Typically observed immunohistochemical staining patterns for RUNX3 are shown in . Using a staining intensity of
2 as the threshold, RUNX3 expression was expressed in the nucleus of 631/849 (74%) tumours and in the cytoplasm of 431/849 (51%) tumours. Both nuclear and cytoplasmic expressions were present in 352 (41%) tumours, nuclear expression only in 279 (33%), neither nuclear nor cytoplasmic expression in 139 (16%) and cytoplasmic expression only in 79 (9%) tumours. In this study, tumours with ‘any nuclear' staining were considered to express ‘active' RUNX3 (n
=631, Group A) (Lau et al, 2006
). Tumours with no nuclear or cytoplasmic staining (Group B) or with expression in the cytoplasm only (Group C) were considered to have ‘inactive' RUNX3.
Representative images showing expression of RUNX3 in the nucleus only (A), in the cytoplasm only (B) and in both the nucleus and cytoplasm (C). Images are at × 40 magnification.
The associations between RUNX3 expression and clinicopathological and molecular features of the CRC series are shown in . No significant associations were seen with patient age or gender, or with the histological grade of the tumour. Tumours located in the proximal colon displayed a trend for less frequent expression of nuclear RUNX3 compared to those arising in the distal colon and rectum. A likely explanation for this finding is that CIMP+ and RUNX3
methylation are known to occur more often in the proximal colon (Weisenberger et al, 2006
) and are associated with methylation-induced transcriptional silencing. The MSI+ phenotype is also closely associated with the CIMP+ phenotype, thus accounting for the lower frequency of nuclear RUNX3 expression observed in MSI+ tumours ().
Associations between expression of RUNX3 and clinicopathological and molecular features of CRC
In keeping with its postulated role as a tumour suppressor, advanced stage tumours (AJCC stages III and IV) showed less frequent expression of nuclear RUNX3 compared to early stage tumours (). Kaplan–Meier analysis confirmed that patients with nuclear RUNX3 expression (n
=631, Group A) had significantly better survival outcomes (P
=0.025, logrank test) compared to the relatively small number of patients (n
=79, Group C) in which RUNX3 expression was restricted to the cytoplasm (). This result agrees with observations made in oesophageal cancer and is consistent with current understanding of the role of RUNX3 as a tumour suppressor (Sakakura et al, 2007
). RUNX3 expression in the cytoplasm has been reported by our group to represent mislocalised and probably inactive protein in gastric and breast cancers (Ito et al, 2005
; Lau et al, 2006
; Subramaniam et al, 2009
). The present results showing that cytoplasmic expression is associated with worse patient outcome support the contention that RUNX3 is in an inactive state.
Kaplan–Meier survival analysis for CRC patients according to the expression of RUNX3. (A) Any nuclear expression (n=631); (B) no nuclear or cytoplasmic expression (n=139); (C) cytoplasmic expression only (n=79). Logrank test: A vs C, P=0.025.
Interestingly, the subgroup of CRC patients with no or very low expression of RUNX3 in the nucleus or cytoplasm (Group B) showed similar prognosis to those with nuclear RUNX3 expression (Group A, ). This contrasts with several other cancer types (Araki et al, 2005
; Wei et al, 2005
; Sakakura et al, 2007
) and suggests that the presence or absence of RUNX3 may play a lesser role to its cytoplasmic localisation in determining clinical phenotype in CRC. It should be highlighted, however, that two of the earlier studies did not distinguish between nuclear and cytoplasmic staining (Araki et al, 2005
; Wei et al, 2005
In summary, the major findings of this study were that nuclear RUNX3 expression was reduced in advanced stages of CRC and that exclusively cytoplasmic expression of RUNX3 was associated with worse patient outcome.