Abstract
Purpose
Gene expression studies in melanoma have been few because tumors are small and cryopreservation is rarely possible. The purpose of this study was to evaluate the Illumina DASL Array Human Cancer Panel for gene expression studies in formalin-fixed melanoma primary tumors, and to identify prognostic biomarkers.
Experimental Design
Primary tumors from two studies were sampled using a tissue microarray needle. Study 1: 254 tumors from a melanoma cohort recruited 2000-2006. Study 2: 218 tumors from a case-control study of patients undergoing sentinel node biopsy.
Results
RNA was obtained from 76% of blocks. 1.4% of samples failed analysis (transcripts from less than 250 of the 502 genes on the DASL chip detected). Increasing age of the block and increased melanin in the tumor were associated with reduced number of genes detected. The gene whose expression was most differentially expressed in association with relapse free survival in study 1 was osteopontin (SPP1, p=2.11×10−6) and supportive evidence for this was obtained in study 2 used as a validation set (p=0.006) (unadjusted data). Osteopontin level in study 1 remained a significant predictor of relapse free survival when data were adjusted for age, sex, tumor site and histological predictors of relapse. Genes whose expression correlated most strongly with osteopontin were PBX1, BIRC5 (Survivin) and HLF.
Conclusion
Expression data were obtained from 74% of primary melanomas and provided confirmatory evidence that osteopontin expression is a prognostic biomarker. These results suggest that predictive biomarker studies may be possible using stored blocks from mature clinical trials.
Keywords: formalin fixed tissue, prognosis, biomarker
Background
The established predictors of outcome for melanoma patients relate to the histological characteristics of the primary tumor (Breslow thickness, the presence of ulceration (1), mitotic rate (2)), tumor site, sex (2) and age (3). The histological characteristics are used to estimate prognosis as part of the AJCC staging system (1) and in various algorithms, to give a more personalized estimate (2, 3), but much of the variance in survival remains un-explained. In order to identify prognostic and predictive biomarkers, and to better establish the biological pathways of relevance, genomic studies of primary melanomas are necessary.
However, primary melanomas are small and as the histological characteristics conveying prognostic information are often focal within the primary tumor, pathologists are reluctant to cryopreserve tissue. Therefore, genomic expression studies have been relatively few in number and limited in size. Studies published to date have predominantly used polymerase chain reaction (PCR) techniques to look at single genes (4, 5), often using less than 25 samples. Larger scale studies have more recently come from Winnepennincx et al. who identified a 254-gene signature predictive of survival which included minichromosome maintenance genes, in 83 cryopreserved primary tumors (6), and Kauffman et al. who identified increased expression of DNA repair genes in metastatic tumors from 60 tumor samples (7).
Whilst the work by Winnepennincx, Kauffman and colleagues (6) represents a major development in the field, the development of approaches to analyze formalin-fixed tumors would allow studies using large numbers of samples stored from melanoma cohorts/clinical trials with long follow up, and potentially with less bias consequent upon selected sampling of tumors deemed suitable for cryopreservation. We have used Illumina's DASL (cDNA mediated annealing, selection, extension and ligation) assay (8), which has been specifically developed for use in formalin fixed tissue, to investigate prognostic biomarkers in stored primary melanomas, and we report here an evaluation of the technique and confirmation of the significance of increased osteopontin expression in melanoma. In this large study, we also investigated the value of quality control measures and sample characteristics to predict performance of RNA samples with the DASL assay.
Methods
Patient Samples
Formalin-fixed paraffin-embedded primary melanoma blocks were identified from two study sets (Table 1). In Study 1, population-ascertained incident melanoma cases were recruited to a case-control study in a geographically defined area of Yorkshire and the Northern region of the UK (67% participation rate). All patients gave written informed consent to participation. 961 male and female patients (aged between 18 and 76 years) were diagnosed in the period from September 2000 to December 2005 (9, 10). The cases were identified via clinicians, pathology registers and the cancer registry to ensure maximal ascertainment. Between September 2000 and December 2001, and from July 2003 till December 2005, patients with Breslow thickness less than 0.75mm were not invited in order to maximize the value of the sample as a cohort looking at prognostic outcomes. Between January 2002 and June 2003 all patients with invasive melanoma were invited to participate. The first 254 blocks identified from participants within the cohort with tumors thicker than 0.75mm with the longest follow up comprised the test set. In Study 2 (the validation set), patients with melanomas Breslow thickness ≥ 0.75mm having undergone sentinel node biopsy (SNB) were recruited to a multi-centre retrospective case-control study. Five centers from the U.K identified all patients having had SNB November 1994 till 2006. Cases were melanoma patients with a positive SNB and controls were those with a negative SNB. The number of patients with a negative SNB, being greater than the number with a positive result controls were randomly selected to be frequency matched by year of SNB and by the centre at which the SNB was performed. The first 218 blocks identified from participants with the longest follow-up in a study designed to identify predictors of sentinel node positivity and relapse were sampled. In both studies the blocks sampled were representative of the age, sex ratio and site distribution of the whole data sets. The Breslow thickness was higher in the Study 1 samples, as expected, as we were unable to sample very small tumors (table 3s). Both studies were approved by both of the UK national ethics committees (MREC and PIAG). Here we report expression studies carried out in samples from Study 1, which we subsequently sought to validate in samples from Study 2.
Table 1.
Descriptive characteristics of the two sample sets.
| Variable | Study 1 | Study 2 | Test statistic and significance value |
|---|---|---|---|
| Number of patients | 156 | 198 | |
|
Age at diagnosis or at SNB, years (mean and range) |
54.9 (19.9-78.5) | 52.0 (14.4-88.0) | t value 1.87 p-value = 0.06 |
| Sex - female (number and %) | 81 (51.9) | 95 (48.0) | Chi squared 0.54, p-value = 0.46 |
| Sex – male (number and %) | 75 (48.1) | 103 (52.0) | |
|
Site of tumor, number (%) Arm Head-neck Leg Trunk Unknown |
31 (19.9) 24 (15.4) 50 (32.0) 49 (31.4) 2 (1.3) |
44 (22.2) 17 (8.6) 69 (34.9) 68 (34.3) 0 (0.0) |
Chi squared 4.13, p-value = 0.25 |
|
Breslow thickness, mm, median (range) |
1.9 (0.9-12.0) | 2.0 (0.78-24.0) | Mann-Whitney Z: −0.57 p-value = 0.57 |
|
Mitotic rate, number (%) <1 1-6 >6 Unknown |
27 (17.3) 76 (48.7) 34 (21.8) 19 (12.2) |
22 (11.1) 90 (45.5) 68 (34.3) 18 (9.1) |
Chi squared 7.33 p-value = 0.03 |
|
Ulcerated tumors, number (%) |
40 (25.6) | 58 (29.3) | Chi squared 0.58 p-value = 0.45 |
|
Relapsers number (%) |
37 (23.7) | 63 (31.8) | Chi squared 3.48 p-value = 0.06 |
|
Deaths number (%) |
21 (13.5) | 47 (23.7) | Chi-squared 6.17 p-value = 0.01 |
|
Follow-up time, months, median (range) |
49.1 (4.93-94.9) | 38.4 (0.03-111.7) | Mann-Whitney Z: 4.65 p-value <0.001 |
Sample Preparation
Primary tumor blocks were identified and a haematoxylin and eosin (H&E) stained slide was examined to identify the deepest part of the tumor having a diameter of greater than 0.8mm, containing the lowest admixture of inflammatory or stromal cells. This area was marked using a fine tipped permanent marker and a tissue microarray (TMA) needle was then used to sample the tumor block horizontally (supplementary information). TMA core needles were used to sample tumors as an efficient approach to obtain sufficient RNA yields from the deepest part of the tumor, in studies potentially using many hundreds of samples whilst preserving the block for use by pathologists subsequently.
Melanin present in primary melanomas co-purifies with DNA resulting in two major problems. Absorption of UV light, can lead to unreliable spectrophotometric quantification of nucleic acids (11). More importantly, melanin can inhibit DNA polymerases (12). To allow evaluation of the effect of melanin on gene expression analysis of melanoma samples, slides were graded using a system devised to visually score melanin content of TMA cores (0 to 3).
RNA extraction
Tissue cores were de-waxed using xylene and two changes of absolute ethanol. RNA was extracted in batches of 24 tissue cores using the High Pure paraffin RNA kit (Roche Diagnostics Ltd, Burges Hill, UK) according to the manufacturer's protocol and eluted in 25μl nuclease free water. For quality control measures see supplementary information.
DASL expression arrays
The Illumina DASL Human Cancer Panel gene set (Illumina Inc., CA, USA) was used to perform the DASL assay for gene expression profiling of all test and control samples. The Cancer Panel includes 1536 unique sequence specific probes targeting 502 genes. Each gene is targeted in three locations by three separate probe pairs designed by a proprietary algorithm (8) (supplementary information).
Data pre-processing
The data were normalized using Beadstudio software (Illumina, USA) before exporting to STATA version 10 for statistical analyses. The normalisation methods used were background correction, cubic spline smoothing (13) and plate scaling. Normalisation was conducted relative to a synthetic reference array, which was created in each study by averaging all melanoma samples (supplementary information).
Target validation by quantitative real time reverse transcription polymerase chain reaction
Osteopontin (SPP1) expression, identified as significantly associated with relapse or histological variables that predict relapse was further investigated by quantitative real time reverse transcription polymerase chain reaction (qRT-PCR) on samples from Study 1 using probes corresponding to the locations of the DASL probes (Taqman® Gene Expression Assays Hs00960942_m1 (Exons 1-2) and Hs00959010_m1 (Exons 5-6), Applied Biosystems, Warrington, UK) (supplementary information). The comparative Ct (or deltaCt) method was used to compare relative fold change in expression of 2 regions of SPP1.
Statistical methodology
The number of genes detected in each sample (probe signal significantly greater than average signal from negative controls with p<0.05), was used as a measure of the quality of the results. The influence of age of tissue block and melanin level of the tumor on number of genes detected was investigated using Spearman's rank correlation and Kruskal-Wallis tests respectively.
Methods used to measure quality and quantity of RNA prior to use in the DASL assay were assessed by correlating the number of genes detected in samples with the quality measure data using either Spearman's rank correlation or the Kruskal-Wallis test. Samples with less than 250 detected genes were classified as failed and excluded from further analysis (Table 1s). Analysis of sample replicates is detailed in the supplementary information (Table 2s). Mean gene expression was used for the remaining sample replicates.
Differential gene expression and survival analyses were performed using log-transformed normalized data (log2). Within the sample sets, mean expression of each gene was compared between samples with histological features of interest using two-sample t-tests and linear regression. Relapse free survival was defined as the period between diagnosis and date of first relapse at any site. Survival analysis was performed using Cox proportional hazards model to calculate hazard ratios and 95% confidence intervals for each gene. These analyses were performed unadjusted and adjusted for demographic and histological factors of prognostic importance in melanoma. Significance values were ranked to identify genes most differentially expressed between groups of interest. Using the Bonferroni method to correct for multiple testing, the significance level was set at 0.0001 for these analyses. All analyses were undertaken using Stata version 10 (StataCorp 2007, College Station, TX).
Generation of gene networks using Ingenuity software
The combined data from the studies, fold changes and significance levels of genes differentially expressed in tumors from patients with reduced relapse free survival time were analyzed using Ingenuity Pathway Analysis software (Ingenuity Systems, Redwood City, California). Genes >1.2 times over- or under-expressed with significance levels <0.05 were interrogated by Ingenuity to find genes most related to each other and a network of these relationships was generated.
Results
Descriptive statistics on the sample sets
The two samples sets were similar (Table 1). Participants in Study 2 were slightly younger. Study 2 tumors had a significantly higher mitotic rate (Pearson's chi squared, p=0.03) and a correspondingly higher relapse rate (23.7% (Study 1) and 31.8% (Study 2) respectively, Pearson's chi squared, p=0.06) and death rate (13.5% (Study 1) and 23.7% (Study 2) respectively, Pearson's chi squared, p=0.01). The sample sets were broadly representative of the larger study sets from which the samples were derived (Table 3s). Use of a TMA needle precluded sampling of thin tumors which accounts for the higher proportion of thicker tumors in the larger study set 1.
RNA yields obtained from tumor samples
Of the 472 formalin-fixed, paraffin-embedded primary tumor blocks identified, 378 (80%) were selected for sampling. Reasons for not sampling a block included too little residual tumor after sectioning for clinical purposes or other research projects, and tumor cells being mixed with large numbers of normal stromal or inflammatory cells. In 17/378 (4.5%) a core was taken but inadequate quantities of RNA were obtained as measured using the Bioanalyser. Overall, adequate RNA yields were obtained from 361/472 (76%) of blocks.
Quality control measures for DASL
Four hundred and twenty three RNA samples including replicate samples were supplied to the Illumina DASL service provider. Less than 250 genes were detected in 6 (1.4%) samples, which were classified as failed samples. The failure rate was 2.1% in Study 1 and 0.9% in Study 2 (Table 1s).
Table 1s (supplementary information) summarizes the associations between quality control measures, age of block and melanin content of tumor and the number of genes detected. Increased block age was predictive of a reduced number of genes detected in Study 2 but not Study 1 (the range of block age being much greater in Study 2). Increased melanin score was predictive of reduced number of detected genes in Study 1. The best quality control predictors of number of genes detected were RNA concentration, RNA integrity (RIN) score (14), and the CT value from qRT-PCR.
Expression data
Study 1
Genes most differentially expressed in tumors from patients with reduced relapse free survival are presented in Table 2. The gene most predictive of relapse free survival in Study 1 was osteopontin (increased expression was associated with shorter relapse free survival). In Study 1 the hazard ratio for reduced relapse free survival associated with increased expression was 3.17 (unadjusted), significant at the p=2.11×10−6 level. This association persisted when the analysis was repeated adjusted for host variables known to predict relapse (age, sex and tumor site, p=9.19×10−6), and when adjusted additionally for Breslow thickness, mitotic rate and the presence of tumor ulceration (p=0.001) (Table 3). Fold change of expression signal was 1.55 between relapsers and non-relapsers in unadjusted analysis. Increased osteopontin expression was also predictive of overall survival (OS) (hazard ratio 2.63 (95% CI 1.38-5.04), p=0.002) with a fold change of 1.41. This association remains significant after adjusting for age, sex and tumor site (p=0.004). Expression signals from all three osteopontin probes on the array were comparably predictive of reduced relapse free survival in study 1 (hazard ratio range 2.10-2.84, significance value range 0.0002-9.91×10−6).
Table 2.
Top 20 genes differentially expressed in tumors from patients with reduced relapse-free survival in Study 1 (unadjusted analysis).
| Gene | Fold change between relapsers and non-relapsers |
Hazard ratio |
95% confidence interval |
Significance value |
|---|---|---|---|---|
| OSTEOPONTIN | 1.55 | 3.17 | 1.91-5.26 | 2.11 × 10−6 |
| RAD54B | 1.39 | 8.59 | 3.17-23.31 | 9.40 × 10−6 |
| HMMR | 1.43 | 4.36 | 2.08-9.13 | 0.00006 |
| CDKN2B | 0.71 | 0.33 | 0.19-0.56 | 0.0001 |
| DEK | 1.15 | 16.94 | 4.32-66.45 | 0.0001 |
| TK1 | 1.23 | 7.82 | 2.70-22.65 | 0.0002 |
| ITGB4 | 0.74 | 0.35 | 0.20-0.60 | 0.0002 |
| BIRC5 | 1.30 | 4.36 | 1.82-10.41 | 0.0003 |
| DSP | 0.65 | 0.60 | 0.46-0.78 | 0.0003 |
| ING1 | 1.13 | 12.81 | 3.07-53.47 | 0.0003 |
| TOP2A | 1.19 | 6.86 | 2.19-21.49 | 0.0003 |
| AR | 0.63 | 0.25 | 0.11-0.55 | 0.0003 |
| E2F5 | 1.34 | 3.86 | 1.82-8.19 | 0.0004 |
| TGFA | 0.66 | 0.32 | 0.16-0.63 | 0.0004 |
| RECQL | 1.18 | 8.64 | 2.58-29.00 | 0.0004 |
| BLM | 1.24 | 4.35 | 1.86-10.14 | 0.0005 |
| FHIT | 1.26 | 3.99 | 1.75-9.10 | 0.0006 |
| GRB7 | 0.66 | 0.53 | 0.37-0.76 | 0.0006 |
| TFAP2C | 0.75 | 0.36 | 0.20-0.64 | 0.0006 |
| RAD51 | 1.30 | 4.16 | 1.81-9.53 | 0.0007 |
Table 3. The association of osteopontin expression with relapse free survival in Studies 1 and 2.
The raw data are presented in column 1. The association was adjusted for sex, patient age and tumor site (as known predictors of outcome) in column 2 (mean signal values are presented for a 45 year old female with a tumor on her leg). In column 3, further adjustment is made for known histological predictors of outcome: Breslow thickness, mitotic rate and ulceration (mean signal values are presented for a 45 year old female patient with a non-ulcerated tumor on her leg which has a Breslow thickness of 2.5mm and a mitotic rate of 1-6/mm2). Study 2 analyses are similarly presented with mean signal values adjusted for sentinel node biopsy status.
| Study 1 raw data |
Study 1 adjusted for age, sex and site of tumor |
Study 1 adjusted additionally for histological measures |
Study 2 study raw data |
Study 2 adjusted for age, sex and site of tumor |
Study 2 adjusted for age, sex, site of tumor and SNB status |
Study 2 adjusted additionally for histological measures |
Study 2 adjusted additionally for histological measures and SNB status |
|
|---|---|---|---|---|---|---|---|---|
|
Mean signal (SD) relapsers |
5396 (2503) |
5031 (2385) |
4552 (2074) | 5812 (2752) |
6130 (2701) |
4103 (2619) |
5907 (2537) | 4575 (2485) |
|
Mean signal (SD) non- relapsers |
3476 (2255) |
3060 (2248) |
3180 (2127) | 4411 (2818) |
4533 (2775) |
3227 (2622) |
4988 (2460) | 4051 (2389) |
| Fold change | 1.55 | 1.64 | 1.43 | 1.32 | 1.35 | 1.27 | 1.18 | 1.13 |
|
Hazard ratio (95% CI) |
3.17 (1.91- 5.26) |
3.33 (1.96- 5.67) |
2.76 (1.49-5.10) |
1.60 (1.13- 2.27) |
1.67 (1.16- 2.40) |
1.40 (0.97- 2.03) |
1.24 (0.81-1.90) |
1.11 (0.73-1.69) |
|
Significance value |
2.11 × 10−6 |
9.19 × 10−6 |
0.001 | 0.006 | 0.006 | 0.07 | 0.32 | 0.62 |
Study 2
In these samples, increased osteopontin expression was also associated with reduced relapse free survival at the p=0.006 level in unadjusted analyses, with a similar fold change of 1.32 (Table 3). When corrected for age, sex and tumor site the significance of the association was p=0.006. Increased osteopontin expression further more was associated with poorer OS (hazard ratio 1.6 (95%CI 1.1-2.5), p= 0.02) in unadjusted analysis. The fold change between survivors and non-survivors was 1.3. Osteopontin remained associated with OS after adjusting for age, sex and tumour site (p=0.02).
qRT-PCR
qRT-PCR with probes to exons 1/2 and 5/6 showed that increased expression of osteopontin with fold changes of 1.74 and 1.67 respectively was associated with reduced relapse-free survival in Study 1 (compared with a fold change of 1.55 in the DASL analysis).
Co-expression of genes with osteopontin
The expression of genes most closely correlated with osteopontin expression was studied in the pooled data set for both studies (analysis adjusted for study) and the results are presented in Table 4. We have listed 32 genes whose expression was significantly correlated (either positively or negatively) with that of osteopontin at the 1.0 × 10−5 significance level or less, and have correlated this further with relapse status. Genes whose upregulation was associated with osteopontin upregulation and with reduced relapse-free survival were BIRC5, IL-8, TK1, HMMR, TOP2A, CCNA2, CDC2, RAD51, NQO1, PTPRH and MAPK10. We also present a gene network for osteopontin derived using Ingenuity Pathway Analysis (Figure 1 and Table 5). The literature-derived Ingenuity knowledge base identified osteopontin as involved in cell adhesion, cell proliferation and cell migration. The network demonstrates that osteopontin is the terminal component of many pathways and therefore overexpression of osteopontin may reflect combined activity in many of these pathways.
Table 4. Gene expression correlations for osteopontin.
Fold changes of gene expression between relapsers and non-relapsers are also presented.
| Gene | Correlation and p value for pooled data set |
Fold change for gene expression between relapsers and non- relapsers |
Significance level for fold change |
Correlation and p value for Study 1 |
Correlation and p value for Study 2 |
|---|---|---|---|---|---|
| PBX1 | −0.34 (3.1 × 10−11) |
0.93 | 0.07 | −0.36 (2.9 × 10−6) |
−0.32 (5.5 × 10−6) |
| BIRC5 | 0.33 (2.0 × 10−10) |
1.24 | 8.81 × 10−6 | 0.40 (2.7 × 10−7) |
0.25 (0.0004) |
| HLF | −0.32 (5.5 × 10−10) |
0.81 | 0.0002 | −0.35 (7.3 × 10−6) |
−0.28 (0.00007) |
| IL8 | 0.31 (1.4 × 10−9) |
1.26 | 0.04 | 0.41 (9.7 × 10−8) |
0.22 (0.003) |
| HMMR | 0.30 (5.8 × 10−9) |
1.20 | 0.004 | 0.33 (0.00003) |
0.27 (0.00009) |
| TOP2A | 0.29 (1.7 × 10−8) |
1.17 | 0.00001 | 0.30 (0.0001) |
0.28 (0.00007) |
| TK1 | 0.29 (2.0 × 10−8 |
1.17 | 9.91 × 10−7 | 0.28 (0.0004) |
0.30 (0.00002) |
| CTSL | 0.29 (2.8 × 10−8) |
1.06 | 0.35 | 0.34 (0.00001) |
0.24 (0.0006) |
| CCNA2 | 0.28 (5.6 × 10−8) |
1.21 | 0.00003 | 0.33 (0.00003) |
0.24 (0.0006) |
| BCL6 | −0.28 (8.3 × 10−8) |
0.92 | 0.0007 | −0.33 (0.00003) |
−0.24 (0.0007) |
| CDC2 | 0.27 (2.2 × 10−7) |
1.29 | 0.0001 | 0.31 (0.00008) |
0.24 (0.0008) |
| RAD51 | 0.27 (3.3 × 10−7) |
1.28 | 1.03 × 10−6 | 0.29 (0.0002) |
0.23 (0.001) |
| ERCC5 | −0.26 (5.1 × 10−7) |
0.95 | 0.08 | −0.17 (0.03) | −0.31 (9.2 × 10−6) |
| NQO1 | 0.26 (5.8 × 10−7) |
1.14 | 0.02 | 0.28 (0.0005) |
0.26 (0.0002) |
| CBFA2T1 | −0.26 (7.6 × 10−7) |
0.93 | 0.22 | −0.28 (0.0003) |
−0.24 (0.0007) |
| MMP1 | 0.26 (1.1 × 10−6) |
1.11 | 0.61 | 0.44 (1.2 × 10−8) |
0.13 (0.06) |
| PTPRH | 0.26 (1.1 × 10−6) |
1.24 | 0.04 | 0.27 (0.0006) |
0.24 (0.0006) |
| FGFR2 | −0.25 (1.2 × 10−6) |
0.89 | 0.05 | −0.33 (0.00002) |
−0.17 (0.02) |
| EGFR | −0.25 (1.2 × 10−6) |
0.83 | 0.01 | −0.27 (0.0007) |
−0.23 (0.0009) |
| TIMP1 | 0.25 (1.5 × 10−6) |
1.06 | 0.33 | 0.18 (0.02) | 0.29 (0.00003) |
| GAS1 | −0.25 (1.6 × 10−6) |
0.94 | 0.02 | −0.27 (0.0006) |
−0.23 (0.001) |
| FLT3 | −0.25 (1.8 × 10−6) |
0.92 | 0.15 | −0.24 (0.003) |
−0.26 (0.0002) |
| RBL2 | −0.25 (2.1 × 10−6) |
0.99 | 0.53 | −0.27 (0.0007) |
−0.25 (0.0004) |
| ETS2 | −0.25 (2.5 × 10−6) |
0.91 | 0.003 | −0.27 (0.0006) |
−0.20 (0.005) |
| NUMA1 | −0.25 (2.9 × 10−6) |
0.97 | 0.06 | −0.31 (0.00009) |
−0.18 (0.01) |
| EPHA1 | −0.24 (4.3 × 10−6) |
0.86 | 0.03 | −0.23 (0.003) |
−0.21 (0.002) |
| MAP3K8 | −0.24 (4.7 × 10−6) |
0.93 | 0.06 | −0.30 (0.0002) |
−0.20 (0.005) |
| VEGF | 0.24 (5.0 × 10−6) |
1.02 | 0.50 | 0.31 (0.0001) |
0.20 (0.004) |
| CCND3 | −0.24 (7.5 × 10−6) |
0.97 | 0.04 | −0.27 (0.0006) |
−0.21 (0.003) |
| AR | −0.23 (9.4 × 10−6) |
0.83 | 0.008 | −0.43 (2.66 × 10−8) |
−0.15 (0.04) |
| FGFR3 | −0.23 (0.00001) |
0.91 | 0.1 | −0.24 (0.003) |
−0.20 (0.006) |
| MAPK10 | 0.23 (0.00001) |
1.31 | 0.01 | 0.19 (0.02) | 0.29 (0.00003) |
Figure 1.
Gene network involving SPP1 from Ingenuity in pooled data: cell growth, proliferation and death (showing only direct interactions except those involving SPP1). Red denotes upregulation, green denotes downregulation
Table 5.
Correlations and differences in gene expression of genes identified in the Ingenuity network as being linked to osteopontin
| Gene | Correlation and p value for pooled data set |
Fold change gene expression between relapsers and non- relapsers |
Significance level for fold change |
Correlation and p value for Study 1 |
Correlation and p value for Study 2 |
|---|---|---|---|---|---|
| IL8 | 0.31 (1.4 × 10−9) | 1.26 | 0.04 | 0.41 (9.7 × 10−8) | 0.22 (0.003) |
| CDC25C | 0.21 (0.00009) | 1.28 | 0.005 | 0.18 (0.03) | 0.17 (0.02) |
| TERT | 0.13 (0.02) | 1.40 | 0.0005 | 0.19 (0.02) | 0.09 (0.23) |
| RARB | 0.11 (0.04) | 1.12 | 0.80 | 0.07 (0.35) | 0.09 (0.20) |
| IL3 | 0.06 (0.25) | 1.39 | 0.009 | −0.06 (0.42) | 0.15 (0.04) |
| IL6 | 0.05 (0.37) | 1.27 | 0.46 | 0.09 (0.26) | −0.01 (0.84) |
| E2F5 | 0.06 (0.25) | 1.26 | 0.004 | 0.05 (0.50) | 0.02 (0.77) |
| MCF2 | 0.02 (0.70) | 1.45 | 0.03 | −0.01 (0.86) | −0.02 (0.79) |
| CDKN2B | −0.00 (0.93) | 0.86 | 0.004 | −0.10 (0.21) | 0.05 (0.52) |
Discussion
Fortunately, melanoma has a good prognosis in the majority of patients, but advanced disease is extremely difficult to treat. Most chemotherapeutic regimes in use have response rates of 12 to 15% (15) but unfortunately there are no biomarkers in clinical use to identify patients likely to benefit. Poor progress in the development of biomarkers has been at least in part a result of the fact that primary melanomas are physically small, and pathologists are reluctant to cryopreserve tumor. Therefore we have explored the possibility of using the Illumina DASL assay to produce gene expression profiles from formalin-fixed tumor tissue. The strengths of this study are that it represents much the largest study of gene expression in primary melanoma and benefits from a test sample set and a validation set.
Concerns about the use of fixed tumors derive from the degradation of RNA which results from delayed time to fixation (16) and time in formalin (17). Increasingly however it has been suggested that technical modification can allow profiling of gene expression and micro-RNA (18-21). The Illumina DASL assay has been shown in other studies to produce comparable results for formalin fixed and fresh or frozen cells (22, 23). In this paper we report the generation of expression data from 74% of consecutive formalin-fixed melanomas. The paucity of cryopreserved tissue is such that researchers have not previously been able to generate expression data from other than a highly selected proportion of melanoma samples. We have not in this study carried out extensive comparisons between results from frozen or fresh and formalin fixed tumors as previous studies have addressed this (8, 23, 24). Our data suggest that future studies designed to identify predictive or prognostic biomarkers for melanoma would generate results from approximately 75% of samples depending on the age of the blocks, and the proportion that were deeply pigmented. We would anticipate that in predictive biomarker studies that we would achieve results from a higher proportion, as patients undergoing chemotherapy are likely to have larger primary tumors than a significant proportion of the tumors sampled in these studies.
The limitations of this study are related to the presence of a limited number of genes on the DASL Cancer Panel, and to sampling of tumors using TMA cores. Using a TMA core does not allow confirmation of the tumor content throughout the core and therefore there is potentially greater contamination with normal cells than in laser microdissected samples. Furthermore, by using this technique we were unable to sample very small tumors and so there is a bias towards sampling of larger tumors. The use of this technology however, has allowed a far greater range of tumors to be examined than in previous research based upon cryopreserved tumors. The use of microdissection would address some of these concerns but would be very much more time consuming for large-scale studies.
Osteopontin was identified as the gene whose increased expression was most strongly associated with reduced relapse-free survival and the validity of this finding was tested by comparison of test and validation sample sets. Quantitative real-time RT-PCR using probes to the same gene detected similar fold changes associated with relapse. We did not go on to confirm the findings using immunohistochemistry because a large study recently reported that osteopontin staining predicts sentinel node positivity and relapse in melanoma (25). Two small previous studies using cryopreserved tumors also showed a correlation between osteopontin expression and progression in melanoma (26, 27).
Osteopontin is a glycophosphoprotein cytokine with pleotropic effects. In normal tissues it plays a role in inflammation, vascular and bone remodellling and in wound repair. It also has a role in cell adhesion, chemotaxis, prevention of apoptosis, invasion, migration and anchorage-independent growth of tumor cells (28). In terms of inhibition of apoptosis it is of note that in our data set increased expression of osteopontin correlated with increased expression of BIRC5 (Survivin, p=2.0×10−10), which was also over-expressed in tumors from patients with poorer relapse-free survival time. Survivin is recognized as a mediator of resistance to apoptosis, increased cell proliferation and invasiveness in melanoma (29, 30). Osteopontin has a key role in the regulation of cell signaling which controls neoplastic and malignant transformation and has been identified as a possible drug target (31). It is known to modulate several signaling pathways such as growth factor/receptor pathways via interactions with cell surface receptors such as CD44 and integrins and the metalloproteinases (32, 33). Osteopontin regulates αvβ3 integrin mediated P13K/Akt/NFkβ dependent urokinase plasminogen activator and metalloprotein expression, which is associated with tumor cell invasiveness (33). This inter-relation between Osteopontin and NFκβ complex is pictured on the Ingenuity network (Figure 1). Osteopontin also increases epidermal growth factor receptor activation (34) and is thought to provide the molecular link between degradation of the extra-cellular matrix, tumor progression and vascularization (34). In our data set we saw commensurate increased expression of genes involved in the interaction between tumor cells, the extra-cellular matrix and angioneogenesis such as MMP1, IL8 and VEGF.
Increased expression of osteopontin has been demonstrated in a number of different cancers and in some, secreted levels in the blood have prognostic value (33, 35). A proportion of melanomas have NRAS mutations (36) and in these, osteopontin transcription may be transcriptionally-activated by the RAS oncogene (37). Its expression is also regulated by Wnt/Tcf signaling, steroid receptors, growth factors, Ets and AP-1 transcription factors (38). In our data set we saw corresponding increased expression of genes involved in cell cycling (CCNA2, CDC2); DNA replication and repair (TOP2A, RAD51); cell signaling (PTPRH, MAPK10); cell division and proliferation (BIRC5, TK1). HMMR is associated with cell motility and the cell cycle and expression levels have been shown to increase with melanoma progression (39). Increased osteopontin expression was associated with reduced expression of the tumor suppressor gene GAS1, which was also under-expressed in tumors from patients who relapsed. GAS1 was recently suggested to be an important tumor suppressor for melanoma (40).
There are in vitro and animal data which suggest a role for osteopontin in melanoma (32) and in a recent large immunohistochemical study of 345 patients, increased osteopontin expression was associated with reduced relapse free and overall survival and increased probability of sentinel node positivity (25). Our study provides strong corroborative evidence for osteopontin expression as a prognostic biomarker, and possibly a drug target in melanoma.
Supplementary Material
Acknowledgements
This work was supported by Cancer Research UK (project grants C8216/A6129 and C8216/A8168), and program grant (C588/A4994), and by the NIH (R01 CA83115).
The collection of samples in the melanoma cohort study was funded by Cancer Research UK (project grants C8216/A6129 and C8216/8168, and program award C588/A4994), and by the NIH (R01 CA83115). Recruitment was facilitated by the UK National Cancer Research Network. We are grateful to the following research staff who collected or managed data over long periods of time: May Chan, Clarisa Nolan, Susan Leake, Birute Karpavicius, Tricia Mack, Paul King, Sue Haynes, Elaine Fitzgibbon, Kate Gamble, Saila Waseem, Sandra Tovey, Christy Walker, Paul Affleck.
The following clinicians participated in Study 1: Mr M.J. Timmons - Airedale General Hospital; Dr D.J. Barker, Mr I.T.H. Foo, Mr R.M. Antrum, Mr S. Al Ghazal, Prof D.T. Sharpe - Bradford Royal Infirmary; Dr H. Galvin, Dr I. Barbar - Calderdale Royal Hospital; Mr N.B. Hart, Mr P.M. O'Hare, Mr P. Stanley, Mr V. Ramakrishnan, Mr A. Platt - Castle Hill Hospital, Hull; Dr G.P. Ford, Dr G. Taylor; Dr M. Shah, Mr O. Fenton - Dewsbury & District Hospital; Dr D. Seukeran - Friarage Hospital, Northallerton; Dr A. Layton - Harrogate District Hospital; Dr D. Cowan, Dr H. Hempel, Dr J. Holder, Dr M.J. Cheesbrough - Huddersfield Royal Infirmary; Dr S. Shehade, Mr H. Siddiqui, Mr K. Allison, Mr K. Erdinger, Mr M. Coady, Mr Ramanathan, Mr T. Muir - James Cook University Hospital; Dr G Stables, Dr Sabine Sommer, Dr Caroline Wilson, Dr M. Goodfield, Dr R. Sheehan-Dare, Dr S.M. Wilkinson, Dr V. Goulden, Mr C. Fenn, Mr K. Horgan - Leeds General Infirmary; Mr P. Baguley - Middlesbrough General Hospital; Dr B. Pollock, Dr E.D.A. Potts, Dr K. Thomson, Miss O.M.B. Austin, Mr A.R. Phipps, Mr S. Southern, Mr L. Le Roux Fourie, Mr S. Majumder - Pinderfields General Hospital, Wakefield; Dr S. Clark, Dr S. McDonald-Hull - Pontefract General Infirmary; Dr A. Maraveyas, Dr S. Walton, Dr N. Alexander - Princess Royal Hospital, Hull; Mr K.R. Mannur - Scarborough Hospital; Dr A.E. Myatt - Selby War Memorial Hospital; Dr M. Marples, Dr M. Cronk, Dr P. Patel, Mr A. Batchelor, Mr H. Peach, Mr M. Liddington, Mr S.L. Knight, Mr S. Kay - St James's University Hospital, Leeds; Dr A.L. Wright, Dr K. London, Mr S.F. Worrall - St Luke's Hospital, Bradford; Dr J.A.A. Langtry, Dr S. Natarajan, Dr Verlangi - Sunderland Royal Hospital; Mr J. Ausobsky - The Yorkshire Clinic; Mr M. Erdmann, Mr N. McLean, Mr R. Debono, Mr R.B. Berry, Mr S. Rao - University Hospital North Durham; Dr A.S. Highet, Dr C. Lyon, Dr J.M. Stainforth, Mr G.V. Miller, Mr J.M. Hayward, Mr J. Taylor, Mr M.R. Telfer - York District Hospital
The following clinicians participated in Study 2; Prof GT Layer and Mr MW Kissin, Royal Surrey County Hospital, Guildford; Mr JA Smallwood, Southampton University Hospital; Mr PRW Stanley, Castle Hill Hospital, Hull; Mr H Peach, Leeds; Dr H Chong, St Georges Hospital; Dr E Thorne and Dr C Ottensmeier, Southampton and Dr E Marshall, Whiston.
Footnotes
Conflict of interest: None
Translational relevance
This paper reports an investigation of new technologies to perform gene expression studies in small formalin fixed tumor biopsies. The study identified increased expression of osteopontin as a predictor of relapse in melanoma. The study therefore confirms an important potential drug target but its translational relevance is that the study supports the view that high through put gene expression studies are now possible from tumor banks. Stored tissue from mature clinical trials may be accessed to investigate predictive markers using this approach.
References
- 1.Balch CM, Buzaid AC, Soong SJ, et al. Final version of the american joint committee on cancer staging system for cutaneous melanoma. J Clin Oncol. 2001;19:3635–48. doi: 10.1200/JCO.2001.19.16.3635. [DOI] [PubMed] [Google Scholar]
- 2.Elder D, Murphy G. Malignant tumors (melanomas and related lesions) (third series).Atlas of Tumor Pathology: Melanocytic Tumors of the Skin. 1991;2:103–205. [Google Scholar]
- 3.Cochran AJ, Elashoff D, Morton DL, Elashoff R. Individualized prognosis for melanoma patients. Hum Pathol. 2000;31:327–31. doi: 10.1016/s0046-8177(00)80246-4. [DOI] [PubMed] [Google Scholar]
- 4.Kim J, Mori T, Chen SL, et al. Chemokine receptor CXCR4 expression in patients with melanoma and colorectal cancer liver metastases and the association with disease outcome. Ann Surg. 2006;244:113–20. doi: 10.1097/01.sla.0000217690.65909.9c. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Vuoristo M, Vihinen P, Vlaykova T, Nylund C, Heino J, Pyrhonen S. Increased gene expression levels of collagen receptor integrins are associated with decreased survival parameters in patients with advanced melanoma. Melanoma Res. 2007;17:215–23. doi: 10.1097/CMR.0b013e328270b935. [DOI] [PubMed] [Google Scholar]
- 6.Winnepenninckx V, Lazar V, Michiels S, et al. Gene expression profiling of primary cutaneous melanoma and clinical outcome. J Natl Cancer Inst. 2006;98:472–82. doi: 10.1093/jnci/djj103. [DOI] [PubMed] [Google Scholar]
- 7.Kauffmann A, Rosselli F, Lazar V, et al. High expression of DNA repair pathways is associated with metastasis in melanoma patients. Oncogene. 2008;27:565–73. doi: 10.1038/sj.onc.1210700. [DOI] [PubMed] [Google Scholar]
- 8.Bibikova M, Talantov D, Chudin E, et al. Quantitative gene expression profiling in formalin-fixed, paraffin-embedded tissues using universal bead arrays. Am J Pathol. 2004;165:1799–807. doi: 10.1016/S0002-9440(10)63435-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Newton Bishop J, Beswick S, Randerson-Moor J, et al. Serum 25-Hydroxyvitamin D3 levels predict Breslow thickness at presentation and survival from melanoma. JCO. 2009 doi: 10.1200/JCO.2009.22.1135. In press. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Falchi M, Bataille V, Hayward NK, et al. Genome-wide association study identifies variants at 9p21 and 22q13 associated with development of cutaneous nevi. Nat Genet. 2009 doi: 10.1038/ng.410. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Dorrie J, Wellner V, Kampgen E, Schuler G, Schaft N. An improved method for RNA isolation and removal of melanin contamination from melanoma tissue: implications for tumor antigen detection and amplification. J Immunol Methods. 2006;313:119–28. doi: 10.1016/j.jim.2006.04.003. [DOI] [PubMed] [Google Scholar]
- 12.Eckhart L, Bach J, Ban J, Tschachler E. Melanin binds reversibly to thermostable DNA polymerase and inhibits its activity. Biochem Biophys Res Commun. 2000;271:726–30. doi: 10.1006/bbrc.2000.2716. [DOI] [PubMed] [Google Scholar]
- 13.Workman C, Jensen LJ, Jarmer H, et al. A new non-linear normalization method for reducing variability in DNA microarray experiments. Genome Biol. 2002;3 doi: 10.1186/gb-2002-3-9-research0048. research0048. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Schroeder A, Mueller O, Stocker S, et al. The RIN: an RNA integrity number for assigning integrity values to RNA measurements. BMC Mol Biol. 2006;7:3. doi: 10.1186/1471-2199-7-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Atkins MB. The treatment of metastatic melanoma with chemotherapy and biologics. Curr Opin Oncol. 1997;9:205–13. doi: 10.1097/00001622-199703000-00016. [DOI] [PubMed] [Google Scholar]
- 16.Mizuno T, Nagamura H, Iwamoto KS, et al. RNA from decades-old archival tissue blocks for retrospective studies. Diagn Mol Pathol. 1998;7:202–8. doi: 10.1097/00019606-199808000-00004. [DOI] [PubMed] [Google Scholar]
- 17.Bresters D, Schipper ME, Reesink HW, Boeser-Nunnink BD, Cuypers HT. The duration of fixation influences the yield of HCV cDNA-PCR products from formalin-fixed, paraffin-embedded liver tissue. J Virol Methods. 1994;48:267–72. doi: 10.1016/0166-0934(94)90125-2. [DOI] [PubMed] [Google Scholar]
- 18.Hoefig KP, Thorns C, Roehle A, et al. Unlocking pathology archives for microRNA-profiling. Anticancer Res. 2008;28:119–23. [PubMed] [Google Scholar]
- 19.Brizova H, Kalinova M, Krskova L, Mrhalova M, Kodet R. Quantitative measurement of cyclin D1 mRNA, a potent diagnostic tool to separate mantle cell lymphoma from other B-cell lymphoproliferative disorders. Diagn Mol Pathol. 2008;17:39–50. doi: 10.1097/PDM.0b013e318146959a. [DOI] [PubMed] [Google Scholar]
- 20.Knudsen BS, Allen AN, McLerran DF, et al. Evaluation of the branched-chain DNA assay for measurement of RNA in formalin-fixed tissues. J Mol Diagn. 2008;10:169–76. doi: 10.2353/jmoldx.2008.070127. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Linton KM, Hey Y, Saunders E, et al. Acquisition of biologically relevant gene expression data by Affymetrix microarray analysis of archival formalin-fixed paraffin-embedded tumours. Br J Cancer. 2008;98:1403–14. doi: 10.1038/sj.bjc.6604316. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Bibikova M, Yeakley JM, Wang-Rodriguez J, Fan JB. Quantitative expression profiling of RNA from formalin-fixed, paraffin-embedded tissues using randomly assembled bead arrays. Methods Mol Biol. 2008;439:159–77. doi: 10.1007/978-1-59745-188-8_11. [DOI] [PubMed] [Google Scholar]
- 23.Ravo M, Mutarelli M, Ferraro L, et al. Quantitative expression profiling of highly degraded RNA from formalin-fixed, paraffin-embedded breast tumor biopsies by oligonucleotide microarrays. Lab Invest. 2008;88:430–40. doi: 10.1038/labinvest.2008.11. [DOI] [PubMed] [Google Scholar]
- 24.Hoshida Y, Villanueva A, Kobayashi M, et al. Gene expression in fixed tissues and outcome in hepatocellular carcinoma. N Engl J Med. 2008;359:1995–2004. doi: 10.1056/NEJMoa0804525. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Rangel J, Nosrati M, Torabian S, et al. Osteopontin as a molecular prognostic marker for melanoma. Cancer. 2008;112:144–50. doi: 10.1002/cncr.23147. [DOI] [PubMed] [Google Scholar]
- 26.Zhou Y, Dai DL, Martinka M, et al. Osteopontin expression correlates with melanoma invasion. J Invest Dermatol. 2005;124:1044–52. doi: 10.1111/j.0022-202X.2005.23680.x. [DOI] [PubMed] [Google Scholar]
- 27.Jaeger J, Koczan D, Thiesen HJ, et al. Gene expression signatures for tumor progression, tumor subtype, and tumor thickness in laser-microdissected melanoma tissues. Clin Cancer Res. 2007;13:806–15. doi: 10.1158/1078-0432.CCR-06-1820. [DOI] [PubMed] [Google Scholar]
- 28.Bellahcene A, Castronovo V, Ogbureke KU, Fisher LW, Fedarko NS. Small integrin-binding ligand N-linked glycoproteins (SIBLINGs): multifunctional proteins in cancer. Nat Rev Cancer. 2008;8:212–26. doi: 10.1038/nrc2345. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Raj D, Liu T, Samadashwily G, Li F, Grossman D. Survivin repression by p53, Rb and E2F2 in normal human melanocytes. Carcinogenesis. 2008;29:194–201. doi: 10.1093/carcin/bgm219. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Grossman D, McNiff JM, Li F, Altieri DC. Expression and targeting of the apoptosis inhibitor, survivin, in human melanoma. J Invest Dermatol. 1999;113:1076–81. doi: 10.1046/j.1523-1747.1999.00776.x. [DOI] [PubMed] [Google Scholar]
- 31.Johnston NI, Gunasekharan VK, Ravindranath A, O'Connell C, Johnston PG, El-Tanani MK. Osteopontin as a target for cancer therapy. Front Biosci. 2008;13:4361–72. doi: 10.2741/3009. [DOI] [PubMed] [Google Scholar]
- 32.Rangaswami H, Kundu GC. Osteopontin stimulates melanoma growth and lung metastasis through NIK/MEKK1-dependent MMP-9 activation pathways. Oncol Rep. 2007;18:909–15. [PubMed] [Google Scholar]
- 33.Tuck AB, Chambers AF, Allan AL. Osteopontin overexpression in breast cancer: knowledge gained and possible implications for clinical management. J Cell Biochem. 2007;102:859–68. doi: 10.1002/jcb.21520. [DOI] [PubMed] [Google Scholar]
- 34.Rangaswami H, Bulbule A, Kundu GC. Osteopontin: role in cell signaling and cancer progression. Trends Cell Biol. 2006;16:79–87. doi: 10.1016/j.tcb.2005.12.005. [DOI] [PubMed] [Google Scholar]
- 35.Rodrigues LR, Teixeira JA, Schmitt FL, Paulsson M, Lindmark-Mansson H. The role of osteopontin in tumor progression and metastasis in breast cancer. Cancer Epidemiol Biomarkers Prev. 2007;16:1087–97. doi: 10.1158/1055-9965.EPI-06-1008. [DOI] [PubMed] [Google Scholar]
- 36.Edlundh-Rose E, Egyhazi S, Omholt K, et al. NRAS and BRAF mutations in melanoma tumours in relation to clinical characteristics: a study based on mutation screening by pyrosequencing. Melanoma Res. 2006;16:471–8. doi: 10.1097/01.cmr.0000232300.22032.86. [DOI] [PubMed] [Google Scholar]
- 37.Guo X, Zhang YP, Mitchell DA, Denhardt DT, Chambers AF. Identification of a ras-activated enhancer in the mouse osteopontin promoter and its interaction with a putative ETS-related transcription factor whose activity correlates with the metastatic potential of the cell. Mol Cell Biol. 1995;15:476–87. doi: 10.1128/mcb.15.1.476. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.El-Tanani MK, Campbell FC, Kurisetty V, Jin D, McCann M, Rudland PS. The regulation and role of osteopontin in malignant transformation and cancer. Cytokine Growth Factor Rev. 2006;17:463–74. doi: 10.1016/j.cytogfr.2006.09.010. [DOI] [PubMed] [Google Scholar]
- 39.Ahrens T, Assmann V, Fieber C, et al. CD44 is the principal mediator of hyaluronic-acid-induced melanoma cell proliferation. J Invest Dermatol. 2001;116:93–101. doi: 10.1046/j.1523-1747.2001.00236.x. [DOI] [PubMed] [Google Scholar]
- 40.Gobeil S, Zhu X, Doillon CJ, Green MR. A genome-wide shRNA screen identifies GAS1 as a novel melanoma metastasis suppressor gene. Genes Dev. 2008;22:2932–40. doi: 10.1101/gad.1714608. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Bibikova M, Chudin E, Arsanjani A, et al. Expression signatures that correlated with Gleason score and relapse in prostate cancer. Genomics. 2007;89:666–72. doi: 10.1016/j.ygeno.2007.02.005. [DOI] [PubMed] [Google Scholar]
- 42.Abramovitz M, Ordanic-Kodani M, Wang Y, et al. Optimization of RNA extraction from FFPE tissues for expression profiling in the DASL assay. Biotechniques. 2008;44:417–23. doi: 10.2144/000112703. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Galinsky VL. Automatic registration of microarray images. II. Hexagonal grid. Bioinformatics. 2003;19:1832–6. doi: 10.1093/bioinformatics/btg260. [DOI] [PubMed] [Google Scholar]
- 44.Pfaffl MW. A new mathematical model for relative quantification in real-time RT-PCR. Nucleic Acids Res. 2001;29:e45. doi: 10.1093/nar/29.9.e45. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.