Glycine Zipper Motifs Mediate TM Helix Packing Interactions in Channel Proteins. We first noticed glycine zipper packing in homooligomeric channel proteins. The unrelated channel proteins, shown in Fig. 1 [KcsA (potassium channel), MscL (mechanosensitive channel of large conductance), VacA (vacuolating toxin A), and MscS (mechanosensitive channel of small conductance)], all use a stripe of small residues in the packing interface (11-14). Although Gly, Ala, and Thr are all observed in these structures, glycine is the most common interfacial residue. Thus, we call this packing motif a glycine zipper, because it is reminiscent of another common oligomerization motif found in soluble proteins, the leucine zipper (15). Like the leucine zipper, the glycine zipper can lead to a variety of different oligomerization states. The channels shown in Fig. 1 range from tetramers to heptamers. The fact that such a large fraction of the relatively few known membrane channel structures use the glycine zipper suggests that the sequence motif plays a special role in forming these homooligomeric bundles. As discussed below, the glycine zipper is also used in asymmetric helix packings as well.
The Glycine Zipper Motif Is Unusually Common. The glycine zipper packing mode is distinct from the well known GXXXG dimerization motif found in glycophorin A and other symmetric dimers that involve direct packing between the Gly faces (16). In glycine zipper packing, the glycine zipper packs against a different face of the associated helix. Thus, the glycine zipper may be more promiscuous in the nature of the helix face that it packs against, and it seemed possible that the glycine zipper could be used in a more catholic fashion throughout membrane protein structures, not just in symmetric oligomers like the channel proteins.
A method to discover sequence motifs that are important for defining membrane protein structure was introduced by Engelman and coworkers (4), who looked for sequence patterns that were significantly more common than expected by chance. Indeed, Senes et al. (4) found that a perfect glycine zipper motif, GXXXGXXXG, with Gly residues spaced every four positions is one of the most overrepresented triplets in predicted TM helices from all membrane proteins, without restriction to a specific class (odds ratio = 1.92; P = 1.38 x 10-13). Although the heptameric packing in the MscS pore has a slightly different spacing (see Fig. 1), a four-residue spacing is strongly preferred over other possible Gly patterns, reinforcing the significance of the GXXXGXXXG sequence pattern. Nevertheless, other spacings could lead to glycine zipper packing if the Gly residues are placed on the same face of the helix as in MscS.
Glycine Zippers Are Conserved. As another measure of the importance of the glycine zipper motif, we find that glycine zipper sequences are strongly conserved. Nineteen proteins with perfect glycine zippers and their homologs were examined. As shown in Fig. 2A, in all but two of the proteins, Gly residues in glycine zippers are more strongly conserved than random Gly residues. These results again indicate that the glycine zipper motif plays a particularly important role in membrane protein structure.
Small Residues Can Substitute in Glycine Zipper Motifs. Because different small residues were found to substitute for Gly in the channel proteins discussed above, we next examined whether other possible sequence patterns, consistent with the glycine zipper spacing, were overrepresented in the TMSTAT database of Engelman and coworkers (4). Senes et al. (4) previously presented the effects on pattern abundance when either the first or third glycine in the GXXXGXXXG motif was replaced with a small residue. We expanded this analysis to examine the abundance of sequence patterns when the perfect glycine zipper motif was adulterated with one, two, or three small residues at any position in the zipper (Ala, Ser, or Thr). As shown in Fig. 2B, the strongest glycine zipper sequence motifs, with odds ratios >1.5, were GXXXGXXXG, AXXXGXXXG, GXXXGXXXA, SXXXGXXXG, GXXXGXXXS, and GXXXGXXXT. These six motifs contain two or more Gly residues, and all maintain the central Gly residue in the pattern. Not only are these sequence motifs overrepresented in TM helices in general, but Russ and Engelman (5) found that many of these zipper patterns are found in homooligomerizing TM sequences discovered experimentally by using the TOX-CAT assay. Indeed, the GXXXGXXXT sequence pattern was particularly common in these TM helices (5).
Although the pattern GXXXG by itself is overrepresented in TM helices (4), the presence of a GXXXG sequence alone is not necessarily significant in the triplet patterns. In Fig. 2C, the odds ratios for the strong glycine zipper motifs are compared with GXXXG containing triplets with Val, Ile, or Leu in place of Ala, Ser, or Thr [also see Senes et al. (4)]. None of these triplets is significantly overrepresented. Thus, there is apparently special significance associated with the strong glycine zipper sequence patterns, and glycine zippers are apparently part of the increased abundance of the GXXXG subpattern.
Glycine Zippers Strongly Influence TM Helix Packing. The presence of a glycine zipper motif has a profound effect on protein structure. We searched for the preferred glycine zipper motifs in membrane proteins of known structure and found 13. All of these glycine zipper motifs are directly involved in helix packing (Fig. 5, which is published as supporting information on the PNAS web site). Thus, the glycine zipper face apparently provides a magnet for helix packing. Moreover, the glycine zipper motif strongly drives right-handed helix packing, as found in the structures shown in Fig. 1. On average, proteins are right-handed (17). Of the glycine zipper helices in proteins of known structure, however, this preference is completely reversed. Ten of 13 (77%) are involved in right-handed helix packing. As shown in Fig. 2D, a four-residue spacing creates an 
20° angle with respect to the helix axis. Small residues are favored in TM helix packing sites (18), and a right-handed helix packing maximizes contact with the small residues of the glycine zipper motif.
Glycine Zippers Are Common in Membrane Proteins. We searched for the preferred glycine zipper motifs in a database of predicted TM helical proteins from the Swiss-Prot database (6). Sequences with >90% sequence identity were removed from the TM protein database, leaving a total of 23,102 protein sequences. Of this set, we found that 1,726 proteins (7.5%) contain the perfect glycine zipper motif, GXXXGXXXG, and 5,684 (25%) contain one of the preferred motifs (see Table 2, which is published as supporting information on the PNAS web site). Thus, approximately one-quarter of all membrane proteins contain one of the strong glycine zipper motifs, and many others likely use weaker glycine zipper motifs. For example, MscS, which is known to use glycine zipper packing (Fig. 1), would not be counted because of the slightly different spacing of the pattern. The analysis of known protein structures discussed above suggests that virtually all of these glycine zipper motifs are involved in helix packing and that the vast majority of these packings are right-handed.
Prediction of Right-Handed Homooligomeric Bundles. Our finding that a glycine zipper creates a strong driving force for right-handed helix packing suggests that the presence of a glycine zipper motif in a single-pass protein is likely to drive the formation of right-handed homooligomeric bundles like those shown in Fig. 1. To look for proteins that are highly likely to form these homooligomeric bundle structures, we searched for single-pass proteins with extended glycine zipper motifs containing four Gly residues (GXXXGXXXGXXXG). The proteins retrieved in this search are listed in Table 1. We predict that these proteins form right-handed helix bundles to deliver their physiological or pathological function.
For many of the extended glycine zipper proteins, there is already experimental evidence for the importance of the glycine zipper motif in their function. One of the known pore-forming proteins is VacA, a toxin secreted by
Helicobacter pylori, the etiologic agent in roughly half of all stomach ulcers (
19). VacA is known to form a hexameric anion selective channel, and mutations in the glycine zipper motif residues G
47 and G
51 abolish the channel activity and cytotoxicity of VacA (
20). Influenza hemagglutinin, syntaxin 17, annexin A7, and ecto-ATPase are involved in membrane fusion, a process that involves formation of a fusion pore (
21-
23). In the best studied of these proteins, influenza hemagglutinin, the glycine positions are highly conserved, and mutations in the glycine residues or alterations of the glycine spacing in the glycine zipper motif block viral fusion (
24). A number of the extended glycine zipper proteins are involved in cell adhesion. Of these extended glycine zipper proteins, myelin protein zero is known to be a tetramer, and glycine mutations at position G
163 and G
167 in the TM domain are linked to Charcot-Marie-Tooth disease and Dejerine-Sottas syndrome (
25,
26).
Extended glycine zippers are also found in tight junction (TJ) proteins such as CLMP that act as gated intercellular pores regulating the exchange of small molecules, ions, and water between cells (27). Moreover, other major TJ proteins, such as the occludins, also have extended glycine zipper motifs but are not included in Table 1 because they contain multiple TM domains (see Table 2). It is therefore tempting to speculate that glycine zipper pore structures mediate the passage of molecules through TJs.
Two glycine zipper proteins, A
and prion protein (PrP), are involved in the neurodegenerative diseases Alzheimer's disease and spongiform encephalopathies. The glycine residues in these TM domains are completely conserved, which suggests that the glycine zipper motifs are playing an important role in their normal function. Cytotoxic fragments of A and PrP include the glycine zipper motifs and are also known to form pores in artificial membranes in vitro that could play a role in disease etiology (28-33). The fact that these peptides contain glycine zipper motifs suggests that in vitro channel formation by these peptides could be driven by glycine zipper packing (see below).
Experimental Test of a Glycine Zipper Packing Prediction. To test whether the glycine zipper motif in the A42 peptide is important for channel formation in vitro, we made three mutant A42 peptides in which Gly-29, Gly-33, and Gly-37 were substituted individually with Leu, disrupting the glycine zipper packing interface. Fig. 3B shows some representative current traces across a polar lipid bilayer membrane exposed to 3 µg/ml A1-42 peptides. Consistent with earlier studies (31-33), we found that the WT A1-42 readily forms channels. Analysis of the current distributions and current-voltage (I-V) plots shown in Fig. 3C demonstrate that the current increased in quantized units with a conductance of 12 pS. These results confirm that A42 can indeed form unique, well organized channel structures under our measurement conditions. In contrast, none of the Gly-to-Leu mutants showed organized channel behavior. All of the mutants induced variable current spikes, indicating that they can disrupt the membrane to some extent, but they are apparently unable to maintain stable, organized channels (see Fig. 6, which is published as supporting information on the PNAS web site). Although overall delivery to the membrane may be altered by peptide solubility, the character of the current spikes induced by the peptides is clearly altered with the mutants. Moreover, we saw no evidence for insoluble fibril formation in our peptide samples as judged by Congo red staining (not shown). As shown in Fig. 4, the ability to disrupt membranes is correlated with cytotoxicity (Fig. 4). These results suggest that the A42 pore formation depends on the glycine zipper motif. Although we cannot be certain that this dependence is due to direct helix packing involving the glycine zipper motif, the fact that all glycine zippers in known structures are in packing interfaces favors a glycine zipper packing hypothesis. In this manner, the identification of a simple sequence motif defines a limited set of structural templates for hypothesis-driven experiments.
Answers to all your Biology Questions
