Identification of differentially expressed genes in matched formalin-fixed paraffin-embedded primary and metastatic melanoma tumor pairs

Authors


Dr Rosalyn Jewell, e-mail: r.a.jewell@leeds.ac.uk

Dear Editor,

Gene expression studies in melanoma have identified prognostic gene signatures (Winnepenninckx et al., 2006; Riker et al., 2008); however, because of limited availability of cryopreserved tissue, gene expression studies in matched primary and nodal metastases are few. Genetic heterogeneity of primary melanoma, germline genetic variation, and epigenetic influences modify gene expression profiles (Cookson et al., 2009); therefore, assessment of matched primary and metastatic tumors may yield critical insights into mechanisms of melanoma progression. We have previously reported use of Illumina’s cDNA-mediated annealing, selection, extension, and ligation (DASL) gene expression assay in formalin-fixed paraffin-embedded (FFPE) primaries (Conway et al., 2009). Here, we report a pilot to assess the feasibility of using the DASL assay to identify differentially expressed genes in primaries and extremely small FFPE sentinel node biopsy (SNB) samples.

Twenty-five FFPE SNB samples, with metastatic deposits ranging from eight cells to 14 mm in diameter, were identified from two study sets (Appendix S1). Specimens were sampled using laser capture microdissection of either new tissue sections, diagnostic hematoxylin and eosin (H&E) slides or archival unstained sections (Appendix S1). RNA was extracted, and expression of 502 cancer-related genes was assessed using the DASL assay. RNA quantity and quality were assessed using spectrophotometry and quantitative real-time PCR (qRT-PCR) of the housekeeping gene, RPL13a.

In the DASL assay, the number of genes detected in each sample is generally used as a measure of result quality (Ravo et al., 2008). The mean number of genes detected from nodal samples was less (mean 242, range 36–369) than detected from RNA extracted from a core of primary tumor (Study 1: 434, Study 2: 457 (Conway et al., 2009), Figure S1).

As presented in Table 1, neither the area microdissected, RNA concentration or Ct values from qRT-PCR of RPL13a correlated with number of genes detected. However, the age of the FFPE tumor block or tissue section negatively correlated with detected genes, and more genes were detected in newly cut sections or H&E slides than from archival unstained uncovered sections.

Table 1.   Characteristics of nodal samples and extracted RNA. Associations between these factors and the number of genes detected in the DASL assay are presented
 Nodal samplesAssociations between sample factors and number of genes detected (Test statistic and significance value)
  1. DASL, cDNA-mediated annealing, selection, extension, and ligation.

Area of tissue microdissected, median (range), μm2223124 (157–6 051 771) 
Number of genes detected for each quartile of area dissected, mean (range)
 <25%262 (182–365)Spearman’s rho −0.18, P = 0.41
 25–50%229 (167–275)
 50–75%260 (209–314)
 >75%237 (166–369)
RNA concentrations, median (range), ng/μl30.3 (0.4–73.1)Spearman’s rho 0.18, P = 0.39
Ct value, median (range)24.1 (20.8–25.2)Spearman’s rho −0.05, P = 0.83
Age of tissue block, median (range), yrs3.8 (1.1–6.9) 
Number of genes detected for each quartile of block age, mean (range)
 <2.29 yrs299 (249–365)Spearman’s rho −0.50, P = 0.01
 2.29–3.76 yrs253 (193–369)
 3.77–4.36 yrs217 (166–290)
 >4.47 yrs227 (182–325)
Age of tissue slide (before RNA extraction), median (range), months41.5 (0.3–75.5)Spearman’s rho −0.51, P = 0.01
Slide type, number (mean number of genes detected)
 New slide8 (280)Kruskal–Wallis chi-squared 8.79, P = 0.01
 Old unstained slide14 (219)
 Diagnostic H+E slide3 (301)
Number of genes detected, mean (range)242 (36–369) 

One nodal sample, with a low number of detected genes, was excluded from further analysis. Of the remaining samples, 22 had matched primary samples (Conway et al., 2009). Gene expression for 21 of the matched pairs was correlated (ρ = 0.15–0.80, P ≤ 0.001, Figure S2); however, a number of genes were differentially expressed (Figure 1, fold changes Table S1). For genes over-expressed in nodal tumors, 9/10 genes were over-expressed in all nodal samples from all pairs assessed. Genes differentially expressed included genes involved in breakdown of extracellular matrix (MMP2) and cellular proliferation and survival (FGF3, FGF5, FGF6 and FGF8).

Figure 1.

 Genes most differentially expressed across matched primary and sentinel node biopsy samples. Box plots are presented for genes most over-expressed in primary (A) and nodal (B) samples. Outside values are excluded for clarity from graph B.

This pilot study was designed to identify the lower limits of sample volume that can be used to generate gene expression data using the DASL assay. A deposit reported as being only eight cells in size provided data, suggesting that very small FFPE tumor samples can be used to assess gene expression.

The number of genes detected in nodal samples was less than from larger primary tissue cores. However, the number of genes expressed in metastases is likely to be smaller, consistent with Fidler’s hypothesis that metastases occur from subclones of the primary (Fidler and Kripke, 1977). Therefore, number of genes detected may not be the best measure of assay performance, and consequently, we have included all nodal samples with a reasonable number of detected genes in this analysis (Appendix S1).

Current quality control measures were unable to predict which samples would successfully yield data. Our results indicate that age and type of tissue sample influence assay performance with freshly cut sections from newer tissue blocks being most likely to produce data.

There was broad correlation between gene expression profiles from the majority of matched primary melanoma and SNB samples as has been reported previously (Augustine et al., 2010; Harbst et al., 2010). Fibroblast growth factors (FGFs), with their broad mitogenic and angiogenic activities, were differentially expressed across matched pairs. FGFs are frequently over-expressed in a number of tumor types (Turner and Grose, 2010), and this may represent an important step in the early metastatic processes of malignant melanoma, warranting further investigation in a larger sample set.

This work is clearly limited by the small number of samples assessed and additional factors in the study design that may influence performance with the assay, such as age of tissue block and differences in tissue age between matched samples. However, this study demonstrates that very small FFPE SNB samples can be used for gene expression analysis enabling more extensive studies using primary and early metastatic tumors from the same patient.

Ancillary