Swimming through the hydrophobic sea: New insights in protein translocation

It’s like Moses and the Red Sea, only backwards. Moses had to get his people across a hostile aqueous environment; membrane proteins and secreted proteins also have to get across a hostile hydrophobic environment: the lipid bilayer of biological membranes. Because numerous proteins reside in exactly that nonpolar milieu or are forced across it in the process of secretion, there must exist mechanisms to overcome the unfavorable free energy of transferring the charges on proteins from aqueous phases into and through the bilayer. Eukaryotic cells have developed elaborate machinery to transfer proteins across membranes during synthesis. Many prokaryotic toxins use an alternate approach: These toxins carry with them the machinery they require to slither into or through the membrane.

A host of bacterial toxin proteins must take on the challenge of bilayer crossing to effect their deadly intent. Diphtheria toxin, for example, kills target cells by enzymatically altering intracellular components. To gain access to its target, the enzymatic portion of diphtheria toxin must enter the cytoplasm. The required translocation is mediated by another portion of the toxin; thus, diphtheria toxin apparently carries its own translocation apparatus. Similarly, some of the colicin toxins, which kill strains of Escherichia coli, defeat their victims by forming ion channels in their membranes. These deadly pores dissipate the transmembrane ionic gradients required for bacterial survival. As with diphtheria toxin, a prelude to colicin activity is the insertion of the protein into the target membrane. What is the physical basis of this transmembrane insertion? Are the mechanisms involved in colicin translocation general or specific to this group of toxins? In this issue of Proceedings, Jakes et al. (1) analyze a series of mutants of colicin Ia that shed further light on the transmembrane meanderings of this protein. They show that a hydrophilic region in the colicin can carry highly charged, foreign protein sequences with it as it translocates across the membrane.

Electrostatic Analyses of Charge-Membrane Interaction.

A major component of the free energy of transferring proteins into membranes is the cost of introducing charged and polar residues into the poorly polarizable lipid bilayer. In a classic analysis, Parsegian (2) calculated the electrostatic energy required to introduce charged spheres into a low dielectric slab (a model for the plasma membrane). He examined several situations that could lower the energy of introducing a charged moiety into a thin hydrophobic medium, including pairing charges of opposite sign as well as “carriers” and “pores,” both with high local dielectric. Pairing with unlike charges did not significantly reduce the transfer energy of a charge, whereas both a carrier mechanism and a pore did lower the energy. In the context of protein translocation, these results imply that polar interactions between charged residues and polarizable parts of a protein could stabilize significantly the membrane-inserted form. Parsegian (2) also noted the possible “electrostrictive” effects of introducing a charge into the bilayer: A charge in the middle of a low dielectric slab will tend to pinch it, shortening the distance to the high dielectric surroundings. Emerging data from experiments examining gramicidin channels in bilayers of different thickness or with different tension suggest that protein–lipid interactions do indeed have significant effects on bilayer geometry (3, 4).

Honig and Hubbell (5) analyzed the thermodynamics of inserting charged amino acids from water into a low dielectric medium. They found less unfavorable interaction energies between charged pairs and the bilayer than had Parsegian (2), on the order of 10–20 kcal/mol/charged pair (ɛ = 2). Unfavorable free energies of this magnitude potentially could be compensated by a few hydrogen bonds in a real protein. Furthermore, major shifts in pK_a, related to the low dielectric surroundings of a membrane, may favor the insertion of neutral forms of titratable groups.

Current theoretical analyses are limited by the complexity of the real system; a lipid bilayer membrane is not simply a slab of low dielectric but rather is a dynamic system including polar headgroups and hydrocarbon tails on each lipid molecule. Indeed, it is difficult even to estimate the dielectric constant in the membrane. Molecular dynamics calculations are not yet up to the problem of protein translocation across membranes. In this light, the study of model systems of membrane–protein interaction will be fundamental for further understanding.

The simplest possible model system for studying protein translocation across membranes includes only a short peptide and a membrane. Maduke and Roise (6) examined a peptide corresponding to a mitochondrial presequence (MLSLRQSIRFFKPATRTLCSSRYLL-NH_2, net charge +6 at neutral pH). By using fluorescently labeled peptide and protease protection, they showed that the peptide is taken up by lipid vesicles only when a negative internal potential is established. The requirement for a negative transmembrane potential hints that at least some of the residues in the presequence remain charged as they transfer through the membrane. The exact mechanism by which full length proteins are imported into mitochondria remains unclear, but these experiments showed that a hydrophilic, charged peptide can bring itself across a lipid bilayer, perhaps initiating the transfer. Further insight may be gained from proteins that translocate spontaneously and reversibly: the colicins.

Colicin Channels.

The colicins are a group of protein toxins synthesized by some types of E. coli for defense against competing strains. Colicins act by a variety of mechanisms; as mentioned, several form ion-conducting channels in the inner membranes of their target bacteria (7). These toxins, colicins A, Ia, and E1, also spontaneously insert into planar lipid bilayer membranes, forming voltage-dependent ion channels. The wild-type channels generally turn on at positive voltage and off at negative voltage. The channel-forming colicins have three distinct domains: an amino-terminal domain required for transport of toxin through the outer membrane, a middle domain required for receptor binding, and a carboxy-terminal domain that forms the ion-conducting channel. Within the channel-forming region, a hydrophobic stretch inserts through the membrane before channel opening, anchoring the colicin to the bilayer (8). An additional hydrophilic region in the channel-forming domain is the focus of the paper discussed here.

In previous work, Finkelstein (9) and compatriots elegantly demonstrated that a portion of the colicin protein, at least 68 amino acids long, moves back and forth through the lipid bilayer as the channel opens and closes (Fig. 1a). In those experiments, Qiu et al. (9) introduced single cysteines in the hydrophilic region of colicin Ia, then site-specifically biotinylated those residues. When the channels were closed, the biotinylated residues remained on the cis side, and streptavidin added to the cis compartment prevented the channels from opening. (cis is defined as the side of the membrane to which toxin was added, trans is defined as V = 0.) On the other hand, when the channels were open, cis streptavidin had no effect, but streptavidin in the opposite compartment locked the channels in the open position. It was surprising that the experimenters also found that the translocated region of colicin Ia is plastic: The portion of the protein actually spanning the bilayer can change depending on exactly what residue is held on the cis side of the membrane by streptavidin.

Figure 1.

Schematic illustration of colicin Ia gating. The hydrophobic C-terminal region (red) is inserted in the membrane in both open and closed states of the channel. (A) The hydrophilic portion of the channel-forming domain (green) reversibly translocates through the membrane when the channel opens. (B) A highly charged peptide (blue) is carried across the membrane as the colicin channel gates.

In a paper appearing in this issue of Proceedings, Jakes et al. (1) extend their analysis of the translocated region of colicin Ia. Having noted the plasticity in which part of the protein traverses the membrane, the group boldly inserted groups of foreign (hydrophilic) amino acids into the translocated region of the toxin. As heterologous peptides, Jakes et al. chose epitope tags recognized by mAbs; thus, anti-peptide mAb could be used to assess the presence of the epitope on one or the other side of a bilayer containing colicin. If the added epitope remains on the cis (toxin-containing) side of the bilayer, then binding to cis mAb should prevent translocation of the epitope and hence opening of the channel. Conversely, if the epitope is translocated across the bilayer, trans antipeptide antibody should lock the channels into the open state.

Jakes et al. (1) found that two inserted peptide epitopes are indeed dragged across the lipid bilayer. They chose the site of insertion for these epitopes based on the crystal structure of colicin Ia in solution (10), placing the heterologous sequences in the loop connecting two α-helices in the translocated segment. The first of these is the “HA” epitope, from the influenza virus hemaglutinin protein, with the sequence YPYDVPDYA. The colicins containing this polar epitope behaved normally, retaining most in vivo bactericidal activity as well as normal in vitro channel-forming activity. When anti-HA antibody was added to the trans solution with channels open, they were locked into the open state, an effect reversed by adding excess HA peptide to the same solution. However, if the antibody was present in the trans solution only when channels were closed, this did not occur. Anti-HA antibody added to the cis solution, in contrast, prevented channels from opening; this effect, too, was reversed by adding HA peptide to the cis compartment. By their own account, Jakes et al. were emboldened by the success of these experiments, so they threw caution to the wind and added, in the same location, the FLAG epitope, a peptide sequence of DYKDDDDK—an amazing seven out of eight residues charged! (net charge −3). Nevertheless, this sequence also was translocated across the membrane (Fig. 1b). The energy barrier for translocation of the hairpin containing the FLAG peptide is apparently larger than that for native toxin; larger positive and negative voltages are required to open and close these channels. That the FLAG peptide is translocated across the bilayer is a surprising result, flying in the face of both common sense and theoretical predictions.

It is difficult to understand the contributions of the electrostatic mechanisms discussed above to the translocation of the colicin Ia hydrophilic region, but the system may help shed light on the situation. What is the actual charge on the peptide as it crosses the bilayer? Measurement of the charge movement associated with gating will begin to address these questions. In the past, these measurements have been difficult because of difficulties in achieving the steady-state conductances necessary for obtaining accurate measuring voltage-dependence. Recent work has suggested means of obtaining steady-state data, however (8). Information also may come from inserting peptides with untitratable charges like arginine. If the hairpin still translocates with multiple fixed charges attached, then shifts in pK_a are not essential for transport. Another question that can be addressed partially in this system is the effect of changes in bilayer thickness. Although electrostrictive effects cannot be measured directly, it would be interesting to know whether the colicin region would still be translocated through a thicker bilayer.

That colicins can translocate apparently arbitrary peptide sequences across cell membranes may have commercial application. Diphtheria toxin derivatives with altered receptor binding properties have been used to selectively kill cancer cells (11). Perhaps colicins could be used to introduce therapeutic peptides that have specific actions short of cell murder. Another intriguing direction stems from the recent discovery that a protein involved in apoptosis, Bcl-XL, has a structure analogous to those of the channel-forming bacterial toxins (12). Indeed Bcl-XL, as well as another apoptosis-related protein, Bcl-2, has been reported to form ion-conducting channels in lipid bilayers (13, 14). Although their function in apoptosis remains unclear, if the channels formed by these proteins are important to cellular function, they will have to translocate into the membrane. In all of the years since Moses’ time, we still have not figured out how he got the Red Sea to part; with a little luck, we will understand the energetics of protein translocation more quickly.

ABBREVIATION

HA

hemaglutinin