Centriole biogenesis is seeded by CEP152-CEP63-PCNT aggregates propagating outside the centriole through the Alström...

While subcellular structures like centrioles assemble flawlessly from protein components, it is unknown whether instructions from outside the components are involved. Centrioles build on the cartwheel scaffold that assembles in interphase and disassembles in mitosis following centriole growth, yet what underlies these cartwheel dynamics remains obscure. Here we identify ALMS1-an intrinsically disordered, disease-linked protein-as an external cofactor that preserves cartwheel-forming capacity without itself entering the cartwheel. Introducing disease-linked mutations disrupted cartwheel dynamics, enabling cartwheels to expand and shed from growing centrioles, and, in turn, form additional centrioles, leading to reciprocal amplification. Molecularly, we identified CEP152, CEP63, and PCNT as ALMS1-interacting proteins that form aggregates devoid of ALMS1, which function as seeds for cartwheel assembly independently of centrioles. These cartwheel seeds (CSs) form in interphase and exhibit a nanoscale, concentric ring comprising CEP152 and CEP63 from which the cartwheel grows. Upon mitotic entry, CSs transiently recruit ALMS1 before disassembling into components that associate with ALMS1 in proximity, matching the cartwheel dynamics. Depleting ALMS1 abolished CS assembly from components, thereby eliminating centrioles, while reintroducing ALMS1 generated de novo centrioles with diverse shapes that were inheritable. Thus, centriole biogenesis is grounded on ALMS1-mediated CS assembly and disassembly that operates outside the cartwheel-centriole relationship, a system, we hypothesize, involving memory.