Ecuación Planteada para el Modelo de Gestión para la toma de

The creation of two DNA arrays for the analysis of KSHV gene expression adds to the growing number of viruses for which this technology has been applied (Chambers et al, 1999, Bresnahan and Shenk, 2000, Stingley et al, 2000, Tenner et al, 2001, Paulose- Murphy et al, 2001, Ahn et al, 2002). The two arrays are designed for different tasks. The KSHV array provides a means for the simultaneous analysis of expression of almost every virus gene. This will greatly increase the number of genes whose time of expression is known and provide a window through which one can observe virus genomic activity. The KSHV-human microarray, containing probes for over 5000 human genes, is the first array allowing simultaneous global measurements of gene expression for both host and virus. The results from such studies will reveal host-pathogen interactions that may underlie disease pathology. The microarray also includes probes for EBV, allowing virus gene expression measurements from EBV infected cells and, therefore, from each of the three genomes contained within the majority of PEL cells.

3.3.1 The sen sitiv ity an d dynam ic ran g e of bo th KSHV a rra y s c o m p are well with o th e r array sy s te m s

Both the KSHV array and KSHV-human microarray comprise purified DNA probes produced from sequence-verified clones. The vast majority of probes are specific by in silico prediction and by experimental testing against human RNA. In addition, signals from negative control elements on each array are consistently absent. Such tests are vital for allowing gene expression measurements from arrays to be trusted (Schena et al,

1995, Lockhart et al, 1996, Schena et al, 1996, Stingley et al, 2000).

Both arrays are able to detect low copy-number transcript controls. The KSHV array can detect RNA spiked into total RNA at 1:1,000,000, equivalent to around 10 copies/cell. The KSHV-human microarray can detect RNA spiked into mRNA at 1:10,000, equivalent to around 6 copies/cell. Expressed as absolute measurements of RNA abundance, both arrays are of equivalent sensitivity (around 1-3x10^ RNA molecules). These limits are in the ranges previously reported for microarrays (Schena et al, 1996, Wilson et al, 1999) and Affymetrix GeneChips (Lockhart et a l, 1996). Using spiked control RNA, Schena et al, (1996) reported a sensitivity of 1:500,000 in total RNA by weight and Lockhart et al, (1996) reported a sensitivity of 6x10^ RNA molecules. This

sensitivity for low copy-number RNA allows the detection of latent KSHV transcripts (Chapter 4). Analyses of variation between experimental replicates in the context of signal strength shows weak signals tend to contain more experimental noise. This is in agreement with previous findings (Cohen et al, 2000, Geiss et a l, 2000, Ross et al, 2000). Filtering the data by signal to background ratio removes these experimental sources of variation thus increasing the reproducibility between gene expression measurements.

Hybridisation efficiency is a function of the concentration of probe DNA in the spot and concentration of labelled target cDNA in solution. Experiments with the KSHV array show that it is the concentration of labelled cDNA in solution that is limiting. A linear increase in signal over a range of 3 logs is observed for a 4-log increase in labelled cDNA without saturation. This allows measurements of gene expression independent of probe concentration (Duggan et al, 1999). Together with the excellent correlation in the amount of probe spotted between arrays, this allows reproducible measurements of gene expression using the KSHV array.

Both KSHV arrays show proportional increases in signal as RNA concentration increases. However, the phosphor imager used to scan the nylon membrane KSHV arrays can detect RNA concentrations at 1:120 (around 80,000 copies/cell) with signal intensities two logs below the saturation limit, whereas the Axon array scanner saturates at concentrations of 1:100 (around 2000 copies/cell). This does not limit the detection of human RNA transcripts, as only 0.5% of elements saturate, consistent with known RNA distributions in human cells (Lewin, 1992). The dynamic range of the KSHV array allows the quantification of the high concentrations of T l.l RNA that occur in cells undergoing lytic replication (above 10,000 copies/cell; Staskus et a l, 1997, Zhong and Ganem, 1997, section 4.2.2). Comparison of data from each array hybridised with biological replicates of the same sample shows that detection of (3-radiation on nylon membrane offers a greater signal to background ratio and a greater data range than detection of fluorescence on glass. The greater signal to background ratio may be because of the increased surface area of nylon membrane. The difference between the dynamic ranges is a reflection of the dynamic range of the PMTs in the two scanner types. Even with these caveats, the data produced by the two arrays are highly correlated. This has a number of implications. It suggests firstly that the data produced by both

arrays are true representations of actual transcript abundance and secondly, that results from different array systems are cross-comparable.

3.3.2 Experim ental replication d e m o n s tra te s iow variability betw een e x p re ssio n m e a su re m e n ts

The KSHV arrays show reproducible measurements of expression both within arrays and between arrays. The variation in signal between duplicate spots is ±3.7%. This compares well with the variation of 3.4% previously measured between duplicate spots printed on membrane arrays (Chen et al, 1998a). The average correlation between technical replicates for the KSHV array and the KSHV-human microarray are 0.94 and 0.96 respectively. For biological replicates this drops to 0.92 and 0.90 respectively. These compare favourably with correlations between previously published technical replicates (0.87 (DeRisi et al, 1997), 0.96 (Cohen et al, 2000), 0.84-0.86 (Mayne et al, 2001), 0.91 (Ahn et al) and biological replicates (0.83-0.92 (Ross et al, 2000), 0.97 (Scherf et al, 2000), 0.90 (Zhang et al, 2001b). As also shown by others, low variability between experimental replicates creates unique expression profiles that can be used to identify samples (Alizadeh et al, 2000, Pérou et al, 2000, Ross et a l, 2000).

The experiments discussed in this chapter provide controls for the biological interpretation of the array-based expression measurements within the following chapters. The KSHV array was used to examine KSHV gene expression during latency and after induction of lytic replication in the PEL cell line BC-3 (Chapter 4). The KSHV-human microarray was used to examine to cellular expression profile of PEL in comparison to other B-cell lymphomas and to find correlations between KSHV and host gene expression (Chapters 4 and 5).