We read with interest the recent research article by Lisa Simms and her colleagues (Gut 2008;57:903–10). While we agree with Simms et al that the NOD2 genotype is an important variable in the study of ileal Crohn’s disease (CD), we disagree with her conclusion that α-defensin expression is related to inflammation rather than NOD2 status. This genotype is especially important in the analysis of Paneth cell antimicrobial expression, given the prominent expression of NOD2 in Paneth cells.1 In a previous investigation, we reported a significant association of the NOD2 genotype with HD5 mRNA expression levels (Wehkamp et al,2 fig 1D), but, importantly, the effect was found only with the 1007fs mutation (SNP13). The decrease of HD5 levels with the 1007fs NOD2 genotype was confirmed at the protein level by western blot analysis (Wehkamp et al,2 fig 1E). The significance of the western blot data was emphasised by our observation that the same 1007fs samples had no decrease in levels of either sPLA2, lysozyme or α-1-antiprotease, three other Paneth cell proteins (Wehkamp et al,2 fig 1G). Of special note, we saw no association of either the R702W or G908R genotypes with HD5 mRNA levels (Wehkamp et al,2 fig 1D). In their recent paper, Simms et al did not stratify their genotype data to investigate if there was a difference in HD5 expression levels associated with any particular NOD2 genotype. Since eight patients in their cohort had a 1007fs mutation, we encourage the analysis of this relevant subgroup before concluding that the defensin decrease is independent of NOD2 genotype. In addition, the conclusion of Simms et al is somewhat incongruous with the reduced defensin expression in NOD2-deficient mice.3

A second point made by Simms et al also contradicts the published data from our previous studies.2 4 We were aware that tissue inflammation is an obvious potential confounding variable for any investigation of altered gene expression in CD. In our 2005 investigation, an experienced gastrointestinal pathologist (Dr Robert E Petras) evaluated histological sections of the specimens used for our Paneth cell expression analysis. To minimise any possible bias, the histological sections were labelled only with an arbitrary numerical identifier, and Dr Petras had no prior knowledge of other data on any of these samples (or data from other arms of the study). We found a significant association of histological inflammation assessed by Dr Petras with interleukin 8 (IL8) mRNA levels, our surrogate molecular marker of inflammation (Wehkamp et al,2 fig 2A), confirming the expected link between morphology and expression of this inflammatory chemokine. In contrast, and most significantly, we observed no association between inflammation and either HD5 expression levels or antibacterial activity in these samples (Wehkamp et al,2 figs 2B and 3). When comparing samples with little or no inflammation with those with either moderate or extensive inflammation, we found no association (Wehkamp et al,2 fig 2B). Moreover, in a 2007 study that was not cited by Simms et al,4 we found reduced expression of TCF-4 in ileal biopsy specimens from patients with ileal CD, and linked expression of this transcription factor to reduced HD5 mRNA (Wehkamp et al,4 figs 1A–C and 2). The levels of HD5 mRNA were also independent of IL8 mRNA levels in these experiments (Wehkamp et al,4 figs 4, 1C, and further data that were not shown).

To examine further if HD5 expression was independent of tissue inflammation, we sought to examine inflamed intestinal mucosa not affected by CD. Here, options are rather limited, since inflamed human ileal specimens are not readily available other than in CD. Therefore, in the 2005 study, we examined ileal mucosa from surgical pouches, with or without inflammation (Wehkamp et al,2 fig 2C and D, respectively). We saw no difference in HD5 mRNA levels when comparing pouchitis and normal pouch mucosal biopsies, supporting the conclusion that inflammation alone does not cause a decrease in HD5 expression. In summary, the data cited support the hypothesis that the reduced expression of HD5 in ileal CD is intrinsic to the disease and not a result of tissue inflammation present in the specimens. Furthermore, these data do not support the conclusion of Simms et al that levels of HD5 expression were dependent on tissue inflammation, and that the decrease in HD5 seen in CD was attributable to tissue inflammation, not disease status.

A possible confounder explaining the difference between the results of Simms et al and our findings is their surprising use of a mix of resection specimens and biopsies in the CD cohort, but only biopsies in the controls. We were intrigued by the data represented in the 3D plots showing the effects of inflammation (defined according to histology) and levels of HD5 and HD6 mRNA (fig 2A and B, respectively). It appears that that there is little or no effect of inflammation among the biopsy specimens (shaded), but striking differences in the surgical resection specimens (white). In contrast, our studies detected no effect of inflammation for either surgical resection2 or biopsy4 specimens. Simms et al did not account for the apparent differences in their two sources of specimens but, since essentially all effects of inflammation were observed in the surgical resection specimens, the group that also had reduced villin expression (fig 2C), a satisfactory explanation is warranted.

A hint as to why the data from Simms et al differ from ours is evident in their fig 1, which shows comparable villin expression (fig 1E), but significantly lower expression of lysozyme (fig 1B) and sPLA2 (fig C) in controls as compared with the Crohn’s samples (both NI and I subgroups). Since we agree with Simms that lysozyme and sPLA2 expression probably reflects Paneth cell mass, the lower expression in the controls would suggest that their approach might have systematically underestimated the Paneth cell gene expression, including defensins, in this group. A potential bias generating this systematic error may result from their method of data normalisation, using β-2-microglobulin (B2M) expression levels. This explanation is consistent with data from our own studies that show that seven other Paneth cell products (including lysozyme and sPLA2) are unchanged in CD compared with controls (Wehkamp et al,2 figs 1G, 3B, and Supplementary fig 5), while defensins were specifically reduced in ileal CD. An alternative explanation for the data from Simms et al is that a loss of surface epithelium leads to a relative enrichment of Paneth cells in the CD specimens (compared with controls) in fig 1, but this is unlikely because villin levels were unchanged. We suggest that it may be helpful to know the absolute transcript numbers for all of these gene products (including B2M and their other five candidate housekeeping genes) to clarify these issues further and allow a meaningful comparison with our data.

Therefore, we feel that the title chosen by Simms et al, ‘Reduced α-defensin expression is associated with inflammation and not NOD2 mutation status in ileal CD’, is an overstatement and not supported by their data. While we certainly value studies that rigorously test our proposed hypotheses on reduced α-defensin expression as a primary defect in CD, we are disappointed that Simms et al did not address numerous details that would help clarify why the findings of the two studies led to discrepant conclusions. We welcome further clarification of data in their study and look forward to the response to our letter.