Article Text

Download PDFPDF

Differential expression of cyclooxygenase 2 in human colorectal cancer
  1. M HULL
  1. Division of Medicine and Molecular Medicine Unit
  2. University of Leeds, Leeds, UK
  3. Email: medmah{at}
  4. University of Birmingham, Birmingham, UK
  1. Department of Natural Science and Biomedicine
  2. University College of Surgery
  3. Faculty of Health Sciences
  4. Jönköping, Sweden
  5. Email: dija{at}
  6. Department of Biomedicine and
  7. Division of Cell Biology
  8. Faculty of Health Sciences
  9. Linköping, Sweden
  10. Email: petso{at}

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Editor,—We were puzzled by the recent paper by Dimberg and colleagues (Gut1999;45:730–732) which reported that upregulation of cyclooxygenase 2 (COX-2) protein expression was prominent in rectal adenocarcinomas compared with that in adenocarcinomas arising from the colon. “Low or undetectable levels of COX-2 protein expression” were demonstrated in 15 of 19 colonic adenocarcinomas located proximal to the rectum. Overall, upregulation of COX-2 protein expression was reported in only 56% of colorectal cancers.

Previous reports,1-7 which include one by the current authors on a not dissimilar case series,1 and two in the joint authorship of the accompanying commentary writer,2 ,3 have shown consistent upregulation of COX-2 expression in colonic and rectal adenocarcinomas (in 85–90% of cases) compared with matched normal colonic mucosa using different techniques, including northern blot analysis, RT-PCR, western blot analysis, and immunohistochemistry. Furthermore, four of these studies refer to the distribution of adenocarcinomas throughout the colon without showing evidence of differential COX-2 expression between rectal and more proximal tumours.1 ,2 ,4 ,5 In the one previous study which analysed COX-2 protein expression in human colorectal cancers by western blot analysis,7 immunoreactive COX-2 was detected in 76% of cases with a 10-fold increase in median tissue COX-2 concentration compared with normal colonic mucosa.

In our view, the authors should attempt to explain the discrepancy between their results and previously published data. It is interesting to note that, in the study of Kargman et al, five of six patients taking NSAIDs had low or undetectable COX-2 protein expression.7 Moreover, aspirin has recently been shown to suppress induction of COX-2 mRNA and protein in interleukin-1β and phorbol ester stimulated human endothelial cells and fibroblasts.8 Do the authors have data on NSAID use in their cohort of patients prior to surgery?



Editor,—We agree with Drs Hull and Langman that we found upregulation of COX-2 protein expression in a lower fraction of colorectal cancers (CRC) than previously reported. In part, this may simply be explained by the composition of different tumour types within CRC—that is, the number of colonic versus rectal tumours in our cohort compared with others. In the papers referred to it is difficult to assess the fraction of the different tumour types studied. The differences may also be dependent on the genetic basis for the CRCs studied, which we also have indicated in our report but perhaps not emphasised sufficiently. CRCs with a defective mismatch repair capability, recognised by microsatellite instability (MSI), are accompanied by reduced COX-2 levels.1-1 At present, we do not know the fraction of MSI type tumours in our series and therefore cannot assess this possibility. An indirect estimate may be achieved since the Min mouse model and human studies provide direct evidence that COX-2 expression may be related to loss of APC function.1-2 APC and β-catenin mutation analysis of our tumour series shows a good, although not perfect, correlation with COX-2 protein upregulation. Among 18/20 rectal tumours with COX-2 protein upregulation, 12 contained mutations in the APC/β-catenin genes. In contrast, only one of three APC/β-catenin mutated colon tumours revealed COX-2 protein induction and among the remaining 15 non-mutated tumours, two displayed COX-2 protein upregulation. Thus the fraction of APC/β-catenin mutated tumours was also slightly lower (21/38—55%) than previously reported and in accordance with the differential COX-2 induction observed. This may indicate that a larger fraction of CRCs in our cohort are of the MSI type.

Other possibilities for the differences in the fraction of COX-2 upregulation in our tumour series may be the definition of “induction”. In our case, a tumour/normal ratio from densitometric scanning of western blots must exceed 2 in two successive independent experiments of the same sample pair to be considered true induction. The use of different methodologies may also influence relative expression of COX-2, for example RT-PCR is sensitive to the quality of isolated mRNA and PCR is not really a quantitative method but needs to be carefully controlled to allow quantitative estimations.

It is also correct, as stated by Drs Hull and Langman, that NSAIDs may suppress COX-2 mRNA induction and COX-2 protein expression in some of the patients in our series. Normally, all drug treatments are withdrawn at least one week prior to surgery at our hospital, making most patients drug free at the time of surgery. However, we do not know if patients self-administer these types of drugs during the waiting period. We have reviewed the medical records for all CRCs included in the study and found that 7/39 patients were receiving aspirin for cardiovascular protection. Five of these patients had rectal tumours and displayed 7–40-fold induction of COX-2 protein: the two patients with colonic cancer and aspirin treatment revealed <2- and 10-fold induction, respectively. Thus COX-2 suppression caused by NSAID cannot explain the low prevalence of COX-2 induction in the colonic tumours.

Regulation of COX-2 is not fully understood. Because of the close correlation of upregulated COX-2 with mutations in APC/β-catenin genes it has been hypothesised that there is a regulatory link and that the chemopreventive effect of NSAIDs can be attributed to inhibition of COX-2. However, in a recent paper by He and colleagues1-3 it was demonstrated that PPARδ (peroxisome proliferator activated receptor δ) is a target of both APC and NSAIDs resulting in suppressed PPARδ activity and promotion of apoptosis. In addition, COX-2 null mouse embryo fibroblast cells remain sensitive to the antiproliferative and antineoplastic effects of NSAIDs1-4; hence there seems to be other important mechanisms for NSAID mediated tumour suppression.

The samples in our series were obtained consecutively without any selection. At present, we believe that the observed differential expression of COX-2 may be due to underlying differences in genetic alterations and/or that rectal tumours may represent a biologically distinct subtype of bowel cancer. However, we cannot exclude the possibility that the next 39 CRCs collected will display the opposite COX-2 expression pattern, although we believe this is unlikely.


  1. 1-1.
  2. 1-2.
  3. 1-3.
  4. 1-4.