(A) Simplified diagram of the concatenated tree for TSET, APs, and COPI, based on Figure 4—figure supplement 8. Numbers indicate posterior probabilities for MrBayes and PhyloBayes and maxium-likelihood bootstrap values for PhyML and RAxML, in that order. (B) Schematic diagram of TSET. (C) Possible evolution of the three families of heterotetramers: TSET, APs, and COPI. We propose that the earliest ancestral complex was a likely a heterotrimer or a heterohexamer formed from two identical heterotrimers, containing large (red), small (yellow), and scaffolding (blue) subunits. All three of these proteins were composed of known ancient building blocks of the membrane-trafficking system (Vedovato et al., 2009): α-solenoid domains in both the large and scaffolding subunits; two β-propellers in the scaffolding subunit; and a longin domain forming the small subunit. The gene encoding the large subunit then duplicated and mutated to generate the two distinct types of large subunits (red and magenta), and the gene encoding the small subunit also duplicated and mutated (yellow and orange), with one of the two proteins (orange) acquiring a μ homology domain (MHD) to form the ancestral heterotetramer, as proposed by Boehm and Bonifacino (12). However, the scaffolding subunit remained a homodimer. Upon diversification into three separate families, the scaffolding subunit duplicated independently in TSET and COPI, giving rise to TTRAY1 and TTRAY2 in TSET, and to α- and β′-COP in COPI. COPI also acquired a new subunit, ε-COP (purple). The scaffolding subunit may have been lost in the ancestral AP complex, as indicated in the diagram; however, AP-5 is tightly associated with two other proteins, SPG11 and SPG15, and the relationship of SPG11 and SPG15 to TTRAY/B-COPI remains unresolved, so it is possible that SPG11 and SPG15 are highly divergent descendants of the original scaffolding subunits. The other AP complexes are free heterotetramers when in the cytosol, but membrane-associated AP-1 and AP-2 interact with another scaffold, clathrin; and AP-3 has also been proposed to interact transiently with a protein with similar architecture, Vps41 (Rehling et al., 1999; Cabrera et al., 2010; Asensio et al., 2013). So far no scaffold has been proposed for AP-4. Although the order of emergence of TSET and COP relative to adaptins is unresolved, our most recent analyses indicate that, contrary to previous reports (Hirst et al., 2011), AP-5 diverged basally within the adaptin clade, followed by AP-3, AP-4, and APs 1 and 2, all prior to the LECA. This still suggests a primordial bridging of the secretory and phagocytic systems prior to emergence of a trans-Golgi network. The muniscins arose much later, in ancestral opisthokonts, from a translocation of the TSET MHD-encoding sequence to a position immediately downstream from an F-BAR domain-encoding sequence. Another translocation occurred in plants, where an SH3 domain-coding sequence was inserted at the 3′ end of the TSAUCER-coding sequence. See also Figure 4—figure supplements 1–10.