3 .1 Introduction
This chapter describes analyses of cDNA from basal cell carcinomas and skin by hybridisation to a gridded keratinocyte library. An analysis of the reproducibility of results was made, and then bearing this in mind, differentially expressed genes were determined. Several genes were differentially expressed in this analysis. Some of these genes, such as
the cytokeratins had already been documented in the literature (Stoler et ah, 1988; Markey
et a l, 1992). MRP-14 and 8, genes represented multiple times on the filters under test were determined to be up-regulated in BCC. RT-PCR was used to verify the higher expression in tumours. Subsequently, immunohistochemisty and immunofluoresence staining were performed.
3 .2 Construction of the Arrayed cDNA Library
The main resource in this analysis was the arrayed keratinocyte library constructed by B.Lin in the laboratory of Anna-Maria Frischauf. In brief, cDNA was synthesised from mRNA obtained from cultured primary kératinocytes. The cDNA was size selected, and cDNAs larger that 400 bp were cloned into the pSPORTl vector. The library was plated, and single colonies were picked by robot into 384 well plates. The library was thus unamplified but was not normalised. These plates were then used to spot 22x22cm nylon filters in a duplicate 3x3 pattern, incorporating as the central guide spot, the neomycin resistance cassette of the sCos vector. These filters, with 9218 duplicate clones on them would be a representative sample of the genes that are expressed to moderate or high levels in primary cultured kératinocytes. Hybridisation of complex cDNA probes to these filters allows the possibility of determining differential expression of those genes represented on the filter.
3 .3 Differential Hybridisation and Image Analysis
Differential hybridisation set-up
normal, non-site matched
mRNA was extracted from one skin sample and two BCC samples (Skin, BCCl and BCC2) kindly provided by T. Quinn (Skin tumour laboratory, I.C.R.F.). Double stranded cDNA was synthesised from these mRNA samples. 65ng of the cDNA was labelled with a^^PdCTP by the random hexamer method. The resulting complex probe was hybridised overnight to the arrayed keratinocyte library in the presence of labelled sCos neomycin cassette DNA, which hybridised to the guide spots. The filters were washed to high stringency, before being exposed to a Phosphorlmager plate (Molecular Dynamics). One batch of each probe from skin, BCCl and BCC2 was hybridised to three identically spotted filters, and the filters scanned to provide raw image of all nine hybridisations.
Using Xdigitise to analyse the raw images
Xdigitise, a program developed by H Griffiths and H Lehrach (Max Planck Institute - Berlin) combines image analysis and quantitation programs. It was used for the analysis of all microtitre plate format high-density filters. The raw images described above and the images from the Unigene filters described in the next chapter were analysed using this program. The Xdigitise software can find and quantitate each of the 9218 spots, assigning the correct spot positions by detecting each of the guide spots. This was possible even when the filter was distorted as a result of multiple hybridisations. Figure 3.1 illustrates the quantitation process. Briefly the user manually assigned the comer guide spots.. From this guide the program placed a basic array layout over the image. Then the program fitted aU the spots in the layout to the correct spots on the filter. The filter was then quantitated on a spot by spot basis. The output of Xdigitise was a list of clone co-ordinates, clone positions and their corresponding signal intensity
Chapter 3 BCC gene expression
The lists of intensities needed to be normalised in order to compare them. This was done by dividing each intensity value in the list by the total signal on the filter, i.e. the sum of the intensity on all the spots. The total sum of the signal on all spots was a rough indication of the specific activity of the probes. These normalised datasets were either compared individually or as the average of the normalised signals from the skin hybridisations and all the BCC hybridisations when differentially expressed genes were being determined. These comparisons were performed using the Cgen program.
Cgen
Cgen, a program written by Wolfgang Lehrach (www.Lehrach.com/wolfgang/Cgen), plotted the intensity data from one hybridisation against the data from another. The product was a scatter plot with a line of best fit through the data. The plot was interpreted as follows. The data from hybridisation A (hyb A) were compared with the data from hybridisation B (hyb B). The intensity of a particular clone in hyb A was the x-axis value, and the corresponding value for the same clone in hyb B was the Y-axis value. In this hypothetical example, any data point that was below the best fit line has a higher signal in hyb A. Conversely, any point above the line was a clone with a higher signal in hyb B . Any points on the line were clones that show no differences in signal intensity. See Figure 3.2
3 .3 Evaluation of the Hybridisations
Comparison o f the same probe on different Filters
Comparing the hybridisation of an identical probe on different filters is the best indication of variation due to DNA content on the spots. The expectation was that the signals from differentially expressed clones in Cgen would still vary, but that variation would be within the cloud of non-differentially expressed genes. In almost all cases this was true, DNA variation was not enough to produce many false positives. However, some of the potentially differentially expressed clones showed an apparent reduction in DNA content on
filter 108, where some of the signals associated with differentially expressed clones were in fact outside the cloud; Yellow spots in figure 3.3 B, C and D.
Comparison o f dijferent BCC samples
(Miller, 1991a; Miller, 1991b; Preston and Stem 1992) Different genes are differentially expressed in different BCCs, so it was necessary to see how the differentially expressed clones varied in the two BCC samples. So DNA loading variation on the spots was not a consideration, comparisons were made between the normalised data from probes BCCl and BCC2 on the same filter (Figure 3.3 A). If the clones in question were differentially expressed to the same degree in both BCCs, all the spots would vary in signal, but only within the cloud of no variation in signal. Almost aU those signals varied systematically between BCCl and BCC2, with almost all the signals being higher in BCCl. What are the explanations for this? One could be that all genes in question are all more highly expressed in BCCl than BCC2, reflecting some difference in the degree of progression of the tumour or some other phenotypic difference. However other explanations are possible. The BCCl probe could have produced a higher background in the hybridisations. This would increase every signal on the filter and would not be affected by normalisation. Possibly the non-BCC components, such as the stroma were contributing in some way to the higher levels of expression of these genes.
3 .5 Determination of Differentially Expressed Genes
Differences in gene expression were determined using Cgen and the normalised summed datasets from BCCl, BCC2 and skin. Candidates were those clones that showed a 3-fold difference in signal between the summed normalised data sets that also showed consistent duplicate signals in Cgen, i.e not one duplicate signal above the cloud, and one below the below the cloud. These were the highlighted spots in Figure 3.4. These clones were single pass sequenced to determine their identity. The table below shows the list of the sequenced differentially expressed genes and their difference in expression. Some of these clones were considered for further analysis. Cytokeratin 6A and 17, which have been previously
Chapter 3 BCC gene expression
described as being over-expressed in basal cell carcinoma, were also upregulated in the 2 BCCs examined (Stoler et al.„ 1988; Markey et al.„ 1992). Under-expression of Galectin-
7 (Magnaldo e ta l, 1995; Magnaldo et a l, 1998) was also seen. Of interest was the over
expression of the genes MRP 8 and MRP 14. These genes were represented multiple times on the filter and so were regarded as excellent candidates for further study. Many of the clones that were down-regulated in BCC were rRNA clones. The rRNA clones were due to contamination of non-transcript RNA in the cDNA synthesis and subsequent library construction. The signal of these clones would be depend on the degree of rRNA contamination in the mRNA prepared from each of the probes. It is possible there was a greater degree of contamination in the skin sample compared to both the BCC samples, making these clones appear more highly expressed in skin.