The assembly and structure of membrane proteins are intimately related. Targeting and insertion of membrane proteins into the endoplasmatic reticulum (ER) membrane involve many components that facilitate the folding of the native structure. General rules for assembly mechanisms and structure determinants in the protein sequence, can be deduced from a number of techniques that are available to study membrane proteins.

The focus of the thesis is on two aspects of membrane protein assembly: the membrane insertion of so-called N-tail proteins, and on how the overall length and the precise amino acid sequence of transmembrane helices (TMHs) affect their location and topology in the membrane. The work is largely based on the use of N-linked glycosylation, both as a topological marker and as a 'molecular ruler'.

The first study demonstrates that the translocation of N-tails in eukaryotic polytopic membrane proteins can depend on downstream, synergistically acting TMHs. In the second study, the effect of charged residues on the location of a model TMH in the ER membrane was studied. The side-chains of positively charged residues were found to be able to 'snorkel' from a location within the hydrophobic core of the membrane up towards the lipid head-group region. Furhter studies on TMHs that are either too short or too long to fit the width of the ER membrane ('hydrophobic mismatch') suggest that short TMHs are positioned approximately perpendicular to the membrane, while long TMHs may compensate for their positive mismatch by bending. Finally, the formation of 'helical haipins', i.e., two TMHs connected by a tight turn, was studied by introducing each of the 20 naturally occurring amino acids near the middle of a very long hydrophobic segment. Charged residues and residues with high helix-breaking propensity were found to induce the formation of helical hairpins of which the shortest possible was found to consist of 31 hydrophobic residues including a turn promoting proline. A TMH turn propensity scale was derived, which hopefully can be used to improve current prediction programs for membrane protein topology.