A Conserved Domain in Axonal Targeting of Kv1 (Shaker) Voltage-Gated Potassium Channels

See allHide authors and affiliations

Science  01 Aug 2003:
Vol. 301, Issue 5633, pp. 646-649
DOI: 10.1126/science.1086998


Axonal voltage-gated potassium (Kv1) channels regulate action-potential invasion and hence transmitter release. Although evolutionarily conserved, what mediates their axonal targeting is not known. We found that Kv1 axonal targeting required its T1 tetramerization domain. When fused to unpolarized CD4 or dendritic transferrin receptor, T1 promoted their axonal surface expression. Moreover, T1 mutations eliminating Kvβ association compromised axonal targeting, but not surface expression, of CD4-T1 fusion proteins. Thus, proper association of Kvβ with the Kv1 T1 domain is essential for axonal targeting.

Axonal potassium channels in vertebrates and invertebrates control the fidelity of action-potential propagation, especially at sites having a low safety factor such as axonal branch points, varicosities, and nerve terminals (14). Transmitter modulation of these channels regulates synaptic efficacy by altering the likelihood of action-potential invasion (5). In the mammalian nervous system, voltage-gated potassium channels in the Kv1 (Shaker) family also reside in axons (6); loss of their channel function results in behavioral abnormalities (7, 8) and symptoms such as limb-shaking in patients with episodic ataxia (9).

Notwithstanding their long history as the first potassium channel discovered (1) and cloned (10), Kv1 axonal targeting signal remains a mystery. Manipulations blocking channel appearance in axons prevent channel surface expression (11), which compounds analyses because of their impact on the channel exit from the endoplasmic reticulum (ER) (12). Studies of chimeras of Kv1.4 and Kv4.2 identify a C-terminal dileucine motif important for Kv4 dendritic targeting, but not the axonal targeting signal for Kv1 (13).

Cultured hippocampal neurons preserve polarized Kv channel expressions as in vivo and thus provide a good assay for axonal targeting (14). Endogenous Kv1.2 and Kv4.2 exhibited largely nonoverlapping distributions in distal neuronal processes after 21 days in vitro (DIV) (Fig. 1A). Kv4.2 was restricted to microtubule-associated protein type 2 (MAP2)-positive dendrites and concentrated on or near the dendritic surface (Fig. 1C and fig. S1A), whereas Kv1.2 was primarily associated with the surface of Tau1-positive axons and only weakly expressed in intracellular compartments in dendrites (Fig. 1B and fig. S1, B and C).

Fig. 1.

Endogenous Kv1.2 but not Kv4.2 is targeted to the axon. (A) Double labeling revealed segregation of Kv1.2 and Kv4.2 in mature hippocampal neurons (21 DIV). Immunofluorescence intensity profiles, along a 40-μm line (black, lower left) were measured from the original 16-bit TIFF images with NIH Image/J (lower right). AU, arbitrary unit. Further analysis of the boxed area is shown in fig S1A. (B) Kv1.2 was in MAP2-negative axons. (C) Kv4.2 was in MAP2-positive dendrites. Arrow, the main axon; arrowheads, secondary axonal branches. Scale bar, 50 μm.

To monitor surface expression of Kv1.2, we transfected hippocampal neurons at 7 DIV, shortly before endogenous Kv1.2 appears, with Kv1.2 carrying an extracellular hemagglutinin (HA) tag, and examined its expression 2 to 3 days later. Like other axonal membrane proteins (14), HA-tagged Kv1.2 resided in dendrites and axons, but its surface expression was primarily axonal (Fig. 2A). By quantifying the polarized surface expression (fig. S2), we determined the axonal polarity index (API), the ratio of average fluorescence intensity for major axonal versus dendritic branches, to be 3.2 ± 0.4 (n = 16) for HA-tagged Kv1.2, and 3.6 ± 0.3 (n = 15) for HA-tagged Kv1.2 fused to green fluorescence protein (GFP) at its N-terminus (Fig. 2C and fig. S3).

Fig. 2.

T1 is necessary for axonal targeting, whereas a tyrosine-based endocytic signal controls axonal surface expression level of Kv1.2. Arrows and arrowheads point to axons and dendrites, respectively. Asterisks indicate untransfected neurons (they stain for the dendritic marker MAP2 but show no staining of transfected protein carrying HA tag and/or GFP). (A) Preferential expression of Kv1.2 with an extracellular HA tag (Kv1.2HA) on the MAP2-negative axonal surface revealed by HA antibody staining of nonpermeabilized neurons (left panels), as compared with staining of total channel proteins in permeabilized neurons (right panels). (B) Y458A increased Kv1.2 axonal surface expression beyond the proximal portion of the axon (arrow), as revealed by HA antibody staining of nonpermeabilized neurons. (C) GFP-tagged Kv1.2HA (GFP-Kv1.2HA) localized preferentially to axonal surfaces, revealed by immunofluorescence from HA reactivity (Cy5) of channels on the surface of nonpermeabilized neuron, compared with GFP fluorescence for total channel protein. (Quantification given in fig. S3.) Replacing T1 domain with GCN4-L1 coiled-coil tetramerization domain eliminates axonal targeting but not surface expression of Kv1.2. Scale bar, 50 μm.

To search for Kv1 axonal targeting signal, we first mutated the conserved C-terminal PDZ-binding motif. The V499A mutation (in which Val499 is replaced by Ala) reduced surface expression (fig. S4A) but not axonal targeting (API for GFP-Kv1.2HA-V499A: 2.8 ± 0.5, n = 11) (15). We also fused the IRK1 (Kir2.1) forward trafficking signals (16) to the Kv1.2 C-terminus, to disrupt the PDZ binding motif without compromising surface expression, which revealed robust axonal targeting (API for Kv1.2HA-FCYENE: 4.0 ± 0.5, n = 10). Thus, the PDZ binding motif is important for surface expression but not axonal targeting of Kv1.2.

Some axonal membrane proteins achieve their polarized distribution via preferential endocytosis from dendritic membranes (14). We tested this possibility by mutating three potential tyrosine-based endocytic motifs (YXXΦ) (15) of Kv1.2. To our surprise, instead of reducing API, when Tyr is replaced by Ala, the Y458A mutation further increased it as compared with the wild type (API: 6.2 ± 0.4 for Y458A, n = 35; 3.0 ± 0.3 for Y429A, n = 10; 3.7 ± 0.4 for Y489A, n = 13) (Fig. 2B and fig. S4, B to D). As expected for an endocytic signal, Y458A tripled channel surface expression without affecting total protein level in COS-7 cells (fig. S4A). This tyrosine is important for both muscarinic cholinergic modulation and channel interaction with the dynamin- and actin-binding protein cortactin (17), which raises the intriguing possibility of transmitter modulation of axonal Kv1.2 channel density.

We then tested whether the highly conserved Kv1 N-terminus could mediate axonal targeting. Deleting the first 30 amino acids had no effect (API: 3.3 ± 0.5, n = 15), but removing the first 150 amino acids including the T1 tetramerization domain (18) abolished both axonal targeting and surface expression (fig. S5, n = 10). Circumventing T1's requirement for channel folding and assembly (19, 20) by replacing T1 with the GCN4-L1 coiled-coil tetramerization domain to restore surface expression (21), we found the channels without T1 only on the somatodendritic surface (API: 0, n = 10) (Fig. 2C). Thus, T1 is necessary for Kv1.2 axonal targeting.

To test whether the C-terminal domain acted in concert with T1 in axonal targeting, we examined C-terminal truncation mutants. However, removal of amino acids 430 to 499 abolished surface expression. Therefore, we fused either the N- or the C-terminal domain of Kv1 to CD4, which normally exhibits equal distribution on axonal and dendritic surfaces (Fig. 3A), to see whether these domains could confer axonal targeting. Fusion of the C-terminal domain of Kv1.1, Kv1.2, or Kv1.4 to the cytoplasmic C-terminus of CD4 did not enhance axonal surface expression (Fig. 3C and fig. S6A). By contrast, fusion of the Kv1.2 N-terminal domain including T1, but not fusion of either half of T1, tripled the axonal polarity index (Fig. 3, B and C, and fig. S6B) and thus revealed that T1 could promote CD4 axonal targeting.

Fig. 3.

Kv1 but not Kv4 T1 confers axonal targeting to unpolarized CD4 and dendritic transferrin receptor. (A) CD4 surface expression (staining with CD4-specific antibody before permeabilization) on dendrites (staining with MAP2-specific antibody after permeabilization) and axons. Fluorescence intensities were similar for the four 25-μm segments along the axons (a and b, blue) and dendrites (c and d, red) after background subtraction (mean and standard error). (B) Similar staining protocol (top) revealed elevated axonal surface expression of CD4 fused to Kv1.2 T1 (bottom). The straight line in the graph indicates background fluorescence level. Asterisks indicate untransfected neurons. Scale bars, 50 μm. (C) Nonpolarized CD4 (API: 0.9 ± 0.1, n = 24) became polarized upon fusion with Kv1.2 T1. (D) Dendritic TfR-GFP (API: 0.1 ± 0.0, n = 21) was targeted to axons by T1 from three Kv1 channels. For (C) and (D), the number of transfected neurons is given in parentheses.

To test whether T1 from Kv1 channels confers axonal targeting to the dendritic transferrin receptor (TfR), we fused T1 from Kv1.1, Kv1.2, or Kv1.4 to the cytoplasmic N-terminus of TfR and also a GFP tag to its extracellular C-terminus. The GFP-tagged TfR was restricted to the somatodendritic surface (22). Remarkably, fusion of T1 from Kv1.2 or other Kv1 family members, but not from Kv4.2, caused TfR to appear on the axonal surface (Fig. 3D and fig. S7).

To search for clues to the mechanism underlying T1-mediated axonal targeting, we relied on the crystal structure of Kv1.2 T1 tetramer (21) and mutated residues on the “top” or “sides” of T1 tetramer (Fig. 4, A and B). We predicted that mutations of residues “pointing out” (white side chains in Fig. 4, A and B) would disrupt T1 interaction with cellular targeting machinery. However, these five mutations had no effect on axonal targeting of CD4-T1 fusion proteins, neither did mutation of a putative phosphorylation site (Y132F) at the end of T1 (Fig. 4C). We then tested T1 mutations designed to disrupt interaction with Kvβ (23) and found that these mutations (F69V, F69V/D70A, L72W) compromised axonal targeting (Fig. 4C). We further examined five mutations at the interface between T1 monomers (E45K, T46V, and Y90A/Y91A on side A; S40W and R82D on side B), and found these mutations also impaired axonal targeting (Fig. 4C). A majority of these mutations (green side chains in Fig. 4, A and B) may have prevented Kv1.2 channel maturation and/or surface expression. However, they did not affect surface expression of CD4-T1. The use of the CD4 reporter thus enabled us to circumvent the difficulty in dissociating the requirement of T1 (and Kvβ) for ER export of Kv1 channels (12, 16, 19, 20) from their possible involvement in axonal targeting.

Fig. 4.

T1 mutations implicate Kvβ in axonal targeting. (A and B) T1 tetramer viewed from the membrane (top) (A) and the side (B) shows C-terminal helices (red) and N-terminal β sheet (blue) and side chains of mutated residues in one of the four T1 monomers. Mutations of those in green but not those in white affected axonal targeting of CD4-T1 fusion. (C) API of mutant and wild-type CD4-T1 (the same sequence from left to right in the graph, and from top down on the key) determined by analyzing at least 10 transfected neurons in three independent transfections. (D) Coimmunoprecipitation of CD4-T1 and Kvβ2 expressed in HEK 293 cells by monoclonal hCD4-specific antibody, as revealed by the 40-kD band on immunoblot probed with monoclonal Kvβ2-specific antibody (top panel; three bands marked with * are IgG heavy- and light-chain and protein G). The middle and lower blots show input controls for Kvβ2 and CD4-T1 fusions.

Given that all three mutations at T1-Kvβ interface greatly reduced axonal targeting, we tested CD4-T1 association with Kvβ and found all T1 mutations that still targeted CD4 to axons were associated with Kvβ2 coexpressed in HEK 293 cells (Fig. 4D). Remarkably, in addition to mutations of the T1-Kvβ interface, the two mutations on side B of the T1-T1 interface and even the T46V mutation on side A that stabilizes T1 tetramer (21) failed to associate with Kvβ2 (Fig. 4). The other two mutations on side A (E45K and Y90A/Y91A) still associated with Kvβ2. Thus, simply having Kvβ bound to CD4-T1 was not sufficient; rather, proper organization of T1 and Kvβ was important for axonal targeting. Transfected GFP-Kvβ2 showed prominent axonal targeting (fig. S8). In the mammalian brain, Kv1 channels are associated with SNARE proteins (the receptors for soluble NSF attachment proteins), probably because of direct interaction between Kvβ and syntaxin 1A (24). Our study thus lays the groundwork for future analyses of the cellular machinery for Kv1 axonal targeting and channel density regulation.

Supporting Online Material

Materials and Methods

Figs. S1 to S8


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article