The eukaryotic subtilisin-like endoprotease furin is found predominantly in the trans-Golgi network (TGN) and cycles between this compartment, the cell surface, and the endosomes. There is experimental evidence for endocytosis from the plasma membrane and transport from endosomes to the TGN, but direct exit from the TGN to endosomes via clathrin-coated vesicles has only been discussed but not directly shown so far. Here we present data showing that expression of furin promotes the first step of clathrin-coat assembly at the TGN, the recruitment of the Golgi-specific assembly protein AP-1 on Golgi membranes. Further, we report that furin indeed is present in isolated clathrin-coated vesicles. Packaging into clathrin-coated vesicles requires signal components in the furin cytoplasmic domain which can be recognized by AP-1 assembly proteins. We found that besides depending on the phosphorylation state of a casein kinase II site, interaction of the furin tail with AP-1 and its 1subunit is mediated by a tyrosine motif and to less extent by a leucine-isoleucine signal, whereas a monophenylalanine motif is only involved in binding to the intact AP-1 complex. This study implies that high affinity interaction of AP-1 or 1 with the cytoplasmic tail of furin needs a complex interplay of signal components rather than one distinct signal. The trans-Golgi network (TGN)1 constitutes the sorting station where soluble and membrane proteins, targeted to different post-Golgi compartments, are packed into distinct carrier vesicles. Three TGN exit routes are known: constitutive transport to the cell surface, transport to secretory granules in cells with a regulated secretory pathway, and transport to endosomes (for review, see Ref. 1). The last pathway is followed primarily by lysosomal enzymes bound to the mannose 6-phosphate receptors (MPRs) by their mannose 6-phosphate residues (for review, see Ref.2). Vesicles leaving the TGN for subsequent transport to the endosomes are clathrin-coated. Formation of these clathrin-coated vesicles (CCVs) involves binding of AP-1 Golgi-specific assembly proteins to the TGN membrane. AP-1 binding is regulated by the ADP-ribosylation factor ARF-1, a small GTP-binding protein (3, 4), and requires the presence of transmembrane proteins, among which MPRs are the major constituents (5-7). CCVs are transport intermediates of vesicular traffic not only from the TGN to endosomes, but also from the plasma membrane to endosomes. Golgi-and plasma membrane-derived CCVs can be distinguished by their different sets of assembly proteins; AP-1 complexes are restricted to coated buds and vesicles of the TGN, whereas AP-2 complexes act in endocytosis at the plasma membrane. Both adaptor complexes are heterotetrameric. The AP-1 complex is composed of two 100-kDa subunits, ␥ and ␤1 adaptin, a medium subunit 1 (47 kDa), and a small subunit 1 (19 kDa). The AP-2 complex is of similar size and composition with two large subunits, ␣ and ␤2 adaptin (100 kDa), in association with a 2 (50 kDa) and a 2 subunit (17 kDa) (for r...

Epithelial cells fulfill important functions at the borderline between intracellular and extracellular compartments. To meet these requirements, they have a polarized organization that involves the separation of the plasma membrane into an apical and a basolateral portion (reviewed in Ref. 1). The two membrane domains have a different protein and lipid composition, which is maintained by a diffusion barrier made up from junctional complexes. A specialized sorting apparatus exists to ensure that proteins and lipids specific for either of the two surface domains are targeted correctly. As far as sorting signals are concerned, more information is available for basolateral proteins than is for apical proteins. Sequence motifs for basolateral targeting are usually located in the cytoplasmic tail and often contain a tyrosine residue (2, 3). Many studies have reported a close relationship between basolateral sorting signals and determinants for coated pit localization (3, 4). Indeed the majority of basolateral proteins that have been identified so far and characterized in more detail are subject to endocytosis. An exception is membrane cofactor protein (MCP), 1 which is not endocytosed though it is efficiently targeted to the basolateral plasma membrane of several polarized epithelial cells (5, 6).MCP is a widely distributed regulatory protein that interacts with complement factors C3b and C4b deposited on self-tissue (7). It promotes the degradation of these factors by plasma serine protease factor I and thus protects the cell from complement-mediated damage. It also serves as a receptor for vaccine strains of measles virus (8, 9). The binding sites for the complement factors as well as for measles virus are located within four cysteine-rich repeating domains at the NH 2 terminus of this type I membrane protein (10, 11). The cytoplasmic portion of MCP consists of 10 membrane-proximal, predominantly basic amino acids designated as intracytoplasmic anchor that is followed by a carboxyl-terminal cytoplasmic tail. The latter domain is subject to alternative splicing resulting in isoforms with two different tails, a shorter one comprising 16 amino acids and a longer one comprising 23 amino acids (reviewed in Ref. 7).Though the two different cytoplasmic tails found in MCP isoforms do not show any sequence similarity, they both contain sequence motifs for basolateral targeting of MCP (5). The sorting signal of the shorter cytoplasmic tail has been characterized in more detail and several interesting features have been revealed (6): (i) a tetrapeptide at the very carboxyl-terminal end (-FTSL) is essential for basolateral targeting; (ii) compared with a common basolateral sorting signal (Tyr-Xaa-Xaaaliphatic amino acid), a phenylalanine is substituted for the tyrosine; (iii) MCP containing the FTSL sequence is not endocytosed; (iv) substitution of a tyrosine residue (-YTSL) results in efficient internalization. By providing sorting information and by avoiding endocytosis, the carboxyl-terminal tetrapeptide ensures constitutiv...