The expression of cytokeratin 5/6 in invasive lobular carcinoma of the breast: evidence of a basal-like subset?

Analysis of gene expression profiling data on breast cancers has revealed "molecular subclasses" that may have prognostic significance. The "basal-like" breast cancers, one of these molecular subclasses, have been associated with a significantly worse overall and disease-free survival as compared with most of the other subclasses. Previous studies on basal-like cancers have been performed predominantly on the ductal histotype. This study was designed to evaluate the significance of the expression of cytokeratin (CK) 5/6, a commonly used surrogate marker for the basal-like phenotype, in invasive lobular carcinomas (ILCs). The immunohistochemical expression of CK5/6, estrogen receptor (ER), progesterone receptor (PR), HER2/neu, and E-cadherin was determined in a group of 82 consecutive archived ILCs diagnosed in 82 women (age range, 29-73 years; mean, 51.9 years). All cases were E-cadherin negative. CK5/6 was positive in 14 (17%) of 82 cases and was entirely negative in the remaining 68 cases (83%). In 8 of the 14 CK5/6[+] cases, staining was diffuse and intense. In the remaining 6 cases, staining was patchy (>1 low-power field between positive areas) but still of high intensity. CK5/6[+] cases were significantly more likely than CK5/6[-] cases to be ER[-] (43% versus 0%, respectively, P < .0001). CK5/6[+] cases were also significantly more frequently of modified Scarff-Bloom-Richardson histologic grade 3, as 7 (50%) of the 14 CK5/6[+] cases were of histologic grade 3, as compared with only 6 (8.8%) of 68 of the CK5/6[-] cases (P = .0009). Notably, the average mitotic index in the CK5/6[+] group was 11/10 high-power fields, as compared with 7/10 high-power fields in the CK5/6[-] group (P = .07). Overall, there were no distinct morphological differences between the 2 groups, and both displayed the well-characterized architectural and cytologic features of ILCs. CK5/6[+] and CK5/6[-] cases did not significantly differ with respect to patient age, frequency of PR expression, tumor size, rate of axillary node involvement, or HER2/neu overexpression. In summary, the present study demonstrated that 17% of ILCs express CK5/6, and that CK5/6[+] cases are more likely to be ER[-] and have a high modified Scarff-Bloom-Richardson histologic grade. Because these findings are a characteristic of ductal basal-like breast cancers, our results suggest that there is a basal-like subset for ILCs with potentially distinct clinicopathologic characteristics. Future studies are required to define the prognostic significance of CK5/6 expression in ILCs.