Tissue characterisation

3.2.1 Haemotoxylin and Eosin staining of tissue sections

All skin samples were stained with Haemotoxylin and Eosin. Paraffin embedded sam-ples were washed in xylene and alcohol, followed by staining then repeat washes in reverse order. Cryo sections were first washed in PBS, stained and then dehydrated in alcohol. Full details of the staining steps can be seen in Appendix. Stained sections were allowed to dry for 10 minutes and sealed off with cover slips.

3.2.2 Sample preparation for Raman analysis Intact skins

For Raman measurements on intact native skin, fresh skin samples from theatres were cut into 0.5-1 cm² pieces with a scalpel blade. The pieces were placed in a custom made stainless steel well block and enough PBS was added inside of the well so as to

keep the underside of the skin hydrated. A sapphire window of 2.5 cm diameter and 0.5 cm thickness (UQC Optics, UK) was used to seal off the well. The same procedure was also followed for some intact HSE measurements though approximately half of all HSE profiles were performed as is in their cell culture plates: Raman depth profile was collected with the lid off just before fixing in formaldehyde. 785 or 633 nm lasers and a x50 long working distance (LWD) objective (Olympus, N.A. 0.50) was used for all intact skin measurements. Collection times varied from 3x50 s per point to 1x60 s per point with total map times ranging from 120 minutes to 44 minutes. Typically three profiles were collected from three random locations on each skin sample.

Tissue sections

Data from tissue sections were collected using 532, 633 and 780 nm lasers and a 50x LWD objective. A 50-µm pinhole aperture was used for most of the data except for occasional thin sections (10 µm) which often benefitted from a 25-µm pinhole aperture.

Collection times varied from a single 30 s exposure to 5x60 s exposures depending on sample, substrate and laser. For normal skin samples, spectra were collected from random points on each of the tissue layers: SC, epidermis and dermis. These layers were easy to identify on white light images and then confirmed from the adjacent H&E stained sections. On average, 5-10 spectra were collected from each layer of 2 to 3 sections of three samples. For melanoma samples, location of every sample point was recorded on the white light image of the sample area. These points were then compared to adjacent H&E stained sections to determine the corresponding tissue type/layer (e.g.

tumourous or normal looking, melanoma invaded versus non invaded dermal compo-nents). Melanoma samples were divided into areas of different categories (Chapter 6) with a minimum of 30 spectra collected from each area.

3.2.3 Raman instruments

Instrument 1: Thermo DXR Raman Microscope equipped with a 532 nm diode laser, a 633 nm HeNe laser and 780 nm diode laser. Spectral resolution was 6 cm^-1 for each laser setup.

Instrument 2 (EIP Loan Pool): Renishaw InVia Raman Microscope equipped with a 785 nm diode laser. Spectrometer aperture was set to 64 µm. The grating was 1200 lines/mm producing a reported maximum spectral resolution of up to 2 cm^-1.

3.2.4 Processing and Analysis of Raman data

Substrate subtraction, fluorescence correction and peak position measurements were done on OMNIC and TQ Analyst (Thermo, U.S.A.). Substrate spectra were sub-tracted from almost 90% of the sample spectra collected from tissue sections on glass and quartz. Polynomials were subtracted from all spectra collected with the 633 nm laser (Chapter 5, Section 5.2.2) to correct for baseline distortions. Degree of the polyno-mial used varied from 3 to 6. Peak positions were measured as the position of maximum height of a peak within a defined range (e.g. Max height in the range 1520-1360 cm^-1 gives Amide II position.).

Unscrambler X (CAMO, Sweden) was used for the MVA methods Cluster Analysis (CA), Principle Component Analysis (PCA) and Linear Discriminant Analysis (LDA).

The data were first baseline corrected and Unit Vector Normalised (UVN). Wards method of clustering was used for CA. This method computes the difference between (i) each sample in a group and (ii) between each sample to the groups centroid. The parameters for PCA setups varied. Every PCA was setup with a minimum of 12 or-thogonal variables depending on the spectral region used, where the number of PCs chosen for each setup described >99% of the variation. All LDA models were setup over the full spectral range. In each LDA, five random spectra from the data set were left out of the training models at each run. Each set of five -unlabelled- random spectra were tested against the corresponding training model until 20 such spectra were tested.

Model sensitivity and specificity were calculated by comparing predicted class to the official class obtained through H&E staining.

Peak height measurements were done on OriginPro (OriginLab), TQAnalyst (Thermo, U.S.A.) and WIRE (Renishaw, UK). Measurement of Phenylalanine intensity was done on OriginPro. Spectra were normalised to Amide I, one peak was fitted to a fixed position of 1002 cm^-1, and height was measured from the baseline. Signal loss versus depth for intact skin was presented as the percentage loss from the maximum intensity point (Chapter 4). Height measurements of the high wavenumber region (3200-2600 cm^-1) of tissue sections were performed on TQAnalyst. WIRE or OriginPro was used for intact skins. For tissue sections, the height of peaks at the fixed locations of 2930, 2880 and 2840 cm^-1 were measured from the baseline. Charts that track the protein to lipid content in sectioned skin layers (Chapter 5, Section 5.2.1) use these measures.

For intact skins, 5 peaks were fitted to this band (fixed centres at 3047, 2980, 2930,

2880 and 2840 cm^-1). Lipid profiles through intact skin layers (Chapter 5, Section 5.2.2) use the intensities of these component bands. OriginPro was used to measure the intensity of the Tryptophan doublet (Chapter 6, Section 6.2.3). This was done by fitting two peaks with fixed centres representing the 1360 and 1340 cm^-1 peaks to the first derivative spectra.