Sensitivity and absolute quantification

In a previous study molecular quantification (218) of uidA genes in CD ileal and colonic tissue generated 0.5-5x106 uidA genes per million human cells in ileal biopsies and approximately 1.25x104 cells in colonic tissue, with 0.75x104uidA genes in healthy tissue. The biopsies used in this work, which also involved mucus removal, unlike the previous study (218), overall produced an average of 187000 18S rRNA genes per biopsy, or 937 human genomes (cells) per biopsy (as there are 200 18S rRNA genes per human genome). Basing calculations on the results of (218) which they expressed as uidA genes /million human cells, one would expect 469–4690 uidA

genes from DNA extracted from ileal tissue and 11.7 uidA genes per biopsy of colonic tissue, with 7 uidA genes in control tissue for average biopsy size. When human biopsy tissue had been analysed, standard biopsy mass (wet weight) was found to be

between 8-115 mg at the extremes of sampling in clinical trials evaluating CD and control tissue biopsies (n=54, mean mass 48.4 ± 3.8 mg SEM (Chapter 7).

In terms of quantifying bacteria using CFU, the numbers likely to be of interest in CD, identified from previous laboratory work were 0 – 3000 CFU per biopsy (232) but these figures include all aerobic bacterial species cultured and not simply the E. coli identified specifically by the uidA gene.

In one study looking at uidA genes in twins (226), the uidA gene numbers are expressed per 106 16S rRNA genes, at levels of log 1.3-4.5 (or 20-31600), but the authors state that this is when E. coli was detectable, and that levels of E. coli were significantly higher in patients with ileal CD than colonic CD. In support of this work however, the authors also conclude that when E.coli was detectable for either colonic or ileal disease it was present at all areas of the gastrointestinal tract (ileum, ascending colon, transverse colon, descending colon and rectum). This serves to illustrate that often different studies use a different denominator, which makes direct comparison difficult for microbial ecology and absolute quantification, and also suggests that the

uidA genes were sometimes undetectable. Without knowing the average 16S rRNA gene numbers present in a biopsy in CD tissue, it was difficult to know what E. coli

numbers to expect and plan assays for. In healthy control tissue in the same (226) study uidA genes were detected but only, where detected, at levels of log 1.2 -2.5 (or 15.8- 316) per 106 16S rRNA genes. Significant variance was clearly seen as in one twin pair with ileal CD, one individual did not have uidA genes identified at all. 16S rRNA gene copies were also expressed in the twin study per β-globin gene present. Whilst it is valid to use a housekeeping gene, the differing denominators for each gene quantity expressed make direct comparison difficult, and ultimately one of the

improvements of the assay in this work is that it could be directly compared with most studies as each component examined was independently absolutely quantified.

Utilising FISH gives at best semiquantitative answers. One study (311) produced FISH images for various categories of mucosal bacterial concentrations, but in the results the authors remarked that accurate quantification could not be achieved, particularly in the tissues with high bacterial concentration. In the same study 16S rRNA genes were looked for with 103-106 CFU/l found, but not in all patient groups and generally in less than 83% of patients. Analysis of the spatial arrangement of bacteria and a visual representation of their abundance is certainly possible (218), but the data in the Swidsinski study are not obtained using absolute quantification, and are categorical data. This data gives only a broad indication of potential bacteria present (<1000, 1000-10000, and >10000 CFU/µl). Other authors have attempted to quantify

E. coli and other bacterial numbers by specifying the % of biopsies examined with the specified bacterial species present within the tissue, without quantifying the actual quantity within each biopsy (200). This is also therefore a categorical measure. Eventually it was decided that the best solution was to aim for the maximum sensitivity possible for an absolute quantification assay (three bacteria), but that acceptable figures would be those approximating 1000 bacteria with acceptable reproducibility or CV (aim <20%). In terms of PCR performance targets were efficiency between 90-110%, error <0.2 and correlation coefficient >0.95, which were all eventually achieved. This was very difficult and pushed the limit of detection for qPCR to the extreme.

The first quantification method resulted in a very poor standard curve. This may have been due to the crude method of DNA extraction used, which would have given even lower yield when each bacterial suspension was diluted. The fact that CFU

counting can be inherently variable did not help this attempt at the assay. The assay was clearly improved using method two when an efficient commercial kit for DNA extraction was used, but this still resulted in an unacceptable assay that was not sensitive enough, with poor efficiency and error, key factors in PCR.

With a switch to purely molecular techniques there was removal of some of the variability seen when using methods involving CFU counting. The published genome size of E. coli, along with accurate quantification of extracted DNA using the Nanodrop ™, allowed good approximation of bacterial quantities, with a technique similar to (306). Unfortunately, this ultimately produced an assay with unacceptable error despite good efficiency and lower limits of detection (3530 copies).

6.6.2 The successful development of a PCR method for bacterial quantification in