Background: Oropharyngeal squamous cell carcinoma (OPSCC) is diagnosed in over 45,000 Americans annually, with a five-year survival of less than 60%. The disease is currently diagnosed and staged by expert histopathologists. This process is largely qualitative and can result in inter-observer discordance. Raman spectroscopy is a high information content, laser-based technology that can be used to interrogate biological specimens. It has been demonstrated that Raman spectroscopy can discriminate OPSCC tissues from normal mucosa in both in-vivo and ex-vivo settings. Tissue microarrays (TMA) are commonly assembled for high throughput screening of molecular markers of disease and may be amenable to spectroscopic investigations. Once optimized, Raman spectroscopy may be a valuable analytical method for evaluation of challenging histopathological questions, such as OPSCC disease progression.
Hypothesis: We hypothesize that a TMA can be utilized to generate Raman spectra with minimal intra-core and inter-core variability. We further hypothesize that Raman spectroscopy can be used to differentiate metastatic OPSCC from non-metastatic OPSCC.
Methods: A TMA representing triplicate primary tumor biopsy cores from 46 OPSCC cases was utilized for this experiment. A 4 μm thin section was cut from the TMA and affixed to an aluminum-coated glass slide for Raman investigation. Tissues were deparaffinized immediately prior to measurement using a xylene protocol. Dispersive Raman spectra were collected using a Falcon II™ Wide-Field Raman Chemical Imaging System (ChemImage Corp., Pittsburgh, PA). 3 spectra, spaced at approximately 250 μm intervals, were collected from each core for a total of 9 spectra per case. Tissues were returned to distilled water every 60 minutes to preserve tissue hydration. Data was processed and analyzed using ChemImage Xpert® software and MATLAB R2014B®. Spectral variability was assessed based on total spectral variability (TSV = Σ σi), the sum of calculated standard deviation (σ) at each relative wavenumber i across the spectrum, and the coefficient of variation (CV = σi / μi). Partial least squares discriminant analysis (PLS-DA) was performed using proprietary algorithms.
Results: A total of 354 Raman spectra were collected from 118 cores. 20 of 138 cores were lost to sectioning of the TMA. 16 spectra were rejected from analysis because of a low SNR or excessive fluorescence background. Analysis included data from 43 OPSCC cases. Intra-core TSV was < 0.020 for all cores with a maximum CV of 0.14%. Inter-core TSV was < 0.068 for all cases with a maximum CV of 0.45%. PLS-DA using spectra from 23 cases with metastatic disease and 20 patients with non-metastatic disease discriminated the two populations with sensitivity and specificity > 95%.
Conclusions: Preliminary data suggests that Raman spectroscopy can detect differences between primary tumors that did and did not metastasize as measured from a TMA. Further work is ongoing to evaluate the significance of this result.