NanoString gene expression panels have emerged as a cost-effective solution for mid-level multiplexing of expression datasets. These panels combine the ease of setup that is a hallmark of NanoString experiments with expertly curated gene targets, ensuring rapid generation of expression measurements for targets of significance. However, while experimentally straightforward, analysis of NanoString datasets is significantly more complex. In order to facilitate the interpretation of these datasets, we developed NavSIVRAC a custom collection of analysis scripts designed to automate the entirety of the analysis process. Using NavSIVRAC, we have probed the expression landscape for a cohort of primary and metastatic BRAF-mutated melanoma patients collected in a controlled setting at an initial screening time point and, for a subset, at second visits. Our analysis reveals an overall diverse tumor population that exhibits clear patterns of expression correlating to the cancer’s progression, providing new potential targets for alternative therapeutic approaches.
RNA was isolated from melanoma samples of primary origin or metastasis using the Qiagen AllPrep FFPE kit. A maximum of 200 ng of RNA was input into a NanoString hybridization reaction containing 800 custom-designed target probes for genes associated with cancer progression. Raw .rcc files obtained from the nSolver system for each sample were uploaded to NavSIVRAC for identification of outliers, normalization, and differential expression analysis.
Preliminary analysis by the NavSIVRAC software identified a small number of samples as outliers that were removed from subsequent analysis. Using the raw expression results for samples that passed the initial quality assessment, NavSIVRAC automatically generated a minimum subset of optimal housekeeping genes (HKGs) for the melanoma dataset. The software then compared a number of normalization strategies for the data based upon the custom HKGs. Multiple approaches yielded normalized data of similar quality, but ultimately the strategy that maximized the uniformity of the pre-defined HKGs was selected. Following normalization of all endogenous counts based on the HKG expression, gene signatures were identified using differential expression and principle component analyses.
Tumor content for the n = 678 samples assayed was 65 ± 24% with a mean area of 135 ± 137 mm2, and all samples yielded sufficient RNA for testing. Of the 800 targets in the NanoString panel, on average, 60 ± 24% yielded quantifiable signal. High, stable expression across the dataset was found for targets involved in cell proliferation and cytoskeletal rearrangements, indicative of rapid growth. Additionally, elevated CD74 and MIF transcript levels were detected for a majority of biopsies, suggesting an active immune response.
Analysis of primary or metastatic samples obtained at the initial visit with NavSIVRAC yielded a 31-gene signature that provided clear separation between the two populations. Metastatic samples exhibited significantly reduced levels of transcripts found in normal skin tissue, including SPINK5 and KRT7, while containing elevated levels B and T cell markers concurrent with increased CD22 expression. Notably, CTAG1B, a known tumor antigen, was highly expressed in a subset of the metastatic population.
Substantial changes in overall expression patterns between screening and second biopsies were observed for the primary melanoma samples that were less evident in the metastatic population. Second biopsies in the primary melanoma sample subset had markedly reduced levels of immune-system markers and increased levels of insulin-like growth factors, signifying a shift to a less-differentiated, more aggressive cellular state. Conversely, samples of metastatic origin showed little change in overall expression over time, implying that without external selective pressures via a targeted therapeutic or increased host-immune response, a dominant and static tumor cell population emerges.
NavSIVRAC is a powerful solution for NanoString data analysis. When applied to a representative melanoma-sample dataset, the automated NavSIVRAC pipeline generated robust, high-quality analyses of expression data. Within primary and metastatic melanoma samples, a gene signature was identified that provides clear separation between the melanoma types for the tested population. Furthermore, application of NavSIVRAC within the primary tumor biopsies revealed marked changes to expression patterns over time, providing potential biomarkers for monitoring of tumor progression. Finally, the absence of changes over time in the metastatic population is illustrative of the eventual progression of all cancers to an undifferentiated state. By integrating outlier analysis, data normalization, and result interpretation into a single, easy-to-use workflow, NavSIVRAC provides an efficient solution for high-complexity RNA datasets.
Nathan Riccitelli– Navigate BioPharma Services, Inc., Carlsbad, California
Melissa Lynch– Navigate BioPharma Services, Inc., Carlsbad, California
Jesus Zaragoza-Alvarez– Navigate BioPharma Services, Inc., Carlsbad, California
Reinhold Pollner– Director, Molecular Diagnostics Assay Development, Navigate BioPharma Services, Inc., Carlsbad, California