Determinants Involved in Kv1 Potassium Channel Folding in the Endoplasmic Reticulum, Glycosylation in the Golgi, and Cell Surface Expression*

Kv1.1 and Kv1.4 potassium channels are expressed as mature glycosylated proteins in brain, whereas they exhibited striking differences in degree of trans-Golgi glycosylation conversion and high cell surface expression when they were transiently expressed as homomers in cell lines. Kv1.4 exhibited a 70%trans-Golgi glycosylation conversion, whereas Kv1.1 showed none, and Kv1.4 exhibited a ∼20-fold higher cell surface expression level as compared with Kv1.1. Chimeras between Kv1.4 and Kv1.1 and site-directed mutants were constructed to identify amino acid determinants that affected these processes. Truncating the cytoplasmic C terminus of Kv1.4 inhibited its trans-Golgi glycosylation and high cell surface expression (as shown by Li, D., Takimoto, K., and Levitan, E. S. (2000) J. Biol. Chem. 275, 11597–11602), whereas truncating this region on Kv1.1 did not affect either of these events, indicating that its C terminus is not a negative determinant for these processes. Exchanging the C terminus between these channels showed that there are other regions of the protein that exert a positive or negative effect on these processes. Chimeric constructs between Kv1.4 and Kv1.1 identified their outer pore regions as major positive and negative determinants, respectively, for both trans-Golgi glycosylation and cell surface expression. Site-directed mutagenesis identified a number of amino acids in the pore region that are involved in these processes. These data suggest that there are multiple positive and negative determinants on both Kv1.4 and Kv1.1 that affect channel folding,trans-Golgi glycosylation conversion, and cell surface expression.

Potassium (K ϩ ) channels are expressed by most cells and play important roles in cell physiology, including setting the resting membrane potential and repolarizing/modulating action potential waveforms in excitable tissues (1)(2)(3). Molecular cloning methods have identified a large, diverse family of voltage-gated ␣-K ϩ channels that includes the four subfamilies Shaker (Kv1), Shab (Kv2), Shaw (Kv3), and Shal (Kv4) (4,5). Functional channels appear to be composed of noncovalently associated homo-and/or heteromeric tetramers of subunits within a subfamily, and numerous amino acid determinants involved in channel operation have been identified (4,5).
Kv1.1 and Kv1.4 (6) are ␣-subunits of the brain dendrotoxin-binding protein (7), are heavily glycosylated proteins (7)(8)(9)(10) that form heteromers (7,11,12), and are associated with cytoplasmic Kv␤-subunits (12,13). These glycoproteins contain 20 -25 kDa of carbohydrate per denatured monomer that was PNGase F-sensitive but Endo H glycosidase-insensitive, and they were also sialidase-sensitive (10). Endo H cleaves only high mannose-type immature N-linkages found on ER 1 glycoproteins and some plasma membrane proteins, whereas PN-Gase F cleaves all N-linkages (high mannose-, hybrid-, and complex-type linkages) (14). These findings imply that these glycoproteins were efficiently processed and sialidated in the trans-Golgi in native brain tissue. Glycosylation is a common posttranslational modification of membrane proteins and is thought to aid in their proper folding/oligomerization and, in some cases, targeting and function. Studies have suggested that glycosylation modified the expression or function of some channels (15)(16)(17)(18)(19)(20)(21)(22), whereas in other cases, no change was reported (23,24). The amino acid determinants responsible for the efficient oligomerization, glycosylation, processing, and targeting of K ϩ channels to the cell surface are not well understood (25,26). Tetramers appear to oligomerize in the ER, and motifs involved in this include the amino acid regions of the distal N terminus and the S1 and the S2 membrane-spanning domains (25,26). A number of studies using cell lines transfected with Kv1 subfamily cDNAs have noted some differences in the extent of glycosylation of the expressed homomeric channel proteins versus those in native brain tissue, where the channels are all expressed as mature glycoproteins (7, 10, 15, 23, 28 -30). Other work indicates that glycosylation and/or cell surface expression of some Kv␣-subunits may be governed by a cytoplasmic C terminus VXXSL motif (28), that various Kv␣-subunits may be differentially glycosylated and expressed on the cell surface (10,30), and that cytoplasmic Kv␤-subunits promote the processing of some Kv␣-subunits (31). The above-mentioned findings imply that for both efficient trans-Golgi glycosylation and high surface expression of heteromeric Kv1 subfamily channel members, there must be a favorable heteromeric subunit composition and/or stoichiometry. A recent report has identified regions on Kir-type K ϩ channels that appear to be possible ER export signals (32). Ion channel trafficking defects associated with human disorders have been summarized and point to the importance of correct protein folding, oligomerization, and targeting to the cell surface for normal function (33).
Given the above-mentioned findings, the aims of this report were to use Kv1. 4 and Kv1.1 chimeras and site-directed mu-tants to map amino acid determinants involved in their proper folding in the ER, efficient trans-Golgi glycosylation conversion, and high cell surface expression.

EXPERIMENTAL PROCEDURES
Cell Lines, cDNAs, and Transfections-Chinese hamster ovary (CHO) pro5 cells were obtained from American Type Culture Collection (Manassas, VA) and maintained in Dulbecco's modified Eagle's medium, or in ␣-minimum Eagle's medium, supplemented with 0.35 mM proline, with 10% fetal bovine serum at 37°C under 5% CO 2 . Cath-adifferentiated (CAD) cells were obtained from Dr. J. Wang (34), and a subclone from these cells was used. CAD cells are a subclone of Cath-a cells that were derived from mouse central nervous system catecholaminergic brain stem neurons by SV40 large T immortalization. CAD cells have lost the expression of large T antigen but remain immortalized and neuronal-like. Undifferentiated CAD cells were maintained in Dulbecco's modified Eagle's medium/F-12 with 10% fetal bovine serum as described above for CHO cells. Rat brain Kv1 cDNAs (6) were engineered using the polymerase chain reaction (PCR) to contain a 5Ј Kozak enhanced ribosomal binding sequence (CCACC) before the start methionine and no endogenous 5Ј-untranslated regions or 3Ј-untranslated regions. C terminus truncation mutants were constructed by PCR by engineering a stop codon and a unique restriction enzyme site on the 3Ј end of the cDNAs. Chimeras were constructed by PCR by using sense and antisense oligonucleotides. As an example, the strategy to exchange the C terminus of Kv1.1 and Kv1.4 will be described. The identical amino acid sequence NYFYH (just past the C terminus end of the S6 transmembrane segment) in both Kv1.1 and Kv1.4 has an identical nucleotide sequence for both channels (the oligonucleotide pairs used were AACTATTTCTACCAC/SP6 and GTG-GTAGAAATAGTT/T7. Appropriate fragments from the first PCR reaction were then gel-isolated/purified and used in a second PCR reaction to construct the Kv1.4-C terminal Kv1.1 and Kv1.1-C terminal Kv1.4 chimeras. Other chimeras were constructed using a similar approach. Site-directed mutagenesis was performed by PCR using sense and antisense oligonucleotides by standard methods. cDNAs were subcloned into eucaryotic expression vector pcDNA3, and the integrity of constructs was confirmed by DNA sequencing. Transient transfections were performed using LipofectAMINE Plus (Life Technologies, Inc.) per the manufacturer's protocol on cells plated in a 35-mm dish, using 0.5 g of cDNA/dish unless otherwise stated. The different mRNAs produced from these cDNAs in cell lines would have identical 5Ј-untranslated regions and 3Ј-untranslated regions derived from only the vector; therefore, their mRNA stabilities and translation abilities would be predicted to be similar.
Membrane Isolation, Glycosidase Treatment, and Immunoblot Analysis-Membrane proteins from rat brain were prepared as described previously (15). Cell lines were processed 20 -24 h after transfection. Cells were scraped and homogenized in ice-cold hypotonic media containing protease inhibitors (K ϩ phosphate buffer (pH 7.4) containing 2 mM EDTA, 1 mM pepstatin A, 1 mM 1,10-phenanthroline, and 0.2 mM phenylmethylsulfonyl fluoride). The homogenate was centrifuged (16,000 ϫ g) at 5°C for 1 h to pellet membranes, which were stored at Ϫ85°C. Endo H and PNGase F glycosidases (Roche Molecular Biochemicals) were used per the manufacturer's protocol, with the final concentration of glycosidases at 0.16 unit/ml and 13 units/ml, respectively, for 20 -24 h at 37°C. ϳ25% of the membranes recovered from a 35-mm well were used to run SDS-9% gels (ϳ20 g of crude membrane protein/gel lane). The proteins were electrotransferred to nitrocellulose (Bio-Rad), and the filter was blocked with 5% nonfat milk in PBS and then incubated in primary antibody overnight (Kv1.1 rabbit polyclonal N terminus Ab (to amino acids 4 -27; dilution, 1:1000) or Kv1.4 mouse monoclonal N terminus Ab (to amino acids 13-37; dilution; 1:1000) (Upstate Biotechnology)). Specificities have been shown for the monoclonal (Trimmer laboratory,Ref. 35) and the polyclonal (Lotan laboratory (Ref. 36) and Thornhill laboratory (Ref. 15)) antibodies. After washing, horseradish peroxidase-linked anti-rabbit secondary or antimouse secondary antibodies were added, and the bound antibodies were detected using enhanced chemiluminescence (ECL detection kit; Amersham Pharmacia Biotech) and Kodak XAR5 film. Filters were exposed to preflashed x-ray film for various lengths of time. Densitometry analysis of signals was performed as outlined under "Cell Surface Biotinylation." The trans-Golgi glycosylation conversion percentage equaled the upper band signal/total band signal (upper ϩ lower band).
Cell Surface Biotinylation-Cells in 35-mm wells were transfected with 0.5 g of plasmid containing channel cDNA and 0.1 g of plasmid containing green fluorescent protein (GFP) as described above. After a 20-h incubation, the cells were prepared for biotinylation of membrane proteins that contained carbohydrates following the Pierce protocol. All procedures were carried out at 5°C unless stated otherwise. Cells were washed in PBSϩ (PBS with 0.1 mM CaCl 2 and 1 mM MgCl 2 ) and incubated in 10 mM NaIO4 in PBSϩ in the dark for 20 min to oxidize cell surface carbohydrates. After washing in PBSϩ, the cells were incubated with 2 mM hydrazide-LC-biotin (Pierce) in 100 mM sodium acetate (pH 5.5) for 20 min in the dark, followed by washing in PBSϩ. The cells were incubated in 1 ml of lysis buffer (150 mM NaCl, 1% Nonidet P-40, 0.5% deoxycholate, 50 mM Tris (pH 7.4), and the protease inhibitors listed in the previous section) for 30 min and then centrifuged to pellet any insoluble material. The supernatants were transferred to another tube, an aliquot was saved for actin and GFP immunoblotting, 75 l of streptavidin-agarose beads (1:1 slurry; Pierce) were added, and the tube was rocked overnight. The beads were washed, eluted with SDS buffer, and run on SDS gels for immunoblotting/detection with Kv1.1 N terminus Ab or Kv1.4 N terminus Ab as described above. Aliquots of the original lysis supernatant were run on SDS gels for immunoblotting with actin Abs (Sigma) or GFP Abs (CLONTECH) to normalize Kv1 signals to control for possible cell density differences or transfection efficiency differences, respectively. A film image was captured with a microtex 8700 scanmaker (dynamic range of 0 -3.5), and densitometry analysis of the differences among lanes from the same film was performed with National Institutes of Health Image 1.6 software that had its own internal calibration. In addition, further external calibration was performed from a scan of an external Kodak 21-step optical density tablet. A standard curve using different dilutions of cell membranes from transfected cells indicated that immunoblot signals were linear over at least a 1-10-fold concentration range. Values are expressed normalized to the Kv1.4 signal taken as 100.0.
Electrophysiological Recordings and Analysis-CHO cell lines were plated on glass coverslips and cotransfected with a plasmid containing channel cDNA (0.5 g) and a plasmid containing GFP cDNA (0. Whole cell currents were recorded and analyzed with an Axopatch 200B amplifier (Axon Instruments) and the pclamp8/Digidata 1200 data acquisition/analysis system (Axon Instruments) 20 -30 h after transfection at room temperature (23°C to 25°C) (37). Cells were held at Ϫ80 mV and depolarized to a maximum activating voltage of 50 -80 mV. Nontransfected CHO pro5 cells showed either no or at most 100 pA of endogenous K ϩ current at maximum activating voltages of 50 mV (15). Because Kv1.4 shows accumulative inactivation (38), the Ϫ80 mV holding potential duration was Ͼ30 s before a test potential was applied. Mean linear membrane leak current was subtracted by the P/4 protocol. Membrane capacitance and series resistance (Rs) were estimated by the pclampex software from a transient capacitance current elicited by a 10-mV hyperpolarizing voltage step from Ϫ80 mV. Membrane capacitance was compensated using the amplifier controls, and Rs was compensated by patch clamp circuitry by 85-95%. The voltageinduced currents were filtered at 5 kHz (Ϫ3dB cutoff) and digitized at 50 s. Maximum peak conductance values (G) were obtained from the mean value of the peak leak-subtracted current (I) using Ohm's law (G ϭ I/(Vp Ϫ EK)) and a predicted Nernst K ϩ equilibrium potential (EK) of Ϫ83 mV. A test voltage value was corrected for the change of V introduced by the residue uncompensated Rs (change of V ϭ I Rs; I is the net voltage-gated current). An unpaired t test was used to assess statistical differences of a control value compared with another one. p Ͻ 0.05 was considered significant.

Trans-Golgi Glycosylation and Cell Surface Expression
Differences-In rat brain, Kv1.4 and Kv1.1 are expressed as mature glycoproteins of 110 kDa (appears to be a doublet) and 80 kDa, respectively (Fig. 1B, lanes 1 and 6, respectively). Both p110 and p80 were Endo H-insensitive glycoproteins that contained 20 -25 kDa of Endo F-sensitive carbohydrate (data not shown), which implied they were efficiently processed and glycosylated in the Golgi/trans-Golgi at their one extracellular N-glycosylation site (Fig. 2, A and C). cDNA analysis showed that Kv1.4 and Kv1.1 have high amino acid identity (79% identical from their cytoplasmic N terminus T1 region through their S6 transmembrane domain), and their calculated core protein molecular masses are 73.5 kDa (655 amino acids) and 56.4 kDa (495 amino acids), respectively (6). Kv1.4 has a longer cytoplasmic N terminus than Kv1.1. When the homotetramers were expressed in CHO cells, Kv1.4 exhibited efficient trans-Golgi glycosylation with significantly more mature p110 glycoprotein band than the immature p85 glycoprotein band, whereas Kv1.1 predominantly expressed an immature p60 glycoprotein band that was Endo H-sensitive (data not shown), indicating ER processing, and there was little indication of a mature p80 band (Fig. 1B, lanes 2 and 7, respectively). Evidence suggests that these differences are due to properties of the Kv1 channel core proteins and that cotransfection with Kv␤-subunit cDNAs did not promote their trans-Golgi glycosylation (29). Kv1.4 homomers expressed in CHO cells also exhibited higher cell surface expression than Kv1.1 homomers as assayed by patch clamping (Fig. 5A, representative whole cell current traces; Fig. 5B, maximum conductance per capacitance). Because Kv1.4 has a smaller single channel conductance (Kv1.4 ϭ 4 picosiemens and Kv1.1 ϭ 10 picosiemens), fast N-type inactivation, and fast C-type inactivation compared with Kv1.1 (6), we predict that the ϳ10-fold higher difference for Kv1.4 versus Kv1.1 recorded in Fig. 5B was more likely closer to ϳ20-fold difference in surface protein expression, and biotinylation of surface glycoproteins suggests that this is the case (Fig. 5, C and D). Thus Kv1.4 and Kv1.1 exhibited very different degrees of efficient trans-Golgi glycosylation and high cell surface expression when expressed as homomers in cell lines, and for both processes, Kv1.4 showed a higher level of glycosylation conversion and surface expression as compared with Kv1.1. Therefore, they are ideal candidates to be used to examine which regions of each protein contain positive or negative determinants for these two processes.
Effects of Truncating or Exchanging the Cytoplasmic C Terminus on Trans-Golgi Glycosylation and Cell Surface Expression-C terminus truncation mutants (Fig. 1A) and chimeras with their cytoplasmic C termini exchanged (Fig. 2B, chimera 1 and 2) were constructed, and their cDNAs were transiently transfected into CHO cells.
Full truncation of the C terminus of Kv1.4 (see Ref. 28) inhibited its trans-Golgi glycosylation dramatically for the Kv1.4 T1 mutant, whereas Kv1.4 T2 and T3 mutants with more of the C terminus showed some trans-Golgi glycosylation conversion of ϳ40% versus ϳ70% for wt Kv1.4 (Fig. 1B, compare lanes 1 and 2 with lanes 3-5). C-terminal truncation also dramatically inhibited cell surface expression of the Kv1.4 T1 mutant as monitored by patch clamping (Fig. 5, A and B). Because Kv1.1 exhibited very inefficient trans-Golgi glycosylation, we truncated its C terminus to test whether it was a negative determinant for this. Kv1.1 truncation mutants T1, T2, and T3 still exhibited inefficient trans-Golgi glycosylation (Fig. 1B, compare lane 6 with lanes 7-10), and the Kv1.1 T1 mutant also showed low cell surface expression levels similar to those of wt Kv1.1 (Fig. 5, A and B). For the truncation mutants, the predominant lower band for Kv1.4 and the bands for Kv1.1 (Fig. 1B, lanes 3-5 and 8 -10, respectively) were Endo H-sensitive, which indicated that they were immature glycoproteins that were not trans-Golgi glycosylated (data not shown).
We next tested the effects of exchanging the cytoplasmic C terminus of Kv1.4 and Kv1.1 on trans-Golgi glycosylation and cell surface expression. If the C terminus of Kv1.4 is required for both of these processes, then exchanging it with the C terminus of Kv1.1 is predicted to inhibit them. Chimera 1, which is Kv1.4 with the C terminus of Kv1.1, exhibited a significant trans-Golgi glycosylation conversion of 34% (Fig.  3A, lane 1), but it was less than the trans-Golgi glycosylation conversion of ϳ70% for wt Kv1.4. The intermediate band in lane 1 of Fig. 3A was not trans-Golgi glycosylated because it was sensitive to Endo H digestion (Fig. 3A, lane 2), and this band was responsible for the lower trans-Golgi glycosylation conversion. It appears that this band had an additional posttranslational modification(s) that may be due to phosphorylation of the C terminus of Kv1.1 at S446 in chimera 1 (36). In contrast, there was a dramatic reduction in cell surface expression of chimera 1, and it was expressed at low levels like wt Kv1.1 using patch clamping (Fig. 5, A-C) and biotinylation of cell surface glycoproteins (Fig. 5, C and D). The biotinylation method gave a lower expression level than patch clamping for this chimera (Fig. 5C), which suggests that the Kv1.1 C terminus on Kv1.4 resulted in current enhancement. Also note that Kv1.4 and C1 cell surface biotinylated proteins in the immunoblot in Fig. 5D are enriched in trans-Golgi glycosylated proteins when compared with other whole cell membrane immunoblots (see the other figures). Will the Kv1.4 C terminus transplanted to Kv1.1 stimulate efficient trans-Golgi glycosylation and high cell surface expression? It appeared that it did not. Chimera 2, which is Kv1.1 with the C terminus of Kv1.4, displayed no trans-Golgi glycosylation (Fig. 3A, lane 4) as well as low cell surface expression levels like wt Kv1.1 as monitored by patch clamping (Fig. 5, A-C) and biotinylation methods (Fig.  5, C and D).
These results suggest that trans-Golgi glycosylation and cell surface expression for Kv1.4 are differentially affected by transplanting the Kv1.1 C terminus on it. Although chimera 1 exhibited inhibition of both trans-Golgi glycosylation and cell surface expression versus wt Kv1.4, there was a much more dramatic effect on the inhibition of cell surface expression. In addition, the C terminus of Kv1.4 transplanted to Kv1.1 did not stimulate its efficient trans-Golgi glycosylation or its high cell surface expression.
Effects of Exchanging N-terminal Regions or Pore Regions on Trans-Golgi Glycosylation and Cell Surface Expression-The results above suggest that there are other dominant-negative and dominant-positive determinants in the N-terminal region (the N terminus through the S6 region, see Fig. 2) of these channels involved in these processes, and a number of chimeras between Kv1.4 and Kv1.1 were constructed to map them. We first transplanted the extracellular S1-S2 loop of Kv1.1 to Kv1.4, which contains the only N-linked site that is glycosylated, to test whether efficient trans-Golgi glycosylation was affected by the different amino acids in this region (Fig. 2, B  and C). No evidence was found that the extracellular S1-S2 loop of Kv1.4 was required for high trans-Golgi glycosylation or high cell surface expression because the loop of Kv1.1 on Kv1.4 chimera 3 was efficiently glycosylated (Fig. 3B, lane 2), and the chimera exhibited similar surface expression levels as wt Kv1.4 (Fig. 5B). Progressively replacing the N-terminal region of Kv1.4 with Kv1.1 (Fig. 2B, chimeras 4 -6) inhibited trans-Golgi glycosylation somewhat versus wt Kv1.4 (Fig. 3B, lane 3-5), but the conversion was still much greater than Kv1.1; cell surface expression for chimeras 4 and 5 showed little change, whereas there was some inhibition for chimera 6 (Fig. 5B).
We next constructed chimeras between Kv1.4 and Kv1.1 in their pore region (Fig. 2B, chimeras 7-16). In the S5-pore-S6 region of Kv1.4 and Kv1.1, there are only nine amino acid differences, five in the outer pore region a, three in the outer pore region b, and one in the deep pore region c (Fig. 2D). Chimera 7, which is Kv1.4 with both the outer pore regions a and b of Kv1.1, exhibited little or no trans-Golgi glycosylation (Fig. 3C, lane 2; Fig. 5D, lane 3) and low cell surface expression levels similar to those of wt Kv1.1 using both patch clamping (Fig. 5, B and C) and biotinylation methods (Fig. 5, C and D). Chimera 16, which is Kv1 .1 with regions a, b, and c of Kv1.4, exhibited a dramatically efficient trans-Golgi glycosylation (Fig. 3C, lane 12) and a high cell surface expression that was similar to that of wt Kv1.4 by both patch clamping (Fig. 5 and C) and biotinylation methods (Fig. 5, C and D). Also note that C16 cell surface biotinylated proteins in the immunoblot in Fig. 5D are enriched in trans-Golgi glycosylated proteins when compared with other whole cell membrane immunoblots (compare with Fig. 3C, lane 12).
Having identified the pore region as containing positive and negative determinants for Kv1.4 and Kv1.1, respectively, for both trans-Golgi glycosylation and cell surface expression, we next sought to identify which of these subregions were involved. The outer pore region a or b was exchanged between Kv1.4 and Kv1.1 (Fig. 2B, chimeras 8 -12). For Kv1.4, replacing its a or b region with those of Kv1.1 (chimeras 8 and 9, respectively) inhibited both trans-Golgi glycosylation (Fig. 3C, lanes 3  and 4) and cell surface expression (Fig. 5B) versus wt Kv1.4, but the inhibition in both cases was less than that recorded for chimera 7, which has both of these regions of Kv1.1. For Kv1.1, replacing its a and b regions with those of Kv1.4 (chimera 10) did not stimulate trans-Golgi glycosylation (Fig. 3C, lane 6), and no detectable currents were recorded from this chimera (Fig. 5B). Kv1.1 with the a or b region of Kv1.4 (chimeras 11 and 12, respectively) did not rescue trans-Golgi glycosylation (Fig. 3C, lanes 7 and 8) or high cell surface expression for chimera 11 (although it was expressed slightly better than wt Kv1.1), whereas no detectable currents were recorded from chimera 12 (Fig. 5B). Kv1.1 with the a and c regions of Kv1.4 (chimera 13) exhibited trans-Golgi glycosylation (Fig. 3C, lane  9) and cell surface expression (Fig. 5B) that was greater than those of Kv1.1 but less than those of Kv1.4. Kv1.1 with the c region of Kv1.4 (chimera 14; S to T mutation) showed trans-Golgi glycosylation (Fig. 3C, lane 10) that was greater than that of Kv1.1 (with filter/film overexposure) but much less than that of Kv1.4, whereas cell surface expression levels (Fig. 5B) were greater than those of Kv1.1 and only somewhat less than those of Kv1.4. Kv1.1 with the b and c regions of Kv1.4 (chimera 15) exhibited efficient trans-Golgi glycosylation, although somewhat less than that of Kv1.4 (Fig. 3C, lane 11), and high cell surface expression (Fig. 5B) that was similar to that of Kv1.4 but less than that of chimera 16. All lower bands in Fig.  3, B and C, were core N-glycosylated in the ER because they were sensitive to treatment with Endo H glycosidase (data not shown).
These results suggest that the pore region of Kv1. , and either Kv1.1 pore region a or b alone (chimeras 8 and 9, respectively) inhibited these processes, but to a lesser extent than they did when together.

Effects of Point Mutations in the Pore Region of Kv1.4 on Trans-Golgi Glycosylation and Cell Surface Expression-A
number of point mutations were then constructed in the outer pore regions of Kv1.4 to map the amino acid determinants involved in its efficient trans-Golgi glycosylation and high cell surface expression. All of the amino acids in the outer pore regions a and b of Kv1.4 were individually mutated to the corresponding ones in Kv1.1 (Fig. 2D). Cells transfected with these constructs were analyzed by immunoblotting, patch clamping, and biotinylation. For trans-Golgi glycosylation conversion, our results suggest that the mutants can be arranged into three groups: 1) mutations that did not alter conversion (T507E, T508S, and I535V, Fig. 4A, lanes 4, 6, and 9, respectively); 2) mutations that had an inhibitory effect, but some conversion could be detected with filter/film overexposure (D504E, P506A/T507E, Q511S, and V537I, Fig. 4A, lanes 2, 5,  7, and 10, respectively); and 3) mutations that had a dramatic inhibitory effect, and little or no conversion could be detected with filter/film overexposure (P506A and K533Y, Fig. 4A, lanes  3 and 8). For cell surface expression levels estimated by patch clamping (Fig. 5, E and F) and biotinylation (Fig. 5, F and G), the mutants did not always fall into the same groups as described above. For example, group 1 members had high cell surface expression similar to wt Kv1.4, but so did some members of group 2, e.g. D504E and V537I. Group 3 members (P506A and K533Y) exhibited low surface expression versus Kv1.4. Patch clamping and biotinylation gave similar estimates for cell surface expression for most of the mutants. A major exception appears to be the Q511S mutant, which exhib- ited a low surface expression level when assayed by biotinylation (Fig. 5, F and G) but exhibited a high surface expression level by patch clamping (Fig. 5, E and F), which suggested that the mutation changed single channel parameters, leading to current enhancement.
We next transfected CAD neuronal cells derived from the mammalian central nervous system (34) with some selected chimeras and point mutants and essentially recorded similar results for trans-Golgi glycosylation conversion (Fig. 4B) as we did with CHO cells (Figs. 3C and 4A), suggesting that the bulk of the intracellular glyco-processing machinery appears to be similar in both a CHO fibroblast-like cell line and a neuronal cell line. However, CAD cells consistently produced a mature Kv1.4 glycoprotein that had a slightly smaller molecular mass as compared with CHO cells (upper band in Fig. 4B, lane 1 is p105, whereas the upper band is p110 in CHO cells in Fig. 4A, lane 1), suggesting that there are some differences in the glycoprocessing enzymes between the two cell lines.
These results suggest that a number of amino acid substitutions in the outer pore region of Kv1.4 to corresponding ones in Kv1.1 have a dramatic effect on efficient trans-Golgi glycosylation and high cell surface expression. The Kv1.4 determinant site was maximally disrupted by the mutation P506A or K533Y in pore region a and b, respectively; either mutation inhibited trans-Golgi glycosylation and reduced cell surface expression.
Immunofluorescence Cellular Localization of Selected Kv1.4 and Kv1.1 Constructs-Selected Kv1 proteins were localized in fixed/permeabilized transiently transfected cell lines by using specific antibodies and immunofluorescence microscopy to determine a channel construct's predominant location (data not shown). Specific antibodies to an ER protein (BiP) and a Golgi protein (p230 or GM130) were also used to help determine whether channels were localized at high or low levels to these internal organelles. Kv1.1 showed a reticulate-like signal that radiated out from the nucleus, suggestive of high retention in the ER, whereas Kv1.4 did not show this signal but rather exhibited a diffuse cell surface signal and a high signal at the cell's perimeter. Similar results have been shown for Kv1.1 and Kv1.4 in transfected COS7 cells (35). Chimera 1, Kv1.4 with cytoplasmic Kv1.1 C terminus, exhibited a unique pattern of intracellular localization. Chimera 2, Kv1.1 with the cytoplasmic Kv1.4 C terminus, showed a pattern similar to Kv1.1 suggesting that it continued to exhibit high retention in the ER. Chimera 16, Kv1.1 with the pore of Kv1.4, showed a localization pattern similar to Kv1.4, whereas chimera 7, Kv1.4 with the outer pore regions of Kv1.1, exhibited a pattern similar to Kv1.1. Overall, these data suggest that our patch clamp and biotinylation data are a good reflection of cell surface expression levels and that mutations here that caused reduced voltage-gated currents are due to partial intracellular organelle retention and not high cell surface expression of dysfunctional channels.

DISCUSSION
Evidence has been presented in this report that Kv1.4 and Kv1.1 have multiple positive and negative determinants that affect channel folding and retention in the ER, trans-Golgi glycosylation conversion, and cell surface expression. It has been suggested that Shaker-like potassium channels are oligomerized in the ER by dimerization of monomers followed by dimerization of dimers to produce properly folded tetramers (26). It also appears that Shaker-like channels in the ER are functional because K ϩ channel activity has been recorded from planar lipid bilayers that had fused rough microsomes isolated from a cell-free protein processing system programmed with Shaker B1(6 -46) cRNA (Ref. 17; see Ref. 25 for review). Thus, it appears that Shaker-like channels in the ER are only transported to the Golgi as properly oligomerized tetramers, and additional glyco-processing may occur in the Golgi (including trans-Golgi glycosylation).

The Kv1.4 C Terminus Is Required for High Cell Surface Expression but Not Trans-Golgi Glycosylation on Kv1.4, but It
Does Not Affect These Processes When Transplanted to Kv1.1-Truncating the C terminus of Kv1.4 inhibited both efficient trans-Golgi glycosylation conversion and high cell surface expression as first reported by Li et al. (28). In addition, we found that truncating the C terminus of Kv1.1 did not affect either of the processes, which indicates that it does not play a role as a negative determinant on the wild type Kv1.1. Exchanging the C termini between Kv1.4 and Kv1.1 produced results that were informative. Chimera 1, Kv1.4 with the C terminus of Kv1.1, still exhibited a trans-Golgi glycosylation conversion of ϳ35% but showed low cell surface expression versus wt Kv1.4. Moreover, chimera 2, Kv1.1 with the C terminus of Kv1.4, did not exhibit either efficient trans-Golgi glycosylation or high cell surface expression. Chimera 1 and 2 results suggest that: 1) 35% efficient trans-Golgi glycosylation can be obtained without high cell surface expression, so that these two processes appear to be directed by different determinants on Kv1.4; 2) the Kv1.4 C terminus contains a determinant for high cell surface expression that is presumably the VXXSL motif as described by Li et al. (28), and the C terminus of Kv1.1 cannot substitute for it; and 3) the C terminus of Kv1.4 transplanted to Kv1.1 cannot rescue either efficient trans-Golgi glycosylation or high cell surface expression; it appears that the outer pore region of Kv1.1 contains dominant-negative determinants for these processes. What explains the finding that truncating the C terminus of Kv1.4 (the T1 mutant) blocked trans-Golgi glycosylation but that transplanting the C terminus of Kv1.1 to Kv1.4 did not? The simplest explanation is that the majority of Kv1.4 truncation mutants were partially retained in the ER or shuttled off to the cytoplasm for degradation or to newly described cytoplasmic structures called aggresomes (39). This would result in both inefficient trans-Golgi glycosylation and low or no cell surface expression. In contrast, chimera 1, which is Kv1.4 with the C terminus of Kv1.1, was folded and oligomerized properly in the ER and transported to the Golgi, where it was somewhat efficiently glycosylated. However, once in the trans-Golgi or trans-Golgi network, it was inefficiently delivered to the cell surface because it did not have the C terminus of Kv1.4.
Thus, it appears that the Kv1.4 C terminus contains determinants recognized by the trans-Golgi processing machinery involved in the high cell surface delivery of Kv1  10 -16). The outer pore regions a and b of Kv1 channels may be in close proximity to each other because these equivalent regions in the Ksca K ϩ channel (Fig. 2D) are in close juxtaposition in the three-dimensional crystal structure of the protein (40). The a region forms a protruding turret on the perimeter of the external face of each Ksca subunit, whereas the b region is somewhat lower and is positioned inward from that position. If the a and b regions of Kv1 channels fold in a similar fashion, then our mapped determinant is composed of regions that are nonadjacent in the primary sequence but are situated close to one another in the folded three-dimensional state. Kv1.1 with pore regions b and c or a, b, and c of Kv1.4 (chimeras 15 and 16, respectively) exhibited both high trans-Golgi glycosylation and cell surface expression without having the identified high expression cytoplasmic C terminus VXXSL motif (28) of Kv1.4. Therefore, Kv1.4 and Kv1.1 have positive and negative determinants, respectively, for both of these processes, but Kv1.4 also requires its own cytoplasmic C terminus for the process of high cell surface expression.
Point mutations in both outer pore regions a and b of Kv1.4 to identical amino acids in Kv1.1 identified a number of determinants critical for Kv1.4's proper folding in the ER, efficient trans-Golgi glycosylation, or high cell surface expression. All of the point mutations exhibited cell surface expression as assayed by patch clamping or surface biotinylation. Group 1 mutants T507E or T508S in region a and I535V in region b did not affect any of these processes and were indistinguishable from wt Kv1.4. Group 2 mutations such as D504E in region a and V536I in region b resulted in inefficient trans-Golgi glycosylation of channels but did not affect their high cell surface expression versus wt Kv1.4; these mutations appear to be determinants only for efficient glycosylation in the Golgi, and they do not affect the channel's proper folding and rapid exit from the ER. Group 3 mutations P506A and K533Y in region a and b, respectively, presumably caused partial retention in the ER of channels and reduced both trans-Golgi glycosylation and surface expression versus wt Kv1.4; these mutations inhibited proper folding and rapid exit from the ER, which is the probable cause of their reduced cell surface expression, but they also disrupted the determinant for efficient glycosylation in the Golgi.
Thus, it appears that wild type Kv1.4's outer pore regions a and b promote its efficient folding and rapid exit from the ER as well as containing positive determinants for efficient trans-Golgi glycosylation, but its cytoplasmic C terminus is required for high cell surface expression. For wt Kv1.1, outer pore regions a and b inhibit efficient folding and rapid exit from the ER, which appears to be the cause of its low cell surface expression, and contain negative determinants for trans-Golgi glycosylation. However, a chimeric Kv1.1 with the pore of Kv1.4 was completely rescued and exhibited rapid exit from the ER, efficient trans-Golgi glycosylation, and high cell surface expression.
Although the trans-Golgi glycosylation of Kv1.1 in transiently transfected cells was extremely low, we have selected a stably transfected CHO cell clone expressing Kv1.1 that exhibited a higher conversion ratio (15), but this ratio was still much less than the conversion of Kv1.4. Furthermore, the Kv1.1 trans-Golgi glycosylation conversion in Kv1.1 cRNA-injected oocytes was also low, but a mature p80 band (similar to that found in brain) was enriched in plasma membranes (36). Apparently, some expression systems are capable of trans-Golgi glycosylation of exogenously expressed homomeric Kv1.1, but they do so inefficiently.
Other Mutagenesis Studies on the Outer Pore Regions of K ϩ Channels-Mutagenesis studies in Kv1 a and b outer pore regions, which have the greatest amino acid sequence differences in the pore in a subfamily, suggest that they are important in defining different properties among closely related channels such as external IC 50 block for tetraethylammonium ion and various toxins, slow C-type inactivation, modulation by external pH, and, in some cases, modulation by external K ϩ and single channel conductance (4,5,27). Our study also suggests that these regions are important determinants on channels involved in the effects outlined in this report. However, it is not apparent from inspection of equivalent a and b sequences in other Kv1 channels or in Kv2, Kv3, and Kv4 channels that there is a highly conserved motif in these regions. Kv1.5 exhibits high cell surface expression that is only somewhat less than that of Kv1.4 (28), and its a region is distinct from Kv1.4. Shaker B also exhibits high surface expression and trans-Golgi glycosylation, although its a and b regions are as different from Kv1.4 as Kv1.1 sequences are. It may be that the secondary structures of the a and b regions are more important than their primary sequence in forming this combined region motif.
N-terminal Regions-Progressively replacing the N-terminal region of Kv1.4 with Kv1.1 regions (chimeras 4 -6) inhibited efficient trans-Golgi glycosylation but did not affect high cell surface expression for chimeras 4 and 5. We propose that this finding is most likely due to these chimeras (N terminus and S1-S4 helixes of Kv1.1 and the S5-pore-S6 region of Kv1.4) folding in such a manner as to change the conformation of the positive determinants in the outer pore Kv1.4 regions a and b required for efficient glycosylation. This is more likely than the possibility of negative determinants for glycosylation in the N-terminal region of Kv1.1 because chimera 16 (Kv1.1 with the 1.4 pore region), which identified the outer pore region as a major determinant for these processes, exhibited both efficient glycosylation conversion and high cell surface expression.
Possible Role of Determinants in Affecting Steady-state Levels of Kv1 Channels in the Plasma Membrane-It appears that there are multiple determinants on both Kv1.4 and Kv1.1 that are involved in inhibiting or stimulating trans-Golgi glycosylation conversion and cell surface expression of these very closely related channels. These determinants may positively or negatively affect the folding, oligomerization, and stability of a homomeric or heteromeric protein complex, and/or they may associate transiently with non-Kv1 proteins involved in these processes in cellular organelles. Determinants on wt Kv1.1 or mutant constructs that cause partial ER retention may influence intrasubunit folding and/or intersubunit dimerization processes, and if inefficient oligomerization occurs, then chaperone-mediated retention may take place, or they may interact directly with the ER processing machinery through specific amino acids. Because Kv1 channels do not contain canonical ER retention or retrieval signals signals (e.g. RXR(R), KDEL) it appears more likely that high ER retention of a given subunit used in this study is due to inefficient folding or insufficient stability causing chaperone-mediated partial retention. Furthermore, Kv1.4 does not contain the putative ER export signals that have been mapped on Kir-type K ϩ channels (32). In contrast, the positive determinants for trans-Golgi glycosylation on Kv1.4 outer pore regions may exert their influence by interacting directly with a molecule(s) in the processing machinery pathway that promotes glyco-processing or by not interacting with a molecule(s) in the pathway that inhibits glyco-processing. Negative determinants for trans-Golgi glycosylation in the outer pore region of Kv1.1 may act by either of these modes. These negative determinants may be involved in preventing a channel from coming in efficient contact with the Golgi glyco-processing enzymes, possibly due to the channel tetramer moving through the Golgi in a lipid subdomain that avoids them.
Kv1.1 is efficiently trans-Golgi glycosylated in native brain tissue. This finding and the work in this report suggest that Kv1.1 must be expressed as a heteromer in brain with other Kv1 and Kv␤-subunits in an unknown stoichiometry that leads to its efficient trans-Golgi glycosylation and cell surface expression. Kv1.1 may be preferentially retained in brain neurons in the ER due to its outer pore a and b regions, which appear to act like a "partial retention signal," until other Kv1 subunits are coexpressed with it, which would give robust cell surface expression of the heteromer. Thus, Kv1 cell surface expression and steady-state levels can be influenced by the channel's subunit composition. The determinants mapped in this study are important in understanding the amino acids involved in proper channel folding, efficient trans-Golgi glycosylation conversion, and high cell surface expression. Furthermore, we have suggested that Kv1 channels with truncated carbohydrate trees can affect their functional properties (15). Thus, the expression level of a channel in the plasma membrane and its functional properties may be affected by these determinants and contribute to the functional diversity of K ϩ channels in different tissue.