Claudin-Low Is the Deepest Breast Cancer Subtype When Measured Correctly: EZH2-Free Principal Component Analysis Resolves the Apparent TNBC Depth Artefact and Confirms Claudin-Low at Distance 6.572 vs TNBC 6.063 from Normal Breast Epithelial Identity

Standard multi-gene Principal Component Analysis (PCA) of breast cancer subtype data consistently places TNBC at greater distance from normal breast epithelium than claudin-low (CL), implying TNBC is the most de-differentiated subtype. This document identifies this ordering as a measurement artefact and reports the corrected result using EZH2-free PCA. The artefact: TNBC has EZH2 elevated +189% above normal — the highest of all subtypes. When EZH2 is included in PCA, TNBC's high EZH2 inflates its apparent distance from the normal reference. EZH2 is the lock mechanism in TNBC (PRC2/EZH2 silences FOXA1, GATA3, and ESR1), not a measure of differentiation loss per se. CL has lower EZH2 (+67%) not because it is less deeply arrested, but because it is a Type 4 root lock attractor — the cell never committed to a luminal programme requiring EZH2 to silence. There was nothing to suppress. EZH2-free PCA of 19,542 single cancer cells (GSE176078) corrects the ordering: Standard PCA (EZH2 included): TNBC 5.951, CL 4.973 (TNBC appeared 19.7% deeper — artefact) EZH2-free PCA: CL 6.572, TNBC 6.063 (CL correctly 8.4% deeper — confirmed) CL moves outward by +32.1% when EZH2 is removed. TNBC moves outward by only +1.9%. The entire apparent TNBC depth advantage disappears and reverses when the confounding variable is excluded. The mechanistic explanation: TNBC is an Attractor Type 2 (epigenetic lock — luminal identity programme present in the genome but silenced by EZH2/PRC2). EZH2 inhibition (tazemetostat) can restore FOXA1 and ESR1 in TNBC because the programme is present but locked. CL is an Attractor Type 4 (root lock — luminal identity programme never activated; cell arrested at progenitor root). EZH2 inhibition cannot restore a programme that was never activated. CL has more total differentiation loss than TNBC; TNBC has more active epigenetic silencing machinery. The therapeutic implication: CL is further from recoverable luminal identity than TNBC. The immune compartment is therefore the primary — not secondary — therapeutic axis for CL. Anti-TIGIT → anti-PD-1 in memory-low CL patients (CS-LIT-19/20, DOI: 10.5281/zenodo.18894331) is not a fallback when endocrine therapy fails; it is the correct first and primary intervention. The general methodological principle: in any cancer type where two subtypes differ in arrest mechanism (Type 2 = epigenetic silencing, high EZH2; Type 4 = root lock, low EZH2), including EZH2 in a depth PCA will systematically overestimate Type 2 depth and underestimate Type 4 depth. EZH2-free PCA using shared identity markers (FOXA1, GATA3, ESR1, SPDEF, CDH1) is the required methodology. No published paper has described EZH2 as a confound in CL vs TNBC depth comparisons, applied EZH2-free PCA to resolve this ordering, or shown CL as the deepest breast cancer subtype by quantitative measurement. Verdict: CONVERGENT-NOVEL.