Introduction Protein targeting towards the proper subcellular compartments is vital for the cell. For compartments surrounded by two membranes, nuclear-encoded proteins contain an N-terminal leader sequence that allows their targeting and post- translational translocation through both envelope membranes (Bruce, 2000; Strub et al., 2000; Vothknecht and Soll, 2000). In the case of higher plant chloroplasts, the acquisition of a chloroplast transit sequence allows the return to the plastid of proteins encoded by genes that were present in the original prokaryotic endosymbiont and were subsequently transferred to the eukaryotic host cell nucleus (Martin and Schnarrenberger, 1997; Martin et al., 1998). In some algae, however, plastids may be surrounded by either three or four membranes. These are thought to result from secondary endosymbioses (where a host cell acquires an eukaryotic endosymbiont already equipped with plastids), a hypothesis supported by differences in the molecular phylogeny of plastid and nuclear genes (Delwiche and Palmer, 1996; Morden et al., 1992). The transfer of genes with a chloroplast transit sequence from the nucleus of the primary endosymbiont to the nucleus of the secondary endosymbiont requires additional signals for correct targeting, which has led to the suggestion that the final composition of the leader sequence might recapitulate the evolutionary history of the plastid (McFadden, 1999). This is supported by analyses of leader sequences of plastid-directed proteins in the four membrane-bound plastids of haptophytes, crysophytes, apicomplexans and diatoms, all of which contain an N-terminal hydrophobic signal peptide (to target proteins to the ER) followed by a transit sequence which can direct proteins synthesized in vitro to purified plant plastids (van Dooren et al., 2001). Several mechanisms are known for directing proteins to four membrane-bound plastids. In one mechanism, such as that used by haptophytes and diatoms, ribosomes are attached directly to the outer plastid membrane. After passage of the first (outermost) membrane, termed the chloroplast ER (Gibbs, 1981), proteins either cross the second membrane through large pores, or traverse the subsequent intermembrane space inside transport vesicles (van Dooren et al., 2001). In contrast to this mechanism, targeting to the four membrane-bound plastids of the apicomplexans does not involve ribosomes bound to the outer membrane, and indeed, many aspects of the mechanism are still unknown (van Dooren et al., 2001). It is known that plastid proteins enter the ER membrane system using an N-terminal signal peptide, since constructs lacking the signal peptide produce a cytoplasmic protein (Waller et al., 2000). It is also known that a transit peptide, exposed after cleavage of the signal peptide, is required for entry into the plastid, as proteins lacking this transit peptide are secreted from the cell (Waller et al., 2000). However, several mechanisms have been suggested for targeting proteins to the outermost plastid membrane from the ER (van Dooren et al., 2001). One proposal is that all proteins entering the ER 2867 Eukaryotic cells contain a variety of different compartments that are distinguished by their own particular function and characteristic set of proteins. Protein targeting mechanisms to organelles have an additional layer of complexity in algae, where plastids may be surrounded by three or four membranes instead of two as in higher plants. The mechanism of protein import into dinoflagellates plastids, however, has not been previously described despite the importance of plastid targeting in a group of algae responsible for roughly half the ocean’s net primary production. Here, we show how nuclear-encoded proteins enter the triple membrane-bound plastids of the dinoflagellate Gonyaulax. These proteins all contain an N- terminal leader sequence with two distinct hydrophobic regions flanking a region rich in hydroxylated amino acids (S/T). We demonstrate that plastid proteins transit through the Golgi in vivo, that the first hydrophobic region in the leader acts as a typical signal peptide in vitro, and that the S/T-rich region acts as a typical plastid transit sequence in transgenic plants. We also show that the second hydrophobic region acts as a stop transfer sequence so that plastid proteins in Golgi-derived vesicles are integral membrane proteins with a predominant cytoplasmic component. The dinoflagellate mechanism is thus different from that used by the phylogenetically related apicomplexans, and instead, is similar to that of the phylogenetically distant Euglena, whose plastids are also bound by three membranes. We conclude that the protein import mechanism is dictated by plastid ultrastructure rather than by the evolutionary history of the cell. Key words: Dinoflagellate, Plastid, Nuclear-encoded protein, Protein import, Signal peptide Summary Plastid ultrastructure defines the protein import pathway in dinoflagellates Nasha Nassoury, Mario Cappadocia and David Morse* Département de Sciences Biologiques, Université de Montréal, 4101 Sherbrooke est, Montréal, Québec, Canada H1X 2B2 *Author for correspondence (e-mail: david.morse@umontreal.ca) Accepted 26 March 2003 Journal of Cell Science 116, 2867-2874 © 2003 The Company of Biologists Ltd doi:10.1242/jcs.00517 Research Article