microRNA from brush biopsy to characterize oral squamous cell carcinoma epithelium

Abstract

Few cancers are diagnosed based on RNA expression signatures. Oral squamous cell carcinoma (OSCC) is no exception; it is currently diagnosed by scalpel biopsy followed by histopathology. This study sought to identify oral tumor epithelial microRNA (miRNA) expression changes to determine if these changes could be used to diagnose the disease noninvasively. Analysis of miRNA profiles from surgically obtained OSCC tissue, collected under highly standardized conditions for The Cancer Genome Atlas, was done to determine the potential accuracy in differentiating tumor from normal mucosal tissue. Even when using small 20 subject datasets, classification based on miRNA was 90 to 100% accurate. To develop a noninvasive classifier for OSSC, analysis of brush biopsy miRNA was done and showed 87% accuracy in differentiating tumor from normal epithelium when using RT‐qPCR or miRNAseq to measure miRNAs. An extensive overlap was seen in differentially expressed miRNAs in oral squamous cell carcinoma epithelium obtained using brush biopsy and those reported in saliva and serum of oral squamous cell carcinoma patients in several studies. This suggested that nonselective release of these miRNAs into body fluids from tumor epithelium was largely responsible for the changes in levels in these fluids seen with this disease. Using a variation in mirRPath we identified the KEGG pathway of neurotrophin signaling as a target of these miRNAs disregulated in tumor epithelium. This highlights the utility of brush biopsy of oral mucosa to allow simple acquisition of cancer relevant miRNA information from tumor epithelium.

Improving accuracy of RNA-based diagnosis and prognosis of oral cancer by using noninvasive methods

“…Table 2 shows this finding. Most notable is the agreement with the work of Maclellan et al who examined serum miRNAs in cases of OSCC and Zhou et al who acquired samples by brush biopsy of tumor but used the same platform to measure miRNA, removing that as a source of variation [27, 64]. Of 10 miRNAs upregulated with p < 1 × 10–7 in serum, 8 were tested in an independent study of brush biopsy miRNA from OSCC and 6 were shown to be similarly enriched (75% agreement) [64].…”

“…[46, 50]. We have found substantial agreement in OSCC miRNA markers identified in brush biopsy and body fluid samples in several studies [27]. Table 2 shows this finding.…”

“…These, of course, would be obtained, dissected and prepared by the same group [25]. When this was done using RNA profiles from a subgroup of TCGA database, external validation with a miRNA-based classifier revealed near 100% accuracy in differentiating normal tissue from tumor tissue [27]. The accuracy of class prediction in two somewhat smaller studies was almost as good [8, 28], suggesting that within a single study group it is possible to construct a highly accurate classifier to differentiate OSCC from normal.…”

“…Descriptions of acceptable levels of stromal tissue in tumor samples are varied between studies, if present at all. The TCGA head and neck squamous cell carcinoma (HNSCC) study indicated that when sample acquisition is uniform, differentiation of normal versus malignant based on RNA is extremely accurate [25, 27]. When we tested marker RNAs identified by Lager et al for European OSCC, mainly tobacco users, versus those miRNAs measured in the TCGA study, over 30 percent were no longer differentially expressed [8, 25].…”

“…Of the 12 miRNAs tested, 6 were tested in both studies and 3 showed the same statistically significant changes [29]. One might conclude that by removing stroma as a contaminant fewer samples are needed to produce a reproducible miRNA signature predictive of OSCC [27]. While the simple acquisition of nearly pure cells from the epithelium is a possible large advantage when using brush biopsy, there are insufficient studies to verify clinical application of this method to tumor diagnostics .…”