Infrastructure and Technology overview

It is unfortunate that unless there is a specific change of note, the literature assumes prior knowledge of the various techniques and technologies and an appreciation of the relevant issues influencing data generation and presentation. This includes knowledge of apparently peripheral issues such as the time between specimen retrieval and processing (fresh [6] vs autopsy specimens [14]), sample processing (paraffin embedded vs frozen [10]), the advantages and disadvantages of various analytical systems, and the choice of the statistical platform used to determine statistically significant changes from the 'norm¹. Another is the choice of which data to present. For example, cMYC at 8q24 and p53 at 17p13 are examples of alterations to regions and genes that appear quite frequently (>20%) across many different study samples and across a broad range of cancer grades. Examination of this subset of patients with regard to other clinically useful molecular or physiological similarities (such as age, PSA or ethnic origins) has not uncovered any other obvious common denominator. Unfortunately, it should also be noted that most publications do not list all the clinical information for each specific patient making it impossible to attribute the true worth of such data. How then to know whether this seemingly discordant data be regarded as experimental artefacts, a result of genetic drift in the different sample populations, or an indicator of another element/process of the disease?

The pros and cons of established DNA techniques such as Comparative Genomic Hybridisation (CGH), Southern Blot hybridisation (SB), Polymerase Chain Reaction(PCR) and in situ Hybridisation (ISH) are relatively well known and documented (see Glossary). With the exception of CGH, these hybridization- based techniques have two common major limitations. The first, primer/probe selection, is controlled by a number of factors including identification of unique DNA sequence, DNA secondary and tertiary structures, length of primer/probes and melting temperatures. A number of software programs exist that can assist in the choice and construction of the primers/probes, but they are not infallible nor do they always agree. The final choice is still reliant on the experience of the investigator. This selection is crucial as it is a major factor determining the theoretical level of hybridization that can be attained between the sample DNA and the probe. Given the optimum choice it is possible to adjust the experimental parameters to where the difference of a single base pair between the probe and the sample DNA results in non-hybridisation. However, this is rarely seen or attainable. Thus for any given region there can be as many different primers/probes as publications, each with its own unique set of possible outcomes. This can lead to apparently contradictory data, exemplified by the numerous publications debating the correlation of p53 tumor suppressor gene mutations and prostate cancer progression [32 - 39].

The second major limitation is the number of primers/probes used. Time, laboratory economics, and the quantity and quality of sample DNA available for investigation all govern the depth and breadth of genomic investigation. In short, only a relatively small selection of regions can be studied at any one time. Tissue microarray (TMA's) have been developed to alleviate some of the problems governing these issues. Unlike transcriptional microarrays, TMA's adhere patients¹ samples to an 'array¹ and then introduce the probe to the array of sample tissues. In the case of Bowen et al [40] for example, cores tissue biopsies were removed from the selected patients' paraffin blocks and arrayed in a new recipient paraffin block. This 'array¹ of tissue samples is then investigated using an appropriate technique such as immunohistochemistry. The volume of data that can be attained from a single experiment is dramatically increased however; it still has the limitations of whatever problems are inherent with the investigative technique that includes the probe/primer selection and number that can be studied at any one time. This is perhaps one of the more attractive feature of CGH; the use of the entire genome as a probe enabling scrutiny of the entire genome simultaneously with every region informative. A relatively high degree of hybridisation can be attained enabling the detection and mapping of single allele copy losses and gains. It cannot however, investigate internal small-scale mutations- the provenance of techniques such as PCR and SB.

The limitations listed for the genomic DNA techniques also apply to the older RNA techniques such as Northern blot hybridization (NB cf glossary) and Immuno-histochemisty (IH). As with TMA¹s, cDNA and oligonucleotide microarrays and their variations have been developed to mitigate some of these limitations. Simultaneous detection of thousands of gene transcripts is possible because the technique attaches the different probes to the array and then hybridises the sample material to the arrayed probes. Luo et al microarray study [11] for example introduces the patient sample material to a microarray of 6500 probes representing 6112 unique genes (4573 of which have been characterized to some extent). The benefits are obvious. Not so obvious however, are the technology issues. For example there is the physical limitations associated with the microarray template manufacture. The material upon which the probes are attached and how they are attached, dictate the number of patient samples that can be tested per array. This in turn can create an economic issue; as one of the most expensive techniques currently in use, the price is generally prohibitive for a large sample number.

Another aspect is commercial versus unique microarray templates. Commercial arrays generally have a pre-determined selection of probes (either oligonucleotides or cDNA) and this selection is specific to each company. Generally there are a number of topical regions that will be represented on all the arrays, but the number relevant to prostate oncogenesis is not guaranteed. The study by Stamey et al [8] for example used a HuGeneFL oligonucleotide array representing 6800 genes (1 probe =1 gene), however the number of relevant genes to prostate cancer oncogenesis is unclear as the details of every probe is not listed. Groups such as Bull et al [13] and Dhanasekaran et al [5] have 'created' their own microarrays by providing all the primers/probes to be adhered to the microarray. This has the advantage of ensuring that all regions represented have a known or suspected affiliation with prostate cancer oncogenesis, but this can also be considered a disadvantage. With so much yet to learn about oncogenesis, a shotgun approach to the entire genome can have its benefits.

Further, the selection of the primer/probes to be used on the microarray is still liable to all the issues discussed above for the selection of primer/probe with genomic DNA studies. Another factor is the choice of sample genetic material. This can be either labeled cDNA as in Luo's study [11], or labeled cRNA created in a further PCR process from cDNA as used by Stamey [8 cf glossary]. How well the samples material will hybridize with the probes of the arrays can be determined by pre-testing. For example in the Stamey study the sample transcripts (cRNA) were pre-tested using another commercially available kit (GeneChip Test3 Array (Affymetrix)) to determine the ratio of 3' and 5' glyceraldehydes 3-phosphate dehydrogenase (GAPDH) and to assess which of the original 22 patient samples collected could produce a detectable signal for at least 40% of the transcripts represented on the array. This does not appear to be a particularly common practice, or perhaps represents yet another detail of presumed knowledge.