Study Design and Sample Collection

Funding for this work had been secured prior to commencing the project; the initial 4 months were taken up with reviewing the literature which was carried out in 2 parts, a review of screening for CRC and SELDI-TOF as a platform for biomarker discovery.

After becoming competent at basic laboratory techniques (tissue culture, protein quantification, gel preparation, making stock solutions) and then attending a training course run by Ciphergen in order to use the machine, the main body of work was begun. During this time ethical approval was sought. Sample collection could then begin and work on tumour analysis and a method of analysing stool samples was developed.

Initially we hypothesised that the proteins that we might detect in the stool or in the serum may perhaps be present in the tumours themselves. Also, this was a way to gain experience with the technique and how to optimise conditions for analysis and use the statistics.

This revealed markedly different spectra between cancer and normal and following further review of the literature, it was decided to try and combine LCM with SELDI as this was a purer sample as there is no contamination by connective tissue and other cells. The literature has many studies comparing 5 tumours with 5 areas of normal tissue, the reason for this being that you need at least 2000 individually dissected cells to achieve an adequate amount of protein to get just usable spectra. This was prohibitive from a time point of view as to dissect 2000 cells for one spot takes many hours of work.

In designing this study sample collection and handling were considered carefully. When planning the number of patients and samples to include it was not possible to use a traditional power calculation as there is no available power calculation for SELDI profiling.

Sample size was justified based on the time available for sample collection and the number of cancers that were seen each week in Ninewells Hospital.

Stool sample collection proved one of the most difficult aspects of this project, collection was performed for the cancer group over a 10 month period and despite best efforts to recruit as many patients as possible, the initial anticipated number of 100 cancer samples was not achieved. Many of the cancer samples were collected pre-operatively when patients were undergoing bowel preparation. This resulted in a sample that was very high in water content and many of these contained too low levels of stool to give a usable profile and so were excluded from analysis. Other reasons for the failure to collect a larger number of stool samples included; samples wrongly being sent to the laboratories for microbiological testing for microbiological testing, patients forgetting to collect the samples and samples going missing from the ward.

The control stool samples were of higher quality in terms of the amount and were of normal consistency but as they were mailed in samples of fresh stool the age of these samples, what storage conditions and transport conditions were like for each was not known. None exhibited any macroscopic mould or bacterial overgrowth. Although this method of sample collection presented a greater degree of variation for each sample it is also the basis of the current screening programme and so was used as it was already known to be acceptable to patients. This time delay may account for some variation between the samples.

The control patients were recruited from the screening programme and so do represent a selected group of patients in that they were all recently FOB positive. Some patients

reported when they were consenting to take part in the study that they were reluctant to handle their bowel movements although this did not prevent anyone in the cancer group taking part in the study. The amount of stool returned by the control patients was in general very low, perhaps as only a small amount is required by the screening programme. In contrast the hospital acquired stool samples were usually larger, perhaps as these patients had been interviewed face to face and were undertaking bowel preparation.

It would have been challenging to collect fresh stool samples from each group, in particular normal controls and to immediately process and store them and this was considered to be beyond the resources available. The bowel preparation for colonoscopy within the screening programme is usually excellent and although taking the samples at this time was considered as a potential source of stool samples it would not have reflected a normal stool. The preparation of the samples for SELDI analysis was developed based on methods of stool preparation used for traditional proteomic faecal analysis as there was no manufacturer protocol or previously published method for this. Development of this was time consuming and initial methods of mixing stool in various buffers in much the same way as serum is handled were not successful at obtaining useable spectra.

The quality of the spectra obtained, in terms of the signal to noise ration and the number of peaks seen and the amounts of each is less clean than the profiles achieved when examining serum and this has perhaps resulted in only the largest and most detailed abundant peaks being examined. Perhaps with techniques of sample fractionation or pre analysis processing, cleaner peaks could have been achieved. In particular, the samples could have been dehydrated and then rehydrated as a more concentrated solution of protein.

It was possible to better control the variables in the collection of serum, the control samples came from patients undergoing screening colonoscopies. They were all fasted pre procedure, had undergone the same bowel preparation and had proven normal colons.

Fasting itself produces a metabolic response and this has been extensively studied in the medium and long terms; there may be changes in the serum proteome during the period that the colonoscopy group are fasting. Cancer serum samples were collected pre-operatively at pre-admission clinic or on the ward prior to theatre, some were fasted but most were not.

This level of homogeneity within the control group is perhaps a disadvantage as it does not reflect the real life conditions that a screening test would need to overcome in order to detect disease.