Background: TNBC is recognized to be heterogeneous based on gene expression profiling. An AR+, androgen-dependent subset of TNBC benefits from antiAR drugs. FOXA1, a nuclear mediator of steroid receptor signaling is involved in the regulation of ER and AR. FOXA1 expression in ER+ BC correlates with better prognosis; the significance of FOXA1 in TNBC is not established. We report the rates of FOXA1 expression in relationship to AR, basal markers and outcome in a retrospective cohort of pts with TNBC. Methods: We identified 1,032 pts with resectable, TNBC (ER/PR<1%; HER2 0-2+/FISH <2) who had surgery at MSK (1998-2006). Exclusion criteria: neoadjuvant chemotherapy, radiation, inflammatory/metastatic BC. We constructed tissue microarrays from 311 primary tumors (>1cm, post 2002) with three 0.6mm cores per tumor. Standard IHC methods were used. Scored positive: FOXA1 >3 per Badve CCR 2007; AR and EGFR paralleled the ASCO/CAP criteria for ER and HER2 respectively; CK5/6 and CK14 any cytoplasmic stain. Associations were analyzed with Fischer Exact/Spearman correlation tests. Log Rank test was used for survival analysis. Results: 303 pts met eligibility criteria. # of positive/evaluable cases: AR 16/283 (6%), FOXA1 79/273 (29%), CK5/6 185/290 (64%), CK14 128/278 (46%) and EGFR 234/289 (81%). There was a significant correlation between FOXA1 and AR (p=1.19E-10). There were inverse correlations between FOXA1 and CK14 (p=0.003), AR and CK14 (p=0.004) and AR and EGFR (p=0.018). 264/291 (91%) pts met Nielsen criteria (NC) for basal-like BC (BLBC). 94% (15/16) of AR+ tumors and 98% (121/124) of CK14+ tumors co-expressed NC. 46% of BLBC by NC expressed CK14. CK14 expression was associated with worse DFS (p=0.003, 5-year DFS 69% vs. 83%) and OS (p=0.01, 5-year OS 71% vs. 85%). FOXA1 expression was not associated with a significant difference in DFS (HR 0.98, p=0.26) or OS (HR 0.99, p=0.52). Conclusions: CK14 expression was the only prognostic marker for survival in our cohort. A high proportion of the AR+ tumors are BL according to NC, which is contradictory to gene expression data. There may be differences between these subsets or discordance between protein expression of AR and molecular subtype of TNBC (luminal AR vs. BL).