Every female mammalian cell carries two X chromosomes, but only one may stay transcriptionally active — a dosage-compensation rule discovered by Mary Lyon (1961). Early in embryogenesis, each epiblast cell independently and randomly silences one X:
P(maternal X active) = p (the skew probability)
P(paternal X active) = 1 − p
choice is made once per progenitor cell, then is CLONALLY inherited
by every descendant → patches, not single-cell speckle
The silenced chromosome is coated in cis by Xist, a long non-coding RNA transcribed from the X-inactivation centre (XIC). Xist spreads outward along the chromosome and recruits PRC2 (depositing H3K27me3), DNA methyltransferases and histone deacetylases, converting open euchromatin into dense, transcriptionally silent heterochromatin — the compact Barr body visible in interphase nuclei.
Expected clonal patch size follows directly from how many progenitor cells existed at the moment of inactivation:
patch area ≈ tissue area / F (F = founder/progenitor cells)
- Founder cells — fewer founders at the time of XCI ⇒ each clone expands further ⇒ larger, coarser patches (this is the same math behind calico-cat fur patches and human skewed-XCI disease mosaics).
- Maternal-X skew — biases the coin flip each founder makes; extreme skew (near 0% or 100%) models pathological skewed XCI seen in X-linked carrier females.
- Xist spreading speed — how fast the inset chromosome's Xist coating (small dots) sweeps from the XIC outward and the chromatin condenses into a Barr body; the wave starts at one end and both the RNA coating and the compaction lag behind it, exactly as ChIP-seq spreading profiles show in real cells.
Real-world relevance: this random, clonally-fixed silencing is why carrier females of X-linked disorders (e.g. red-green color blindness, some forms of hemophilia) show partial, patchy phenotypes instead of an all-or-nothing outcome — and why calico/tortoiseshell cat coats are visual proof of X-inactivation mosaicism.