Shocking Disparities in Voltage-Dependent Channels

STRANGE FEATURES

The controversy largely surrounds one part of the structure. The ion selectivity, or pore region, fits well with MacKinnon's previously published structure for KcsA, a pH-gated channel. But with abundant experimental evidence dating back to the 1950s, the other, more novel part of the structure, the "voltage sensor," does not fit well with previous data, nor with the conventional voltage-gating model that has become well established.

"When you looked at the structure of KcsA, there was no doubt in your mind that it had to be right because it fit all of the work that had gone before," says Diane Papazian, who studies voltage-gated channels at the University of California, Los Angeles. "It was a very different experience looking at KvAP." Francisco Bezanilla, also of UCLA, agrees: "The selectivity problem was solved beautifully by the KcsA structure," he says, but the KvAP structure has some "very strange features."

Voltage-gated potassium channels comprise four subunits, each containing six transmembrane spans (TMs). Biochemical, electrophysiological, and mutational analysis of these channels long ago identified the fourth TM, known as S4, as the key element in voltage sensing. S4 is an unusual TM in that it carries, depending on channel type, from four to seven positively charged residues, usually arginines.

In the conventional model, S4 is entirely surrounded by other TMs, which shield the positively charged residues from the hydrophobic lipid bilayer. When a membrane voltage is imposed, the resulting electrical field is thought to pull these charges in different directions, causing a series of modest conformational changes that result in channel gating. But in MacK-innon's KvAP structure, S4 and part of nearby TM3 form a paddle-like structure that juts out into the middle of the bilayer, charged residues and all. MacKinnon proposed a mechanism in which this paddle produces large conformational changes by swinging up and down through the membrane in response to voltage.

This highly unconventional model met with a strong reaction from ion channel researchers. " [These papers] had a huge impact because the result was something that people in the field were looking for for a long time," says Richard Horn of Jefferson Medical College in Philadelphia. "This was the Holy Grail."

CRYSTALLIZING MODELS

As might be expected, the quest for this lofty prize proved difficult. Youxing Jiang, first author on the KvAP papers and now an assistant professor at University of Texas Southwestern Medical Center in Dallas, describes a long struggle: "For almost three years we didn't see even a trace of crystals. We started to think about ways to stabilize the protein." Hartmut Michel's laboratory at the Max Planck Institute of Biophysics in Frankfurt had succeeded years before in crystallizing cytochrome c oxidase, another integral membrane protein, by using antibodies as stabilizers. Jiang raised antibodies to the KvAP channel and included portions of them in crystallization trials. "The effect was dramatic," he says. "Now we had something to work with."

©2003, Elsevier

In (A), accessibility and disulfide bonding sites determined experimentally in the Drosophila Shaker channel are mapped onto the paddle model and onto an alternative model with the voltage sensing domain (S1–S4) re-oriented. Small purple spheres are K+ ions in the pore of the channel. Other spheres: green = always external; blue = not accessible; red = internally accessible; yellow and yellow circle = cysteine pair with rapid disulfide bonding indicating <13 Ångstrom distance. The paddle model does not account for either Shaker accessibility or the short distance between Shaker's S4 and S5, but the re-oriented model does. (B) shows voltage sensing motions of S4 in the re-oriented model. (From C.S. Gandhi et al., Neuron, 40:515–25, 2003.)

Others, however, point to these same antibodies as the likely source of the structure's flaws. "It was an extremely hard problem to crack," says Perozo. "These proteins are designed to be floppy," so they can perform their functions. "But floppiness is your main enemy in structure determination." Previously, antibodies had been used only on relatively rigid proteins. " [MacKinnon] had no inkling of how much these antibodies could distort the protein," says Perozo. Adds Horn, "These antibodies opened the protein like a flower."

Nonetheless, Ehud Isacoff of UC, Berkeley says that the data is "extremely valuable." MacKinnon's group also expressed and crystallized the voltage sensor alone, apart from the pore domain. "The crystal structure of the isolated S4 region is very reasonable," Isacoff says.

WITHOUT A PADDLE

Many favor a new model in which the S4 region, as folded alone, is docked against the side of the pore domain. This peripheral location for S4 still challenges the conventional idea that S4 is buried among other TMs. Perozo points out, however, that "no direct, very reliable, hard data [has proven] that S4 is entirely surrounded by protein." He has just published a paper showing that one side of S4 almost certainly lies in contact with membrane lipids.⁶ "The common conclusion to all of us," says Isacoff, "is that the crystal structure needs to be adjusted." Perozo agrees: "It is simply a matter of deconvoluting a distorted protein into something that makes sense."

Perhaps because of the controversy, neither Jiang nor MacKinnon were willing to answer specific questions about these issues, although both indicated that they stand by the structure and model as published. Jiang is reluctant to enter the fray because he no longer works on voltage-gated channels. MacKinnon refused multiple requests for an interview, stating that all he has to say on the matter can be found in the literature.

MacKinnon's group continues to publish papers supporting the model. They recently published evidence that at least part of the voltage sensor must be accessible from within the lipid bilayer.⁷ This paper shows that certain tarantula venom toxins must enter the bilayer to reach their target, which is well known to be the voltage sensor itself. Another paper uses electron microscopic analysis to support a location for the voltage sensor at the protein-lipid interface, rather than surrounded by protein as in the conventional model.⁸

Despite the controversy, the KvAP papers are stimulating a great leap forward in understanding how the voltage sensor works. Before these papers, says Garcia, "Everyone was designing experiments to support the old model. Now we have to consider new ideas." Papazian concurs. "We all got thinking in ways we hadn't before. At this point, nobody's ideas are 100% correct."

Researchers in the field continue to praise MacKinnon generously. "In my lifetime I have never encountered a scientist who has had so much of an impact," says Horn. "He's in a class by himself. Nobody can touch him." Echoes Bezanilla, "His impact on the field has been incredible."