with the SALSA MLPA KIT P027-B1 UM (MRC-Holland, Amsterdam, The Netherlands) followed by capillary electrophoresis on a Genetic Analyzer 3130xl (Applied Biosystems, Life Technologies, Carlsbad, CA, USA). Data were quantified using GeneMapper (Applied Biosystems) and subsequently subjected to normalization and statistical analysis using an in-house developed software (Gerdes et al. 2005). In brief, normalized probe signals were used to calculate mean ratios between patient and reference samples for each of the chromosomes 1p, 3, 6 and 8. Chromosomes with mean ratios below 0.87 or above 1.13 and with a corresponding p-value of 0.025 or less (one-tailed t-test) were considered abnormal with significant losses or gains, respectively. Reference samples were either DNA extracted from normal choroidal archival tissue samples or from blood drawn from healthy donors. Bioinformatic analyses
analysis
and
statistical
For the analysis of microRNA expression data, images of hybridized arrays were analysed by GenePix Pro 6.0.1.25 package and converted to text files (gpr.) (Molecular Devices, Sunnyvale, CA, USA), which were imported into R environment (R Development Core Team 2013). LIMMA package (Smyth 2004) was used for importing and preprocessing of the data. The mean pixel intensities were taken to estimate spot intensities (foreground) and the median pixel intensities to estimate background intensity. Spots with quality flags less than 49 were excluded from the analysis. The spot intensities were normalized by quantile normalization implemented in LIMMA package (Smyth 2004), log2 transformed, and three intra-array replicates of each probe were averaged. The probes with a variance of more than 0.5 were used for further analysis. To identify subgroups based on miRNA expression, a principal component
analysis (PCA) and a hierarchical clustering were performed, and a heatmap was generated. To identify miRNAs that were differentially expressed between metastatic and non-metastatic UM, a supervised analysis of differential expression was performed using a moderated t-test as implemented in LIMMA. Additionally, to identify miRNA associated with poor survival, we performed Cox regression analysis for each miRNA as implemented in survival package (Therneau & Grambsch 2000; Therneau 2013). The p-values from differential expression and Cox regression analyses were adjusted for multiple testing by the Benjamini and Hochberg method. The patient with unknown metastatic status was not included in differential expression and survival analyses. MicroRNA clusters identified from heatmap analysis, chromosomal aberrations identified by MLPA, and clinicopathological features were compared with metastasis and survival using Kaplan–Meier analysis and logrank test or log-rank test for trend. The primary end-point for survival was defined as the time from the date of diagnosis to the date of death from all causes or metastatic death. The influence of clinicopathological characteristics and chromosomal aberrations in all-cause mortality was also assessed in Cox regression models (with age as underlying time scale; Korn et al. 1997). These analyses were conducted both unadjusted and adjusted for possible confounding by gender and age dichotomized at 65 years at diagnosis through inclusion of these as variables in multivariable Cox regression models. Statistical analyses were performed using SPSS (SPSS version 20; SPSS Science, Chicago, IL, USA) and SAS (SAS statistical package ver. 9.1, Cary, NC, USA), and p-values 65 years 15 (42) TNM size category T1 (small) 3 (8) T2 (medium) 10 (28) T3 (large) 14 (39) T4 (very large) 9 (25) Ciliary body involvement Present 12 (33) Absent 24 (67) Extraocular extension Present 4 (11) Absent 32 (89) Cell type Spindle 15 (42) Mixed 9 (25) Epithelioid 12 (33) Closed loops Present 23 (64) Absent 13 (36) Mitotic count§ 0–1 13 (36) ≥2 23 (64) TNM stage at diagnosis I 2 (6) IIA 6 (17) IIB 11 (30) IIIA 8 (22) IIIB 1 (3) IV 8 (22)
Survival status
M
NM
p value
Dead
Alive
p value
9 11
6 9
0.87†
10 13
5 8
0.97†
10 10
10 5
0.28†
13 10
8 5
0.56†
1 4 6 9
2 6 7 0
0.002‡
2 5 7 9
1 5 7 0
0.004‡
5 15
6 9
0.62†
8 15
4 9
0.81†
4 16
0 15
0.001†
4 19
0 13
0.003†
7 5 8
8 3 4
0.07‡
7 6 10
8 3 2
0.007‡
14 6
8 7
0.33†
16 7
7 6
0.30†
5 15
8 7
0.08†
6 17
7 6
0.07†
1 2 3 5 1 8
1 4 8 2 0 0
