Optimising microarray print quality

Generally the quality of printing by the Liverpool Microarray Facility (The University of Liverpool) was very poor and batches of slides were regularly discarded as they were unusable. Initially, test microarrays were prepared using cDNA amplified from 96 clones from the brain oligo subtracted library. The cDNAs were printed repeatedly on the same array which was then hybridised with brain cDNA from a pool of samples. Analysis of a

scanned slide at different levels of magnitude revealed various defects. Spot quality was poor at the top left of each replicate which was caused by a defective print tip (Figure 4.16.a i). Spots were also shown to be misaligned as shown by the kinks in the columns of spots (Figure 4.16.a ii). Although spot size was non-uniform (Figure 4.16.c iii) each replicate spot in the same position in each of the three replicates exhibited consistent size.

R ep lica te 1 R ep lica te 2 R ep lica te 3 i _ii iii

Figure 4.16.a. Test array, printed with 96 cDNAs viewed at three levels of magnification. Slide dimensions are 25 x 75 mm. At low

magnification (i) the poor print quality of the top left sub-array of each replicate is indicated by the arrows. Medium magnification of the array (ii) shows misaligned columns of spots indicated by an arrow.

Despite the imperfections in the spot printing, reasonable quality data could be expected from arrays of this standard. When the arrays were printed using the entire libraries, however, the print quality deteriorated further. Defects were diverse and included overlapping spots, inconsistent printing across the array, high or uneven background levels, and poor spot quality manifested as black-holes or doughnuts. Overlapping spots were found at the borders of adjacent sub-arrays (Figure 4.16.b). Spots with a doughnut

conformation were found on some arrays (Figure 4.16.c). This normally occurs when the cDNA solution spotted onto the microarray dries too quickly following printing.

Figure 4.16.b. An example of overlapping spots shown at the border between two sub-arrays as indicated by arrow.

Figure 4.16.c. An example of spots on the brain array with a doughnut conformation. The row indicated by the arrow has three such spots.

The uneven printing was observed as banding across each row of sub- arrays on hybridised slides (Figure 4.16.d). The bands of higher fluorescence are comprised of spots which either have more material printed per feature, or the features have different binding properties. An investigation was

undertaken to test whether the uneven banding shown in figure 4.16.d was due to the amount of cDNA printed per feature or if it could be attributed to variable binding characteristics across the array. A spot check was carried out to try and troubleshoot what the Liverpool Microarray Facility were doing wrong. A spot-check involves hybridising Cy3 labelled random 20-mers to a microarray, which will cause in turn cause all cDNA spots on the array to fluoresce. The array was not pre-hybridised before hybridisation to the fluorescent 20-mers and thus the array will show high background fluorescence. This is also useful as areas of differential background are highlighted. The scan of the spot-check array revealed a similar banding pattern to that shown in figure 4.16.d. Clear bands of low background fluorescence are visible on the scan of the spot-check array, indicated by arrows on Figure 4.16.e. These areas of low background contrast with the generally high background. These bands of low background fluorescence correspond to the areas of increased spot intensity shown in Figure 4.16.f.

Figure 4.16.d. Uneven microarray printing on the brain array. The array was co-hybridised with fluorescently labelled eel brain cDNA from the 5 month silver FW and 7 day yellow FW groups. Arrows indicate bands of spots exhibiting higher levels of hybridisation to labelled material.

Figure 4.16.e. Scan showing a single replicate from the spot-check array. A brain array, not subjected to prehybridisation, was hybridised with fluorescently labelled (Cy5) random 20-mer oligonucleotides. The arrows indicate four horizontal bands of low background which

correspond to the bands of high spot intensity shown in figure 4.3.d. At higher resolution, three bands per sub-array are visible (Figure 4.16.f). The top 4 rows show high spot fluorescence and low background. The central 11 rows of each sub-array have spots which fluoresce at a slightly lower level but have a high background. The bottom 5 rows of each sub-array have spots which show low fluorescence combined with a high background.

High spot fluorescence and low background Moderately high spot Fluorescence and high background

Low spot fluorescence and high background

Figure 4.16.f. Sub-array from the brain array subjected to the spot- check. Three levels of print quality are visible

We communicated our findings to the Liverpool Microarray Facility who revealed that arrays were printed over the course of three days and that the three levels of banding would correlate to their spots printed on each of the three days. The disparity of quality between spots printed on separate days was attributed to the gradual degradation of the lysine coating of the slides which occurs once they are removed from the packaging which contains a protective atmosphere. Subsequently the arrays were remade with all spots being printed in a single day which improved quality to an adequate level (figure 4.16.g) and these arrays were used in the final hybridisation

experiments. As a consequence of the various defects found with the printing the brain arrays were reprinted four times and the SSH arrays more than 10 times. Despite repeated attempts to aid the Liverpool Microarray Facility with troubleshooting problems the printing was never optimised. To keep the project progressing, sub-standard arrays were used which compromised the experiments as many spots were completely absent, reducing both the quantity and quality of data.

Figure 4.16.g. A brain oligo array exhibiting relatively even fluorescence across the slide. The array was co-hybridised with fluorescently labelled eel brain cDNA from the silver 5 month FW acclimated group and the yellow 7 day FW acclimated group. The array was printed in one continuous run on a single day.