Raman micro-spectroscopy for accurate identification of primary human bronchial epithelial cells

www.nature.com/scientificreports OPEN Received: 19 February 2018 Accepted: 20 July 2018 Published: xx xx xxxx Raman micro-spectroscopy for accurate identification of primary human bronchial epithelial cells Jakub M. Surmacki 1,2, Benjamin J. Woodhams Bruce A. J. Ponder2 & Sarah E. Bohndiek 1,2 1,2 , Alexandria Haslehurst2, Live cell Raman micro-spectroscopy is emerging as a promising bioanalytical technique for label-free discrimination of a range of different cell types (e.g. cancer cells and fibroblasts) and behaviors (e.g. apoptosis). The aim of this study was to determine whether confocal Raman micro-spectroscopy shows sufficient sensitivity and specificity for identification of primary human bronchial epithelial cells (HBECs) to be used for live cell biological studies in vitro. We first compared cell preparation substrates and media, considering their influence on lung cell proliferation and Raman spectra, as well as methods for data acquisition, using different wavelengths (488 nm, 785 nm) and scan protocols (line, area). Evaluating these parameters using human lung cancer (A549) and fibroblast (MRC5) cell lines confirmed that line-scan data acquisition at 785 nm using complete cell media on a quartz substrate gave optimal performance. We then applied our protocol to acquisition of data from primary human bronchial epithelial cells (HBEC) derived from three independent sources, revealing an average sensitivity for different cell types of 96.3% and specificity of 95.2%. These results suggest that Raman microspectroscopy is suitable for delineating primary HBEC cell cultures, which in future could be used for identifying different lung cell types within co-cultures and studying the process of early carcinogenesis in lung cell culture. Raman spectroscopy is a powerful bioanalytical technique that reveals the chemical constituents of a given sample based on the inelastic scattering properties of molecular bonds. Despite the relatively weak nature of the Raman effect (fewer than 1 Raman scattering event occurs for every 107 elastic scattering events)1, the advent of confocal Raman micro-spectroscopy methods that allow 3D localization of signals together with highly sensitive detectors have enabled this label-free technique to be applied in living cells over the past decade, as extensively reviewed2–8. In particular, the ability to monitor the concentration of lipids, proteins and nucleic acids enables interrogation of a wide range of cellular processes. Examples range from identification and spatial localization of the main cellular components9–12, through time lapse studies of live and apoptotic cells13–15, to discrimination of normal and cancer cells16–20. The aim of this study was to determine whether confocal Raman micro-spectroscopy would show sufficient sensitivity and specificity for identification of primary human bronchial epithelial cells (HBECs) as well as immortalized cell lines to be used for live cell biological studies in vitro. Despite the apparent promise for Raman spectroscopy in this application8, several challenges must be overcome to apply Raman spectroscopy in live cell studies, which have also been identified by recent reviews2,4,6. For example, the low probability of Raman scattering leads to a direct trade-off for live cell imaging between high signal-to-noise ratio, requiring long acquisition times, and adequate temporal resolution, required for longitudinal imaging of biological dynamics in live cells. Therefore, to achieve our aim, we first established a detailed protocol for imaging of live human lung cells in vitro by directly comparing methods for sample preparation and data acquisition. We performed these initial optimization studies in live human lung cancer (A549) and fibroblast (MRC5) immortalized cell lines and compared the imaging results qualitatively with fluorescence imaging. We then applied the optimized protocol to acquire data from primary HBECs from several different sources. Using partial least squares discriminant analysis, we achieved an average sensitivity of 96.3% and specificity of 95.2%, suggesting that Raman micro-spectroscopy 1 Department of Physics, University of Cambridge, Cavendish Laboratory, JJ Thomson Avenue, Cambridge, CB3 0HE, United Kingdom. 2Cancer Research UK Cambridge Institute, University of Cambridge, Li Ka Shing Centre, Robinson Way, Cambridge, CB2 0RE, United Kingdom. Correspondence and requests for materials should be addressed to S.E.B. (email: ) SCIentIfIC REPOrtS | (2018) 8:12604 | DOI:10.1038/s41598-018-30407-8 1 www.nature.com/scientificreports/ may indeed be suitable for differentiating between HBEC primary cell cultures and could in future be applied to identification of different lung cell types within co-cultures and studying the process of early lung carcinogenesis in cell culture. Results Comparison of cell preparation and data acquisition methods for delineating cancer and fibroblast cell lines. Firstly, we evaluated the impact of different cell preparation conditions. Raman spectroscopy of cell substrates and culture media was performed at 488 nm and 785 nm (Supplementary Fig. 1). These results indicated that, in line with previous work21, a quartz substrate provides the best compromise for live lung cell imaging. In addition to the expected strong Raman peaks due to water at around 1640, 3250 and 3430 cm−1, cell culture media contributes additional peaks at around 1046, 1305 and 1454 cm−1, however, compared to physiological buffered solutions (HBSS, LCIS and PBS) it does not have a detrimental impact on the proliferation of the cell cultures over extended time periods (up to 48 hours). Secondly, we compared results obtained using different data acquisition methods. Photothermal and photochemical reactions to laser illumination can rapidly induce cell death22. To avoid extended dwell time and allow more frequent Raman spectroscopy data acquisitions (technical replicates) from more cells (biological replicates) when studying primary HBECs, we examined the potential of using a line-scan rather than an area-scan data acquisition. We started by performing area-scans of lung A549 cancer cells and MRC5 fibroblast cells at 488 nm excitation using both K-means clustering and sum filters to generate Raman images (Fig. 1A). The associated cluster spectra are presented in Supplementary Figure 2 after background (cluster 1) subtraction. Epi-fluorescent imaging of the same A549 cell stained with NucBlue (nucleus) and Nile Red (lipids) after the Raman experiment are also shown in Fig. 1, which allowed us to perform a qualitative comparison of the lipid rich regions and nuclei location as described below. As the MRC5 cells are migratory, fluorescence staining and comparison could not be performed due to live cell motion. The main differences observed between the clusters from the two cell types (examined in Supplementary Fig. 2) were in the cytoplasm (cluster 3), nucleus (cluster 4) and (...truncated)