Dr Yovtchev from our group studied the CK 19 in breast cancer. You can see the study results here:

https://www.researchgate.net/publication/232720872_Epithelial_cells_expressing_cytokeratins-19_and_bone_marrow_micrometastases_in_patients_with_breast_cancer_at_the_time_of_primary_surgery_clinical_outcome_during_long-term_follow-up?ev=prf_pub

Article Epithelial cells expressing cytokeratins-19 and bone marrow ...

Cyytokeratin-specific immuohistochemical stains have been used in histopathology for quite sometime to aid diagnosis, for example to discern if cells are indeed epithelial in origin, or in the case of breast cancer ,whether they are of the luminal or basal-like subtypes. Recently overlap has been discovered between some members with the polar opposite molecular subtypes in breast cancer - luminal and basal-like. However, when compared to the gold standard definition of molecular subtypes using gene expression microarray expression signatures (either Intrinsic gene list or PAM50), they have not been found to be specific enough as luminal cytokeratins are expressed in basal-like breast cancers as well, and not all basal-like breast cancers express basal cytokeratins. Thus cytokeratin expression patterns in practice cannot be considered to be characteristic or pathognomonic of either subtype, but rather can be said to be highly associated with each entity. Basal-like cytokeratins DO possess prognostic significance, but less than more specific biomarkers for basal-like breast cancer such as FOXC1, which have been shown to be superior prognostic markers on the basis of comparing prognostic models using either cytokeratins or FOXC1 [Ann Surg Oncol. 2011 Dec;18(13):3839-47. doi: 10.1245/s10434-011-1657-8. Epub 2011 Mar 18.

Basal-like breast cancer defined by FOXC1 expression offers superior prognostic value: a retrospective immunohistochemical study. Ray PS, Bagaria SP, Wang J, Shamonki JM, Ye X, Sim MS, Steen S, Qu Y, Cui X, Giuliano AE].

Dr Ray is right gene expression profiling is a better way to classify mammary tumours into different molecular sub-types to access the prognosis. However, all the laboratories may not be equipped to carry out gene expression profiling. Alternatively an extended surrogate panel of antibodies are used to study tissue expression of various markers by IHC viz. hormone receptors,Her -2 and cytokertins, EGFR etc to classify mammary tumours into molecular sub-types.The tumors which are negative for hormone receptors and Her 2 are subjected to cytokeratin 5/6 and/or EGFR staining to classify them into basal sub-types. The tumours which are positive for hormone receptors are classified as luminal sub-types. Basal sub-types are associated with poor prognosis as compared to luminal subtypes both in humans as well as canines. So the cytokeratins which are considered as diagnostic markers for epithelial tissues can be used to access prognosis of mammary tumours both in humans and canines,moreover, for studying tissue expression of above markers by IHC one does not a specialized set up as is required for studying the gene expression profiling.

I agree with Dr .Gupta that while gene expression profiling may be the academic gold standard for molecular subtyping in breast cancer, it is not really a feasible method for widespread clinical implementation - labs around the world are not equipped to perform such tests, definitely not on a regular basis in a clinical capacity. Thus IHC, which is a more readily accessible technique, is more useful. Multimarker panels such as cytokeratins + EGFR have been shown to be useful for diagnosing basal-like breast cancer, however single marker IHC using FOXC1 has been shown to be atleast equivalent, in many cases superior to the multimarker panels. Recently it has been shown to be essentially equivalent to the PAM50 gene expression panel in its diagnostic accuracy for basal-like breast cancer. Thus IHC for FOXC1 promises to be very useful as it can be performed by any lab that routinely performs IHC and would involve evaluation of only a single marker. Currently work is underway to validate the FOXC1 molecular diagnostic test in a 1000 patient multi-institutional trial funded by the NCI/NIH. In the meantime, the custom developed monoclonal antibody specialized for use on FFPE sections is available for research use only.