2.2.1.1 DNA Comparative Genomic Hybridization (CGH)
DNA CGH experiments were all performed using a common reference design with TIGR4 as the reference pneumococcal genome. Each experiment included fluorochome labelling
of DNA with a dye swap step so that for each test isolate of pneumococcal DNA (biological replicate), two microarrays were hybridized (technical replicates) – one with the test DNA labelled with fluorochrome Cy3 (GE Healthcare, United Kingdom) and one labelled with Cy5 (GE Healthcare, United Kingdom) (Churchill, 2002) as shown in Figure 2-1.
Figure 2-1 Results of microarray CGH dye swap experiments for isolate 06-1805 (ST227). In both microarrays illustrated above, DNA from the test clinical isolate 06-1805 (used in Chapter 10) is competitively hybridized with DNA from the laboratory reference strain, TIGR4. On the left, the TIGR 4 genes are labelled with the fluorophore Cy3 and 06-1805 genes are labelled with the fluorophore Cy5. This labelling is reversed on the microarray on the right. When genes are present in TIGR4 but not the test clinical isolate they appear as red dots (left) or green dots (right). When genes are present in 06-1805 but not the TIGR4 genome they appear as green dots (left) and red dots (right) – these genes are identifiable in the laboratory reference strain R6 but are absent from TIGR4. Where genes are present in both 06-1805 and TIGR4 there is competition to hybridize on the array and the dots appear yellow.
For each microarray, one Cy3 labelled DNA sample and one Cy5 labelled DNA sample were prepared by heating (for 5 minutes at 95°C) 5µg of DNA and 1µl of random primers (Invitrogen, United Kingdom) made up to 41.5µl with nuclease free water (Ambion/ Applied Biosciences, United Kingdom). This was then snap cooled on ice and briefly centrifuged. To each was added 5µl 10xREact 2 buffer (Invitrogen, United Kingdom), 1µl dNTPs (5mM dATP, 5mM dGTP, 5mM dTTP and 2mM dCTP), 1.5µl Cy3 or Cy5 dCTP (GE Healthcare, United Kingdom) and 1µl Large fragment DNA Polymerase I (Klenow) (3-9U/µl) (Invitrogen, United Kingdom) and the solution was incubated at 37°C in the dark for 90 minutes in a Techegene thermal cycler (Bibby Scientific, United Kingdom).
Microarrays were prepared by soaking in a prehybridization solution preheated to 65°C (composed of 8.75ml 20xSSC (Ambion/Applied Biosciences, United Kingdom), 250µl
20% SDS (Ambion/Applied Biosciences, United Kingdom), 5ml (100mg/ml) Bovine Serum Albumin (Sigma-Aldrich, United Kingdom) and made up to 50ml of sterile double distilled water) for 20 minutes in a Coplin jar (Fisher Scientific, United Kingdom) and placed in a Techne Hybridizer HB-1D (Bibby Scientific, United Kingdom). Prehybridized arrays were rinsed in 400ml double distilled water for 1 minute and then in 400ml propan- 2-ol (VWR International, USA) for 1 minute. Each array was placed in a 50ml centrifuge tube and centrifuged at 1500 rpm for 5 minutes then stored in a dust free box until ready for hybridization.
Cy3 and Cy5 labelled DNA samples were combined into a Qiagen Minelute purification column (Qiagen, United Kingdom). 500µl of Buffer PB was added and centrifuged at 13000 rpm for 1 minute (Eppendorf centrifuge 5417C, USA). The flow through was discarded and 500µl of Buffer PE was added and centrifuged at 13000 rpm for 1 minute. Again the flowthrough was discarded then 250µl of Buffer PE was added to the same column and centrifuged at 13000 rpm for a minute. Flow through was discarded and then the column was spun again for 1 minute at 13000 rpm then placed in a fresh 1.5ml collection tube. 15.9µl of nuclease free water (Ambion/Applied Biosciences, United Kingdom) was added to the membrane and centrifuged for 1 minute at 13000 rpm to elute DNA for hybridization.
14.9µl of the Cy3/Cy5 labelled DNA sample was added to 4.6µl of filtered 20xSSC and 3.5µl of 2% SDS and the resulting solution heated for 2 minutes at 95°C in a Techegene thermal cycler (Bibby Scientific, United Kingdom). Lifter slips (Erie Scientific Company, USA) were placed on the pre-hybridized microarrays and the DNA solutions pipetted onto the bottom left corner to allow the solution to be drawn across the microarray by capillary action. Prepared microarrays were placed in a hybridization cassette, sealed and then submerged in a water bath at 65°C in a Techne Hybridiser HB-1D (Techne, USA) in the dark for 20 hours.
After hybridization, arrays were transferred to a slide rack and washed with agitation in a pre-heated solution of 20ml 20xSSC, 1ml 20% SDS made up to 400ml with sterile double distilled water for 2 minutes and then transferred to a further solution of 1.2ml 20xSSC made up to 400ml with sterile double distilled water and washed with agitation for 4 minutes. Microarrays were then placed in 50ml centrifuge tubes and dried by centrifugation at 1500 rpm for 5 minutes. The hybridized microarrays were then scanned using ScanArray Express ™ (Packard Biosciences Biochip Technologies, Perkin Elmer).
2.2.1.2 RNA Comparative Hybridization
RNA expression experiments were also performed using a common reference design using TIGR4 RNA grown to midlog phase (Conway and Schoolnik, 2003). Unlike the DNA CGH experiments, 3 biological replicates of each test isolate RNA (i.e. three independent RNA extractions from different broth cultures) were hybridized against TIGR4 RNA (from the same batch of TIGR4 RNA) with only one technical replicate per biological replicate as required for statistical validity (Foster and Huber, 2002). The choice of using an RNA reference control grown to midlog phase was made as this is an established means of determining a baseline for gene expression. It is appreciated that this is not the only possible method and that alternatives such as using genomic DNA or a mixture of reference RNA from several sampling conditions have also been used (Conway and Schoolnik, 2003).
For each microarray, one Cy3 labelled cDNA sample (2-10µg) and one Cy5 labelled cDNA sample (2-10µg) were prepared by heating for 10 minutes at 70°C with 1µl of random primers (Invitrogen, United Kingdom) and made up to 11µl with nuclease free water (Ambion/ Applied Biosciences, United Kingdom). This was then snap cooled on ice and briefly centrifuged. To each was added 5µl 5xFirst strand buffer (Invitrogen, United Kingdom), 2.5µl DTT (100mM), 2.3µl dNTPs (5mM dATP, 5mM dGTP, 5mM dTTP and 2mM dCTP), 1.7µl Cy3 or Cy5 dCTP (GE Healthcare, United Kingdom) and 2.5µl SuperScript II (200U/µl) (Invitrogen, United Kingdom) and the solution was incubated at 25°C in the dark for 10 minutes then 42°C for 90 minutes in a Techegene thermal cycler (Bibby Scientific, United Kingdom).
Microarrays were prepared in an identical way to that described above in Section 2.2.1.1. Cy3 and Cy5 labelled DNA samples were combined into a Qiagen Minelute purification column (Qiagen, United Kingdom). 250µl of Buffer PB was initially added and centrifuged at 13000 rpm for 1 minute (Eppendorf centrifuge 5417C, USA). The rest of the procedure for preparing the hybridization solution, hybridizing the microarray and scanning is identical to that described in Section 2.2.1.1.
2.2.1.3 Microarray Normalization for DNA Comparative Genomic
Hybridization
Tagged Image File Format (TIFF) images of the scanned microarrays created by ScanArray Express ™ (Packard Biosciences Biochip Technologies, Perkin Elmer) were
entered into Bluefuse for Microarrays 3.5 © with the Cy3 labelled image in Channel 1 and the Cy5 labelled image in Channel 2. The array gridmap files provided by the Bacterial Microarray Group at St George’s Hospital, University of London (BµG@S, United Kingdom) were utilised. The post processing protocol was devised by Dr Jason Hinds of BµG@S and comprised of initial exclusion of unreliable results due to poor quality hybridizations with a confidence estimate of less than 0.1. Controls spots on the array were identified using an array gridmap GenePix Array List (GAL) file (SPv1_1_0_CGH_Gridmap.bcf) and data pertaining to control spot hybridizations was automatically removed from the analysis. To correct for spatial, intensity and dye related effects, normalization was performed using the option, “Global Lowess excluding all with text.” Confidence flags were set at their default settings. Replicates of each dye swap were combined by fusion.
CGH was performed in Bluefuse for Microarrays 3.5© using a protocol devised by Dr Jason Hinds through the identification of a normal distribution of experimental variability and by identifying variability which was two standard deviations from the mean of this normal distribution for all the results for the TIGR4 and R6 genes represented on the microarray. Automated classification of regions of variability was performed by setting a ratio threshold for amplification as 1.0 and ratio threshold for deletion at -1.0 with the minimum number of clones included in the region in order for it to be classified as an amplification or a deletion set at 1. Dye swap processing was enabled.
Data analysis was completed in Genespring GX 7.3.1 (Agilent Technologies, USA) again using protocols devised by Dr Jason Hinds. Output_fused_CGH files were imported into Genespring GX 7.3.1 and further normalization was performed after data transformation to account for dye swaps. This normalization was performed using the, “Per spot and divided by control channel,” protocol with a cross gene error model using the error model for one- colour data. The error model was based on deviation from 1. The generation of gene lists using Genespring GX 7.3.1 (Agilent Technologies, USA) was accomplished by importing the Bluefuse for Microarrays 3.5© generated output_fused.xls files to create an experiment whereby the microarray dye swaps for each strain could be analysed using the, “Filter on data file,” option. To generate each gene list, a search was performed using the, “Type,” column employing the search criteria, “Column values must be not equal to NO CHANGE,” and “Value must appear in at least 1 of the selected columns.” The resulting gene list could be saved or exported into Microsoft Office Excel 2003, Microsoft®, United Kingdom for comparison with further strains.
2.2.1.4 Microarray Normalization for RNA Expression
For RNA expression experiments, normalization was performed (again using protocols written by Dr Jason Hinds) by importing the Output_fused.xls files into Genespring GX 7.3.1 (Agilent Technologies, USA) for the 3 biological replicates of each isolate. When TIGR4 cDNA was labelled with Cy3 an initial “dye swap” normalization step was used for RNA expression experiments. An additional, “Per gene,” normalization step was applied to specific samples (serotype 1 isolates) where the isolates being compared involved more than one MLST of S. pneumoniae to take into account clustering by MLST rather than by the clinical condition being investigated.
Statistical analysis of RNA expression data generated by Genespring GX 7.3.1 (Agilent Technologies, USA) was performed using the statistical analysis (ANOVA) tool. This performs a 1-way parametric test without assuming variances are equal. The false discovery rate was set at 0.05 resulting in a false discovery rate of about 5% of genes. Multiple testing correction was performed using a Benjamini and Hochberg False Discovery Rate. No post hoc tests were used.
The class prediction function on Genespring GX 7.3.1 (Agilent Technologies, USA) was used to generate lists of genes which were predictive of invasiveness, brain abscess or complicated pneumonia respectively, with a predictive strength calculation based on Fishers Exact Test. These lists were then imported into Microsoft Office Excel 2003, Microsoft®, United Kingdom for comparison.